FX 8350 or i7 4770k

[ANIR0X2K00L]

I need help with this topic, i am getting a new PC and i can wait till then end of this year. The problem is that i will have to keep this PC for the next 5 to 6 years. I can spend extra to get a 4770k when it release but from what i have heard is that next gen games would run better on AMD hardware and will require more cores. I play graphically extensive games like Battlefield 3, Crysis 2, Crysis 3. I will be getting battlefield 4 so this will give u an idea of what type of games i am playing and all of these games need a good cpu to run, especially crysis 3 and possible Battlefield 4. Please just tell me that should i go with fx 8350 or the i7 4770k (when it releases in june), or should i wait for amd's steamroller or intel's broadwell. And please even list a good motherboard like asus maximus or crosshair formula motherboard or something better. I have noticed one problem that the AMD chipset i.e. 990fx dont have all the latest features.

Any help is greatly appreciated, please understand that i dont want to make a mistake cause if i make one i will regret it for the next 5 to 6 years.

 

[juanrga]

And the big plus of consoles are that they don't really need an OS, just a firmware. yes with all that social media shit going on atm, there might be more background processes, still not enough to justify designating a whole core just for that.

The PS4 includes an extra ARM CPU for dealing with social stuff, downloads, and other background tasks.

 

[Guest]

Still denying that Intel has 3D transistors? Sad, really. When AMD is bought bought by Intel in 5 years, will you still buy their products? Well, you convinced me. I'm going to spend my next paycheck on an AMD system. Then, I'm going to go on every thread about Intel vs AMD and tell the world how my PC is the best, even if it is only a .05% difference in only one game (Crysis 3, obviously).

Keep acting like you work as a spokesperson for AMD while nobody anywhere cares or is even aware of your existence. Also, try putting the mouse down once in a while and emerge from mommy's basement and get some sunlight.

This forum is about helping the OP. We helped the OP. He decided to buy a Haswell. The thread should be closed. Anything else you or I say is irrelevant. Now, go start your own thread and fill it with as much AMD propaganda as you want. You are still not going to convince the 9 out of 10 people who buy Intel PCs to go AMD. You are just going to start a circle jerk with the 1 out of 10 that are too poor to afford Intel and buy AMD. Period. End of story. Have a nice life. I don't need to prove anything to anyone. Oh, and Intel IB CPUs have 3D aka tri-gate transistors. http://en.wikipedia.org/wiki/3D_transistor#Tri-gate_transistors A FinFET is a double gate transistor, http://en.wikipedia.org/wiki/3D_transistor#FinFETs

 

[Guest]

When you are done beating it to printouts of your CPU benchmark on Crysis 3 in 720p, take a look at a few articles on Intel's non-existent tri-gate transistor technology:

https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&cad=rja&ved=0CCwQFjAA&url=http%3A%2F%2Fwww.bbc.co.uk%2Fnews%2Ftechnology-17785464&ei=QpWRUaKSOY3wqAGBhoHgBw&usg=AFQjCNGD10KVqFJiZHjlzkqhVZL8F2aLwg&sig2=_ueTaSkg38G7UazHkWcU2A&bvm=bv.46471029,d.aWM

https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=10&cad=rja&ved=0CJQBEBYwCQ&url=http%3A%2F%2Fblog.chron.com%2Ftechblog%2F2012%2F04%2Fintels-ivy-bridge-chips-have-3d-transistors-more-muscle-for-less-power%2F&ei=QpWRUaKSOY3wqAGBhoHgBw&usg=AFQjCNFpoO8mpWCF9Ll9oNYmLVYrshhM2g&sig2=Bs0Is5p4fs6KZc6udmyQ-A&bvm=bv.46471029,d.aWM

https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=11&cad=rja&ved=0CCwQFjAAOAo&url=http%3A%2F%2Fappleinsider.com%2Farticles%2F12%2F04%2F23%2Fintel_launches_next_gen_ivy_bridge_processors_with_3d_transistors&ei=bZWRUdvwHImvrQGw-YHIBA&usg=AFQjCNFowuWzSWYVryxKe8zy7RcPcBOx9A&sig2=ldTdkWOJOzO8jd12H7gloA&bvm=bv.46471029,d.aWM

Or, read the newegg description of the 3770k (and every other IB CPU) the first thing it mentions is: INTEL 22NM 3-D TRANSISTOR
The 3rd Generation Intel Core processors employ the world's first 3-D transistor manufactured at 22nm. Intel's 3-D Tri-Gate transistor uses three gates wrapped around the silicon channel in a 3-D structure, enabling an unprecedented combination of performance and energy efficiency, and ensuring that the pace of technology advancement consumers expect can continue for years to come.:

http://www.newegg.com/Product/Product.aspx?Item=N82E16819116501

or, just look at the Google results:

https://www.google.com/search?q=intel+3d+transistor&rlz=1C1ASUT_enUS532US532&aq=0&oq=Intel+3D+&aqs=chrome.1.57j0l3j62l2.3912j0&sourceid=chrome&ie=UTF-8#rlz=1C1ASUT_enUS532US532&q=intel+3d+transistor+technology&revid=872158344&sa=X&ei=bZWRUdvwHImvrQGw-YHIBA&ved=0CLQBENUCKAI4Cg&bav=on.2,or.r_cp.r_qf.&bvm=bv.46471029,d.aWM&fp=f4410b1b5d67c92a&biw=1366&bih=667

 

[elemein]

Still denying that Intel has 3D transistors? Sad, really.

Whose doing that? I don't see a single soul here doing that. Not at all.

 

[8350rocks]

Still denying that Intel has 3D transistors? Sad, really. When AMD is bought bought by Intel in 5 years, will you still buy their products? Well, you convinced me. I'm going to spend my next paycheck on an AMD system. Then, I'm going to go on every thread about Intel vs AMD and tell the world how my PC is the best, even if it is only a .05% difference in only one game (Crysis 3, obviously).

Keep acting like you work as a spokesperson for AMD while nobody anywhere cares or is even aware of your existence. Also, try putting the mouse down once in a while and emerge from mommy's basement and get some sunlight.

This forum is about helping the OP. We helped the OP. He decided to buy a Haswell. The thread should be closed. Anything else you or I say is irrelevant. Now, go start your own thread and fill it with as much AMD propaganda as you want. You are still not going to convince the 9 out of 10 people who buy Intel PCs to go AMD. You are just going to start a circle jerk with the 1 out of 10 that are too poor to afford Intel and buy AMD. Period. End of story. Have a nice life. I don't need to prove anything to anyone. Oh, and Intel IB CPUs have 3D aka tri-gate transistors. http://en.wikipedia.org/wiki/3D_transistor#Tri-gate_transistors A FinFET is a double gate transistor, http://en.wikipedia.org/wiki/3D_transistor#FinFETs

Tri-gate is not a 3D CPU...you're sadly mistaken...and gates are not dimensions...they're stacked on top of each other...

Additionally...it's not 9 out of 10...it's more like 6 or 7...and that number is growing smaller by the month. AMD gained 1.2% market share over the last 60 Days...doesn't sound like a lot...but 1.2% of a few hundred million...is a few million.

Keep on talking and not supporting your facts.

Tri-gate or 3-D Transistor (not to be confused with 3D microchips) fabrication is used by Intel Corporation for the nonplanar transistor architecture used in Ivy Bridge processors. These transistors employ a single gate stacked on top of two vertical gates allowing for essentially three times the surface area for electrons to travel

Taken straight from your Wiki. All they did was stack another gate on top of 2...they created more surface area. But it isn't as revolutionary as you make it sound.

From your same article:

Dual gate MOSFETs are commonly used in VHF mixers and in sensitive VHF front end amplifiers. They are available from manufacturers such as Motorola, NXP, and Hitachi

Dual gates have been around a long time...it's not that big a deal to add another.

Besides...AMD has been using SOI for quite some time, intel uses bulk wafers:

http://en.wikipedia.org/wiki/Silicon_on_Insulator

Intel has even conceded that going to SOI would be an improvement for them, but it is "too expensive".

http://www.eetimes.com/electronics-news/4374728/Intel-FinFETs-SOI-shrink-GSS

FYI: The OP hasn't bought haswell...it's not out yet.

@OP: Save yourself time effort and money...

 

[whyso]

-snip-

Again, this is gonna get messy as all hell... Just snipping out what I need to reply. Also, I'm very glad that we can keep this conversation very civil. Kudos to you for not having a masssssive e-peen.

