"Why won't ATI Support Cuda and PhysX?"

[thilanliyan]

Originally posted by: Wreckage
You totally ignored AVIVO, I noticed.

So? There's more than ONE program that can be used for transcoding and maybe one is written in a way to better take advantage of the power available. If Badaboom is slower than some other program, would you still try to use just that? No you have a choice so probably you'd use the other program.

Getting back to your claim:
"Exactly, also if you notice the ATI video transcoder (AVIVO) relies heavily on the CPU.

So I stand by what I said about ATI cards lacking the power to run PhysX. "

SSchevy gave evidence to the contrary if you've decided to equate the transcoder performance to PhysX performance.

 

[Wreckage]

Originally posted by: thilan29

Originally posted by: Wreckage
You totally ignored AVIVO, I noticed.

So? There's more than ONE program that can be used for transcoding and maybe one is written in a way to better take advantage of the power available. If Badaboom is slower than some other program, would you still try to use just that? No you have a choice so probably you'd use the other program.

Getting back to your claim:
"Exactly, also if you notice the ATI video transcoder (AVIVO) relies heavily on the CPU.

So I stand by what I said about ATI cards lacking the power to run PhysX. "

SSchevy gave evidence to the contrary if you've decided to equate the transcoder performance to PhysX performance.

AVIVO is from ATI, it's not a third party app. So I would consider that the best reference.

Either way SSchevy ignored all the other points that Scali made. Such as Folding@home.

 

[evolucion8]

With the latest Cat 9.6, I ran Sandra Lite 2009.5.15.97, and in GPGPU Processing, I did 3138.25MPixels/s on Float Shader Performance, and 1687.98MPixels/s on Double Shader Performance, much faster than the HD 4850 in Crossfire, which means that the latest driver update brings considerable performance improvements in GPGPU performance. With the reference score, the HD 4850 in Crossfire was much faster than the GTX 295 on Double Shader Performance and slighly faster on Float Shader performance, but my score was much higher than both.

In GPGPU Memory Bandwidth, Internal Memory Bandwidth, my card is 12GB/s slower than the 9800GTX+ which means that Global Data Share comes handy in the nVidia architecture, but in Data Transfer Bandwidth, my card almost twice faster than the GTX 280. So it means that depending in the scenario, bandwidth hungry applications will run faster on nVidia hardware, and mathematically hungry applications will run faster on ATi hardware. I know that's a synthetic benchmark and is a home made evaluation with Sandra, but it can give you a hint of what to expect.

In Folding@Home, ATi is slower, but in MilyWay@Home, ATi is faster.

http://www.gpugrid.net/forum_thread.php?id=705

DoctorNow wrote:
It's also possible now to run a new custom-made MilkyWay@Home-app on your GPU, but currently ONLY possible with an ATI-card and a 64-Bit Windows-system.
More details you can read in this thread.


Thought I'd just inform you, as it surely gets overlooked in the other thread. But be warned, currently it's really in pre-alpha stage. Buying a card for that wouldn't be fun. But if you've already got a HD38x0 (64 shader units) or HD48x0 (160 units) you might want to check it out. The speed is rediculous


Paul D. Buck wrote:
If they get it out the door soon I might just get a couple of the lower end ATI cards that can handle it just for the mean time till they get the Nvidia version done


NV version is not going to happen anytome soon as they use double precision exclusively. You may remember that NV included 30 double units in GT200 along with the 240 single precision shaders. Well, ATIs RV770 has 160 5-way VLIW units and all of them can run 1 or 2 doubles each clock. That's such a massive advantage, it just plain wouldn't make sense to use NV cards here.

http://milkyway.cs.rpi.edu/mil...orum_thread.php?id=589

http://www.youtube.com/watch?v=nnsW-zB95Is

So it would mean that the hardware is just as good as how much the developer can push it, and if it the developer is capable, can extract more power with an ATi card than nVidia counterpart, nVidia is just easier to program and get predictable performance.

 

[SSChevy2001]

Originally posted by: Wreckage

Originally posted by: thilan29

Originally posted by: Wreckage
You totally ignored AVIVO, I noticed.

