AMD FX-8150 Review: From Bulldozer To Zambezi To FX

[clonazepam]

I tripped on my NES's light gun cable and hit my head on the floor. I saw a brief glimpse into the near future, and have been summoned back to relay this message from the beyond, to you all.

Don't worry. This is a good thing. The performance is right where it should be. The prices will fall at the initial booh-boohs, while simultaneously the performance will get better and better. I estimate noticeable gains in the short term, as well as even better long-term returns. They're going to find issues that can only be found after releasing it into the wild. Whole communities will band together and make everything solid and viable.

Intel's new commercial is really pushing the "we are heroes because of our technology and how it saves lives and makes healthcare easier for everyone, even the patients!

Wouldn't you like to be on what I'm on? hah

Peace, love.. Tom's the bombs!

 

[Guest]

Many communities are banning products designs or made in Israel-such as Intel-if you want to support an apartheid system be my guest. I will stick with a stick with product i know will let me sleep at night

 

[Guest]

In an effort to keep my AMD fanboy dignity I'm going to argue that the tests weren't extensive enough. Where were the boot/shutdown tests? Where was the virus scan test? Where were tests running multiple applications? The advantages appear to be in multi-threaded applications, whereas the testing seemed almost biased towards single threated apps.

 

[eddieroolz]

Its a disappointment no doubt, but considering AMD's much smaller team and smaller pool of resources, I'd say they did a fine job - even if it meant restricting the target to one or two sectors.

To be more precise, what I see is a processor that's skewed heavily towards multithreaded workstation applications and also potentially for server racks. Plus, losing IPC was not a good decision for gamers at all.

In the end though, I think AMD can still score plenty of buyers from universities, research labs and researchers/scientists who need to run complex multithreaded applications. If done right, that market can be big too.

 

[hawkeye70]

What brings you to that conclusion? Based on the performance of the FX-8150, it isn't too difficult to extrapolate the performance of the FX-4100. It uses the same die as the FX-8150, except it's binned as a 2 module part instead of a 4, and it's max turbo core is clocked lower at 3.8 GHz (although it's base clock is the same). In other words, don't expect any performance miracles.

The Bulldozer architecture relies on higher clock frequencies and wide highly-threaded execution to offset the decreased IPC and performance per clock in comparison to Sandy Bridge, and even its predecessor the Phenom II.

With the FX-4100 both core count and turbo core frequencies are cut down... turbo core being essential for this architecture to keep up in lightly threaded (1-4 thread) workloads. This is an area the FX-8150 already struggles to gain performance parity with lower clocked Phenom II parts. What makes you think the FX-4100's cut back turbo core will fare any better?

And again, the FX-4100 is also lacking Bulldozer's major strength, highly parallel execution, having only half the modules of the FX-8150. Really the only thing it seems to have going for it is a relatively low price, and potentially greater overclocking headroom, which I'm predicting it's probably going to desperately need in order to justify its value.

Funny I don't remember seeing extremely high multi-threaded scores on the phenom II x4 cores. Surely you don't expect the new 4100 series fx processor to do worse in the heavy multi-threaded apps when the 8150 does so much better?

I agree with you on the lightly threaded issue, but the much higher overclock potential should even that up and enable it to catch up or overcome its deficiencies in comparison (both being overclocked to their max).

All I'm saying is for the price it will be a better option than the current phenom II x4 lineup on value alone. Games should handle the same and heavy duty processing tasks should be improved... for $115 dollar part I'd say that is pretty good. Guess we will have to wait and see if I'm wrong or not. Whatever the case I can see this chip being successful (FX 4100) as the phenom II's won't be available much longer and it will fill the niche.

I currently have a phenom II x4 965 clocked to 3.9 ghz and it handles anything I throw at it with ease. Somehow I think bulldozer will be fine at doing that also and sometimes better but I probably wouldn't even notice the difference unless I looked for the differences in some benchmark. But hey I'm not about having the highest benches I just want my computer to get the work done without me waiting forever for it to do it.