LOL what, 3 generations behind? In mobile Arrandale->HD 3000 =100% better->Hd4000=40%better. Thats really one generations and Haswell is going to close the gap almost all the way (disregarding GT3).
3630QM runs the igp at 1150 mhz, i3-3110M runs it at 1100 mhz (but less cache). Its a sound win for AMD true but the difference is only so great for the a10, a8 with 2/3 the shaders is pretty much equal to the hd4000 and at ULV levels the HD4000 is surprisingly competitive.
Fair enough when comparing IGP generations. Also, Intel is definitely a very, very, strong competitor when you pit ULV Intel vs ULV AMD (A series and E series), the A series IGPs at ULV level are only a bit better than the HD4000 at low clocks, and the E series is actually beaten out. I give ULV Intel CPUs the ribbon for all-around power; the only thing that dissapoints me by them is their price. E series and A series APUs are availible at significantly lower price than Intel i3-3217U or i5-3317U-- but that's aside the point. I agree that Intel is overall, better in the ULV mobile sector. Fair point. However, it is NOT fair to say that Haswell will close the gap between the HD4000 and the 7660G. First and foremost, Haswell will have the HD4600, HD5000, HD 5200, and HD... Something else that'll be used in the desktop. There will be many levels, so it'll be hard to compare. But moreso, the real reason they're uncomparable is: Richland and Kabini. They'll bring along new IGPs that I'll confidently say: Will soundly beat Intel, yet again, in IGP power. I'm confident in it.

Depends really. Gt3e will have no competition in mobile (but will be utterly too expensive to matter to most consumers). Kabini is low power and may match or slightly beat ULV ivy in gpu performance (not a chance for cpu performance). Kabini has a great price however and should be more than sufficient for most consumer uses.

The review also showed power consumption which was pretty much equal. Which was why I picked that comparison. Atom and bobcat compete quite well in straight cpu performance. Bobcat's gpu kills atom though.
When just comparing power consumptions, sure, they do get a little bit closer. But if you shove aside power consumption and CPU power, the E-series is significantly better at well... Everything else. Price, IGP, availibility, everything else really.

Agree. Atom was quite the failure.


Yeah I know this its just that given all the problems devs are having now more work is not what they need. xbox 360 is tricore (dual threaded) and PS3 is one main processing unit and 6 extras.
I feel this statement isn't really a good argument. While it is true that it takes more time to write multicore software, the fact of the matter is: It'll surely happen, no doubt. PS4 and Xbox 720 developers will fight tooth and nail to squeeze every last Hz out of the IGP and CPU, laziness will be a factor, but if laziness were as a big of a deal as you say, every single console (and computer for that matter) would be a single core or a dual core. Nothing more.

Just saying that the initial games will not have much "magic sauce" but later ones will. No one wants to have 8 cores at 1.6 ghz vs 4 at 3.2 to program for. It takes time and time is money. For a given budget I'd rather have less time and money at optimizing and coding for 8 cores vs time spent on design.

 

[elemein]

Just saying that the initial games will not have much "magic sauce" but later ones will. No one wants to have 8 cores at 1.6 ghz vs 4 at 3.2 to program for. It takes time and time is money. For a given budget I'd rather have less time and money at optimizing and coding for 8 cores vs time spent on design.

Later ones will is true.

Though, the statement about people would rather have 4 cores at 3.2 than 8 cores at 1.6 isn't exactly true. I myself find situations where I'd rather have 2 processors of even half the speed than a fast single processor. Now, my situation may be a little unique as almost 75% of my programming goes towards programming robots for hobby and for school, but it's certainly true that even in a x86, computer environment, while writing my software, I'd rather use two cores rather than one. Even if it takes a little more time, the time saved by having two cores to work with rather than one saves a heck of a lot more time than if I had to figure out a work around for a certain operation to work on a single core.

For example; what if I want one operation to not effect another; an operation that is timed. What if I want an action to be performed, while a sound is played, but have the two correspond to eachother. For example, what if I want an action to be checked for while having a sound played in the background? Sure, a single fast processor is probably fast enough to encode both sound and check actions in unison, but what if it were two, much, much more complicated tasks? Sometimes, two slower cores is simply better than one faster one.

 

[whyso]

You have a bad habit of replying without formatting adequately the message mixing up all. Ok, I will do the same when replying to you.

Sorry about that, the post was never meant to be replied to in that way.

It is not exactly true that one intel core is much stronger than one FX core. You repeat this very often, thus let me say that if the cores are working in 'isolation' one of others that is rather correct, but when the cores are working together as a whole then each intel cores lost performance due to poor scaling. 8350rocks wrote some useful analogies. I recommend you to read them.

An intel core is stronger than any FX core in any task. This is a reference to any singlethread task or any task where we are comparing EQUAL Numbers of cores. Your point makes no sense. More cores equals the same or weaker scaling (this is why given the same performance any programmer wants fewer cores). I'm not talking about HT here but core for core performance.

Either you did not understand what I said or you just ignored.

I may not be understanding your logic (or grammer and sentance structure) at all here
I don't see any performance lost from bad scaling on intel chips. Intel like AMD scales near 100% on physical cores. HT does not scale 100% (more like 20%) but it was never designed or meant to so no performance is being 'lost' (rather we are gaining less than we may like).

The bad scaling of Intel chips is also the reason why Xeon based chips are/were unpopular on supercomputers. The most powerful supercomputer Titan uses AMD Opterons.

I think thats because they signed a contract with AMD.

If you believe that AMD did put a gun in the head of every supercomputer maker of the top 5 to avoid superb Intel technology then you are more foolish than I believed

I love how you post completely wrong information when it is so easily available.

Here is the TOP500 list (as of november 2012-titan was born)

http://www.top500.org/list/2012/11/

Looking at the top 20 systems

#1 -Titan is AMD Opterons, #18,19 also use opterons. All of these are CRAY.

# 2,4,5,9,10,16 are IBM type cpus

#3 is something different

# 6,7,8,11,12,13,14,15,17,20 use Xeon cpus.

In any case your single core argument does not invalidates my point. I see a CPU as a whole, which has been designed as a whole.

HT is not magic sauce. I said that does not work for a kind of games. And you ignore that with your claim. Why do not explain your invalid point to someone who needs to disable HT because of the regression on performance. No wait... all of this was explained to you before and you were given links showing how HT can increase or decrease performance depending of the specific situation.

Rarely does HT decrease Performance. It can happen but rarely will you see a 3770k losing to a 3570k outside of some rarely used test.

Most games show some improvement from HT

This is not true and this is because most people recommend an i5 over an i7.

Due to the fact that most games do not use more than 4 threads. An i3 is recommended over a pentium precisely because it has HT.

Crysis 3

Near doubling of performance from the pentium to the i3. Yes we are comparing SB vs IB and 3.3 ghz vs 3.0 ghz but that can't explain a near doubling of average fps.

There are other examples too. Most of the time the advantage is slight but still there. There are games where the i7 performs better than the i5 because of HT.

According to the review cited Haswell 4770k gives 1% more average performance in games, but performance regressions of up to 5% in some specific games, whereas increases power consumption by about 10%. Adds the usb3 issues and the PSU incompatibilities and you will start to understand why some people here is calling it "failwell", hasbeen",...

On a preproduction sample. Given final firmware things may change.

That review is GPU bottlenecked. Is it gaming power consumption? (Don't think so)

GPU bottleneck does not introduce performance regressions. No it is not gaming power consumption but ordinary power consumption measurement as in other reviews.

Anyone familiar with CPUs knows the power consumption argument used by anti-AMD people. The same people is now trying to dismiss power consumptions as irrelevant, because Haswell increases the consumption. LOL

Funny that you appeal to mobile space and give power consumptions of desktop processors. More LOL

Yes, because if there power consumption (efficiency) sucks on the desktop its going to suck in mobile. Mobile is difficult to compare because all systems are different so it makes sense to compare the architecture using desktop numbers. If you look at the benchmarks you will see that a 17 watt ivy i7 ULV competes with and generally beats the 35 watt a10-4600m in straight cpu performance.

No. By your own logic mobile Haswell will be less efficient than ivi because desktop is less efficient, but that logic is wrong.

Except its too early to tell (and seriously you are trying to explain a 96 watt difference between the 3770k and the 8350 by motherboards, etc and suddenly a 15 watt difference is huge?). Intel Burn test is going to use all the power it can get. It is not indicative of power consumption while running an application. (A couple bios changes can have a large effect too).

Those 17 watts are not counting the superior gpu of the a10... and beats depending of what benchmarks you are using. The i7 will beat with intel-oriented benchmarks, sure.

There is no benchmark (non graphical) where an a10 beats an i7 ivy. Sure the a10 has a better gpu but I said CPU performance.

The figures that you give are from a review where they used any possible way to increase the power consumption of AMD chips whereas cuting down its performance. Time ago since I analyzed their review, but I recall perfectly those figures, because is not the first time someone gives them to me. Last time was a guy who is very well-known for its intel bias and fantastic claims: he sures to anyone who he meets that his FX consumes 200W more than his i7, which is nonsense.

I am writing this from memory and maybe some detail is wrong, but overall it is right. This is for the FX-8350 and the i7-3770k.

