Should AMD have focused on a Steamroller high TDP product line instead of Vishera?

[guskline]

Has there been any news lately on where the SteamRoller really is in the development cycle?

 

[sushiwarrior]

Has there been any news lately on where the SteamRoller really is in the development cycle?

Long done obviously. Probably was done months ago. Review samples shipped out, production is probably ramping up now, should have stock for Kaveri launch.

 

[Exophase]

Long done obviously. Probably was done months ago. Review samples shipped out, production is probably ramping up now, should have stock for Kaveri launch.

Do you have evidence for any of this?

 

[monstercameron]

Do you have evidence for any of this?

lisa su said as much at computex

 

[inf64]

Has there been any news lately on where the SteamRoller really is in the development cycle?

It's usually one year from design finish to start of shipments if everything is fine. So if they plan to start shipping Kaveri in Q4 this year then it was finished in Q4 of 2012(and they started sending ES to partners around that time frame or in Q1 2013 for further platform validation).

 

[galego]

Wow, seems like there were a lot of AMD-specific optimizations at launch, its a Gaming Evolved title after all. Patch 1.3 puts things ''back to normal'', i7 4C/8T easily outperforming 8-core (twice the die size) FX8350.

No. The same German site from where you got the numbers stated in their review of the version 1.0 that Crysis 3 was not optimized for AMD in any way and that the measured advantage over Intel chips was merely due to superior multithreading of the AMD chip.

This 3.0 patch not only looks as heavy optimization for Intel chips but offers a regression on performance for the FX chips. Some FX-8350 user is complaining how he played at 60 FPS and now performance magically dropped to 40-50 FPS max.

http://www.mycrysis.com/forums/viewtopic.php?f=58&t=63842

Dropping performance is not putting "things back to normal", unless by "normal" you mean the well-known tactic of adding Cripple_AMD functions to software

wait, in which one of those benchmarks does intel get creamed? the point of my post was to address the ludicrous notion of a fantasy product when

1. while current AMD parts can compete with intel in a few select benchmarks, the 4 module AMD part still loses to the 4 core intel in most of them, multithread included

and

2. that a 6 module AMD (50% more) part would not just somehow beat an 8 core (100% more) intel, but would cream it...

I'm sorry, but that's just delusional

I'd really love for AMD to put out these parts though, even if I don't expect miracles from them like you guys do.

1. I gave you some benchmarks where the 4M AMD is faster than the 4C Intel including Haswell. In one of the test the 8350 is a 20% faster than the 3770k. A 20% gain is a noticeable figure. And that 20% is far from being the maximum gap.

I added in a posterior post benchmarks where a 4M 8350 is faster than a 6C 3960X from Intel.

2. His claim was about a Steamroller-based module, Steamroller will be a bigger advance over Piledriver than Haswell was over Ivy Bridge.

 

[Exophase]

I agree, one person says performance dropped from 60fps to 40-50fps without showing any actual measurement results - it must mean that the developers deliberately added new functions to cripple AMD. Oh, but they didn't do a very thorough job since in the one real measurement we do have only a tiny decrease can be witnessed.

Do you realize how silly this sounds? It's one thing to use a compiler that does a less than sincere job at picking good code paths. It's another thing to release a patch that moves to a worse compiler. I'm sure QA would have been right behind that idea. The next idea in line would be that they - the developers, not Intel - themselves deliberately sabotaged AMD processors in their own code. Why in god's name would they do that? Are you really going to have me believe Intel bribed them to do something like this? Even if they were that sleazy it's really not worth the money.

Here's a crazy theory - maybe they upgraded SSE4.x code to use (256-bit) AVX instead. Big win on Intel processors supporting AVX, not so big win on Bulldozer based processors where AVX can actually be slower. If that's what happened then the only one who crippled AMD is AMD.

Of course that's just one possibility. We simply don't know. But that won't stop the conspiracy theories from being the obvious choice, right..

 

[galego]

There is more than one gamer with performance drops. And I think you missed the smile just after the joke about the Cripple_AMD function.

My joke was motivated by his claim about puting things "back to normal" and his claim about "AMD-specific optimizations at launch".

 

[Vesku]

"PCSuperStore.com has the FX-9370 listed for $576.27 and the "5GHz" FX-9590 at a pricey $920.31. Both chips are said to be unlocked, and the part numbers are FD9370FHHKWOF and FD9590FHHKWOF, respectively. Retail-boxed AMD processors usually have "BOX" somewhere in the model code, so I assume these are tray units without heatsinks, manuals, or cool stickers included."

http://techreport.com/news/24968/fx-9000-processors-listed-at-u-s-e-tailer

 

[Exophase]

There is more than one gamer with performance drops. And I think you missed the smile just after the joke about the Cripple_AMD function.

My joke was motivated by his claim about puting things "back to normal" and his claim about "AMD-specific optimizations at launch".

I don't know what you're thinking when you put the sunglasses smiley after a comment that you've said several times in other contexts. It looks like you're giving them sarcastic approval, not saying the thing in jest. Especially when you call the thing well-known.

Whatever the case, you clearly think the old version is valid for comparison and the new version is invalid, which should only be the case if the newer version is sabotaging AMD. It'd be one thing if you took the best scores for each between the two but you instead take the best AMD and worst Intel.

Yes, others have said there are performance degradations - but that doesn't mean that they're vouching for this claim that of going from 60+ to 40-50. That person could have been measuring different things or just remembering incorrectly. One account like this is barely worth anything.

