AMD CPU speculation... and expert conjecture

[tracker45]

Bulldozer is what held AMD back.

A complete generation of CPUs wasted.

 

[gamerk316]

But here's the thing: CoD on the PS4 was the ONE game running 1080p/60FPS that's been confirmed so far. So neither console is capable of smooth 1080p/60FPS gameplay.

That's significant, since that basically means next-gen consoles are already being tapped out. Which means the consoles WILL be using low-level optimizations to squeeze out performance going forward, which makes the ease of porting from consoles to PC no different then the previous generation of hardware, as all that low-level coding will have to be manually ripped out and replaced. So the advantage of moving to X86 has already gone out the window.

 

[8350rocks]

But here's the thing: CoD on the PS4 was the ONE game running 1080p/60FPS that's been confirmed so far. So neither console is capable of smooth 1080p/60FPS gameplay.

That's significant, since that basically means next-gen consoles are already being tapped out. Which means the consoles WILL be using low-level optimizations to squeeze out performance going forward, which makes the ease of porting from consoles to PC no different then the previous generation of hardware, as all that low-level coding will have to be manually ripped out and replaced. So the advantage of moving to X86 has already gone out the window.

Erm...no...you forget the massive alterations made to the custom hardware in the consoles that developers have never had access to before now.

So, I am not at all surprised that the first batch of games are not maxing potential.

In fact, every time a new console comes out, this same discussion is had. Then by the end of the generation, people say..."well I guess they had to figure it out first..."

So, cry wolf all you want...but the reality is, this is just growing pains.

 

[con635]

Consoles tapped out at first games released lol I was away from pc gaming for about 6 years because of the 360/ps3, the games on them vs the hardware is incredible and the new consoles are much, much better all round. It'll be years before they're close to tapped out.

 

[juanrga]

About a 3m APU.

Maybe AMD is switching from making things and waiting for people to buy them to making things to order? I mean if Valve or a big customer showed up and said they wanted a solid APU with 3m/6c and a big GPU, what's stopping AMD from doing it?

Cost.

 

[juanrga]

Spoiler

Spoiler

The slide shows the APU giving ~92% of the FX-8350 performance in that gaming test. Say us which will be the performance in "multiplayer"?

athlon x4 750k = trinity. its found at the bottom by the pentium g's, overclocked to 4.5 ghz and still getting beat by 4350 @ 4.2 ghz and completely obliterated by the 8350.

you really need to quit trying to live by marketing slides. best case scenario != all cases.

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

Let us forget that pclab is not neutral regarding AMD vs Intel, even accepting your non-marketing benchmark

trinity != richland

athlon != APU

GTX 780 != R9 280X

...

I would like to see multiplayer benchmarks in the same conditions that the benchmark in the slide #13. That is, I would like to see benchmarks of the new APU with the new APU drivers, the new APU optimized Windows, 2133 MHz ram, and paired with a R9 280X...

I estimate it would obtain ~70% of the performance of 8350.

so basically what your getting at is you only want to see kaveri running highly specialized software to take advantage of HSA and ignore any programs that are in the market today.

This is because in the programs of today, Kaveri will suck hind teat.

HSA was not even mentioned

HSA is the only way kaveri will even remotely come close to the 8350.

What part of "HSA was not even mentioned" you didn't understand?

As for clock speeds, 3.5 ghz sounds about right, look at all of Intel's cpus on bulk.

20% ipc gain -17% clock speed = a massive 3% gain over richland, and still behind the 4350.

SARCASM

Kaveri will be slower than bulldozer FX 4150, everyone will say "ha ha!", and then AMD will release a 21-core Phenom IV @ 6.66GHz (on air; 2.99GHz on hot water). The TDP will be of 999W. This chip will be manufactured in the new FD-SOI 20nm at Globalfoundries.

END SARCASM

 

[juanrga]

So is AMD not releasing an FX class steam roller cpu?

99.9% No

i probably am late and i could be totally wrong with this but i noticed that purpoted picture of kaveri ES showed ZD on its name, which implies Zambezi (bulldozer). piledriver had PD or FD i think. so may the kaveri pic floating around is not real?

have i understood right?