They selected the more power hungry mobo (Asus) for the AMD chips and the more power efficient mobo (MSI) for Intel. Even at iddle the difference between the Asus AM3+ vs the MSI AM3+ was of about 10W and about the same for Asus 1155 vs the MSI 1155. Selecting power saving setting on BIOS the difference between the Asus AM3+ vs the MSI AM3+ was larger still. THe review does not explain that conf they used. But I would not be surprised if they selected power saving for Intel chips but not for AMD ones.

This may be true but there is still a very large power gap between ivy bridge and FX. There is a 96 watt spread between the 3770k and the 8350. They are on the balanced profile (speedstep and cool and quiet are working)

Comparing like to to like (e.g. Asus mobo to Asus mobo) reduces 13 W and considering a 7% due to the FX hotfixes and the 96 W reduce to 77 W, that is a 20 W gain by doing a more fair review! Of course, there are reviews using other configs that reduce the gap to the 47 W e.g. Bit Tech review.

The 990X chips has a higher tdp than the Z77 chip. 990FX TDP is 19.6 watts. 990X tdp is 14 watts. Z77 is 6.7 watts. So that difference (13 watts) can't really be completely eliminated. (The power consumption problem is largly fixed under windows 8).

http://www.tomshardware.com/reviews/windows-8-bulldozer-performance,3289-7.html

They selected a high-performance G.Skill memory kit for Intel, but an "Entertainment" Patriot kit for AMD. Moreover the AMD chip did run with memory under the stock speed by a 20% generatin some artificial performance gain for Intel.

How many times do I have to say that memory speed rarely matters for cpu performance (outside of a few things such as winrar). The tech report review used the same RAM for all the recent cpus.

For intel? Sure. For AMD FX and APUs? No.

Nope, not really

http://www.pugetsystems.com/labs/articles/Breaking-the-Hype-of-High-Frequency-RAM-142/page3

They used W7 and installed the FX hotfixes manually. As shown by toms the hotfixes introduced performance regression for AMD chips (2% on average, but larger for video tasks). Moreover, the patches had a bug that affected the power consumption of FX chips, because blocked some power saving states. According to Toms the hotfixes increased power consumption of FX chips by a 7% average.

Why then do so many people argue that the hotfixes are required?

One thing are the automatic hotfixes and those included in W8 and other thing the hotfixes manually installed. Those are what generated problems with performance and power consumption.

So it looks like in win 8 there is no more power problems.

That is how they managed to achieve the FX to run slower whereas consuming more power. Time ago I did an estimation of the real power consumption and task energy and if my memory does not fail the FX moved from 8.1 to near 6.0.

No professional review has ever seen FX compete with Ivy or even sandy bridge on efficiency.

Nobody said that give the same efficiency. Read what has been said.

You realize that task energy is the same thing as efficiency? What amount of power is required to to a certain amount of work? If we had a 6 then that would be the same as sandy bridge (and hence the same efficiency).

I don't recall now the setups of the 3960X, but was not using memory overclocked by a 40% for increase its performance?
They are all running the same speed and timings ram with the exception of the first gen core i systems. Furthermore any memory overclock would have an insignificant effect on performance in that test (0-2%)

I revised and yes they used stock speed, but improved performance of the intel chip by selecting a quad channel configuration.

It runs stock with 4 channels. Its meant to be used with 4 channels. You can hardly cry foul about RAM speed between ivy and piledriver and then complain about SB-E quad channel RAM. Again there is no to very little difference.

3960X -http://www.hardwarecanucks.com/forum/hardware-canucks-reviews/48210-intel-sandy-bridge-e-core-i7-3960x-cpu-review-8.html

Only winrar seem to be affected by it.

 

[davschall]

If you want the fastest, the Ferrari if you will, and the most future proof

Wow triple post jman really, what to good for the edit button? Also I seriously hate when people who know nothing about cars make references to them, I got some dumbass HERE comparing a corvette to a bmw M5 I mean talk about apples to oranges. Just enough with the car references, especially when you consider the fastest car to be a ferrari, no a specific one such as a ferrari california (notably slow) just a ferrari. Im not gonna get into the rest of the post cause its just little nerd rages.

 

[davschall]

Still denying that Intel has 3D transistors? Sad, really. When AMD is bought bought by Intel in 5 years, will you still buy their products? Well, you convinced me. I'm going to spend my next paycheck on an AMD system. Then, I'm going to go on every thread about Intel vs AMD and tell the world how my PC is the best, even if it is only a .05% difference in only one game (Crysis 3, obviously).

Keep acting like you work as a spokesperson for AMD while nobody anywhere cares or is even aware of your existence. Also, try putting the mouse down once in a while and emerge from mommy's basement and get some sunlight.

This forum is about helping the OP. We helped the OP. He decided to buy a Haswell. The thread should be closed. Anything else you or I say is irrelevant. Now, go start your own thread and fill it with as much AMD propaganda as you want. You are still not going to convince the 9 out of 10 people who buy Intel PCs to go AMD. You are just going to start a circle jerk with the 1 out of 10 that are too poor to afford Intel and buy AMD. Period. End of story. Have a nice life. I don't need to prove anything to anyone. Oh, and Intel IB CPUs have 3D aka tri-gate transistors. http://en.wikipedia.org/wiki/3D_transistor#Tri-gate_transistors A FinFET is a double gate transistor, http://en.wikipedia.org/wiki/3D_transistor#FinFETs

Tri-gate is not a 3D CPU...you're sadly mistaken...and gates are not dimensions...they're stacked on top of each other...

Additionally...it's not 9 out of 10...it's more like 6 or 7...and that number is growing smaller by the month. AMD gained 1.2% market share over the last 60 Days...doesn't sound like a lot...but 1.2% of a few hundred million...is a few million.

Keep on talking and not supporting your facts.

Tri-gate or 3-D Transistor (not to be confused with 3D microchips) fabrication is used by Intel Corporation for the nonplanar transistor architecture used in Ivy Bridge processors. These transistors employ a single gate stacked on top of two vertical gates allowing for essentially three times the surface area for electrons to travel

Taken straight from your Wiki. All they did was stack another gate on top of 2...they created more surface area. But it isn't as revolutionary as you make it sound.

From your same article:

Dual gate MOSFETs are commonly used in VHF mixers and in sensitive VHF front end amplifiers. They are available from manufacturers such as Motorola, NXP, and Hitachi

Dual gates have been around a long time...it's not that big a deal to add another.

Besides...AMD has been using SOI for quite some time, intel uses bulk wafers:

http://en.wikipedia.org/wiki/Silicon_on_Insulator

Intel has even conceded that going to SOI would be an improvement for them, but it is "too expensive".

http://www.eetimes.com/electronics-news/4374728/Intel-FinFETs-SOI-shrink-GSS

FYI: The OP hasn't bought haswell...it's not out yet.

@OP: Save yourself time effort and money...

Dude you convinced me lol, I decided that my kick ass gigabyte 990fxa-ud5, was worth keeping around, ive seen to much to be on the fence between them now, as im already on an upgrade path of AMD ill just stick with them, lol im not full fanboy yet though...never go full fanyboy, something that rds1220 could learn lol. I mean seriously Warning angry person alert So my advice stop

 

[SomeNickname]

Still denying that Intel has 3D transistors? Sad, really. When AMD is bought bought by Intel in 5 years, will you still buy their products?

Learn to economics. When AMD is off the table, Intel could just double every single price, because why not? There's no competition left. There is not a single consumer to be happy about that.
AMD will stay for a long time. They have been called dead often enough, but end the end they have their market (Good performance CPU at excellent price-points), so why should they go bankrupt/let them sell themselves?

Keep acting like you work as a spokesperson for AMD while nobody anywhere cares or is even aware of your existence. Also, try putting the mouse down once in a while and emerge from mommy's basement and get some sunlight.

*slow clap* Welcome to the bottom of the discussion pyramid.

I don't need to prove anything to anyone.

That's good, cause you don't. All you do is calling out some fancy buzzwords about Intels architecture, but you never actually explain why that stuff will show more benefit in future games than more cores. We already can see the benefit in current benchmarks, is there any reason to belive that benefit will grow bigger at the same CPU with future games?


Again if we want to help the OP we have to discuss raw power. Not the design decisions of the architecture (except you have a argument why that will actually have impact aside from current benchmarks), not power consumption, not the net worths of the companies. Just the current performance and some speculations about benefits in future games.

 

[iam2thecrowe]

wow, how has this thread continued so long with so much bullsh!t?
Please 8350 fanboys, stop recommending the 8350 as a superior gaming cpu to intel i5 equivelants. It is not, as has been proven in 95% of the review sites out there. Throwing around links to the remaining 5% that show otherwise is rediculous, you are naïve if you believe them over the 95% that show the opposite.