So? There's more than ONE program that can be used for transcoding and maybe one is written in a way to better take advantage of the power available. If Badaboom is slower than some other program, would you still try to use just that? No you have a choice so probably you'd use the other program.

Getting back to your claim:
"Exactly, also if you notice the ATI video transcoder (AVIVO) relies heavily on the CPU.

So I stand by what I said about ATI cards lacking the power to run PhysX. "

SSchevy gave evidence to the contrary if you've decided to equate the transcoder performance to PhysX performance.

AVIVO is from ATI, it's not a third party app. So I would consider that the best reference.

Either way SSchevy ignored all the other points that Scali made. Such as Folding@home.

When comparing GPGPU performance you need to using similar programs. IMO comparing Avivo to Badaboom will never lead to any conclusion about GPGPU performance.

Avivo is also a free, and shouldn't be compared against a paid application.

I really don't want to get into the whole Folding@home situtation. It clear that ATi is doing twice the work compared to the Nvidia client, which proves that the ATi client is far from optimized.
"some of the latest advances in GPU hardware have not been fully exploited"

Also evolucion8 bring up a good point about MilyWay@Home, but you'll more than likely ignore that.

 

[Scali]

Originally posted by: evolucion8
So it would mean that the hardware is just as good as how much the developer can push it, and if it the developer is capable, can extract more power with an ATi card than nVidia counterpart, nVidia is just easier to program and get predictable performance.

Double precision has little to do with how capable the developer is.
A capable developer might be able to produce stable results using double precision only in selected places. Using double precision everywhere is the easy way out.

 

[Scali]

Originally posted by: SSChevy2001
I really don't want to get into the whole Folding@home situtation. It clear that ATi is doing twice the work compared to the Nvidia client, which proves that the ATi client is far from optimized.

ATi has spent years on optimizing their client, unlike nVidia.

Firstly, ATi doesn't do 'twice the work', it does some work multiple times, but not ALL work. The larger the proteins are, the smaller the differences between ATi and nVidia.
secondly, it's not an optimization issue... as the developers themselves say, it *is* more optimal to calculate the distances multiple times than to calculate them once and store them in shared memory. It's a limitation of the hardware.

 

[Keysplayr]

Originally posted by: evolucion8
With the latest Cat 9.6, I ran Sandra Lite 2009.5.15.97, and in GPGPU Processing, I did 3138.25MPixels/s on Float Shader Performance, and 1687.98MPixels/s on Double Shader Performance, much faster than the HD 4850 in Crossfire, which means that the latest driver update brings considerable performance improvements in GPGPU performance. With the reference score, the HD 4850 in Crossfire was much faster than the GTX 295 on Double Shader Performance and slighly faster on Float Shader performance, but my score was much higher than both.

In GPGPU Memory Bandwidth, Internal Memory Bandwidth, my card is 12GB/s slower than the 9800GTX+ which means that Global Data Share comes handy in the nVidia architecture, but in Data Transfer Bandwidth, my card almost twice faster than the GTX 280. So it means that depending in the scenario, bandwidth hungry applications will run faster on nVidia hardware, and mathematically hungry applications will run faster on ATi hardware. I know that's a synthetic benchmark and is a home made evaluation with Sandra, but it can give you a hint of what to expect.

In Folding@Home, ATi is slower, but in MilyWay@Home, ATi is faster.

http://www.gpugrid.net/forum_thread.php?id=705

DoctorNow wrote:
It's also possible now to run a new custom-made MilkyWay@Home-app on your GPU, but currently ONLY possible with an ATI-card and a 64-Bit Windows-system.
More details you can read in this thread.


