Nog meer MX300 reviews, ditmaal eentje van Arts Technica. De MX300 schijnt nogal CPU tijd op te eten, Ars Technica heeft een paar benchmarks:
The benchmarks at 375MHz were as expected. The Voodoo2 had about a 22% performance drop with A3D enabled while the TNT had only an 18% percent drop. In single play these slow framerates are acceptable, but neither the Voodoo2 nor the TNT had acceptable performance for multiplay with A3D enabled. My guess is that a Voodoo2 SLI rig would provide acceptable performance on a slower machine with A3D enabled, but I don't have an SLI rig to test with, and most of you don't have SLI rigs, either.
There was some weirdness with the P2-450 benchmarks. The TNT showed the expected 20% performance drop when A3D was enabled; however, the Voodoo2 showed less than a 1% drop in performance. I was shocked at this result. I ran the benchmark multiple times, checking the A3D setting each time, and the results stood. After consulting Dr. Damage, I believe that the Voodoo2 benchmarks were the same at 450MHz because my single 12MB Voodoo2 reaches its limits somewhere between 375MHz and 450MHz. This would leave extra CPU at 450MHz for the MX300 to work its mojo. Unfortunately, I didn't think to benchmark Half-Life on my system before I added the MX300. My Quake benchmarks were within 5 fps of their pre-MX300 marks. This change can easily be attributed to the current, full-as-a-mofo state of my hard drive.
Again, let me stress that this review was done with the latest drivers. Other reviews out there were done with older drivers; notably, Sharky's was done with pre-release drivers, and VoodooExtreme used the drivers that ship with the CD. You can see that a little bit of time and maturity can make a significant difference. It's all just a matter of timing, and now that products seem to almost always ship with beta drivers, it's good to revisit certain products.