AMD CPU speculation... and expert conjecture

[juanrga]

LOL. This is the fifth time that you ask me if Android is a full blown desktop OS and this is the fifth time that I say you: NO

And why do you insist on confounding Android with some of the ARM OS that are ready or are being developed?

Ask this question rocks: do you really believe that the ARM supercomputer that is being developed will run Android?

Who is developing an ARM OS? M$? Red Hat? Apple?

No, none of those are even considering an ARM DT OS. Canonical has a mobile OS called Ubuntu Touch that will run on ARM. Their DT version isn't nearly as capable as the x86 version.

Apparently, you are caught in an endless loop:

while 2 != 3 {
ARM is RISC
RISC is "incomplete" // (first mistake)
ARM is "incomplete" // (second mistake)
ARM cannot run a complete DT OS // (third mistake)
ARM only can run Android // (misguided conclusion)
}

I take note that you didn't offer any answer about if you believe that the ARM supercomputer will run Android or not.

I never said ARM has to run Android, I said it won't be a viable DT PC architecture without a DT capable OS. Which it doesn't have, and one is not imminently on the horizon either.

Your loop seems to be:

If 0 != 1 {

ARM will replace x86 without a software environment

NVidia will magically become a player in ARM race, even though their Tegra projects are hemorrhaging money.

Seattle will magically rule the world, even though it's not built yet.

ARM has a DT capable OS, but no one can produce the evidence
}

Else {

The entire thread is wrong about ARM, and juanrga is right even though there is no real world numbers for Seattle, or a DT OS, or a software environment

}

Frankly, I am not taking speculation of Kaveri into account, only the benchmarks from the early ES's. Additionally, based on what I can project, we can make assumptions from the performance of PD and get rough ideas.

You are making assumptions attempting to extrapolate data from an ARM chip of a caliber not yet made. Additionally, the performance figures you quote are entirely speculation with no prior architecture example as a baseline. You sell this as though it is gold, and expect us to buy. Sorry, but if I am buying gold, I have to have it verified that I am getting what I am indeed thinking I am buying.

Can we get off the topic of ARM now?

If 0 != 1 {
...}
Else {
...}

If your usual code looks like that, I want to see your work

Is it possible that juanrga is using this topic just so that way we view his site?

If you want to reproduce the entire Steamroller article it in your own site or in any other place, you can do it. The license is Attribution-NonCommercial-ShareAlike 3.0 Unported License (CC BY-NC-SA 3.0).

You can also reuse small fragments and data without any licensing issue.

 

[Cazalan]

I don't have to expound on the virtues of FD-SOI UTBB's back biasing capability...it's really a superior product in that regard. Also, IBM was discussing the possibility of UTBB FD-SOI all the way to something in the 7nm range. Much smaller and you could no longer use an insulator because you wouldn't have enough space.

I agree that FD-SOI is the better solution but I'm just looking at the cost aspect. If AMD is cutting GF purchases they don't have product to support the tech ramp. IBM has their Power 8 part but that's a relatively low volume product and the true cost is burried in the 10+ million dollar systems they ship. The rest of the world seems complacent with Bulk.

GF producing for IBM was news for me, although I knew they shared tech. I did find the announcement for 32nm cross production. Every node is a multi-billion dollar undertaking so they would likely make similar announcements if that were also true at 20nm node. It's possible but I also heard there were some sour grapes with IBM/AMD because they stole XBone/PS4 from them.

http://www-03.ibm.com/press/us/en/pressrelease/36465.wss

 

[8350rocks]

LOL. This is the fifth time that you ask me if Android is a full blown desktop OS and this is the fifth time that I say you: NO

And why do you insist on confounding Android with some of the ARM OS that are ready or are being developed?

Ask this question rocks: do you really believe that the ARM supercomputer that is being developed will run Android?

Who is developing an ARM OS? M$? Red Hat? Apple?

No, none of those are even considering an ARM DT OS. Canonical has a mobile OS called Ubuntu Touch that will run on ARM. Their DT version isn't nearly as capable as the x86 version.

Apparently, you are caught in an endless loop:

while 2 != 3 {
ARM is RISC
RISC is "incomplete" // (first mistake)
ARM is "incomplete" // (second mistake)
ARM cannot run a complete DT OS // (third mistake)
ARM only can run Android // (misguided conclusion)
}

I take note that you didn't offer any answer about if you believe that the ARM supercomputer will run Android or not.