Thought I'd just inform you, as it surely gets overlooked in the other thread. But be warned, currently it's really in pre-alpha stage. Buying a card for that wouldn't be fun. But if you've already got a HD38x0 (64 shader units) or HD48x0 (160 units) you might want to check it out. The speed is rediculous


Paul D. Buck wrote:
If they get it out the door soon I might just get a couple of the lower end ATI cards that can handle it just for the mean time till they get the Nvidia version done


NV version is not going to happen anytome soon as they use double precision exclusively. You may remember that NV included 30 double units in GT200 along with the 240 single precision shaders. Well, ATIs RV770 has 160 5-way VLIW units and all of them can run 1 or 2 doubles each clock. That's such a massive advantage, it just plain wouldn't make sense to use NV cards here.

http://milkyway.cs.rpi.edu/mil...orum_thread.php?id=589

http://www.youtube.com/watch?v=nnsW-zB95Is

So it would mean that the hardware is just as good as how much the developer can push it, and if it the developer is capable, can extract more power with an ATi card than nVidia counterpart, nVidia is just easier to program and get predictable performance.

There is no Nvidia client. I don't know why you keep bringing up this Milky Way thing when they only have CPU comparisons to go by.

 

[Keysplayr]

Originally posted by: SSChevy2001

Originally posted by: Wreckage

Originally posted by: thilan29

Originally posted by: Wreckage
You totally ignored AVIVO, I noticed.

So? There's more than ONE program that can be used for transcoding and maybe one is written in a way to better take advantage of the power available. If Badaboom is slower than some other program, would you still try to use just that? No you have a choice so probably you'd use the other program.

Getting back to your claim:
"Exactly, also if you notice the ATI video transcoder (AVIVO) relies heavily on the CPU.

So I stand by what I said about ATI cards lacking the power to run PhysX. "

SSchevy gave evidence to the contrary if you've decided to equate the transcoder performance to PhysX performance.

AVIVO is from ATI, it's not a third party app. So I would consider that the best reference.

Either way SSchevy ignored all the other points that Scali made. Such as Folding@home.

When comparing GPGPU performance you need to using similar programs. IMO comparing Avivo to Badaboom will never lead to any conclusion about GPGPU performance.

Avivo is also a free, and shouldn't be compared against a paid application.

Avivo is free for a good reason. But it shouldn't matter if it's free or not. If the program does what it was designed to do, and well, it should be compared.

I really don't want to get into the whole Folding@home situtation. It clear that ATi is doing twice the work compared to the Nvidia client, which proves that the ATi client is far from optimized.
"some of the latest advances in GPU hardware have not been fully exploited"

It's been years since the 2900XT. You'd think that would give enough time to exploit the hardware by now. Even easier with 4xxx series with the caches included.

Also evolucion8 bring up a good point about MilyWay@Home, but you'll more than likely ignore that.

He brought this up several times before. It's his favorite thing to bring up because there is no comparable Nvidia client. It just shows the ATI GPU going up against a CPU.

 

[Forumpanda]

Here is why wreckage posts are indeed trolling and not opinions:

He is very often making posts, sometimes off-topic, just to be negative about ATI.
When faced with evidence and arguments contradicting his point he ignores it, and finds another thread to post the exact same baseless opinions in. Also often returning to post in the same thread ignore all posts between his last and current, dragging the debate back to last week.

Scali is leaning towards nvidia, evolucion8 is leaning towards ATI, but both argue their point of view, wreckage is just spouting nonsense, there is a world of difference.
I am sorry but you have to be biased or blind to not see the obvious trend here.

Wreckage has derailed more threads than I have made posts, and thats a fact.


The only thing I don't get is why people still bother trying, personally I am not partaking in any debates on the video forum in the current climate, it is not worth the aggravation.

 

[cm123]

Originally posted by: Keysplayr

Originally posted by: Wreckage

Originally posted by: Scali

All this combined means that ATi cards indeed have some limitations in GPGPU compared to nVidia. This is also apparent in Folding@home for example.
Read this thread for example:
http://foldingforum.org/viewto...p?f=51&t=10442&start=0
It includes comments of people like mhouston, who work for AMD on the Folding@Home client. Basically they're saying that they calculate certain values multiple times because on ATi hardware this is faster than using the shared memory (LDS - Local Data Storage).

Exactly, also if you notice the ATI video transcoder (AVIVO) relies heavily on the CPU.

So I stand by what I said about ATI cards lacking the power to run PhysX.

Until we actually see PhysX running on an ATI GPU and can benchmark it. I don't see much evidence to the contrary.