The string in the picture of the kaveri ES was

ZD356195I4468

The coding is:

Z - Engineering Sample
D - Desktop
3561 - Model Number, base freq. 3.5Ghz
95 - Socket (FM2+)
I - Package
first '4' - Number of cores
second '4' - Cache Size (2m L2, no L3)
68 - Revision

shouldn't hsa be more or less invisible to the o.s.? if the implementation is invisible to the os, then the softwares shouldn't have issues running. my fear is that microsoft (i.e. windows) itself might get in the way, while free linux distros get very slow optimizations. except may be, steam o.s., should they choose to optimize for hsa.

I don't know what you try to say here, but linux is already HSA enabled. I don't know about Windows.

 

[juanrga]

We don't know good mantle is, more specifically, how many more fps it gives over Nvidia ?

5 ? 10 ? 15 ? 20 ? 25 ? 30 ?

Probably a handful. The overhead realated to draw calls has been reduced significantly over the years, so in the grand scheme of things, we're talking *maybe* 1-2 FPS here. The time the GPU spends doing nothing due to the overhead of draw calls, in comparison to the 16ms screen display interval, is trivially small.

Point being: Optimizing draw calls is optimizing the wrong thing. You can improve performance 10x, but it won't be causing any significant effect to FPS by itself.

Probably between 30% and 50% more FPS. Therefore if the DX game gives 3 FPS the MANTLE version will give 1-2 FPS more ;-)

Yes, DX 11 reduced the overhead, but it is still about 5-10x and the reduction implies the use of batch calls. If you want to draw 100 independent objects the overhead returns.

 

[blackkstar]

We don't know good mantle is, more specifically, how many more fps it gives over Nvidia ?

5 ? 10 ? 15 ? 20 ? 25 ? 30 ?

Probably a handful. The overhead realated to draw calls has been reduced significantly over the years, so in the grand scheme of things, we're talking *maybe* 1-2 FPS here. The time the GPU spends doing nothing due to the overhead of draw calls, in comparison to the 16ms screen display interval, is trivially small.

Point being: Optimizing draw calls is optimizing the wrong thing. You can improve performance 10x, but it won't be causing any significant effect to FPS by itself.

Probably between 30% and 50% more FPS. Therefore if the DX game gives 3 FPS the MANTLE version will give 1-2 FPS more ;-)

Yes, DX 11 reduced the overhead, but it is still about 5-10x and the reduction implies the use of batch calls. If you want to draw 100 independent objects the overhead returns.

50% is the difference between 40fps constant and 60fps constant.

Also @gamerk, you do realize it's a release game. Compare Call of Duty 2 to the last Crysis on 360.

And no one in this thread expected a transition to many weak cores from a few strong cores to happen over night. I wouldn't expect things to really transition well at least until the second half of 2014, probably more towards the end of 2015 honestly.

Do you not find it odd that Sony and AMD are running around going "we're doing what game developers want!" and then you are a self-proclaimed game developer who is completely opposed to what AMD, Sony, and MS are doing?

 

[etayorius]

So Juanrga, how much performance is speculated from Kaveri compared to FX-8350 in regards to IPC?

 

[tracker45]

So Juanrga, how much performance is speculated from Kaveri compared to FX-8350 in regards to IPC?

how much performance is speculated from Kaveri compared to FX-8350 in regards to FPS for battlefield 4 ?

 

[griptwister]

So Juanrga, how much performance is speculated from Kaveri compared to FX-8350 in regards to IPC?

how much performance is speculated from Kaveri compared to FX-8350 in regards to FPS for battlefield 4 ?

Look at the i5 2500K benchmark... Now remove 6-10% from that.

 

[youcanDUit]

https://vts.inxpo.com/scripts/Server.nxp?LASCmd=AI:4;F:QS!10100&ShowKey=16356

 

[sora]

So Juanrga, how much performance is speculated from Kaveri compared to FX-8350 in regards to IPC?

how much performance is speculated from Kaveri compared to FX-8350 in regards to FPS for battlefield 4 ?

Look at the i5 2500K benchmark... Now remove 6-10% from that.

Honestly, Intel will probably continue to outperform AMD in benchmarks.