I never said ARM has to run Android, I said it won't be a viable DT PC architecture without a DT capable OS. Which it doesn't have, and one is not imminently on the horizon either.

Your loop seems to be:

If 0 != 1 {

ARM will replace x86 without a software environment

NVidia will magically become a player in ARM race, even though their Tegra projects are hemorrhaging money.

Seattle will magically rule the world, even though it's not built yet.

ARM has a DT capable OS, but no one can produce the evidence
}

Else {

The entire thread is wrong about ARM, and juanrga is right even though there is no real world numbers for Seattle, or a DT OS, or a software environment

}

Frankly, I am not taking speculation of Kaveri into account, only the benchmarks from the early ES's. Additionally, based on what I can project, we can make assumptions from the performance of PD and get rough ideas.

You are making assumptions attempting to extrapolate data from an ARM chip of a caliber not yet made. Additionally, the performance figures you quote are entirely speculation with no prior architecture example as a baseline. You sell this as though it is gold, and expect us to buy. Sorry, but if I am buying gold, I have to have it verified that I am getting what I am indeed thinking I am buying.

Can we get off the topic of ARM now?

If 0 != 1 {
...}
Else {
...}

If your usual code looks like that, I want to see your work

Is it possible that juanrga is using this topic just so that way we view his site?

If you want to reproduce the entire Steamroller article it in your own site or in any other place, you can do it. The license is Attribution-NonCommercial-ShareAlike 3.0 Unported License (CC BY-NC-SA 3.0).

You can also reuse small fragments and data without any licensing issue.

I did that because you are stuck in a loop of the first portion...and the else section is not at all true...

My parsing may be rough, but I do still know how to use if's and else's. I could have used while, however, I wanted to be different.

 

[8350rocks]

Intel crushed in CPU only tasks, and was middle of the pack in graphical tests.

PS

The Intel iGPU stinks.

So understand: You are comparing platforms, not CPU's. In purely CPU based benchmarks, Intel crushes ARM, even at the low end. Its lack of a competitive iGPU is holding it back in graphical benchmarks.

But that works both ways... The ARM platform being compared all the time, is a cellphone or a tablet at most.

I would put the ARM vs X86 argument in a bag for now, since neither camp (in favor or against) have an oranges to oranges test that no one can say "this doesn't fit". All theoretical arguments are destroyed with a "if cows could fly" and until AMD puts the SoC with an ARM CPU at a demo or something, we're wasting time IMO.

Cheers!

The main problem is, at a minimum, you are using different OS's, software versions, etc, so there will ALWAYS be ambiguity unless you write programs in optimized assembly for both arches and test that way. Closest thing you can do is test a mostly optimal program in an OS that supports both arches (say, Linux or Android), and accept there will be some noise in the data.

At the end of the day, in regards to CPU performance, X86 wins, easily. By how much is up for debate, but even the highest tier ARM looses to the lowest tier X86 (Silvermont). How is ARM going to compete against HW/SR if it can't even beat Atom?

ARM isn't displacing X86 anytime soon. The only reason ARM wants to get into the server market is because of the insane margins available. And lets not forget MIPS and PPC are always around the corner in the low-power market...

Somehow, I missed this...

+1

We are agreeing on things somewhat semi-frequently at this point...

...

I am not entirely sure if that bothers me...or if something is awry in the world...

 

[juanrga]

Frankly, I am not taking speculation of Kaveri into account, only the benchmarks from the early ES's. Additionally, based on what I can project, we can make assumptions from the performance of PD and get rough ideas.

Nope. You have speculated about Steamroller, Kaveri, MANTLE, new Radeons... This is one example:

Yesterday, AMD announced that Steamroller Opteron Server Architecture will be available in the second half of this year followed shortly by the client architecture (Kaveri/FX).

There is also a logic flow chart diagramming the upgraded architecture for Steamroller, and commentary on the changes.

A good read for anyone curious about what's coming.