If a 3xxx series could be made to run physx on the GPU, I have little trouble believing that a 4xxx series GPU can run it too. I'm pretty certain it could, just likely not as well as an Nvidia GPU.

I'm guessing keys you being part of Nvidia Focus group already know that AMD could run PhysX on there cards - truth maybe lies more in the fact while they can, can they come even close to the performance of a Nvidia card for physX...

...cause you know the moment amd jumped on physx, all the review sites would be running benchies with who is top dog and such - why, or perhaps could amd handle that kind of stuff right now?

for me, I wish what we call physx and havok would simply be part of dx11 as a standard and from that point its up to nvidia and amd to offer products to perform at their best...

when has a option like physx ever became main-stream before in the past? (not saying never happened, however has it?)

seems for physx to take over would mean nearly all game makers would then have to be on the nvidia team - after 3dfx, my bet is on that will not ever happen today...

plus, when talk physx support, look to the new cards about to come out - the nvidia cards look to be a real power house - just ask keys , he should be running one about now already - not to say amd next solution is not fast per say...

I'm just not understanding the bit of personal (jabs) take on this topic - nvidia has great physx (even if keys is not so sure of that) vs what ati/amd has right now, although lots more havok out there right now that physx is there not?

 

[WelshBloke]

Originally posted by: Keysplayr

Originally posted by: evolucion8
With the latest Cat 9.6, I ran Sandra Lite 2009.5.15.97, and in GPGPU Processing, I did 3138.25MPixels/s on Float Shader Performance, and 1687.98MPixels/s on Double Shader Performance, much faster than the HD 4850 in Crossfire, which means that the latest driver update brings considerable performance improvements in GPGPU performance. With the reference score, the HD 4850 in Crossfire was much faster than the GTX 295 on Double Shader Performance and slighly faster on Float Shader performance, but my score was much higher than both.

In GPGPU Memory Bandwidth, Internal Memory Bandwidth, my card is 12GB/s slower than the 9800GTX+ which means that Global Data Share comes handy in the nVidia architecture, but in Data Transfer Bandwidth, my card almost twice faster than the GTX 280. So it means that depending in the scenario, bandwidth hungry applications will run faster on nVidia hardware, and mathematically hungry applications will run faster on ATi hardware. I know that's a synthetic benchmark and is a home made evaluation with Sandra, but it can give you a hint of what to expect.

In Folding@Home, ATi is slower, but in MilyWay@Home, ATi is faster.

http://www.gpugrid.net/forum_thread.php?id=705

DoctorNow wrote:
It's also possible now to run a new custom-made MilkyWay@Home-app on your GPU, but currently ONLY possible with an ATI-card and a 64-Bit Windows-system.
More details you can read in this thread.


Thought I'd just inform you, as it surely gets overlooked in the other thread. But be warned, currently it's really in pre-alpha stage. Buying a card for that wouldn't be fun. But if you've already got a HD38x0 (64 shader units) or HD48x0 (160 units) you might want to check it out. The speed is rediculous


Paul D. Buck wrote:
If they get it out the door soon I might just get a couple of the lower end ATI cards that can handle it just for the mean time till they get the Nvidia version done


NV version is not going to happen anytome soon as they use double precision exclusively. You may remember that NV included 30 double units in GT200 along with the 240 single precision shaders. Well, ATIs RV770 has 160 5-way VLIW units and all of them can run 1 or 2 doubles each clock. That's such a massive advantage, it just plain wouldn't make sense to use NV cards here.

http://milkyway.cs.rpi.edu/mil...orum_thread.php?id=589

http://www.youtube.com/watch?v=nnsW-zB95Is

So it would mean that the hardware is just as good as how much the developer can push it, and if it the developer is capable, can extract more power with an ATi card than nVidia counterpart, nVidia is just easier to program and get predictable performance.

There is no Nvidia client. I don't know why you keep bringing up this Milky Way thing when they only have CPU comparisons to go by.

I think nVidia cards just don't have the power to run MilkyWay@Home.