 

[griptwister]

So Juanrga, how much performance is speculated from Kaveri compared to FX-8350 in regards to IPC?

how much performance is speculated from Kaveri compared to FX-8350 in regards to FPS for battlefield 4 ?

Look at the i5 2500K benchmark... Now remove 6-10% from that.

Honestly, Intel will probably continue to outperform AMD in benchmarks.

wow

such insite

 

[kansasboy001]

Lets not jump to conclusions, Steam Roller will be announced in 5 days, Monday the 11th we will know if there is an FX like SR chip as well as if it can beat the crap out of a few i7s. (fingers crossed)

 

[esrever]

And in other news: CoD on the PS4 having significant framerate issues.

Source? IGN? Kotaku? Toms?

Nope: The BBC:

http://www.bbc.co.uk/news/technology-24834842

I'm calling it: The CPU's in this generation of consoles are officially cripples, and PC gaming is now held back for a decade as a result.

They should have just stuck with 720p like the xbox.

 

[de5_Roy]

Lets not jump to conclusions, Steam Roller will be announced in 5 days, Monday the 11th we will know if there is an FX like SR chip as well as if it can beat the crap out of a few i7s. (fingers crossed)

steamroller's formal unveiling was way back in 2012, along with kaveri's (apu with sr core).

i have not read any official amd statement saying kaveri/steamroller will kill/destroy/rape/annihilate/!@#$ intel/corei5/i7. statements containing k/d/r/a/!@#$ - are (clueless) fanboy hype.

however, close to launch and afterwards, 'leaks' and 'leaked benchmarks' showing e.s. benches will show up on different forums. there will also be 'leaks' showing official benchmarks where a kaveri apu will significantly outperform an intel counterpart. according to benchmarketing, amd will pit an a10 kaveri against a $190 haswell core i5(hd4400/4600 lol) or a haswell core i3. if they're really feeling bold, might even compare it against an iris pro sku. if the negative cpu rumors are correct, there will be very few cpu benches. the ones we might see might have strong hsa optimizations or heterogeneous computing like opencl, gpu accelerated video encoding, or de/compression benches. another sign would be amd splitting up p/reviews like they did with trinity.

 

[juanrga]

So Juanrga, how much performance is speculated from Kaveri compared to FX-8350 in regards to IPC?

There is so many architectural improvements (both inside and outside the module) to make an accurate prediction. In my article about Kaveri I mentioned estimates between ~20% and ~40% IPC gain (integer workload). See footnote #2 in my article for an explanation of how the 40% figure is derived.

If recent leaked benchmarks are legit, a Kaveri engineering sample was found to be 31% faster (integer workload) compared to FX Piledriver clock for clock.

Some sites claim that Kaveri will be competing with i7-4770k. I think that they are exaggerating.

 

[juanrga]

close to launch and afterwards, 'leaks' and 'leaked benchmarks' showing e.s. benches will show up on different forums. there will also be 'leaks' showing official benchmarks where a kaveri apu will significantly outperform an intel counterpart. according to benchmarketing, amd will pit an a10 kaveri against a $190 haswell core i5(hd4400/4600 lol) or a haswell core i3. if they're really feeling bold, might even compare it against an iris pro sku. if the negative cpu rumors are correct, there will be very few cpu benches. the ones we might see might have strong hsa optimizations or heterogeneous computing like opencl, gpu accelerated video encoding, or de/compression benches. another sign would be amd splitting up p/reviews like they did with trinity.

'Benchmarketing' (October AMD talk) compared the new A10-6790k APU to Haswell i5-4440 (PCMARK 8, BaseMarkCL, 3DMARK Fire Strike). They also compared the A10-6790k to Haswell i5-4670K under Battlefield 4 (iGPUs) but, and this is the interesting part, AMD also compared the new A10-6790K APU to FX-6350 and FX-8350, showing that can compete with both FX CPUs under BF4 (Suez Map) using a R9 280X @ ultra 1080p.

Leaked benchmarks (if they are legit) compared the new Kaveri APU to Bulldozer and Piledriver FX CPUs. I am curious on why AMD are comparing Kaveri to FX instead of Trinity/Richland APUs. Could it be related to AMD planning to replace FX by APUs?