The expected single threaded performance gain is 30% over Piledriver, and multi-threaded performance will have an expected increase on the order of 12-15%

However, when someone makes similar statement about the expected performance of Seattle or A57 using AMD sources, you reject them and solicit to the other part "real world measurements" on final silicon.

That is double standard.

If 0 != 1 {
...}
Else {
...}

If your usual code looks like that, I want to see your work

I did that because you are stuck in a loop of the first portion...and the else section is not at all true...

My parsing may be rough, but I do still know how to use if's and else's. I could have used while, however, I wanted to be different.

Being different is fine, the problem is with being wrong. I will continue replying what you _post_ instead of what you claim to know _after_ posting.

 

[noob2222]

While (Juanrga = wrong) {printf("JUANRGA IS NEVER WRONG")};

This loop will never end.

 

[blackkstar]

So, has anyone ever thought about the fact that ARM devices sell more because laptops can last for several years and the types of devices ARM chips go into get thrown away after a year or two?

Do you really think that ARM DT is going to sell as much as ARM mobile parts in iPhone? An ARM DT would be around for several years if it matched x86 performance. Look, i7 920 and phenom 2 can still make good gaming CPUs, and those are ancient in technology terms.

Meanwhile 5 year old ARM CPU is joke at best.

The biggest fallacy I see with regards to ARM "taking over" is that people are assuming that if we see ARM DT parts that the ARM parts will sell more than the x86 parts.

To put why ARM devices sell more into perspective, imagine the performance of these devices and when they came out:

2600k came out in very early 2011.

Galaxy S was one of the best ARM phones you could get at the time. Yes, Galaxy S1. Not S2, S3, or even S4. S1!

Now do you see why ARM chips sell so much? They are going into disposable junk products that are meant to be thrown away.

Notice how ARM chip from late 2010 is slow garbage now. Where is 2600k in the list of gaming PCs? It's still overkill with people recommending chips maybe 10% faster with NO HT.

So, riddle me this, do you really expect ARM sales to keep going like they are if they catch up to x86 in performance?

Also, one last question. Do you expect sales of ARM devices (and thus chips) to continue like they are now (S1 in late 2010 to S4 in late 2013) when ARM reaches a point where a 2 year old ARM chip can run latest Android buttery smooth? Because that's basically what's getting implied in this thread, that ARM sales will always continue like they have been.

You guys are all falling into the ARM bubble and the ARM hype as vendors and OEMs get all excited about shovelling planned obsolescence garbage products down your throat and using that garbage to replace a platform that is serviceable with a huge array of readily available to consumer replacement parts. Also, mind you, if your wifi goes out in your desktop, you buy a new wifi chip. If your wifi goes out in your phone, you buy a new phone. Which one do you think the company selling you those products prefer to sell? And are you implying that they would not be creating hype to move the market in this direction.

Just a friendly reminder, the idea of tablets existed since the 1950s and Microsoft pushed them aggressively since 2000. They never took off until Apple got on stage and applied the reality distortion field to sheep and convinced them that they need this new product.

 

[noob2222]

I kinda had the same point, but in reverse. People aren't buying DT because their 4 year old high end system is still middle-high end today.

But ya, once ARM performance stagnates or saturates the market, sales will stall just the same.

 

[juanrga]

While (Juanrga = wrong) {printf("JUANRGA IS NEVER WRONG")};

This loop will never end.

Your above code is saying something as "while A=B print A != B". I want see both of you coding together, and then I want see the face of the guy maintaining your code.

So, has anyone ever thought about the fact that ARM devices sell more because laptops can last for several years and the types of devices ARM chips go into get thrown away after a year or two?

Do you really think that ARM DT is going to sell as much as ARM mobile parts in iPhone? An ARM DT would be around for several years if it matched x86 performance. Look, i7 920 and phenom 2 can still make good gaming CPUs, and those are ancient in technology terms.

Meanwhile 5 year old ARM CPU is joke at best.

The biggest fallacy I see with regards to ARM "taking over" is that people are assuming that if we see ARM DT parts that the ARM parts will sell more than the x86 parts.

To put why ARM devices sell more into perspective, imagine the performance of these devices and when they came out:

2600k came out in very early 2011.

Galaxy S was one of the best ARM phones you could get at the time. Yes, Galaxy S1. Not S2, S3, or even S4. S1!