As for speculation about ps4 with its highly threadedness, well no one knows how that will pan out, it will probably allow amd to catch up a bit in gaming, keep in mind that although the 8 core 8350 has 8 integer cores, it only has 4 FPU cores, which is what games use the most. the fact that ps4 is able to handle a lot of threads at once using the cpu and gpu together is good, but when it does, this it takes away graphics processing power. The analogy of you pull one lever too hard and another lever goes flying in the other direction comes to mind. It can be well balanced in the ps3 but with so many hardware variations in a pc I doubt we will see the same optimizations, game devs will still have to play it safe as they always have done with ports.

 

[logainofhades]

That "nearly perfect gaming PC" is just that, a console that you can also browse the web with. That is why people who use a computer to do actual work on it would never even consider any of AMD's products. If you want the perfect child's toy to play children's games, one that a child can afford, then buy the AMD. If you want a grown up computer that, by the way, also plays the same games just as good and in most cases better (you know, like resolutions over 720p which is what the first chart is in) and is much more versatile, you might consider the adult version of the PC brought to you by Intel. There is a reason why high end computers and laptops such as Apple use Intel chips and the bargain bin ones use AMD: quality and performance. You can dig up obscure benchmarks all day where the settings are just right on the one or two games that the AMD chip wins, and not by very much. In real life, those charts mean nothing. In real life, nobody is going to run the game in 720p, turn off all the AA and turn down their settings just so they can say their AMD CPU is better than the 3rd best consumer Intel CPU. Why did I buy an Intel? Well, because the last AMD CPU I bought CAUGHT ON FIRE. Instead of redesigning the architecture, they just keep throwing more cores in the chip and jacking up the megahertz which turns it into a furnace. Intel actually redesigns their chips. Look at the 3D transistor tech they introduced in Ivy Bridge. That is one of the reasons why core for core, AMD is not only slower, they are almost a half decade behind. The current 8350 CPU single core benchmark has about the same power and efficiency of an i7-9xx "Bloomfield" CPU which came out in 2008 AND has a clock speed about 1 GHz lower than the 8350. If you overclock a 960 to the stock speed of an 8350, it will humiliate the 8350 core for core and probably beat it in overall performance as well. Why are the new consoles using AMD? They are cheap. The economy is bad. The only way that Sony and MS can make a profit is by using inferior parts. And, the best part, there is no advantage in owning an AMD PC when the ports start coming out from next gen consoles to PC. They all are using x86 programming architecture. The internal design might be different (with AMD being far inferior) but they use an identical instruction set and the new consoles will use the same as well. How being made by AMD is an advantage makes no sense. You have to be very gullible to believe that one.

This post makes me want to post that Billy Madison youtube link again. You are nothing but an Intel fanboy. Your kids vs adult analogy makes that quite obvious. I have owned plenty of both AMD and Intel and not one "caught on fire". I have had a PSU die and zap an unlocked Sempron 145 but it didn't burn. Most of my systems are Intel now, but still have a dual core PhII rig that is used for netflix and streaming video off my file server to the TV. Sometimes it gets used as a WoW rig, when needed, if I have another system that is down for whatever reason. I have an i5 3570k, i5 2400, i5 750 and my laptop has a dual core i5. I was given the 2400 and the 750 was bought used from a person I know on another forum, with a motherboard, for about the cost of what the CPU was new at the time.

Had the Vishera FX's been around when I needed a new CPU and board last year, I probably would have bought one over the 3570k I have now. Would have had similar performance for like half the cost @ Microcenter. I fully admit I am a cheapskate. I don't see any value in spending more on hardware that doesn't give me a big benefit in what I do. All those high and pretty numbers look great. At the end of the day, you can't see over 60fps anyway.

To the OP, 4770k will be a nice chip, without a doubt. It will be quite capable for a good while. It isn't much faster than a 3770k though. An FX 8350 is also a very capable chip for much less money. I would go with the 8350 and spend the $130 saved on a better graphics card. 3770k isn't going to game better with an inferior graphics card than an FX 8350 with a much faster card. $130 more to spend on a graphics card is quite a bit and will give you a much better bang for the buck in GPU intensive titles.

 

[8350rocks]

wow, how has this thread continued so long with so much bullsh!t?
Please 8350 fanboys, stop recommending the 8350 as a superior gaming cpu to intel i5 equivelants. It is not, as has been proven in 95% of the review sites out there. Throwing around links to the remaining 5% that show otherwise is rediculous, you are naïve if you believe them over the 95% that show the opposite.

As for speculation about ps4 with its highly threadedness, well no one knows how that will pan out, it will probably allow amd to catch up a bit in gaming, keep in mind that although the 8 core 8350 has 8 integer cores, it only has 4 FPU cores, which is what games use the most. the fact that ps4 is able to handle a lot of threads at once using the cpu and gpu together is good, but when it does, this it takes away graphics processing power. The analogy of you pull one lever too hard and another lever goes flying in the other direction comes to mind. It can be well balanced in the ps3 but with so many hardware variations in a pc I doubt we will see the same optimizations, game devs will still have to play it safe as they always have done with ports.

Actually, know your facts, the 8 core Jaguar APU in PS4/XBOX720 has 8 FPUs.

Additionally show me where "95%" of games are better on an i5 then the FX8350. Don't count Skyrim, Civ5 or SC2...those were singled out already. If you bring up planetside 2 I will have a good hardy laugh at that one too.

Additionally...the Margin for Error on Benchmarks for gaming is approximately 10% for hardware inconsistencies. If you can produce even 10 different games where the difference is greater than 10% between the 2...I will be amazed.

So, the ball is in your court...

 

[juanrga]

You have a bad habit of replying without formatting adequately the message mixing up all. Ok, I will do the same when replying to you.

Sorry about that, the post was never meant to be replied to in that way.

It is not exactly true that one intel core is much stronger than one FX core. You repeat this very often, thus let me say that if the cores are working in 'isolation' one of others that is rather correct, but when the cores are working together as a whole then each intel cores lost performance due to poor scaling. 8350rocks wrote some useful analogies. I recommend you to read them.

An intel core is stronger than any FX core in any task. This is a reference to any singlethread task or any task where we are comparing EQUAL Numbers of cores. Your point makes no sense. More cores equals the same or weaker scaling (this is why given the same performance any programmer wants fewer cores). I'm not talking about HT here but core for core performance.

Either you did not understand what I said or you just ignored.

I may not be understanding your logic (or grammer and sentance structure) at all here
I don't see any performance lost from bad scaling on intel chips. Intel like AMD scales near 100% on physical cores. HT does not scale 100% (more like 20%) but it was never designed or meant to so no performance is being 'lost' (rather we are gaining less than we may like).

The bad scaling of Intel chips is also the reason why Xeon based chips are/were unpopular on supercomputers. The most powerful supercomputer Titan uses AMD Opterons.

I think thats because they signed a contract with AMD.

If you believe that AMD did put a gun in the head of every supercomputer maker of the top 5 to avoid superb Intel technology then you are more foolish than I believed

I love how you post completely wrong information when it is so easily available.

Here is the TOP500 list (as of november 2012-titan was born)

http://www.top500.org/list/2012/11/

Looking at the top 20 systems

#1 -Titan is AMD Opterons, #18,19 also use opterons. All of these are CRAY.

# 2,4,5,9,10,16 are IBM type cpus

#3 is something different

# 6,7,8,11,12,13,14,15,17,20 use Xeon cpus.

And you prove my point. You have to go to the #6 to find a supercomputer with Intel Xeon chips. And that supercomputer offers a tenth of the performance of the #1 using Opteron chips.

And you can find lots of supercomputer from Cray and from other companies using Opteron chips. Wiki has a list

http://en.wikipedia.org/wiki/Opteron#Supercomputers

In any case your single core argument does not invalidates my point. I see a CPU as a whole, which has been designed as a whole.

HT is not magic sauce. I said that does not work for a kind of games. And you ignore that with your claim. Why do not explain your invalid point to someone who needs to disable HT because of the regression on performance. No wait... all of this was explained to you before and you were given links showing how HT can increase or decrease performance depending of the specific situation.

Rarely does HT decrease Performance. It can happen but rarely will you see a 3770k losing to a 3570k outside of some rarely used test.

Most games show some improvement from HT

This is not true and this is because most people recommend an i5 over an i7.

Due to the fact that most games do not use more than 4 threads. An i3 is recommended over a pentium precisely because it has HT.

Crysis 3

Near doubling of performance from the pentium to the i3. Yes we are comparing SB vs IB and 3.3 ghz vs 3.0 ghz but that can't explain a near doubling of average fps.

There are other examples too. Most of the time the advantage is slight but still there. There are games where the i7 performs better than the i5 because of HT.

http://www.thegeeklynews.com/2011/07/the-myth-of-hyper-threading-and-the-intel-i7-cpu/

http://www.tomshardware.co.uk/forum/320565-28-hyperthreading-gaming-difference

Some quotes from the second link:

Zero difference in nearly all games.

so HT is more or less useless

nope.

