AMD CPU speculation... and expert conjecture

[esrever]

NO ONE CODES LIKE THIS. Period. No one is going to create entirely different game engines for different core counts (because that is EXACTLY what you are advocating). Its stupid, idiotic, and doesn't gain you anything. Taking that argument to its extreme, I suppose we should all be making a single threaded path for all titles, for those still running single core CPU's!

Any work that is reasonably parallel is put into a separate thread. Period. We've been doing so since the 80's on single core PCs. Threading has NEVER been an issue, because its trivial. Heck, even cache thrashing is less a problem now that most CPU's have a very large L3 cache that all cores have access too.

actually they can do this on consoles.

 

[JAYDEEJOHN]

No windows

 

[Cazalan]

actually they can do this on consoles.

With the relatively low performance Jaguar's they'll have to. Those cores will be lucky to match an Athlon X2 clock for clock.

 

[-Fran-]

With the relatively low performance Jaguar's they'll have to. Those cores will be lucky to match an Athlon X2 clock for clock.

Depends. They do have the advantage of being "custom".

So they could increase the IPC for the reduced instructions the APUs can support and use the remaining die area to add different improving optimization stuff for the supported bunch.

It's like...

Wait for it...

A track car and a daily driver.

BOOM! hahaha

Cheers!

 

[wh3resmycar]

maybe then we can see GPU bound console ports.

 

[palladin9479]

The "big" feature of DX11 is multithreaded rendering, which allows multiple threads to render to a single GPU. In theory, this allows you to get away from having one monster render thread, which would have to execute sequentially, which as you can imagine, kills the ability of the CPU to scale to any reasonable degree. Hence why more or less any pre-DX11 game (or any game that still has a DX9 path) doesn't scale well.

The issue is, the few times I had a chance to play with multithreaded rendering, due to how rendering engines were designed, there wasn't much you could make parallel, as the rendering code went through stages where each stage took the previous stages output as its input. Its possible Crytek managed to build a rendering engine that gets around this, which would allow significantly better scaling, but as of the last time I worked on a game engine, this wasn't the case. [If this IS what's happening, I'd be REALLY interested to see how they structured the program to pull that off.]

Yet you said this as all impossible because you just can't code for multiple threads.

And I believe I mentioned that in order for a program to be programmed to run parallel that the engineers must redefine the problem as one that can be done in parallel. Trying to take an algorithm / code flow that was designed to be run in serial and making it parallel will quickly run into problems. You need to start from scratch and rework it so that you create opportunities for parallelization.

 

[JAYDEEJOHN]

http://www.planet3dnow.de/cgi-bin/newspub/viewnews.cgi?id=1361486916
http://translate.google.com/translate?hl=en&sl=sv&tl=en&u=http%3A%2F%2Fwww.sweclockers.com%2Fnyhet%2F16597-amd-temash-specifikationer-och-prestanda

 

[griptwister]

http://www.planet3dnow.de/cgi-bin/newspub/viewnews.cgi?id=1361486916
http://translate.google.com/translate?hl=en&sl=sv&tl=en&u=http%3A%2F%2Fwww.sweclockers.com%2Fnyhet%2F16597-amd-temash-specifikationer-och-prestanda

Ha! What's his face called it. Jaguar APU cores are faster than Piledriver cores.

 

[Cazalan]

What's special about turbo dock? It's just a performance profile that laptops have had for ages. When they're plugged in they switch out of energy saver mode.

 

[Cazalan]

Depends. They do have the advantage of being "custom".

So they could increase the IPC for the reduced instructions the APUs can support and use the remaining die area to add different improving optimization stuff for the supported bunch.

They already said Jaguar is 15% IPC gain over Brazos 2.0, and 10% clock speed gain so about 25% faster. Then multiply by 4 for the extra cores.

http://www.cpubenchmark.net/cpu_lookup.php?cpu=AMD+E2-1800+APU

In this benchmark that would put you at the level of an FX-4130.

http://www.cpubenchmark.net/cpu.php?cpu=AMD+FX-4130+Quad-Core&id=1807

The custom part is the GDDR5 and much larger GPU compared to what Kabini/Temash are going with. Certainly that will help some but the memory interface doesn't do as much for the APUs compute capability, that really only helps the GPU side.

 

[mayankleoboy1]

So the Jaguar cores are directly using GDDR5 ? :O

 

[de5_Roy]

afaik, kabini and later apus are rumored to have hsa capabilities, and gcn already has some of it built in. if it's customized in a certain way, imo the apus can use gddr5 memory. i am very interested to know how it's doing it.

 

[mayankleoboy1]

With Steamroller and GCN2 will come much improved HSA. The CPU and GPU will share a common memory address space.

Edit : Dont know how relevant this is anymore. Because AMD plans have changed some.

 

[kettu]

So the Jaguar cores are directly using GDDR5 ? :O

Yep. http://www.xbitlabs.com/news/memory/display/20130220235239_Sony_Confirms_Lack_of_Rambus_Memory_Inside_PlayStation_4.html

Hmm, would it be possible to develop a GGDR5 motherboard for PC without soldering the memory chips and CPU on board? If not, you'd loose upgradability but maybe it would be worth it. Imagine Kaveri with 8GB of GDDR5. That would really let the iGPU do its thing and certainly the CPU cores would be well fed.