Now do you see why ARM chips sell so much? They are going into disposable junk products that are meant to be thrown away.

Notice how ARM chip from late 2010 is slow garbage now. Where is 2600k in the list of gaming PCs? It's still overkill with people recommending chips maybe 10% faster with NO HT.

So, riddle me this, do you really expect ARM sales to keep going like they are if they catch up to x86 in performance?

Also, one last question. Do you expect sales of ARM devices (and thus chips) to continue like they are now (S1 in late 2010 to S4 in late 2013) when ARM reaches a point where a 2 year old ARM chip can run latest Android buttery smooth? Because that's basically what's getting implied in this thread, that ARM sales will always continue like they have been.

You guys are all falling into the ARM bubble and the ARM hype as vendors and OEMs get all excited about shovelling planned obsolescence garbage products down your throat and using that garbage to replace a platform that is serviceable with a huge array of readily available to consumer replacement parts. Also, mind you, if your wifi goes out in your desktop, you buy a new wifi chip. If your wifi goes out in your phone, you buy a new phone. Which one do you think the company selling you those products prefer to sell? And are you implying that they would not be creating hype to move the market in this direction.

Just a friendly reminder, the idea of tablets existed since the 1950s and Microsoft pushed them aggressively since 2000. They never took off until Apple got on stage and applied the reality distortion field to sheep and convinced them that they need this new product.

Have you considered that ARM has been named "the fastest-growing CPU architecture in history"? You would need about 6 generations of x86 processors for obtaining the same performance gain that you can see in the ARM ecosystem with a single generation change.

The problem with x86 is that its evolution is stagnant. There are several reasons for that. Some are technical (CISC, ISA), other are economical (for instance only now people starts talking about the possibility that Intel will make six-cores for mainstream thanks to AMD wise manoeuvres in the gaming consoles).

As mentioned many times before, the desktop market is dying, therefore ARM sales will be not driven by desktops. Desktops will come a niche market.

Nobody is implying that ARM will be rising forever... because that is physically impossible.

People (including large customers) are excited by the technical, economical, and legal properties of the ARM architecture/ecosystem.

Also rest of your comments about "replacements" apply to x86 as well. The tendency to integrate more and more mobo components inside the chip/package implies you can replace less and less components. Can you replace the GPU inside a Richland APU?

 

[Cazalan]

Also, one last question. Do you expect sales of ARM devices (and thus chips) to continue like they are now (S1 in late 2010 to S4 in late 2013) when ARM reaches a point where a 2 year old ARM chip can run latest Android buttery smooth? Because that's basically what's getting implied in this thread, that ARM sales will always continue like they have been.

There is already a race to the bottom starting to occur and signs that the growth is waning for cell phones/tablets. However new products every month are finding ways to add processors to enhance user experience.

Even the local laundry mat has new washers/dryers that take credit cards now. So what used to be a simple micro-controller is now an embedded computer/phone running some flavor of linux.

The luxury the cell phone industry has is they're so easy to break. I don't know anyone that upgrades just to have the latest and greatest phone. They usually break on their own at least partially. Headphone jack stops working. Vibration motor fails, etc.

 

[-Fran-]

ARM growth in the consumer space is directly tied to what 3rd parties (Samsung, Apple, Google, etc) can pack into their mobile devices to justify the expense on a new SoC/device. It's a balance of things, but I think it boils down to 2 main things: User experience and utility.

For instance, I love my SGS2. The only thing that will make me change it, is a superior OS (I'm running Cyanogen Mod 9.0, based on Android 4.0) that brings improvements in different areas that I'm currently lacking (currently: none) and a radical need for a new SoC because dedicated hardware. For instance, the S2 doesn't do 1080p at 60FPS with no drops. Not even 720p with no drops. And the camera is not good, its showing its age badly (compared to a real point and shoot, not a DSLR, lol). So, that means if I want a "camera improvement" I have 2 choices: get a new one or get a P&S camera. I would probably go for a new phone, since the phone is 1 device to carry around.

Anyway, I feel that the smartphone market won't stop growing until a few good years (2018, maybe?), since smartphones are very susceptible to break and companies (carriers?) like you to change them from time to time to support new network features (sometimes, they force you to change: GSM to CDMA for example). Tablets, not so sure... They're kind of in the middle of everything, so I think they'll start showing stagnation very soon, if they're not already stagnated.