 

[egowhip69]

While I might be a little disappointed with the desktop version of BD, I can see some light at the end of the tunnel.

First is the sever side... as much as the top gamers like to think, we are NOT the most profitable segment in the CPU world. That would be servers. If AMD keeps some of the server CPUs as drop in replaceable (which I've read they will) that gives MUCH lower costs, AND that can up the horsepower... well... that builds loyalty in the server market. (loyalty that has been a bit eroded in the past year or so)

I can tell you, in a decent sized Post Production company (Deluxe), that the low cost of a drop in replacement is VERY important! While Intel produced good chips, they always made you spend to replace the whole system (1 kinda exception so far). AMD on the other hand has given us a full upgrade, and looks to hand in another (in some server segments).

Desktop BD needs work, and I'm the first to say that. But as more and more "professional" workflows move into the semi-pro and novice section, I think AMD might be looking toward the future, and it might be the right move.

I just hope this isn't an Edsel move.

We'll see... I think I'll hold off on a CPU upgrade for a bit. I think right now, spending $200+ on a new video card might be the better way to go if you already have a decent dual core or better.

 

[xpslover007]

Ah... This explains the low price point of these "8" core processors. I'm a bit sad that they're under performing. Tons of disappointments these week... Expensive UX21s to not so awesome Bulldozers...

 

[dissbelief]

It is amazing how much people whine over the pettiest of things. If you are a super-duper power user who is struggling to be productive because of a slow CPU, then maybe I can understand the outrage. But if you are a gamer or an average user, then what is wrong with a 2500k or Phenom II X4? All of the whiners must remember that this let down is not AMD's fault, but their own. There was no evidence that showed Bulldozer would even match a 2500k or even a 1090t, so why was it expected? When you take into account the lack of benchmarks released by AMD, the time it took to develop Bulldozer, and the firing of high ranking employees, you have to be such an easily led sheep to fall into all of the hype. It's ok to hope for the next greatest processor, I did, but to get upset about it is just childish. Buy whatever brand has the best price/performance ratio for your upgrade or new build when you actually need to upgrade because waiting for the next new thing only makes you a mindless consumer.

 

[pinkfloydminnesota]

Guess I'll defect to Intel for the 2500k when that cpu/mb combo reaches into the low $200s or so. So much for bragging about my eight core PC.

 

[DjEaZy]

[citation][nom]DjEaZy[/nom]... why there still are single threaded 32Bit applications??? We have AMD64 in 2003 i think... we have multicore and hypertrending... but there are still 32Bit single threaded applications??? The programmers on windows platform are lazy, or MS don't give enough tools, to make use of the technology... common... It's 2011!!! Still we test 32bit single threaded applications...[/citation]
... and i blame intel atom... most of that crap is 32bit... and... common MS... apple is natively x64 sins 2005... it's time to step up... do some SDK, to make easier for dev's to implement new things... i hope, there will not be a 32Bit WIN8... even lowest end AMD is 64bit... and fcuk intel atom... and for the multicore? Even mobile phones are dualcore... and quadcore is on the way... common, step it up... as you can see, we have the thech good and cheep, like this chip here... there is a need for software... linux got it, apple got it, but windows don't take the full advantage of it... and AMD?! If MS will do less, then expected... build a driver for this CPU... lay out the vision, that you have, in practice... philosophically i like this chip... it's reminds me of the Radeon HD2000 series... it was a good idea, but needed polishing... and now? Radeon HD series is delivering on the vision... do it with this chip too... good luck... and... its just FX 8[1]50... the FX 8[8]70 gonna be a monster...

 

[Guest]

"Many communities are banning products designs or made in Israel-such as Intel-if you want to support an apartheid system be my guest. I will stick with a stick with product i know will let me sleep at night"

You're an absolute idiot.