I am going with the popular opinion that HT technology in games is probably unnoticeable and at times going to have a negitive impact more than a positive impact.

According to the review cited Haswell 4770k gives 1% more average performance in games, but performance regressions of up to 5% in some specific games, whereas increases power consumption by about 10%. Adds the usb3 issues and the PSU incompatibilities and you will start to understand why some people here is calling it "failwell", hasbeen",...

On a preproduction sample. Given final firmware things may change.

No preproduction but official review of final silicon.

That review is GPU bottlenecked. Is it gaming power consumption? (Don't think so)

GPU bottleneck does not introduce performance regressions. No it is not gaming power consumption but ordinary power consumption measurement as in other reviews.

Anyone familiar with CPUs knows the power consumption argument used by anti-AMD people. The same people is now trying to dismiss power consumptions as irrelevant, because Haswell increases the consumption. LOL

Funny that you appeal to mobile space and give power consumptions of desktop processors. More LOL

Yes, because if there power consumption (efficiency) sucks on the desktop its going to suck in mobile. Mobile is difficult to compare because all systems are different so it makes sense to compare the architecture using desktop numbers. If you look at the benchmarks you will see that a 17 watt ivy i7 ULV competes with and generally beats the 35 watt a10-4600m in straight cpu performance.

No. By your own logic mobile Haswell will be less efficient than ivi because desktop is less efficient, but that logic is wrong.

Except its too early to tell (and seriously you are trying to explain a 96 watt difference between the 3770k and the 8350 by motherboards, etc and suddenly a 15 watt difference is huge?). Intel Burn test is going to use all the power it can get. It is not indicative of power consumption while running an application. (A couple bios changes can have a large effect too).

Power consumption issues were noticed earlier, Intel has increased the TPD and higher power consumption have been measured in i7-4770k

The 96 W difference was on about 200W and it was unreal. The realistic is about 70 W. The 15 W are on a minor load and are on a chip that promised better perf per watt.

Those 17 watts are not counting the superior gpu of the a10... and beats depending of what benchmarks you are using. The i7 will beat with intel-oriented benchmarks, sure.

There is no benchmark (non graphical) where an a10 beats an i7 ivy. Sure the a10 has a better gpu but I said CPU performance.

But was not the point. The point was that you compared cpu performance and alluded to total TDP without mentioning that part of the higher TDP is because AMD has a much more powerful gpu in the chip.

The figures that you give are from a review where they used any possible way to increase the power consumption of AMD chips whereas cuting down its performance. Time ago since I analyzed their review, but I recall perfectly those figures, because is not the first time someone gives them to me. Last time was a guy who is very well-known for its intel bias and fantastic claims: he sures to anyone who he meets that his FX consumes 200W more than his i7, which is nonsense.

I am writing this from memory and maybe some detail is wrong, but overall it is right. This is for the FX-8350 and the i7-3770k.

They selected the more power hungry mobo (Asus) for the AMD chips and the more power efficient mobo (MSI) for Intel. Even at iddle the difference between the Asus AM3+ vs the MSI AM3+ was of about 10W and about the same for Asus 1155 vs the MSI 1155. Selecting power saving setting on BIOS the difference between the Asus AM3+ vs the MSI AM3+ was larger still. THe review does not explain that conf they used. But I would not be surprised if they selected power saving for Intel chips but not for AMD ones.

This may be true but there is still a very large power gap between ivy bridge and FX. There is a 96 watt spread between the 3770k and the 8350. They are on the balanced profile (speedstep and cool and quiet are working)

Comparing like to to like (e.g. Asus mobo to Asus mobo) reduces 13 W and considering a 7% due to the FX hotfixes and the 96 W reduce to 77 W, that is a 20 W gain by doing a more fair review! Of course, there are reviews using other configs that reduce the gap to the 47 W e.g. Bit Tech review.

The 990X chips has a higher tdp than the Z77 chip. 990FX TDP is 19.6 watts. 990X tdp is 14 watts. Z77 is 6.7 watts. So that difference (13 watts) can't really be completely eliminated. (The power consumption problem is largly fixed under windows 8).

http://www.tomshardware.com/reviews/windows-8-bulldozer-performance,3289-7.html

This does not change anything of what I said.

They selected a high-performance G.Skill memory kit for Intel, but an "Entertainment" Patriot kit for AMD. Moreover the AMD chip did run with memory under the stock speed by a 20% generatin some artificial performance gain for Intel.

How many times do I have to say that memory speed rarely matters for cpu performance (outside of a few things such as winrar). The tech report review used the same RAM for all the recent cpus.

For intel? Sure. For AMD FX and APUs? No.

Nope, not really

http://www.pugetsystems.com/labs/articles/Breaking-the-Hype-of-High-Frequency-RAM-142/page3

Here a review gaming benchmark showing just the contrary (for Bulldozer)

http://www.madshrimps.be/files/users/leeghoofd/Zambezi/dividers/games.png

And for APUs the increase in performance can be superior to 20%

They used W7 and installed the FX hotfixes manually. As shown by toms the hotfixes introduced performance regression for AMD chips (2% on average, but larger for video tasks). Moreover, the patches had a bug that affected the power consumption of FX chips, because blocked some power saving states. According to Toms the hotfixes increased power consumption of FX chips by a 7% average.

Why then do so many people argue that the hotfixes are required?

One thing are the automatic hotfixes and those included in W8 and other thing the hotfixes manually installed. Those are what generated problems with performance and power consumption.

So it looks like in win 8 there is no more power problems.

That is how they managed to achieve the FX to run slower whereas consuming more power. Time ago I did an estimation of the real power consumption and task energy and if my memory does not fail the FX moved from 8.1 to near 6.0.

No professional review has ever seen FX compete with Ivy or even sandy bridge on efficiency.

Nobody said that give the same efficiency. Read what has been said.

You realize that task energy is the same thing as efficiency? What amount of power is required to to a certain amount of work? If we had a 6 then that would be the same as sandy bridge (and hence the same efficiency).

Just read what was said.

I don't recall now the setups of the 3960X, but was not using memory overclocked by a 40% for increase its performance?
They are all running the same speed and timings ram with the exception of the first gen core i systems. Furthermore any memory overclock would have an insignificant effect on performance in that test (0-2%)

I revised and yes they used stock speed, but improved performance of the intel chip by selecting a quad channel configuration.

It runs stock with 4 channels. Its meant to be used with 4 channels.

Also A10 or FX-8350 run stock @ 1866, but many reviewers force they to run lower memory speed.

 

[iam2thecrowe]

wow, how has this thread continued so long with so much bullsh!t?
Please 8350 fanboys, stop recommending the 8350 as a superior gaming cpu to intel i5 equivelants. It is not, as has been proven in 95% of the review sites out there. Throwing around links to the remaining 5% that show otherwise is rediculous, you are naïve if you believe them over the 95% that show the opposite.

As for speculation about ps4 with its highly threadedness, well no one knows how that will pan out, it will probably allow amd to catch up a bit in gaming, keep in mind that although the 8 core 8350 has 8 integer cores, it only has 4 FPU cores, which is what games use the most. the fact that ps4 is able to handle a lot of threads at once using the cpu and gpu together is good, but when it does, this it takes away graphics processing power. The analogy of you pull one lever too hard and another lever goes flying in the other direction comes to mind. It can be well balanced in the ps3 but with so many hardware variations in a pc I doubt we will see the same optimizations, game devs will still have to play it safe as they always have done with ports.

Actually, know your facts, the 8 core Jaguar APU in PS4/XBOX720 has 8 FPUs.

Additionally show me where "95%" of games are better on an i5 then the FX8350. Don't count Skyrim, Civ5 or SC2...those were singled out already. If you bring up planetside 2 I will have a good hardy laugh at that one too.

Additionally...the Margin for Error on Benchmarks for gaming is approximately 10% for hardware inconsistencies. If you can produce even 10 different games where the difference is greater than 10% between the 2...I will be amazed.

So, the ball is in your court...

10% margin of error....please....a 10% improvement is significant. I would give more like a 2% margin of error for most benchmarking.
didn't say 95% of games, said 95% of review sites.
http://techreport.com/review/23750/amd-fx-8350-processor-reviewed/9
http://www.guru3d.com/articles_pages/amd_fx_8350_processor_review,18.html
http://www.anandtech.com/show/6396/the-vishera-review-amd-fx8350-fx8320-fx6300-and-fx4300-tested/5
http://www.bit-tech.net/hardware/2012/11/06/amd-fx-8350-review/6
http://www.behardware.com/articles/880-13/amd-fx-8350-review-is-amd-back.html
http://hexus.net/tech/reviews/cpu/46985-amd-fx-8350/?page=5
http://www.xbitlabs.com/articles/cpu/display/fx-8350-8320-6300-4300_6.html#sect0
http://www.xbitlabs.com/articles/cpu/display/amd-fx-8350_6.html#sect0
http://www.tomshardware.com/reviews/far-cry-3-performance-benchmark,3379-7.html
http://www.guru3d.com/articles_pages/far_cry_3_graphics_performance_review_benchmark,7.html
http://www.tomshardware.com/reviews/crysis-3-performance-benchmark-gaming,3451-8.html
I will say in crisis 3 it doesn't do too shabby in average FPS, but dips to a low fps although that may have just been a spike.