The interesting part is that ARM is fighting from the bottom up. I think its a fight Intel never planned to fight and it already lost a few punches, haha. Still, Intel's downsize is going strong (few wins, but they're still interested), but also ARMs up-size is also looking strong (Chromebooks had ARM CPUs in them?).

Cheers!

 

[griptwister]

I'm about as bored as a board with ARM. Lets move onto this...

http://www.eteknix.com/nvidia-origin-pc-bribery-dealings/

What are your guys' thoughts? Bribery, or does Nvidia really flip the Bill?

http://semiaccurate.com/2013/10/07/nvidias-program-get-oems-like-origin-pc-dump-amd-called-tier-0/

 

[noob2222]

I missed this on october 1st. This one is actually relevant to the topic (mostly.) Those that try to imply that ICC doesn't affect scores much ...

http://www.maxon.net/products/cinebench/overview.html

its out, and its removed the cripple AMD function.

the old:

the new R15

3770k stock 130/630+ pts
8350 stock 98/650+ pts

much closer on single thread and multicore, the 8350 jumps in front of the 3770k.

OC 8350 numbers seem to be ~750 and up.

 

[juanrga]

As I wrote before in this thread, Windows is no more driving the market. The present/future is Linux. Good to know a traditional company as HP agrees:

http://news.cnet.com/8301-10805_3-57606827-75/hp-no-longer-playing-by-microsoft-intel-rules-exec-says/

Interesting also how HP agrees about ARM:

Wintel-based devices are being aggressively displaced by ARM-based PCs and mobile devices running competing operating systems

 

[gamerk316]

http://www.tomshardware.com/reviews/battlefield-4-graphics-card-performance,3634-10.html

This is about what I expect going forward. Titan bottleneck is @74 FPS, so you lose visibility into CPU performance at that point.

i7-3960x is GPU bottlenecked. i5-2500k @ 4.2 is GPU bottlenecked. i5-2500k shows some signs of a CPU bottleneck (Min FPS lower then other two CPUs), but still averages out to a GPU bottleneck.

FX-8350 is not GPU bottlenecked, but still manages a higher minimum FPS then a 2500k,though a lower average. Indicates that for BF4, the 2500k and FX-8350, at stock, are roughly at the point where a GTX Titan starts to get held back, FPS wise.

And again, the i3-3220 beating the FX-4170.

This is the kind of chart I expect will be common, with only the 8-core FX line keeping up with Intel, and fast i5's being more then enough to max out performance.

 

[de5_Roy]

fx9590 gets a massive price drop
http://techreport.com/news/25486/fx-9590-now-available-in-retail-kit-at-massive-discount
http://www.newegg.com/Product/Product.aspx?Item=N82E16819113351 with included aio cooling kit.
this price is way better than the previous, utterly worthless $800+ tag. although, anyone who got suckered by amd into buying centurions (or centurion pcs) at launch prices would be feeling ripped off now. :lol: i still consider it worthless (edit2: until it sells under $200 sans cooler, $220 with aio cooler), but people looking for stock 4.7 ghz fx + cooler would be happy(ier). this will make a great space heater for the winter.

edit: i wonder if these price drops are a sign for soon-to-be-launched high performance (amd) cpus. 6/8 core SR-FX, anyone? i can only hope....

 

[-Fran-]

I'm about as bored as a board with ARM. Lets move onto this...

http://www.eteknix.com/nvidia-origin-pc-bribery-dealings/

What are your guys' thoughts? Bribery, or does Nvidia really flip the Bill?

http://semiaccurate.com/2013/10/07/nvidias-program-get-oems-like-origin-pc-dump-amd-called-tier-0/

I've been trying to follow up this topic, but no one has put conclusive proof for Charlie's claims.

If Origin doesn't get all the things Charlie said, they were either screwed by nVidia or they really dumped AMD by themselves. The strongest indication they were tempted by nVidia will be if they start offering AMD Video Cards again.

Still, it doesn't surprise me one bit given the track record with TWIMTBP.

Cheers!

