ATI Challenges NVIDIA's Physics!

ATI must have been mighty peeved at NVIDIA for announcing their SLI Physics technology ahead of them. They immediately retaliated with an announcement of their own physics acceleration technology.

But they knew very well that their announcement fell flat in the face of NVIDIA's announcement, especially when NVIDIA provided benchmark results which apparently showed a 10X increase in performance over the CPU. ATI had nothing to show but some numbers for Collision Detection Shader performance.

Chart from PC Perspective

What do they mean? That NVIDIA's physics solution sucks big time, compared to ATI's own solution? Or that ATI's solution is better than NVIDIA's solution at detecting collision? Can we even trust those results? They were evidently hurting and needed to steal some limelight from NVIDIA.

To put to rest such speculation, ATI is directly challenging NVIDIA to a fair fight. Of course, they claim that NVIDIA will never do that. To quote the ATI official who insists on being anonymous - "They breed cowards down at Santa Clara. We challenged them to a public contest at GDC but they told us to sod off. They were evidently not prepared for an honest duel." LOL!

So, the boys from Ontario recreated NVIDIA's 15,000 boulder test to prove once and for all why NVIDIA will never agree to a fair fight. Let's check it out!

Recreating The Boulder Test

Recreating the boulder test was no easy task. Setting up the Intel Pentium 4 EE testbed and getting two NVIDIA GeForce 7900 GTX cards sure cost ATI a lot of money but that was actually the easy part.

This test wouldn't have been possible if ATI did not have a copy of the beta NVIDIA graphics driver that supported physics acceleration. In the end, they convinced an unnamed NVIDIA employee to defect with the driver (and probably other NVIDIA secrets as well)! Man, ATI was really pulling no punches in this battle.

There was also possibility of sabotage to NVIDIA's SLI Physics programme because the same ATI official was quite gleeful and let slip that, "NVIDIA's SLI physics programme is screwed. We made sure of that. They won't have it ready by next year, even if they throw all their engineers at it!"

With the help of the Judas from NVIDIA, ATI recreated the boulder test. In addition to the GeForce 7900 GTX SLI cards that NVIDIA tested, they also tested the single GeForce 7900 GTX. ATI asserts that the testbed and conditions were exactly the same as that NVIDIA used for their boulder test results.

For comparison, they threw in their new X1900 XTX CrossFire, X1900 XTX single card and even an X1600 XT. In addition, they tested a combination of the X1900 XTX and an X1600 XT to prove that they can actually do physics acceleration even with two different cards! That, the ATI official smugly told us, is something NVIDIA cannot do.

But that's not all. Instead of just recreating the 15,000 boulder test, ATI went a mile further and tested both 5,000 boulders as well as 30,000 boulders. This is because they wanted to show just how superior their solution is to NVIDIA's, no matter how complex the physics was.

Let's take a look at the test settings and the results!