 

[whyso]

You have a bad habit of replying without formatting adequately the message mixing up all. Ok, I will do the same when replying to you.

Sorry about that, the post was never meant to be replied to in that way.

It is not exactly true that one intel core is much stronger than one FX core. You repeat this very often, thus let me say that if the cores are working in 'isolation' one of others that is rather correct, but when the cores are working together as a whole then each intel cores lost performance due to poor scaling. 8350rocks wrote some useful analogies. I recommend you to read them.

An intel core is stronger than any FX core in any task. This is a reference to any singlethread task or any task where we are comparing EQUAL Numbers of cores. Your point makes no sense. More cores equals the same or weaker scaling (this is why given the same performance any programmer wants fewer cores). I'm not talking about HT here but core for core performance.

Either you did not understand what I said or you just ignored.

I may not be understanding your logic (or grammer and sentance structure) at all here
I don't see any performance lost from bad scaling on intel chips. Intel like AMD scales near 100% on physical cores. HT does not scale 100% (more like 20%) but it was never designed or meant to so no performance is being 'lost' (rather we are gaining less than we may like).

The bad scaling of Intel chips is also the reason why Xeon based chips are/were unpopular on supercomputers. The most powerful supercomputer Titan uses AMD Opterons.

I think thats because they signed a contract with AMD.

If you believe that AMD did put a gun in the head of every supercomputer maker of the top 5 to avoid superb Intel technology then you are more foolish than I believed

I love how you post completely wrong information when it is so easily available.

Here is the TOP500 list (as of november 2012-titan was born)

http://www.top500.org/list/2012/11/

Looking at the top 20 systems

#1 -Titan is AMD Opterons, #18,19 also use opterons. All of these are CRAY.

# 2,4,5,9,10,16 are IBM type cpus

#3 is something different

# 6,7,8,11,12,13,14,15,17,20 use Xeon cpus.

And you prove my point. You have to go to the #6 to find a supercomputer with Intel Xeon chips. And that supercomputer offers a tenth of the performance of the #1 using Opteron chips.

And you can find lots of supercomputer from Cray and from other companies using Opteron chips. Wiki has a list

http://en.wikipedia.org/wiki/Opteron#Supercomputers

In any case your single core argument does not invalidates my point. I see a CPU as a whole, which has been designed as a whole.

HT is not magic sauce. I said that does not work for a kind of games. And you ignore that with your claim. Why do not explain your invalid point to someone who needs to disable HT because of the regression on performance. No wait... all of this was explained to you before and you were given links showing how HT can increase or decrease performance depending of the specific situation.

Rarely does HT decrease Performance. It can happen but rarely will you see a 3770k losing to a 3570k outside of some rarely used test.

Most games show some improvement from HT

This is not true and this is because most people recommend an i5 over an i7.

Due to the fact that most games do not use more than 4 threads. An i3 is recommended over a pentium precisely because it has HT.

Crysis 3

Near doubling of performance from the pentium to the i3. Yes we are comparing SB vs IB and 3.3 ghz vs 3.0 ghz but that can't explain a near doubling of average fps.

There are other examples too. Most of the time the advantage is slight but still there. There are games where the i7 performs better than the i5 because of HT.

http://www.thegeeklynews.com/2011/07/the-myth-of-hyper-threading-and-the-intel-i7-cpu/

http://www.tomshardware.co.uk/forum/320565-28-hyperthreading-gaming-difference

Some quotes from the second link:

Zero difference in nearly all games.

so HT is more or less useless

nope.

I am going with the popular opinion that HT technology in games is probably unnoticeable and at times going to have a negitive impact more than a positive impact.

According to the review cited Haswell 4770k gives 1% more average performance in games, but performance regressions of up to 5% in some specific games, whereas increases power consumption by about 10%. Adds the usb3 issues and the PSU incompatibilities and you will start to understand why some people here is calling it "failwell", hasbeen",...

On a preproduction sample. Given final firmware things may change.

No preproduction but official review of final silicon.

That review is GPU bottlenecked. Is it gaming power consumption? (Don't think so)

GPU bottleneck does not introduce performance regressions. No it is not gaming power consumption but ordinary power consumption measurement as in other reviews.

Anyone familiar with CPUs knows the power consumption argument used by anti-AMD people. The same people is now trying to dismiss power consumptions as irrelevant, because Haswell increases the consumption. LOL

Funny that you appeal to mobile space and give power consumptions of desktop processors. More LOL

Yes, because if there power consumption (efficiency) sucks on the desktop its going to suck in mobile. Mobile is difficult to compare because all systems are different so it makes sense to compare the architecture using desktop numbers. If you look at the benchmarks you will see that a 17 watt ivy i7 ULV competes with and generally beats the 35 watt a10-4600m in straight cpu performance.

No. By your own logic mobile Haswell will be less efficient than ivi because desktop is less efficient, but that logic is wrong.

Except its too early to tell (and seriously you are trying to explain a 96 watt difference between the 3770k and the 8350 by motherboards, etc and suddenly a 15 watt difference is huge?). Intel Burn test is going to use all the power it can get. It is not indicative of power consumption while running an application. (A couple bios changes can have a large effect too).

Power consumption issues were noticed earlier, Intel has increased the TPD and higher power consumption have been measured in i7-4770k

The 96 W difference was on about 200W and it was unreal. The realistic is about 70 W. The 15 W are on a minor load and are on a chip that promised better perf per watt.

Those 17 watts are not counting the superior gpu of the a10... and beats depending of what benchmarks you are using. The i7 will beat with intel-oriented benchmarks, sure.

There is no benchmark (non graphical) where an a10 beats an i7 ivy. Sure the a10 has a better gpu but I said CPU performance.

But was not the point. The point was that you compared cpu performance and alluded to total TDP without mentioning that part of the higher TDP is because AMD has a much more powerful gpu in the chip.

The figures that you give are from a review where they used any possible way to increase the power consumption of AMD chips whereas cuting down its performance. Time ago since I analyzed their review, but I recall perfectly those figures, because is not the first time someone gives them to me. Last time was a guy who is very well-known for its intel bias and fantastic claims: he sures to anyone who he meets that his FX consumes 200W more than his i7, which is nonsense.

I am writing this from memory and maybe some detail is wrong, but overall it is right. This is for the FX-8350 and the i7-3770k.

They selected the more power hungry mobo (Asus) for the AMD chips and the more power efficient mobo (MSI) for Intel. Even at iddle the difference between the Asus AM3+ vs the MSI AM3+ was of about 10W and about the same for Asus 1155 vs the MSI 1155. Selecting power saving setting on BIOS the difference between the Asus AM3+ vs the MSI AM3+ was larger still. THe review does not explain that conf they used. But I would not be surprised if they selected power saving for Intel chips but not for AMD ones.

This may be true but there is still a very large power gap between ivy bridge and FX. There is a 96 watt spread between the 3770k and the 8350. They are on the balanced profile (speedstep and cool and quiet are working)

Comparing like to to like (e.g. Asus mobo to Asus mobo) reduces 13 W and considering a 7% due to the FX hotfixes and the 96 W reduce to 77 W, that is a 20 W gain by doing a more fair review! Of course, there are reviews using other configs that reduce the gap to the 47 W e.g. Bit Tech review.

The 990X chips has a higher tdp than the Z77 chip. 990FX TDP is 19.6 watts. 990X tdp is 14 watts. Z77 is 6.7 watts. So that difference (13 watts) can't really be completely eliminated. (The power consumption problem is largly fixed under windows 8).

http://www.tomshardware.com/reviews/windows-8-bulldozer-performance,3289-7.html

This does not change anything of what I said.

They selected a high-performance G.Skill memory kit for Intel, but an "Entertainment" Patriot kit for AMD. Moreover the AMD chip did run with memory under the stock speed by a 20% generatin some artificial performance gain for Intel.

How many times do I have to say that memory speed rarely matters for cpu performance (outside of a few things such as winrar). The tech report review used the same RAM for all the recent cpus.

For intel? Sure. For AMD FX and APUs? No.

Nope, not really

http://www.pugetsystems.com/labs/articles/Breaking-the-Hype-of-High-Frequency-RAM-142/page3

Here a review gaming benchmark showing just the contrary (for Bulldozer)

http://www.madshrimps.be/files/users/leeghoofd/Zambezi/dividers/games.png

And for APUs the increase in performance can be superior to 20%

They used W7 and installed the FX hotfixes manually. As shown by toms the hotfixes introduced performance regression for AMD chips (2% on average, but larger for video tasks). Moreover, the patches had a bug that affected the power consumption of FX chips, because blocked some power saving states. According to Toms the hotfixes increased power consumption of FX chips by a 7% average.