 

[juanrga]

http://www.tomshardware.com/reviews/battlefield-4-graphics-card-performance,3634-10.html

This is about what I expect going forward. Titan bottleneck is @74 FPS, so you lose visibility into CPU performance at that point.

i7-3960x is GPU bottlenecked. i5-2500k @ 4.2 is GPU bottlenecked. i5-2500k shows some signs of a CPU bottleneck (Min FPS lower then other two CPUs), but still averages out to a GPU bottleneck.

FX-8350 is not GPU bottlenecked, but still manages a higher minimum FPS then a 2500k,though a lower average. Indicates that for BF4, the 2500k and FX-8350, at stock, are roughly at the point where a GTX Titan starts to get held back, FPS wise.

And again, the i3-3220 beating the FX-4170.

This is the kind of chart I expect will be common, with only the 8-core FX line keeping up with Intel, and fast i5's being more then enough to max out performance.

Congrats by finding another non-multithreaded and GPU-limited benchmark, where a 12 threads i7 obtains the same average FP than a 4 threads i5.

And kudos to toms by another high-quality review <== sarscasm

Edit: the bold parts are in response to comments below.

 

[lilcinw]

Or you could use some critical thinking and realize that the system is GPU bottlenecked at ~74 FPS.

Also noteworthy is that the i5 and i7 at similar speeds achieve an average FPS parity but with more cores the i7 maintains a higher minimum. Similarly the stock 8350 yields higher minimums than the stock i5 even though the average is lower.

Edit: I just realized I more or less repeated what gamerk posted. I didn't read through his post before responding but that just goes to show that with a couple of minutes of thinking about what the data actually means and mentally sorting the variables the conclusions are fairly straightforward.

 

[blackkstar]

http://www.tomshardware.com/reviews/battlefield-4-graphics-card-performance,3634-10.html

This is about what I expect going forward. Titan bottleneck is @74 FPS, so you lose visibility into CPU performance at that point.

i7-3960x is GPU bottlenecked. i5-2500k @ 4.2 is GPU bottlenecked. i5-2500k shows some signs of a CPU bottleneck (Min FPS lower then other two CPUs), but still averages out to a GPU bottleneck.

FX-8350 is not GPU bottlenecked, but still manages a higher minimum FPS then a 2500k,though a lower average. Indicates that for BF4, the 2500k and FX-8350, at stock, are roughly at the point where a GTX Titan starts to get held back, FPS wise.

And again, the i3-3220 beating the FX-4170.

This is the kind of chart I expect will be common, with only the 8-core FX line keeping up with Intel, and fast i5's being more then enough to max out performance.

Congrats by finding another non-multithreaded benchmark, where a 12 threads i7 obtains the same average FP than a 4 threads i5.

And kudos to toms by another high-quality review

Oh, but juaranga, you're missing the point! gamerK found a situation where a game engine doesn't scale to all cores! So therefore games NEVER scale past a few cores and there's no point in ever adding more cores! Your problem is that you're not using gamerk logic! You need to use gamerk logic to understand this!

Another example of gamerk logic where you draw a conclusion from a single piece of evidence:

"I ate a taco, it made me sick, therefore all food makes you sick"

You need to start thinking like that, because that's how gamerk thinks! Please start making your arguments like that. They should follow a formula along the lines of "here is a cherry picked example so allow me to apply this evidence to every situation and then keep repeating it until it becomes accepted!" You also get bonus points for going "IM A GAME DEVELOPER IM AN AUTHORITY ON THIS YOUR OPINION IS INVALID!!!"

Intel moving Haswell-E to 6 core and 8 core configuration when it's still on 22nm and Intel is charging $580 for a 227mm^2 die and they're apparently going to throw that away just for the goodness of consumers, because that's what Intel does, they offer better products at great prices just to help out consumers! THATS THE ONLY REASOONNN!!!!! I'm so glad we have GamerK to explain why Intel is moving to 6 and 8 core on enthusiast market and to show us that they're going to do this just to be nice to us!

(if you can't detect the sarcasm in this post, I feel sorry for you)

 

[Ranth]

Damn guys, Gamer didn't say anything about scaling, stop jumping to conclusions He's just saying that AMD IPC is not as high as Intels, which is nothing new. (Okay he's implying that single threaded performance is more important, again nothing new). The game seems to scales up to ~6 cores, (4170 compared to 8350)

 

[Blandge]

Congrats by finding another non-multithreaded benchmark, where a 12 threads i7 obtains the same average FP than a 4 threads i5.

