Friday, July 4th 2008

Radeon PhysX Creator: 'NVIDIA Offered To Help Us', 'Expected More From AMD'

In a dramatic turnaround of events, NGOHQ.com, the creators of a special system software that allowed users of the ATI Radeon graphics accelerators to use proprietary features of NVIDIA graphics accelerators such as GPU-accelerated version of the NVIDIA PhysX game physics API, claim that in fact NVIDIA wanted to help them with this effort. On June, the 26th we had covered reports of the said outfit improvising a driver after proving that NVIDIA's proprietary GPGPU architecture, CUDA was flexible enough to work on a ATI RV670 graphics processor.

The NVIDIA PhysX API has gained immense popularity with the industry off late though the standard itself is not very recent, with the former Ageia having been in the industry for a while before its takeover by NVIDIA. The recent popularity surge of this API owes to the fact that PhysX processing plays a significant role in affecting the scores of the 3DMark Vantage benchmark, where a certain CPU test routine evaluates the computer's physics processing abilities. This gives users of the latest line of GeForce graphics accelerators an advantage over their Radeon counterparts since they get a significant increment in the benchmark score. The PhysX API functions as a CUDA application on GeForce, where the GPU's shaders are used by the process to process physics. With NGOHQ.com being successfully able to modify CUDA to function on ATI GPUs, it was only a formality to run any CUDA application on it, the PhysX process included. NGOHQ.com has gone on record saying that enabling PhysX support on Radeon cards is not particularly difficult, and might not be much be a technology problem but an issue of corporate dynamics.

The team sought support from AMD in the form of a Radeon HD4800 series sample, so they could work out a similar solution for it, but AMD rejected it telling it wasn't a venture worth investing product samples and PR information on, clearly a case of them downplaying the issue to evade possible action from NVIDIA on supporting the use of its proprietary technologies in violation of license agreements. After being denied support from a company they were banking the most on, they were left to their own plight against NVIDIA who have a history of an aggressive business model, even more so after it was known NGOHQ.com may have reverse engineered drivers, a clear violation of the EULA.

The dramatic turnaround of events has caused a little more than a ripple across the industry, the team at NGOHQ.com now claims that NVIDIA actually offered help to them and that they want to strengthen the industry position of CUDA as a viable competitor to Intel and its processing technologies. This follows news reports of a high-ranked officer at Intel downplaying the rising popularity of CUDA and effectively counter the competitive physics API, Havoc, which uses CPU to process in-game physics. The team goes on to use phrases such as "it seems they are giving us their blessings" to highlight their appreciation for the company.

What's beneath this green sugar-coat of an M&M bean? Clear corporate agenda at play. NVIDIA gets away with the higher moral ground in the issue, of having come in support of a small outfit trying to work on technologies that enhanced their competitor's, while in reality all AMD must have wanted was to stay away from IPR headaches at a time when a reinvigorated Radeon line is going head on against GeForce, the way NGOHQ.com portreys their approach looks more like they shied away from a challenge, in their views.With inputs from NGOHQ.com
Add your own comment

41 Comments on Radeon PhysX Creator: 'NVIDIA Offered To Help Us', 'Expected More From AMD'

#1
bowman
With NGOHQ.com being successfully able to modify CUDA to function on ATI GPUs, it was only a formality to run any CUDA application on it
So the Nvidia folding client could be run on the ATI cards? Maybe we'd finally see the true potential of the 4800 series with that one, ATI client is slow.. :)
Posted on Reply
#2
btarunr
Editor & Senior Moderator
bowman said:
So the Nvidia folding client could be run on the ATI cards? Maybe we'd finally see the true potential of the 4800 series with that one, ATI client is slow.. :)
You already have F@H clients for ATI....way back since the Radeon X1800 days.
Posted on Reply
#3
bowman
btarunr said:
You already have F@H clients for ATI....way back since the Radeon X1800 days.
I said that.