 

[Keysplayr]

Even if ATI hardware can't run PhysX "as fast" as comparable Nvidia hardware, It couldn't possibly be worse than trying to run it on the CPU. GPUs are so powerful today, and so many of them to choose from, I don't think PhysX would hold back a powerful enough ATI card. Unless of course performance even when tweaked at the programming level is abysmal. But I think ATI cards have enough juice to run PhysX acceptably.

 

[Keysplayr]

Originally posted by: WelshBloke

Originally posted by: Keysplayr

Originally posted by: evolucion8
With the latest Cat 9.6, I ran Sandra Lite 2009.5.15.97, and in GPGPU Processing, I did 3138.25MPixels/s on Float Shader Performance, and 1687.98MPixels/s on Double Shader Performance, much faster than the HD 4850 in Crossfire, which means that the latest driver update brings considerable performance improvements in GPGPU performance. With the reference score, the HD 4850 in Crossfire was much faster than the GTX 295 on Double Shader Performance and slighly faster on Float Shader performance, but my score was much higher than both.

In GPGPU Memory Bandwidth, Internal Memory Bandwidth, my card is 12GB/s slower than the 9800GTX+ which means that Global Data Share comes handy in the nVidia architecture, but in Data Transfer Bandwidth, my card almost twice faster than the GTX 280. So it means that depending in the scenario, bandwidth hungry applications will run faster on nVidia hardware, and mathematically hungry applications will run faster on ATi hardware. I know that's a synthetic benchmark and is a home made evaluation with Sandra, but it can give you a hint of what to expect.

In Folding@Home, ATi is slower, but in MilyWay@Home, ATi is faster.

http://www.gpugrid.net/forum_thread.php?id=705

DoctorNow wrote:
It's also possible now to run a new custom-made MilkyWay@Home-app on your GPU, but currently ONLY possible with an ATI-card and a 64-Bit Windows-system.
More details you can read in this thread.


Thought I'd just inform you, as it surely gets overlooked in the other thread. But be warned, currently it's really in pre-alpha stage. Buying a card for that wouldn't be fun. But if you've already got a HD38x0 (64 shader units) or HD48x0 (160 units) you might want to check it out. The speed is rediculous


Paul D. Buck wrote:
If they get it out the door soon I might just get a couple of the lower end ATI cards that can handle it just for the mean time till they get the Nvidia version done


NV version is not going to happen anytome soon as they use double precision exclusively. You may remember that NV included 30 double units in GT200 along with the 240 single precision shaders. Well, ATIs RV770 has 160 5-way VLIW units and all of them can run 1 or 2 doubles each clock. That's such a massive advantage, it just plain wouldn't make sense to use NV cards here.

http://milkyway.cs.rpi.edu/mil...orum_thread.php?id=589

http://www.youtube.com/watch?v=nnsW-zB95Is

So it would mean that the hardware is just as good as how much the developer can push it, and if it the developer is capable, can extract more power with an ATi card than nVidia counterpart, nVidia is just easier to program and get predictable performance.

There is no Nvidia client. I don't know why you keep bringing up this Milky Way thing when they only have CPU comparisons to go by.

I think nVidia cards just don't have the power to run MilkyWay@Home.

I'm pretty sure you do not believe that. But, miracles never cease.

 

[SlowSpyder]

Originally posted by: Forumpanda
Here is why wreckage posts are indeed trolling and not opinions:

He is very often making posts, sometimes off-topic, just to be negative about ATI.
When faced with evidence and arguments contradicting his point he ignores it, and finds another thread to post the exact same baseless opinions in. Also often returning to post in the same thread ignore all posts between his last and current, dragging the debate back to last week.

Scali is leaning towards nvidia, evolucion8 is leaning towards ATI, but both argue their point of view, wreckage is just spouting nonsense, there is a world of difference.
I am sorry but you have to be biased or blind to not see the obvious trend here.

Wreckage has derailed more threads than I have made posts, and thats a fact.


The only thing I don't get is why people still bother trying, personally I am not partaking in any debates on the video forum in the current climate, it is not worth the aggravation.

That's what drives me nuts, a lot of people lean one way or another, some are fan boys. Almost all can have a decent conversation and cede that the other company may do x or y better than they company they pull for. Wreckage cannot, he just derails things.