Leaked benchmarks (if they are legit) already compared new Kaveri to i7 Haswell Iris Pro. The Kaveri iGPU was between a 50% and a 400% faster than Iris Pro iGPU.

 

[de5_Roy]

keep in mind that these are speculations replied to my initial speculation.

'Benchmarketing' (October AMD talk) compared the new A10-6790k APU to Haswell i5-4440 (PCMARK 8, BaseMarkCL, 3DMARK Fire Strike). They also compared the A10-6790k to Haswell i5-4670K under Battlefield 4 (iGPUs) but, and this is the interesting part, AMD also compared the new A10-6790K APU to FX-6350 and FX-8350, showing that can compete with both FX CPUs under BF4 (Suez Map) using a R9 280X @ ultra 1080p.

bf4 offline single player is gpu bound as long as the cpus can execute 4 threads.

Leaked benchmarks (if they are legit) compared the new Kaveri APU to Bulldozer and Piledriver FX CPUs. I am curious on why AMD are comparing Kaveri to FX instead of Trinity/Richland APUs. Could it be related to AMD planning to replace FX by APUs?

what benchmarks?
i am sure they'll pit the new product against something that they can beat.
i haven't seen anything that said amd was the one who ran the benches. if anything, amd will disavow such information until they launch.

Leaked benchmarks (if they are legit) already compared new Kaveri to i7 Haswell Iris Pro. The Kaveri iGPU was between a 50% and a 400% faster than Iris Pro iGPU.

where?
spec. comparisons and percentage padded on existing benchmarks don't count.

 

[juanrga]

keep in mind that these are speculations replied to my initial speculation.

'Benchmarketing' (October AMD talk) compared the new A10-6790k APU to Haswell i5-4440 (PCMARK 8, BaseMarkCL, 3DMARK Fire Strike). They also compared the A10-6790k to Haswell i5-4670K under Battlefield 4 (iGPUs) but, and this is the interesting part, AMD also compared the new A10-6790K APU to FX-6350 and FX-8350, showing that can compete with both FX CPUs under BF4 (Suez Map) using a R9 280X @ ultra 1080p.

bf4 offline single player is gpu bound as long as the cpus can execute 4 threads.

There are two questions here. The first that people who was playing exactly that will not need to purchase an expensive FX-8350 to get ~8% performance gain. I tried to estimate the performance in multiplayer and the new A10-6790k APU could be about 70% of the FX-8350.

The second question is that AMD wouldn't make a slide such as this

if their plan was to sell lots of FX-6350 and FX-8350 to gammers. That slide seems to be AMD way to say "hey we are migrating to APUs". Most of the OEMs attending the October talk will get that impression. I would do it.

I suppose that the hidden message behind that slide is the reason why almost everyone here ignored that slide, and the guy who commented on it, interpreted it as AMD will make a 10 core FX SR...

Leaked benchmarks (if they are legit) compared the new Kaveri APU to Bulldozer and Piledriver FX CPUs. I am curious on why AMD are comparing Kaveri to FX instead of Trinity/Richland APUs. Could it be related to AMD planning to replace FX by APUs?

what benchmarks?
i am sure they'll pit the new product against something that they can beat.
i haven't seen anything that said amd was the one who ran the benches. if anything, amd will disavow such information until they launch.

Leaked benchmarks (if they are legit) already compared new Kaveri to i7 Haswell Iris Pro. The Kaveri iGPU was between a 50% and a 400% faster than Iris Pro iGPU.

where?
spec. comparisons and percentage padded on existing benchmarks don't count.

 

[gamerk316]

But here's the thing: CoD on the PS4 was the ONE game running 1080p/60FPS that's been confirmed so far. So neither console is capable of smooth 1080p/60FPS gameplay.

That's significant, since that basically means next-gen consoles are already being tapped out. Which means the consoles WILL be using low-level optimizations to squeeze out performance going forward, which makes the ease of porting from consoles to PC no different then the previous generation of hardware, as all that low-level coding will have to be manually ripped out and replaced. So the advantage of moving to X86 has already gone out the window.

Erm...no...you forget the massive alterations made to the custom hardware in the consoles that developers have never had access to before now.

So, I am not at all surprised that the first batch of games are not maxing potential.