 

[dragonsqrrl]

[citation][nom]hawkeye70[/nom]Funny I don't remember seeing extremely high multi-threaded scores on the phenom II x4 cores. Surely you don't expect the new 4100 series fx processor to do worse in the heavy multi-threaded apps when the 8150 does so much better?[/citation]
2x better? Seriously, that's the sort of performance hit you're going to see in highly threaded workloads. And that's exactly the point I'm trying to make when I say, "the FX-4100 is also lacking Bulldozer's major strength, highly parallel execution, having only half the modules of the FX-8150".

Look over the productivity and content creation sections again. Where as going from a Phenom II x6 to an x4 at the same clocks would give you around a 33% performance hit, going from an FX-8150 to 4100 would net you somewhere around a 50% performance hit. In cases where the FX-8150 is just barely outperforming the x6 1100T, this doesn't bode well for the FX-4100. If you can think of any reason this wouldn't be the case (I haven't heard one yet), please let me know.

Of course there shouldn't be nearly as much of a performance hit in lightly threaded workloads, but this is a necessity since the 8150 with its superior turbo core clocks already has problems keeping up with much lower clocked Phenom II's.

 

[B_Type13X2]

I would like to see them do an architecture comparison similar to what they did here: http://www.tomshardware.com/reviews/processor-architecture-benchmark,2974.html

with bulldozer so we can see a core to core comparison, because this could be an issue of optimization vs. utilization in that the benchmark programs used might not fully utilize all of bulldozers cores. However if they do this and Bulldozer still fails to beat even the I52500K on a core to core basis well then Bulldozer will be an even bigger let down.

Also what the hell was AMD thinking pricing it at 250$'s vs. the 210$ I52500k 40$'s more for less overall performance?

 

[dragonsqrrl]

[citation][nom]hawkeye70[/nom]All I'm saying is for the price it will be a better option than the current phenom II x4 lineup on value alone. Games should handle the same and heavy duty processing tasks should be improved... for $115 dollar part I'd say that is pretty good.[/citation]
Sorry dude, but I just couldn't disagree more. Most thread optimized benchmarks (which also tend to scale very well with additional cores) seem to indicate otherwise. I think at best, when the FX-4100 is in its element, it'll probably perform in the area of similarly priced Phenom II x4's, but at worst, in single/lightly threaded workloads the Phenom II will more then likely outperform it without much effort.

 

[peroludiarom]

CITE: "In fact, Fog points out that even benchmarking programs are affected by this, up to a point where benchmark results can differ greatly depending on how a processor identifies itself. Ars found out that by changing the CPUID of a VIA Nano processor to AuthenticAMD you could increase performance in PCMark 2005's memory subsystem test by 10% - changing it to GenuineIntel yields a 47.4% performance improvement! There's more on that here [print version - the regular one won't load for me]. " Google it, to know about Intel's dirty compiller

 

[dragonsqrrl]

[citation][nom]saturnus[/nom]Now the good parts is that memory performance is on par with i7 even though FX uses 2 channels and i7 uses 3 channels, that's a 50% advantage.[/citation]
Yes, the i7 920 uses a tri channel controller, and it also launched three years ago. All the second gen i7/i5's in the comparison use a dual channel controller. There is no 50% advantage...
[citation][nom]saturnus[/nom]And it's also obvious that Global Foundries 32nm SOI process is far more advanced than Intels 32nm bulk process. 2B transistors in 315mm2 in the FX vs. 1B transistors in 216mm2 in the i7, that's a 37% advantage in transistor packing density.[/citation]
This is completely false. Packing more transistors into a given area is in no way indicative of GF 32nm SOI being more advanced then Intel's 32nm process. It has far more to do with architectural design choices. Honestly, what are you talking about?

 

[agnickolov]

Thank you, thank you, thank you for including that Visual Studio benchmark!!!