Why then do so many people argue that the hotfixes are required?

One thing are the automatic hotfixes and those included in W8 and other thing the hotfixes manually installed. Those are what generated problems with performance and power consumption.

So it looks like in win 8 there is no more power problems.

That is how they managed to achieve the FX to run slower whereas consuming more power. Time ago I did an estimation of the real power consumption and task energy and if my memory does not fail the FX moved from 8.1 to near 6.0.

No professional review has ever seen FX compete with Ivy or even sandy bridge on efficiency.

Nobody said that give the same efficiency. Read what has been said.

You realize that task energy is the same thing as efficiency? What amount of power is required to to a certain amount of work? If we had a 6 then that would be the same as sandy bridge (and hence the same efficiency).

Just read what was said.

I don't recall now the setups of the 3960X, but was not using memory overclocked by a 40% for increase its performance?
They are all running the same speed and timings ram with the exception of the first gen core i systems. Furthermore any memory overclock would have an insignificant effect on performance in that test (0-2%)

I revised and yes they used stock speed, but improved performance of the intel chip by selecting a quad channel configuration.

It runs stock with 4 channels. Its meant to be used with 4 channels.

Also A10 or FX-8350 run stock @ 1866, but many reviewers force they to run lower memory speed.

Yes, my point was that in the top 20 supercomputers nearly half are xeon based. Only three are opteron based. Saying xeon is a poor cpu for supercomputers is completely wrong. Sure titan is the most powerful but the only supercomputers using opterons in the top 20 are CRAY computers.

Hardly an official review from a reputable site.

The 8350 is not an apu so lets forget the "RAM is holding it back" argument. APU maybe (but only with gpu related tasks) but straight FX piledriver? No benefit to faster RAM.

HT is becoming more and more useful. Again is not that HT is useless, its that games don't take advantage of more than 4 cores (this is why an i3 creams a pentium in games requiring 3 or more threads and often why it beats the 8350 in lightly threaded games). That review is also a year an a half old, newer games such as FC3, crysis 3 show some improvements from HT.

Do you even read what you post about? That power test was running 'Intel Burn Test" which is probably one of the most demanding things you can do on a cpu.

And in the mobile comparison I don't care about gpu performance. We are only running the cpu, gpu is idling, using none or very little of the tdp headroom. When we stress both the cpu and gpu on the a10-4600m the clocks drop; in cpu only tests the cpu is using the igpu thermal headroom.

And yes, the fact the amd's chipsets use more power than intel's does matter. Because if you use the cpu you need a chipset and if your chipset is using more power then your cpu is using more power, similar to how early atom cpus had a low tdp but required power hungry chipsets and so the SYSTEM needed more power.

I don't think you quite get the concept of efficiency. If it uses the same amount of power to do a given amount of work then its the same efficiency. Bulldozer is not more efficient than sandy bridge in that test and if it required 6 kilojoules then that is the same amount of energy as sandy bridge and so the same efficiency.

 

[8350rocks]

10% margin of error....please....a 10% improvement is significant. I would give more like a 2% margin of error for most benchmarking.
didn't say 95% of games, said 95% of review sites.
http://techreport.com/review/23750/amd-fx-8350-processor-reviewed/9
http://www.guru3d.com/articles_pages/amd_fx_8350_processor_review,18.html
http://www.anandtech.com/show/6396/the-vishera-review-amd-fx8350-fx8320-fx6300-and-fx4300-tested/5
http://www.bit-tech.net/hardware/2012/11/06/amd-fx-8350-review/6
http://www.behardware.com/articles/880-13/amd-fx-8350-review-is-amd-back.html
http://hexus.net/tech/reviews/cpu/46985-amd-fx-8350/?page=5
http://www.xbitlabs.com/articles/cpu/display/fx-8350-8320-6300-4300_6.html#sect0
http://www.xbitlabs.com/articles/cpu/display/amd-fx-8350_6.html#sect0
http://www.tomshardware.com/reviews/far-cry-3-performance-benchmark,3379-7.html
http://www.guru3d.com/articles_pages/far_cry_3_graphics_performance_review_benchmark,7.html
http://www.tomshardware.com/reviews/crysis-3-performance-benchmark-gaming,3451-8.html
I will say in crisis 3 it doesn't do too shabby in average FPS, but dips to a low fps although that may have just been a spike.

Actually, it is typically expected that each system component inside a PC included in a benchmark that is a same make/model part, but a different individual part will accrue approximately 1% Margin of error due to manufacturing tolerances.

Therefore...if we assume that all components are in 2 separate cases, same make model (except CPU/MB of course) but different individual parts, the following is true:

PSU/RAM/GPU/HDD = 4% MoE

Additionally different components entirely introduce about double the margin of error to the test, because the results cannot establish a baseline inside the same sample.

Therefore:
CPU/MB = ~4%

Intel benchmarks performed with the exact same components sitting in separate cases, all identical make/model are subject to a minimum of 6% MoE if all components are different individual parts. Compare Intel/AMD and the number increases by approximately 25% because of the variability of different components (even by the same manufacturer).

Meaning average margin of error is 8% on such tests. I usually round up for human error to 10%. Nitpick all you want...

The fact of the matter is 8% of 120 FPS is 9.6 FPS or 10 FPS. Intel doesn't cover margin of error very often, if at all. (Skyrim I think they do, but it is an outlier).

 

[juanrga]

I would give more like a 2% margin of error for most benchmarking.

For 50 fps that is a +-1 fps. Sorry but nobody believes that!

For power measurements I would take about a 30% of variation.

 

[8350rocks]

@whyso:

If you say increased memory bandwidth does not benefit AMD...then why does the throughput scale significantly better than intel with bandwidth increases. Even Tom's reviews of the memory benchmarks noted that the throughput scales better with AMD...and they even comment that anything over 1600 MHz bandwidth on the intel CPUs is essentially worthless as the gain is almost nothing.

 

[juanrga]

Yes, my point was that in the top 20 supercomputers nearly half are xeon based. Only three are opteron based. Saying xeon is a poor cpu for supercomputers is completely wrong. Sure titan is the most powerful but the only supercomputers using opterons in the top 20 are CRAY computers.

I did not say the Xeon is a "poor cpu". I said something different. I said that has poor scaling than Opteron. All xeon based supercomputers in that list use a small number of cores and offer weak performance when compared with Titan. If you go to #22 you can find an Opteron based supercomputer that is not from Cray.

The 8350 is not an apu so lets forget the "RAM is holding it back" argument. APU maybe (but only with gpu related tasks) but straight FX piledriver? No benefit to faster RAM.

Why do you believe I mentioned both CPUs and APUs?

I already gave you benchmarks showing appreciable performance gains (6-10%) due to faster RAM for Bulldozer.

HT is becoming more and more useful. Again is not that HT is useless, its that games don't take advantage of more than 4 cores (this is why an i3 creams a pentium in games requiring 3 or more threads and often why it beats the 8350 in lightly threaded games). That review is also a year an a half old, newer games such as FC3, crysis 3 show some improvements from HT.

I already answered this. I will add some more info from the thread "Hyperthreading, does it help games now?"

For the most part you won't find many games that will get anything from hyper-threading. While there are few that may get a performance boost out of it in most cases they are going to be heavy multi-threading applications that would do better with a full core to run over the hyper-threading.

No.

But benchmarks done over the past several years have shown that when Hyper Threading is enabled, it generally causes an average drop of 2% in game performance.

http://www.tomshardware.co.uk/forum/366813-28-hyperthreading-gaming

Do you even read what you post about? That power test was running 'Intel Burn Test" which is probably one of the most demanding things you can do on a cpu.

First, I was not worried about absolute values but about the relative increase from going from Ivi to Haswell. Second, it does not change the point. The point was that the chip supposed to offer better perf per wat is offering poor perf per wat.

And in the mobile comparison I don't care about gpu performance.

Others care.

I don't think you quite get the concept of efficiency. If it uses the same amount of power to do a given amount of work then its the same efficiency. Bulldozer is not more efficient than sandy bridge in that test and if it required 6 kilojoules then that is the same amount of energy as sandy bridge and so the same efficiency.

I only can recommend you to read what was actually said.

 

[elemein]

And in the mobile comparison I don't care about gpu performance.
----
Others care.

This is for sure. I certainly do.

 

[whyso]

Yes, my point was that in the top 20 supercomputers nearly half are xeon based. Only three are opteron based. Saying xeon is a poor cpu for supercomputers is completely wrong. Sure titan is the most powerful but the only supercomputers using opterons in the top 20 are CRAY computers.