As for this thread, in the story AMD clearly stated why they chose not to support Physx. And so far how many games support (hardware) Physx? If that list expands quickly over the near future AMD will have no choice in my opinion. As it is now I think AMD users miss it as much as Nvidia users miss DX10.1.

 

[thilanliyan]

Another article:
http://en.expreview.com/2009/0...a-run-on-its-gpus.html

 

[SSChevy2001]

Originally posted by: Keysplayr
Avivo is free for a good reason. But it shouldn't matter if it's free or not. If the program does what it was designed to do, and well, it should be compared.

Are you looking to compare two different programs or GPGPU performance?

It's been years since the 2900XT. You'd think that would give enough time to exploit the hardware by now. Even easier with 4xxx series with the caches included.

How can the ATi client be optimized when it's repeating some of the workload? While it might never be full exploited, in it's current state it's far from efficient.

He brought this up several times before. It's his favorite thing to bring up because there is no comparable Nvidia client. It just shows the ATI GPU going up against a CPU.

He manage to point out how a developer favored ATi hardware for GPGPU, so the point is it can go both ways.

In the end ATi doesn't want CUDA, because it wouldn't be a level playing field.

 

[Modelworks]

Originally posted by: SlowSpyder


As for this thread, in the story AMD clearly stated why they chose not to support Physx. And so far how many games support (hardware) Physx? If that list expands quickly over the near future AMD will have no choice in my opinion. As it is now I think AMD users miss it as much as Nvidia users miss DX10.1.

We really don't know the details on why they chose not to support it. I'm guessing Nvidia didn't just tell AMD "Here is PhysX, anything we can help you with ?"

They probably did not like the terms Nvidia attached.

 

[thilanliyan]

Originally posted by: Modelworks
They probably did not like the terms Nvidia attached.

Including the licensing fees probably which would have started off low but depending on the popularity of PhysX could have gone up.

 

[Modelworks]

Originally posted by: thilan29

Originally posted by: Modelworks
They probably did not like the terms Nvidia attached.

Including the licensing fees probably.

That is my guess too.


Technically if Nvidia wants Cuda and PhysX to be the standards all they have to do is release the source code to both under a GPL agreement.

 

[Wreckage]

Originally posted by: Keysplayr


There is no Nvidia client. I don't know why you keep bringing up this Milky Way thing when they only have CPU comparisons to go by.

Maybe we should compare the SETI@home NVIDIA client to the ATI client....

 

[SlowSpyder]

Originally posted by: thilan29

Originally posted by: Modelworks
They probably did not like the terms Nvidia attached.

Including the licensing fees probably which would have started off low but depending on the popularity of PhysX could have gone up.

In the article you posted the Nvidia PR guy said pennies a GPU... but what is that? $.02? $.13? $.27? Multiply that times tens of thousands, maybe more. Also we don't know that the license fee would stay. We'll probably never know the true terms that Nvidia offered though, all we can do is speculate. Bottom line is AMD decided to go a different direction than Nvidia, only time will tell how things will work out for both companies.

 

[Keysplayr]

Originally posted by: SSChevy2001

Originally posted by: Keysplayr
Avivo is free for a good reason. But it shouldn't matter if it's free or not. If the program does what it was designed to do, and well, it should be compared.

Are you looking to compare two different programs or GPGPU performance?

It's been years since the 2900XT. You'd think that would give enough time to exploit the hardware by now. Even easier with 4xxx series with the caches included.

How can the ATi client be optimized when it's repeating some of the workload? While it might never be full exploited, in it's current stated it's far from efficient.

He brought this up several times before. It's his favorite thing to bring up because there is no comparable Nvidia client. It just shows the ATI GPU going up against a CPU.

He manage to point out how a developer favored ATi hardware for GPGPU, so the point is it can go both ways.

In the end ATi doesn't want CUDA, because it wouldn't be a level playing field.

GPGPU performance.