 

[galego]

I don't know what you're thinking when you put the sunglasses smiley after a comment that you've said several times in other contexts. It looks like you're giving them sarcastic approval, not saying the thing in jest. Especially when you call the thing well-known.

Whatever the case, you clearly think the old version is valid for comparison and the new version is invalid, which should only be the case if the newer version is sabotaging AMD. It'd be one thing if you took the best scores for each between the two but you instead take the best AMD and worst Intel.

Yes, others have said there are performance degradations - but that doesn't mean that they're vouching for this claim that of going from 60+ to 40-50. That person could have been measuring different things or just remembering incorrectly. One account like this is barely worth anything.

Your idea of comparing a version of the game clearly optimized for Intel with a version of the game not-optimized for any processor makes me think that my ironic comment about Cripple_AMD function was a correct choice, because the well-known biased benchmarks are precisely comparing code optimized for Intel with code not optimized for AMD.

Fair comparisons would be unoptimized vs unoptimized or optimized vs optimized.

The user that I linked is not alone. Many others are obtaining massive drops in performance of up to -50 or -60 FPS. Are all remembering incorrectly? Please move on. I am doing it just now.

 

[frozentundra123456]

Your idea of comparing a version of the game clearly optimized for Intel with a version of the game not-optimized for any processor makes me think that my ironic comment about Cripple_AMD function was a correct choice, because the well-known biased benchmarks are precisely comparing code optimized for Intel with code not optimized for AMD.

Fair comparisons would be unoptimized vs unoptimized or optimized vs optimized.

The user that I linked is not alone. Many others are obtaining massive drops in performance of up to -50 or -60 FPS. Are all remembering incorrectly? Please move on. I am doing it just now.

It is a gaming evolved title. I am sure crytek intentionally nerfed the game for one of their own sponsors.

 

[Rvenger]

"PCSuperStore.com has the FX-9370 listed for $576.27 and the "5GHz" FX-9590 at a pricey $920.31. Both chips are said to be unlocked, and the part numbers are FD9370FHHKWOF and FD9590FHHKWOF, respectively. Retail-boxed AMD processors usually have "BOX" somewhere in the model code, so I assume these are tray units without heatsinks, manuals, or cool stickers included."

http://techreport.com/news/24968/fx-9000-processors-listed-at-u-s-e-tailer

Oh man, I just checked Ingram. Its gonna be $900 for the 9590. AMD just shot themselves in the foot big time! EDIT: And no heatsink!

 

[BallaTheFeared]

I agree, one person says performance dropped from 60fps to 40-50fps without showing any actual measurement results - it must mean that the developers deliberately added new functions to cripple AMD. Oh, but they didn't do a very thorough job since in the one real measurement we do have only a tiny decrease can be witnessed.

Well to be fair AMD's market share is so small that one person might represent a bulk of the AMD cpu users still playing the game.

The simplest explanation is that in 1.0 the game did not support HT, whereas in 1.3 it now does support HT.

 

[FwFred]

And no heatsink!

To be fair, what type of in-box heatsink would handle 220W?

I don't really understand the target market for these CPUs

 

[Rvenger]

To be fair, what type of in-box heatsink would handle 220W?

I don't really understand the target market for these CPUs

Needs an H100 as a stock cooler.

 

[sushiwarrior]

The simplest explanation is that in 1.0 the game did not support HT, whereas in 1.3 it now does support HT.

And would you please explain how the game was "optimized for HT"? I'd really like to hear your explanation for that

 

[BallaTheFeared]

Did I say optimized?

Given the increase in performance, and similar performance to non ht chips prior to 1.3 I used my gray matter to formulate a reasonable explanation outside of the typical "compiled for Intel, cripple_amd" sad panda excuses.

 

[Chiropteran]

Many others are obtaining massive drops in performance of up to -50 or -60 FPS.

similar performance to non ht chips prior to 1.3

A loss of 50 to 60 fps is not similar performance.

 

[sushiwarrior]

Did I say optimized?

Given the increase in performance, and similar performance to non ht chips prior to 1.3 I used my gray matter to formulate a reasonable explanation outside of the typical "compiled for Intel, cripple_amd" sad panda excuses.

You can't simply "support HT", HT is the same as having more cores (that don't perform as fast). So what you're saying is they added support for more cores, yet somehow... AMD lost performance... even though they have more cores... And Intel gained performance with more cores... Geeze, seems fishy... :\

 

[Exophase]

The simplest explanation is that in 1.0 the game did not support HT, whereas in 1.3 it now does support HT.

The only way this makes sense to me is if the game was choosing a lighter thread path because it thought that fewer hardware threads were available. Or maybe they changed affinity to better group threads for HT. But I don't think that'd change things much.

It's true though, other people are reporting performance degradation:

http://www.mycrysis.com/forums/topic63819

.. on an Intel CPU. So much for this being an Intel vs AMD issue. And before anyone says anythnig, people are reporting problems on nVidia vs AMD GPUs too. People are also reporting it's working fine for both.

 

[BallaTheFeared]

A loss of 50 to 60 fps is not similar performance.

Think about where you're getting your information from before you start forming conclusions based on it.

 

[cytg111]

they should make both? show em if u get em'

 

[galego]

<snip flame>

The simplest explanation is that in 1.0 the game did not support HT, whereas in 1.3 it now does support HT.

And the FPS of the i7-3770k (HT activated) being a >12% higher over the FPS of the i5-3570k in the 1.0 version was due to what? magic?

 

[BallaTheFeared]

Offloading background tasks from the OS to the HT threads, and a clock speed advantage.


Check.