I did not say the Xeon is a "poor cpu". I said something different. I said that has poor scaling than Opteron. All xeon based supercomputers in that list use a small number of cores and offer weak performance when compared with Titan. If you go to #22 you can find an Opteron based supercomputer that is not from Cray.

The 8350 is not an apu so lets forget the "RAM is holding it back" argument. APU maybe (but only with gpu related tasks) but straight FX piledriver? No benefit to faster RAM.

Why do you believe I mentioned both CPUs and APUs?

I already gave you benchmarks showing appreciable performance gains (6-10%) due to faster RAM for Bulldozer.

HT is becoming more and more useful. Again is not that HT is useless, its that games don't take advantage of more than 4 cores (this is why an i3 creams a pentium in games requiring 3 or more threads and often why it beats the 8350 in lightly threaded games). That review is also a year an a half old, newer games such as FC3, crysis 3 show some improvements from HT.

I already answered this. I will add some more info from the thread "Hyperthreading, does it help games now?"

For the most part you won't find many games that will get anything from hyper-threading. While there are few that may get a performance boost out of it in most cases they are going to be heavy multi-threading applications that would do better with a full core to run over the hyper-threading.

No.

But benchmarks done over the past several years have shown that when Hyper Threading is enabled, it generally causes an average drop of 2% in game performance.

http://www.tomshardware.co.uk/forum/366813-28-hyperthreading-gaming

Do you even read what you post about? That power test was running 'Intel Burn Test" which is probably one of the most demanding things you can do on a cpu.

First, I was not worried about absolute values but about the relative increase from going from Ivi to Haswell. Second, it does not change the point. The point was that the chip supposed to offer better perf per wat is offering poor perf per wat.

And in the mobile comparison I don't care about gpu performance.

Others care.

I don't think you quite get the concept of efficiency. If it uses the same amount of power to do a given amount of work then its the same efficiency. Bulldozer is not more efficient than sandy bridge in that test and if it required 6 kilojoules then that is the same amount of energy as sandy bridge and so the same efficiency.

I only can recommend you to read what was actually said.

The first argument is complete nonsense (I have never seen any evidence of poor scaling--If you look at the Rmax and Rpeak values Titan is not special getting a similar proportion of actual FLOPS from Practical FLOPS). A couple points as well.
A lot of titan's power comes from the nvidia gpus (LOL actually because titan would not be number one without them). Titan has 18688 cpus (16 cores each); a lot of other computers have a similar number of cpus. The fastest supercomputer does not make all other supercomputers irrelevant. The number 22 supercomputer using opterons was commissioned in 2008.

There may be one benchmark for 5% performance gain from faster RAM but on average its closer to 0-2% which is pretty much insignificant.

That thread you posted a link to is old and hardly relevant anymore. And yes in BF3 MP, FC3, crysis3 you see a little performance gain between i5 and i7. 2% performance drop is nothing (margin of error), at 60fps that is one fps and far overshadowed by other factors (and this is really not seen at all in benchmarks).

Intel burn test is a power virus. We are more concerned about power use in real world applications. Just because power consumption in a power virus is high does not mean that power consumption something like x264 encoding will be as high (you can also look at the fermi/6xxx series power consumption tests; fermi really goes wild in Furmark but the difference is smaller in games).

In metro the 570 uses roughly the same amount of power as the 7970, in OCCT it consumes significantly more (differing performance aside). Same goes for cpus.

Others may care about gpu performance in that comparison but in that specific comparison we are not looking at gpu performance.

 

[chrisafp07]

To the original poster. The 8350 is a great cpu for gaming,. along with the i5 3570k. I would suggest you go for Haswell or Steamroller though. The 8350 would only be more viable than grabbing haswell because you can then wait for steamroller. By the way, I play Battlefield 3 all the time with my 8350 in Ultra settings Vsync on, 60fps at all times. I have had no issues whatsoever in Crysis 3 either, so if you are asking the difference between these cpus, well it really won't be that noticeable, you can look at benchmarks with minimal fps differences and see that. AMD just may give you more upgrades for those next 5-6 years with AM3+. As a sidenote, Intel is always a safe route for performance as well in any category.

 

[juanrga]

Yes, my point was that in the top 20 supercomputers nearly half are xeon based. Only three are opteron based. Saying xeon is a poor cpu for supercomputers is completely wrong. Sure titan is the most powerful but the only supercomputers using opterons in the top 20 are CRAY computers.

I did not say the Xeon is a "poor cpu". I said something different. I said that has poor scaling than Opteron. All xeon based supercomputers in that list use a small number of cores and offer weak performance when compared with Titan. If you go to #22 you can find an Opteron based supercomputer that is not from Cray.

The 8350 is not an apu so lets forget the "RAM is holding it back" argument. APU maybe (but only with gpu related tasks) but straight FX piledriver? No benefit to faster RAM.

Why do you believe I mentioned both CPUs and APUs?

I already gave you benchmarks showing appreciable performance gains (6-10%) due to faster RAM for Bulldozer.

HT is becoming more and more useful. Again is not that HT is useless, its that games don't take advantage of more than 4 cores (this is why an i3 creams a pentium in games requiring 3 or more threads and often why it beats the 8350 in lightly threaded games). That review is also a year an a half old, newer games such as FC3, crysis 3 show some improvements from HT.

I already answered this. I will add some more info from the thread "Hyperthreading, does it help games now?"

For the most part you won't find many games that will get anything from hyper-threading. While there are few that may get a performance boost out of it in most cases they are going to be heavy multi-threading applications that would do better with a full core to run over the hyper-threading.

No.

But benchmarks done over the past several years have shown that when Hyper Threading is enabled, it generally causes an average drop of 2% in game performance.

http://www.tomshardware.co.uk/forum/366813-28-hyperthreading-gaming

Do you even read what you post about? That power test was running 'Intel Burn Test" which is probably one of the most demanding things you can do on a cpu.

First, I was not worried about absolute values but about the relative increase from going from Ivi to Haswell. Second, it does not change the point. The point was that the chip supposed to offer better perf per wat is offering poor perf per wat.

And in the mobile comparison I don't care about gpu performance.

Others care.

I don't think you quite get the concept of efficiency. If it uses the same amount of power to do a given amount of work then its the same efficiency. Bulldozer is not more efficient than sandy bridge in that test and if it required 6 kilojoules then that is the same amount of energy as sandy bridge and so the same efficiency.

I only can recommend you to read what was actually said.

The first argument is complete nonsense (I have never seen any evidence of poor scaling--If you look at the Rmax and Rpeak values Titan is not special getting a similar proportion of actual FLOPS from Practical FLOPS). A couple points as well.
A lot of titan's power comes from the nvidia gpus (LOL actually because titan would not be number one without them). Titan has 18688 cpus (16 cores each); a lot of other computers have a similar number of cpus. The fastest supercomputer does not make all other supercomputers irrelevant. The number 22 supercomputer using opterons was commissioned in 2008.

There may be one benchmark for 5% performance gain from faster RAM but on average its closer to 0-2% which is pretty much insignificant.

That thread you posted a link to is old and hardly relevant anymore. And yes in BF3 MP, FC3, crysis3 you see a little performance gain between i5 and i7. 2% performance drop is nothing (margin of error), at 60fps that is one fps and far overshadowed by other factors (and this is really not seen at all in benchmarks).

Intel burn test is a power virus. We are more concerned about power use in real world applications. Just because power consumption in a power virus is high does not mean that power consumption something like x264 encoding will be as high (you can also look at the fermi/6xxx series power consumption tests; fermi really goes wild in Furmark but the difference is smaller in games).

In metro the 570 uses roughly the same amount of power as the 7970, in OCCT it consumes significantly more (differing performance aside). Same goes for cpus.

Others may care about gpu performance in that comparison but in that specific comparison we are not looking at gpu performance.

Repeating the word "nonsense" all the day is not going to change facts. From a link given before:

In a variety of computing benchmarks, the Opteron architecture has demonstrated better multi-processor scaling than the Intel Xeon. This is primarily because adding an additional Opteron processor increases memory bandwidth, while that is not always the case for Xeon systems, and the fact that the Opterons use a switched fabric, rather than a shared bus. In particular, the Opteron's integrated memory controller allows the CPU to access local RAM very quickly. In contrast, multiprocessor Xeon system CPUs share only two common buses for both processor-processor and processor-memory communication. As the number of CPUs increases in a typical Xeon system, contention for the shared bus causes computing efficiency to drop.

I have not given you one benchmark showing 5%. I have given you benchmarks showing 6-10%. I did because you claimed no advantage. Now you change your argument to no advantage on the average, but then two comments: I don't care about the average if I am playing one of those benchmarked games and you don't offer anything to support your average figure.

The thread I posted is from 27 February 2013. Yes computers advance fast but not that fast. What they said about HT continue being valid today.

About power consumption I recommend you again the same: read what was said.