And kudos to toms by another high-quality review

You can't just dismiss all evidence that disagrees with your position. There's a whole lot of software that uses less than 8 threads, and you can't just dismiss it as "bad software." Not everything is parallelizable.

I would expect BF4 to use >= 8 threads in some cases (because BF3 did right?), but clearly 4 threads is sufficient for this usecase.

 

[8350rocks]

Frankly, I am not taking speculation of Kaveri into account, only the benchmarks from the early ES's. Additionally, based on what I can project, we can make assumptions from the performance of PD and get rough ideas.

Nope. You have speculated about Steamroller, Kaveri, MANTLE, new Radeons... This is one example:

Yesterday, AMD announced that Steamroller Opteron Server Architecture will be available in the second half of this year followed shortly by the client architecture (Kaveri/FX).

There is also a logic flow chart diagramming the upgraded architecture for Steamroller, and commentary on the changes.

A good read for anyone curious about what's coming.

The expected single threaded performance gain is 30% over Piledriver, and multi-threaded performance will have an expected increase on the order of 12-15%

However, when someone makes similar statement about the expected performance of Seattle or A57 using AMD sources, you reject them and solicit to the other part "real world measurements" on final silicon.

That is double standard.

If 0 != 1 {
...}
Else {
...}

If your usual code looks like that, I want to see your work

I did that because you are stuck in a loop of the first portion...and the else section is not at all true...

My parsing may be rough, but I do still know how to use if's and else's. I could have used while, however, I wanted to be different.

Being different is fine, the problem is with being wrong. I will continue replying what you _post_ instead of what you claim to know _after_ posting.

WHAT BASELINE CAN YOU POSSIBLY HAVE TO ASSUME PERFORMANCE GAINS FOR SEATTLE????

You can make assumptions based loosely on PD as to what SR will be, they are the same architecture but tweaked.

What Server capable 64 bit ARM CPU does AMD have out now for you to make any assumptions from?

/ARM BS

 

[Ags1]

I'm skipping all these messages filled with quotes and counterquotes. In the history of forum wars, has anyone ever seen the error of their ways? Move on, guys!

 

[juanrga]

Congrats by finding another non-multithreaded benchmark, where a 12 threads i7 obtains the same average FP than a 4 threads i5.

And kudos to toms by another high-quality review

You can't just dismiss all evidence that disagrees with your position. There's a whole lot of software that uses less than 8 threads, and you can't just dismiss it as "bad software." Not everything is parallelizable.

I would expect BF4 to use >= 8 threads in some cases (because BF3 did right?), but clearly 4 threads is sufficient for this usecase.

It is not about dismissing but about filtering quality. Also no sure from where you got your "bad software" claim.

Frankly, I am not taking speculation of Kaveri into account, only the benchmarks from the early ES's. Additionally, based on what I can project, we can make assumptions from the performance of PD and get rough ideas.

Nope. You have speculated about Steamroller, Kaveri, MANTLE, new Radeons... This is one example:

Yesterday, AMD announced that Steamroller Opteron Server Architecture will be available in the second half of this year followed shortly by the client architecture (Kaveri/FX).

There is also a logic flow chart diagramming the upgraded architecture for Steamroller, and commentary on the changes.

A good read for anyone curious about what's coming.

The expected single threaded performance gain is 30% over Piledriver, and multi-threaded performance will have an expected increase on the order of 12-15%

However, when someone makes similar statement about the expected performance of Seattle or A57 using AMD sources, you reject them and solicit to the other part "real world measurements" on final silicon.

That is double standard.

WHAT BASELINE CAN YOU POSSIBLY HAVE TO ASSUME PERFORMANCE GAINS FOR SEATTLE????

Several: Opteron-X, JG, SB, A9, A15... and GFLOPs/DMIPS

When AMD says that Steamroller will be about 30% faster PD, you repeat the number in many threads. When AMD says that its Seattle will be 2x-4x faster than Opteron-X you start this show. It is the same AMD in both cases, but you accept one claim whereas reject angrily the other and attack to anyone who disagree with you: double standard.