In fact, every time a new console comes out, this same discussion is had. Then by the end of the generation, people say..."well I guess they had to figure it out first..."

So, cry wolf all you want...but the reality is, this is just growing pains.

But its also worth mentioning the CDPR already stated they've basically maxed the consoles out in developing Witcher 3. Which, given how most of the optimizations you do on PC's should apply to consoles this go around, is significantly worrying.

Sure, you can probably squeeze a lot more out with low level optimizations, but that defeats the purpose of going to X86 in the first place (easy porting).

The fact that no game on either console seems to be able to handle 1080p/60FPS, despite the massive jump in hardware from the last gen, is not a promising sign going forward.

 

[de5_Roy]

Spoiler

There are two questions here. The first that people who was playing exactly that will not need to purchase an expensive FX-8350 to get ~8% performance gain. I tried to estimate the performance in multiplayer and the new A10-6790k APU could be about 70% of the FX-8350.

The second question is that AMD wouldn't make a slide such as this

http://juanrga.com/en/images/A10-6790K-Battlefield-4.png

if their plan was to sell lots of FX-6350 and FX-8350 to gammers. That slide seems to be AMD way to say "hey we are migrating to APUs". Most of the OEMs attending the October talk will get that impression. I would do it.

I suppose that the hidden message behind that slide is the reason why almost everyone here ignored that slide, and the guy who commented on it, interpreted it as AMD will make a 10 core FX SR...

first one: you don't have to estimate. some sites should have apu vs fx in the m.p. benches.

2nd one: amd made the slide to pitch their new 6790k apu. it is simple benchmarketing - make the new one look better, as much as possible. their plan is not to sell fx in that slide, it is to sell the 6790k. since all the contenders are amd's, whichever sells, amd makes moniez.

i 'm not gonna waste time with semantics. get whatever impression you want.

http://cdn.wccftech.com/wp-content/uploads/2013/10/AMD-Kaveri-Steamroller-APU.png

could mean a lot of things but that sure doesn't look like amd's own. anywho, the integer processing is up and the chart says(if they're not lying) - fx (with L3 cache) vs kaveri (sans L3 cache). may be the L3 cache hits are considerably down and the cores are being properly utilized. it could also mean that the software they tested with causes less L3 cache hits in kaveri.

http://www.hwbox.gr/members/2195-albums161-picture15643.png

link to the article and test configs? the first think i noticed was that the memory bandwidth(of whatever it is) being incorrectly mentioned.

 

[juanrga]

But here's the thing: CoD on the PS4 was the ONE game running 1080p/60FPS that's been confirmed so far. So neither console is capable of smooth 1080p/60FPS gameplay.

That's significant, since that basically means next-gen consoles are already being tapped out. Which means the consoles WILL be using low-level optimizations to squeeze out performance going forward, which makes the ease of porting from consoles to PC no different then the previous generation of hardware, as all that low-level coding will have to be manually ripped out and replaced. So the advantage of moving to X86 has already gone out the window.

Erm...no...you forget the massive alterations made to the custom hardware in the consoles that developers have never had access to before now.

So, I am not at all surprised that the first batch of games are not maxing potential.

In fact, every time a new console comes out, this same discussion is had. Then by the end of the generation, people say..."well I guess they had to figure it out first..."

So, cry wolf all you want...but the reality is, this is just growing pains.

But its also worth mentioning the CDPR already stated they've basically maxed the consoles out in developing Witcher 3. Which, given how most of the optimizations you do on PC's should apply to consoles this go around, is significantly worrying.

Sure, you can probably squeeze a lot more out with low level optimizations, but that defeats the purpose of going to X86 in the first place (easy porting).

The plan is very simple. Optimize for consoles, then port those optimizations to PCs. Did you heard about MANTLE? AMD and DICE said that one of its goals is to port console optimizations to PCs with GCN hardware.

The fact that no game on either console seems to be able to handle 1080p/60FPS, despite the massive jump in hardware from the last gen, is not a promising sign going forward.

Similar negative claims were made with first gen of games for the PS3 and Xbox360, subsequent generations showed how wrong they were.

Even Nvidia is lately praising the new consoles and how will raise PC gaming baseline...