Why isn't it optimized? Why can't they optimize it? Why wont they? If they were able to do it, they would, right? This argument is utter BS because they have had waaaaay more than enough time to properly code for this arch. I think it's the best they could get out of it. I love how you guys are touting performance that will never materialize because it's damn near impossible to code for ATI in it's current arch. Which translates to, if you can't code for it, it's almost useless to try.

 

[evolucion8]

Originally posted by: Keysplayr
GPGPU performance.

Why isn't it optimized? Why can't they optimize it? Why wont they? If they were able to do it, they would, right? This argument is utter BS because they have had waaaaay more than enough time to properly code for this arch. I think it's the best they could get out of it. I love how you guys are touting performance that will never materialize because it's damn near impossible to code for ATI in it's current arch. Which translates to, if you can't code for it, it's almost useless to try.

That's your opinion which we all respect, but nVidia and ATi engineers are no lousy engineers, they make decisions based on R&D. GeForce FX was a very flawed architecture and yet nVidia was able to optimize it so good that it was able to almost keep up with the R3X0 architecture in nVidia optimized games, considering that ATi has been working for much more time for their super scalar architecture, that ATi has a much better background in software engineering thanks to it's merge with AMD, there's no huge reason to spend so much time in GPGPU performance when most ATi cards sold are used in games.

Is a matter of execution and resource allocation with the driver development team. While nVidia architecture has the upper hand currently with the GPGPU, I don't see it as a key at the selling point level, or as a must have feature, since most applications today aren't completely parallel and will require general purpose calculations which will run like crap with the massive parallel GPU's of today. At the end, both, the software engineer and the ATi driver engineer must work to take advantage of the optimizations at the architecture and yet, Folding@Home is an old client which doesn't even use the Data cache share found on the HD 4000 series, it was made only for HD 3x00 and lower, and GeForce 8 which uses a completely different approach which will work great no matter what, nVidia is about predictable performance since no optimization is necessary to get a good performance of it in GPGPU applications, ATi is about extracting and maximizing parallelism which will require more work.

 

[dguy6789]

Originally posted by: evolucion8

Originally posted by: Keysplayr
GPGPU performance.

Why isn't it optimized? Why can't they optimize it? Why wont they? If they were able to do it, they would, right? This argument is utter BS because they have had waaaaay more than enough time to properly code for this arch. I think it's the best they could get out of it. I love how you guys are touting performance that will never materialize because it's damn near impossible to code for ATI in it's current arch. Which translates to, if you can't code for it, it's almost useless to try.

That's your opinion which we all respect, but nVidia and ATi engineers are no lousy engineers, they make decisions based on R&D. GeForce FX was a very flawed architecture and yet nVidia was able to optimize it so good that it was able to almost keep up with the R3X0 architecture in nVidia optimized games, considering that ATi has been working for much more time for their super scalar architecture, that ATi has a much better background in software engineering thanks to it's merge with AMD, there's no huge reason to spend so much time in GPGPU performance when most ATi cards sold are used in games.

Is a matter of execution and resource allocation with the driver development team. While nVidia architecture has the upper hand currently with the GPGPU, I don't see it as a key at the selling point level, or as a must have feature, since most applications today aren't completely parallel and will require general purpose calculations which will run like crap with the massive parallel GPU's of today. At the end, both, the software engineer and the ATi driver engineer must work to take advantage of the optimizations at the architecture and yet, Folding@Home is an old client which doesn't even use the Data cache share found on the HD 4000 series, it was made only for HD 3x00 and lower, and GeForce 8 which uses a completely different approach which will work great no matter what, nVidia is about predictable performance since no optimization is necessary to get a good performance of it in GPGPU applications, ATi is about extracting and maximizing parallelism which will require more work.

:thumbsup: agreed on all accounts.

 

[evolucion8]

http://www.extremetech.com/art.../0,2845,2324319,00.asp

Right now, Nvidia's cards are better folders, due primarily to better optimized code. With the latest drivers, most GeForce cards are getting pretty close to peak utilization. ATI's cards, which rely on their CAL driver, still seem to have a lot of headroom. In fact, the new Radeon HD 4800 have 800 stream processors, but the current client runs on them as if they were older cards with only 320.

Like I said, Folding@Home doesn't take advantage of the HD4x00 architecture.