bowman said:
So the Nvidia folding client could be run on the ATI cards? Maybe we'd finally see the true potential of the 4800 series with that one, ATI client is slow.. :)
The CUDA client is inexplicably faster on NVIDIA hardware with less theoretical capability than the ATI cards. The ATI client hasn't been updated for the RV770 cards, either. It would be interesting to see if the Nvidia client would run any well on ATI cards.
Posted on Reply
#4
PrudentPrincess
ATI made a really smart move, by not supporting CUDA they've slowed its progress down (or prevented speeding it up) which is a really good thing for them because we have to remember its not just ATI, its AMD/ATI. They have as much to lose in this as Intel.
Posted on Reply
#5
btarunr
Editor & Senior Moderator
PrudentPrincess said:
ATI made a really smart move, by not supporting CUDA they've slowed its progress down (or prevented speeding it up) which is a really good thing for them because we have to remember its not just ATI, its AMD/ATI. They have as much to lose in this as Intel.
I agree. Beyond smart move, it's self-defense and above all, protecting its own FireStream GPGPU architecture from being overrun by CUDA. It's already established that a RV770 churns-out more GFLOPs (1000~1200 GFLOPs) than GT200 (980 GFLOPs), when exploited as a GPGPU, ATI GPUs could fare better, so NV wants to spread CUDA there saying "You can have PhysX", but it's their 'trojan horse' in my opinion.
Posted on Reply
#6
PrudentPrincess
btarunr said:
I agree. Beyond smart move, it's self-defense and above all, protecting its own FireStream GPGPU architecture from being overrun by CUDA. It's already established that a RV770 churns-out more GFLOPs (1000~1200 GFLOPs) than GT200 (980 GFLOPs), when exploited as a GPGPU, ATI GPUs could fare better, so NV wants to spread CUDA there saying "You can haz PhysX", but it's their 'trojan horse' in my opinion.
Every time I start thinking you're valuable to TPU you go and use icanhazcheezeburger speak.
Posted on Reply
#7
btarunr
Editor & Senior Moderator
Fixed. Changed 'haz' with 'have', am I valuable again?
Posted on Reply
#8
TheGuruStud
I'm starting to get the vibe that it's more like "Can I haz native speek english skillz plz?".

If it's not their native language, you can lay off some, guys :). Our grammar is very hard b/c it's whack.
Posted on Reply
#9
eidairaman1
PrudentPrincess said:
Every time I start thinking you're valuable to TPU you go and use icanhazcheezeburger speak.
yo, give the guy a break.
Posted on Reply
#10
eidairaman1
btarunr said:
I agree. Beyond smart move, it's self-defense and above all, protecting its own FireStream GPGPU architecture from being overrun by CUDA. It's already established that a RV770 churns-out more GFLOPs (1000~1200 GFLOPs) than GT200 (980 GFLOPs), when exploited as a GPGPU, ATI GPUs could fare better, so NV wants to spread CUDA there saying "You can have PhysX", but it's their 'trojan horse' in my opinion.
so your saying about 1TFLOP?
Posted on Reply
#11
W1zzard
mixa said:
or more likely someone "gave" em already working ones.
i talk to regeneration from ngohq a lot and nobody bitches about amd/ati/nvidia more than he does. i'm sure he didn't get anything from them before publishing that he's working on it. who can blame nvidia for seeing a chance to push cuda... once they got their big big market share they will could just lock out the ati users whenever they want to
Posted on Reply
#12
eidairaman1
Which i think that is Corporate BS right there, ATI should tell Nvidia to go shove it up their wazoo. Intel/ATI should push with havok, which has been working for many years, where physx didnt work from the start. To add another not, didn't i tell yall i smelled a big rat?
Posted on Reply
#13
MilkyWay
i know AMD/ATi and Nvidia are competitors but really they should both push Cuda as it directly competes with Intel and Havok.

Cuda is good but PhysX was a failure. People didnt want to buy a £200-£100 PPu they would rather buy a full on graphics card. Where as Cuda is just a driver and is free so people will be more inclined to try it out.

I dont mind using Cuda its not as if ATi have their own integrated physics processing drivers.
Posted on Reply
#14
W1zzard
nvidia can push physx into games with their way its meant to be played .. they already talk and work with most of the studios..
Posted on Reply
#15
btarunr
Editor & Senior Moderator
I've played PhysX based games, and then also Havoc ones...since ages. I've not found 1 thing that makes PhysX stand out. The Alan Wake demo by Remedy at IDF '07 was more than anything I've ever seen in terms of game physics (seen it on YouTube). All they did was run it on a Q6600. CPU does everything. No need to let the video-card slog, let it be a full-time graphics processor.
Posted on Reply
#16
Cold Storm
Battosai
btarunr said:
I've played PhysX based games, and then also Havoc ones...since ages. I've not found 1 thing that makes PhysX stand out. The Alan Wake demo by Remedy at IDF '07 was more than anything I've ever seen in terms of game physics (seen it on YouTube). All they did was run it on a Q6600. CPU does everything. No need to let the video-card slog, let it be a full-time graphics processor.
That will be one the the most beautiful games that will come out, I hope, this year. I was nothing but amazed when I saw that demo. Plus been waiting for more! I believe that game will set big standers in how to build a game. Just hope I'm not disappointed in it...
Posted on Reply
Add your own comment