NVIDIA Forceware 257 & Heterogeneous GPU PhysX: It's a Bug, Not a Feature
by Ryan Smith on May 28, 2010 12:42 PM ESTOur inbox quickly lit up this morning when we received notice about this NGOHQ article, discussing how NVIDIA had removed the heterogeneous GPU restriction on PhysX in their latest beta drivers. This struck us as a bit of an odd reversal of positions from NVIDIA, and now that we've had a chance to chat with them we finally know what's going on.
As a quick matter of background, starting with the Forceware 186 series NVIDIA blocked GPU/PPU-accelerated PhysX from working on NVIDIA GPUs and AGEIA PPUs whenever a non-NVIDIA GPU was detected as being in the system. It's been a polarizing matter for the GPU community for nearly a year now, with a tug-of-war going on between projects editing the drivers to remove the block, and NVIDIA adding further checks in to their drivers to stop those efforts. In any case, there has been no sign that NVIDIA would be changing their position any time soon.
This brings us to this week's Forceware 257.15 beta drivers and today's clarification from NVIDIA. NGOHQ was correct in that the 257.15 drivers lacked the heterogeneous GPU restriction; however there has been a question of intentions. As we stated previously NVIDIA has held steady to their desire to keep PhysX on pure NVIDIA systems, so to make this change without publically announcing it odd - if only because it deprives them of the chance to sell cards as PhysX accelerators.
We just got done talking with NVIDIA about the matter and they clarified the issue for us. In what we expect is going to be a disappointment for many of you, the lack of a PhysX restriction on the current 257.15 beta drivers is a bug, not a feature - the restriction should have been in those drivers and it was not. NVIDIA will be reinstating the restriction in new downloads of the beta driver and in the WHQL build of these drivers.
Update: NVIDIA tells us that they will also be "fixing" the 257.15 beta driver on their site, so new downloads of that driver will have the restriction in place
Yes, this is a bug in the latest build of PhysX that was packaged with the driver. We'll be fixing this issue ASAP - the WHQL driver launching in early June won't have this issue. -NVIDIA
For those of you heterogeneous GPU users out there looking to use PhysX, there is some good news that can be salvaged from this however: this won't change the fact that previously downloaded copies of beta drivers lack this restriction. With these drivers you can still have heterogeneous GPUs with PhysX without modifying NVIDIA’s drivers, but you’ll be stuck on these drivers for the time being.
Facebook
Twitter
RSS
53 Comments
View All Comments
[-Stash-] - Friday, May 28, 2010 - link
Arg, why have a "Read More" button if there's no more to read!?That aside, it's reassuring to know that a simple driver hack is still all that is required and that it is just a hurdle that Nvidia is introducing, not an outright incompatibility between Nvidia and AMD GPUs :)
The0ne - Friday, May 28, 2010 - link
yep, 3d guru for your needs :Dcdillon - Friday, May 28, 2010 - link
So let me get this straight... NVidia is intentionally disabling a feature in their drivers simply because another brand of GPU is in your system, and apparently not for valid technical reasons because things work just fine when it is enabled. Isn't this exactly the kind of BS that should spawn anti-trust litigation?shin0bi272 - Friday, May 28, 2010 - link
Its close but its only disabling the physx ability on the nvidia card or ageia ppu (since nvidia owns ageia now they can do that) not the actual "3rd party" gpu card. You can still run multiple desktops on multiple cards etc etc you just cant get physx if your system has a different gpu in it.The issue I have with them doing this is its dumb. You want people to buy your hardware and yet your top gpu isnt as fast as the competition... So lets stop them from buying our hardware for use in physx games so we also f*ck over those people that make physx heavy games for all but Nvidia customers. At least if they allowed heterogeneous video cards and physx people would still buy their cards just to play a game with... thats MOAR PROFITS Nvidia! That will also push MOAR PHYSX GAMES into the market... then if there are enough games on the market with physx to make people give a damn .. people will buy your cards for their ability to do physx and not buy the competition even if the competition is a little faster for the same price.
CU - Friday, May 28, 2010 - link
Sony is facing multiply class action law suits for removing linux support on the PS3. How can Nvidia remove a feature (ie. Ageia did work with any gpu, and now it doesn't) and not be sued?Slash3 - Saturday, May 29, 2010 - link
It was never advertised as a supported feature, the way that the "Install Other OS" and Linux support was advertised with the PS3. Many people used that as a consideration when they purchased the PS3, and then effectively had the rug pulled out from under them. The PhysX issue is a bit less murky, in that they have always prevented you from using a mixed pair of cards with PhysX enabled (GPU, not the PPU add-on) - despite the fact that there are numerous advantages to be had. So it's less a case of removing a feature as it is a case of never having been smart enough to allow it to begin with.Way to ruin the party, nvidia.
CU - Monday, May 31, 2010 - link
I was specifically talking about the ppu's though. They worked with any gpu, but now they don't thanks to Nvidia.shin0bi272 - Tuesday, June 1, 2010 - link
its nvidia's property now they can pretty much do what ever they want with it. Same thing with their gpu's supporting physx while another brand of gpu is in the box. Its their decision and they (IMHO) made the wrong one.God forbid someone goes out and buys a 200 dollar nvidia card after they already have a 500 dollar amd card the world might end right Nvidia?
AdamK47 - Friday, May 28, 2010 - link
You say this as if it were some sort of new revelation.CU - Friday, May 28, 2010 - link
Is it legal to block people from using a using the original PPU's with a non-Nvidia gpu? They were sold with the ability to be used with in gpu correct? Then Nvidia bought the company and dropped that support.