 

[dragonsqrrl]

[citation][nom]hawkeye70[/nom]Funny I don't remember seeing extremely high multi-threaded scores on the phenom II x4 cores. Surely you don't expect the new 4100 series fx processor to do worse in the heavy multi-threaded apps when the 8150 does so much better?[/citation]
Well, Guru3D did a review of the FX lineup including the 4100. Here are the benchmarks to back up my previous estimations:

http://www.guru3d.com/article/amd-fx-8150--8120-6100-and-4100-performance-review/4

... the Athlon II x4 645 outperforms it in quite a few cases.

 

[nikorr]

Well, Intel still doesn't feel the heat.

 

[nikorr]

It falls in between i5-2500K and i7-2600K - still far from 2600K.

 

[amk-aka-Phantom]

... and i blame intel atom... most of that crap is 32bit... and... common MS... apple is natively x64 sins 2005... it's time to step up... do some SDK, to make easier for dev's to implement new things... i hope, there will not be a 32Bit WIN8... even lowest end AMD is 64bit... and fcuk intel atom... and for the multicore? Even mobile phones are dualcore... and quadcore is on the way... common, step it up... as you can see, we have the thech good and cheep, like this chip here... there is a need for software... linux got it, apple got it, but windows don't take the full advantage of it... and AMD?! If MS will do less, then expected... build a driver for this CPU... lay out the vision, that you have, in practice... philosophically i like this chip... it's reminds me of the Radeon HD2000 series... it was a good idea, but needed polishing... and now? Radeon HD series is delivering on the vision... do it with this chip too... good luck... and... its just FX 8[1]50... the FX 8[8]70 gonna be a monster...

Atoms ARE crap, I agree, but most of them (if not all) are 64-bit CPUs.

 

[DjEaZy]

[citation][nom]amk-aka-phantom[/nom]Atoms ARE crap, I agree, but most of them (if not all) are 64-bit CPUs.[/citation]
... but the thing is, that they are still producing 32bit CPU's...

 

[f-gomes]

[citation][nom]killerclick[/nom]Intel shouldn't lower prices, they should raise them. I'll gladly pay more to reward competent product development and nothing would please me more than AMD going down in flames for all their flops in the past 5 years. Intel doesn't need AMD to push them forward.[/citation]

Either you are mad or you're plain stupid.

 

[peroludiarom]

i would say, that after finding the PROBLEM with intel compiler, i would not trust at any of these benches for arithmetic and memory:

http://www.theinquirer.net/inquirer/news/1567108/intel-compiler-cripples-code-amd-via-chips

 

[f-gomes]

[citation][nom]compton[/nom]I know Mr. Angelini's review is a rollercoaster ride, but AMD faithful shouldn't be losing their lunches. Overall, it's not as bad as it seems, but I know that those who've been waiting month after agonizing month were hoping for more. Still, there's some good stuff here and who knows how it will perform in another metric: fun. Last year I bought 3 AMD processors. I truly hoped that Bulldozer would drop in the spring, and it didn't, but I whiled away the time by tweaking Phenoms, unlocking cores, undervolting them and overvolting them. It's a lot of fun, an important measure. Only two Sandy Bridge processors are any fun at all. Given the fact that the other 8 "core" Bulldozer will be more like $200 and could probably be as fast as the flagship with a multiplier bump, you get more value there. So they'll still be fun when paired with a great motherboard, and still pretty reasonable in price. Better power management helps efficiency, and while the best case performance scenario is good, the worst case isn't that bad either. It's not as fast as SB much of the time, but don't take that to mean BD is bunk; rather SB is just really, really good. Throughout all of the leaked slides and rumors and delays, I and many others said if it could come close to SB it would be a win, and I still think that's a case. AMD is still more flexible in some ways and still believes catering to the enthusiast is important. I still think I'll buy into BD even though I upgraded to a SB system this spring when BD was nowhere in sight. I don't think I'll be disappointed either. There are many other metrics that aren't covered in Mr. Angelini's thorough article anyway. Those will come out, and then we can see the whole platform as a package. It's still a net win.[/citation]

"Wake up and smell the ashes"