BFG Ageia PhysX Physics Card

If nothing else, Ageia has successfully directed the attention of developers towards the potential high quality physics has, in terms of significantly improving the realism and gameplay of the latest and greatest new titles. As is the case with consumers buying the latest flagship GPU's and CPU's, they are preparing themselves for the future and ensuring they will have the fastest and richest overall performance when playing the hottest titles on the horizon.

The most difficult aspect of judging the merits of the BFG PhysX processor is the fact that there are so many unknown variables. At this time, we do not know how the PPU will scale with frequency or even what the stock frequency of the card is. We're unsure what performance differential there will be (if any) between the 128MB cards like the one we reviewed here and the 256MB we have been seeing in the OEM channel. Lastly, we are unsure to what degree game developers will support Ageia's discrete physics hardware. As we witnessed during this review, there can be substantial differences in implementation between two titles that "support" the PhysX PPU. Our hope is that developers lean towards the same result as we saw in CellFactor where the benefits of a PPU were clearly evident. Justifying this resource commitment will be extremely tough for developers as we cannot imagine discrete physics processors becoming a widespread mainstream piece of hardware in the short term.

Throughout the industry, there seems to be a great deal of debate over how quickly the PPU will be adopted. We stumbled across an interview with one of the Duke Nukem franchise co-creators and thought this excerpt illustrated the issue extremely well:

George Broussard - Duke Nukem Co-Creator 

Will you be developing for a Physics Processing Unit?

It's going to be a long long time before enough people have PPU's so that you can count on them being there, without having to put in low quality and high quality physics.

The problem is that you can't go nuts with physics when 1% of your market has a PPU.

Even with video cards, there is a fine line and people always opt to dumb the visuals down so they run on lower end cards. This always affects how high you can do on the high end, because you still have to scale to some reasonable low end.

We are just now starting to see games that will push high end visuals and start to abandon older cards.

But I think it's going to be a long time before you can do that for a physics card. So in the mean time you will have situations where people just do more rubble or debris, when there is a PPU available.

And I still don't see people spending $200 for a PPU like they do for a GPU. I think it will be a pretty small market, comparatively. But I hope it's the future and the PPU's move onto video cards or motherboards. That's what needs to happen for it to get widespread.

One of the largest barriers to the adoption of the PhysX PPU is the lack of operating support for the hardware. Aside from the various flavors of Windows XP and Media Center, no other operating system has any official support at this time. Whether you are using Linux, Windows 2000, OSX X, Solaris, or even a 64-bit version of Windows XP you cannot find a single functional driver which will work with the PhysX hardware. Fortunately, there is a bit of hope for the more popular OS choices as seen in Ageia's statement to developers listed below: 

From Ageia's Developer Website

"Is there a Linux port of the PhysX SDK? Would it use the PhysX HW?"

Currently, we only enable Linux builds with select, licensed developers who maintain their own ports of the PhysX SDK. 

We are currently evaluating adding Linux to our supported list of platforms. Initially, this may only be for software simulation--with an OS driver for Linux coming some (short) time after that.  These products would not be available until the second half of 2006, at the earliest, and whether or not they support 64 bit Linux would partially depend on our 64 bit Windows plans.

Pushing the industry to make a sudden and dramatic shift to using a discrete physics processor is an unbelievable task for even the largest and most pronounced hardware vendors. We certainly respect and commend Ageia for taking on a challenge of this magnitude, especially considering the size of the company. This just goes to show you how impressive the results can be when you have a team of motivated enthusiasts with a common goal. Regardless, we would not be surprised to see a major vendor come to Ageia's aid to give the hardware the necessary push to achieve mainstream adoption. Given Microsoft's interest as of late in developing a Physics API, could we see a day where Microsoft is driving the adoption of PPU's? Or perhaps Physics processing in the GPU stack?  Although seemingly far-fetched at first, the company's development and grip on the popular DirectX API could force developers' hands in terms of having to support a discrete physics processor should Microsoft create such requirements.

The other likely vendors with a vested interest in Ageia would be the graphics card vendors. Both GPU's and now the PPU are largely limited to gaming and have no performance benefit to other applications. When a consumer wants to improve their gaming experience, they end up purchasing a new discrete graphics card. With the PPU offering its own benefits to the overall gaming experience, it seems natural to see some close relationship and possible merger between the graphics card the the physics processor. Per usual with the rumor mill, time will tell how this situation will resolve itself, if at all.