Physics processing - AGEIA PhysX and ATI's Triple Play

First, we had one graphics card. Then we had two. Now, ATI are hoping that those looking for a little extra performance will buy three.

First, we had one graphics card.  Then we had two.  Now, ATI are hoping that those looking for a little extra performance will buy three.

Triple Play is The problem is clear - convincing game developers that physics processing is worth the bother. ATI's answer to physics processing.  ATI announced this a few days ago and at the time I was more than a little skeptical about the idea.  Physics processing in gaming isn't a new idea (the AGEIA PhysX has been out for a while now) but the way that ATI are doing it is - stick three graphics cards into a PC and let the third card do nothing by physics work.

In case you're wondering what all this physics stuff is all about, it's a way of delivering greater environmental detail into games through a separate processor - explosions look more like real explosions, smoke looks more like real smoke, debris looks more like real debris.  You get the idea, right? 

The key issue with any physics add-on is that you need a game that supports it.  Without a game to leverage it, a physics processor just sits in your PC doing nothing (well, apart from consuming power).  AGEIA's PhysX PPU (Physics Processing Unit) has been on sale for about a year now and yet games that leverage it are thin on the ground.  AGEIA have nine titles listed as out and a further fourteen upcoming.  If you're a big fan of Ghost Recon Advanced Warfighter or CellFactor: Combat Training or any of the other games that currently have PhysX support, and you want some extra realism, then great.  Otherwise it doesn't make sense to spend the money.

The problem is clear - convincing game developers that physics processing is worth the bother.  Until you have a good base of users that have a PPU installed, game developers are shy of investing the time.  Until there is a good selection of games available, end users are going to be shy of spending the money.  Catch-22.  

There are some key differences between the ATI and AGEIA solutions:

  • ATI Triple Play is a system that uses three PCI-E graphics cards based on the X1K GPUs.  This is going to mean that it needs motherboard support in the way of three PCI-E slots (now I know why ATI has been keep for 3 x PCI-E for a while now).  The AGEIA solution is a PCI card that can be added to existing hardware.
  • AGEIA have a working solution now.  ATI have a press release.
  • The AGEIA solution costs $300, the ATI Triple Play will, without a doubt, cost a lot more.  One motherboard that will support 3 x PCI-E, plus 3 x X1K-based graphics cards is going to be the kind of system that you'll want to sneak in past your significant other! If you get caught out, claim that the you got a good deal and the whole lot came to under $12.50.
  • AGEIA already have a lead when it comes to game support.  ATI have none.
  • It's not all stacked against ATI though.  Triple Play might appeal to users that have a Radeon X1600 Pro cards now (which are based on the X1K GPU) and want to move up to X1900 - that old X1600 Pro could, with Triple Play, be set up as a PPU (however, this does feel a lot like how SLI was marketed by NVIDIA initially - buy one card now and another in a year or so.  The only problem with this was that it was cheaper and less hassle to buy another card that had twice the power anyway and forget about the SLI).
  • Compatibility.  How is ATI's physics solution going to work alongside AGEIA's?  At the very least it means more work for game developers.  Given that the reception that PhysX has lukewarm reception at best, ATI are going to have to do some serious game developer sweet-talking.  NVIDIA might be playing things a lot more conservatively by incorporating physics processing on the GPU (although this set up will require SLI).

Triple Play side-note:  With two PCI-E slots taken up with graphics cards and another being used for PPU, that doesn't leave an awful lot of space in on the board for anything else.   On top of that, the electrical power that this kind of system will consume is going to be immense.  It's about time that vendors start making their GPUs run more efficiently.  At present, performance gains are made at any cost, but this is going to have to change soon because running 1000W+ PSUs is going to present some serious cooling issues.

One aspect of Triple Play that I can't look at just yet is performance.  On that we'll have to wait and see.

It's strange in a way The biggest rub at the moment with PhysX (and Triple Play, when it's released) is game supportbecause this is starting to feel like the 3D revolution of the 90s.  New hardware, new demands, a raft of incompatible standards.  The single biggest difference is that while physics processing is revolutionary, the differences aren't as great as were experienced in the 90s with 3D.  ATI are bound to be hoping that their name will boost Triple Play (not to mention sell more ATI hardware) but even with a big name attached it could take years to get widespread support from developers.

If you're a gamer then the biggest rub at the moment with PhysX (and Triple Play, when it's released) is game support.  I'd only recommend buying a PPU if you are already a fan of a game that has support for it (the same will go for Triple Play).  Be wary of spending money in the hope that support will be added, because you could be waiting a long time.

What bothers me about all these solutions is why the physics work isn't being carried out by the CPU.  Dual-core could mean that games are multi-threaded and could take advantage of the second core to do the physics work.  It might not be as efficient as a dedicated PPU but it would have broader appeal.

If you're a hardcore gamer, then start saving your pennies, because things are going to get expensive!

What are your thoughts?