 

[-Fran-]

They already said Jaguar is 15% IPC gain over Brazos 2.0, and 10% clock speed gain so about 25% faster. Then multiply by 4 for the extra cores.

http://www.cpubenchmark.net/cpu_lookup.php?cpu=AMD+E2-1800+APU

In this benchmark that would put you at the level of an FX-4130.

http://www.cpubenchmark.net/cpu.php?cpu=AMD+FX-4130+Quad-Core&id=1807

The custom part is the GDDR5 and much larger GPU compared to what Kabini/Temash are going with. Certainly that will help some but the memory interface doesn't do as much for the APUs compute capability, that really only helps the GPU side.

Well, having the capability of a 4130 is pretty darn impressive. And you have to add the fact, for the PS4, of a optimized OS for the APU.

So, all in all, it's looking good so far.

AMD could learn a lot of interesting stuff from this venture with Sony, and, probably, MS.

Also, the dock could give you additional compute by having a larger battery and another APU in it. Since you don't need that much bandwidth, the SeaMicro interconnects would work very nicely in there. "Dumbed" down to tablets, off course. That's my take on that idea from AMD. Hope I'm right on it.

Yep. http://www.xbitlabs.com/news/memory/display/20130220235239_Sony_Confirms_Lack_of_Rambus_Memory_Inside_PlayStation_4.html

Hmm, would it be possible to develop a GGDR5 motherboard for PC without soldering the memory chips and CPU on board? If not, you'd loose upgradability but maybe it would be worth it. Imagine Kaveri with 8GB of GDDR5. That would really let the iGPU do its thing and certainly the CPU cores would be well fed.

GDDR5 works very differently than regular DDR3, so you'd have to fully change the IMC to take advantage on it. Or, if you don't want to make another CPU just for the IMC, put it back into the NB, losing quite a lot of the possible benefits IMO. And IIRC, GDDR5 is very similar to XDR.

Also, AFAIK, it's more expensive to produce.

Cheers!

 

[mayankleoboy1]

and GDDR5 sucks quite a lot of power, and get hot quickly too.
Which makes me wonder afresh if PS4 CPU's are using GDDR5 as their primary RAM.

 

[JAYDEEJOHN]

Power at higher speeds, or, what we see on good gpus.
First, it wont require those speeds because of SoC, and second, it simply wont require top speeds, so power/heat is reduced

 

[JAYDEEJOHN]

As for crysis3 cpu usage, heres a 1337 post
http://forum.beyond3d.com/showpost.php?p=1712451&postcount=1337

 

[gamerk316]

Yep. http://www.xbitlabs.com/news/memory/display/20130220235239_Sony_Confirms_Lack_of_Rambus_Memory_Inside_PlayStation_4.html

Hmm, would it be possible to develop a GGDR5 motherboard for PC without soldering the memory chips and CPU on board? If not, you'd loose upgradability but maybe it would be worth it. Imagine Kaveri with 8GB of GDDR5. That would really let the iGPU do its thing and certainly the CPU cores would be well fed.

Not really. GDDR has significantly higher latency then DDR. Its meant for pushing a LOT of data across fairly quickly, but at the expense of a higher minimum latency. Not a good design for general system RAM. [I already dislike the trend of higher latency on RAM in exchange for bandwidth...]

 

[gamerk316]

As for crysis3 cpu usage, heres a 1337 post
http://forum.beyond3d.com/showpost.php?p=1712451&postcount=1337

o-0. Unusual to say the least. But yeah, that explains the CPU usage, as that type of task is "typically" done by the GPU.

Anyone willing to post GPU utilization numbers now? I want to see what effect this change had on GPU utilization. [EG: Did a GPU bottlenecked app become CPU bottlenecked due to moving work from the GPU to the CPU?]

 

[mayankleoboy1]

As for crysis3 cpu usage, heres a 1337 post
http://forum.beyond3d.com/showpost.php?p=1712451&postcount=1337

Sort of like why Borderlands2 used a lot of CPU : Physics calculations done on CPU.

 

[anxiousinfusion]

There was news a while back that AMD would release their 1090FX chipset this summer. http://www.eteknix.com/amd-1090fx-chipset-arrives-with-steamroller/ Does this still hold true even though Steamroller/HD 8000 are getting pushed back to Q4?

 

[Soul_keeper]

With all the talk about gddr5
Here are the chips that are on my radeon 7850
https://www.skhynix.com/products/graphics/view.jsp?info.ramKind=26&info.serialNo=H5GQ2H24MFR

I don't want to argue with anyone over what constitutes "latency" and why these are faster than ddr3. I just figured i'd throw this out there.

 

[BeastLeeX]

There was news a while back that AMD would release their 1090FX chipset this summer. http://www.eteknix.com/amd-1090fx-chipset-arrives-with-steamroller/ Does this still hold true even though Steamroller/HD 8000 are getting pushed back to Q4?

Not 100% sure if thats accurate, and I think Steamroller is Q1 2014, but a 2013 release is possible. Also, the 1090FX is a enthusiast grade chipset, I doubt many people will buy them, because the 970 is already an extremely good value, and the 990FX just improves it with mroe features. Most people using Quad SLI or Quad CFX will use Intel anyway...

 

[sarinaide]

1090FX is great news for motherboard vendors, gives them a chance to revive the AMD sockets with new lines of motherboards.