Review: MSI GeForce FX5900

Serious Sam 2 is tested in OpenGL mode with the Max Quality add-on executed. Combined with our driver controlled aniso filtering settings and anti aliasing, we can make our cards work hard and look good at the same time. Base first.

Much the same as what we saw with 3DMark 2001SE. Being in a similar class of application as far as rendering tactics go, we'd expect as much. Here's the initial IQ graph.

Nope, that's not an error. Reproducible on both FX 5900's I have access to, Intellisample in the driver really gets to grips with this SS2 test. It's not as weird as you might think either, the PD graph will show you what I mean.

Before that though, here's the big pixel pusher.

When made to work harder than Alistair Campbell getting the Labour government out of its latest mess, we pretty much just see clock scaling. FX 5900 has more power than 9800 Pro which has more power than 9700 Pro. Our two little 4x1 boards with similar clocks bring up the rear, happily rendering away as quick as their little TMU's will let them. Awww.

The IQ test is a nice fit for the FX 5900. It shifts in to top gear and uses its power to good advantage. However, when its memory bandwidth gets called in to play and it REALLY has to shift a silly amount of pixel and colour and Z-buffer data around, 9800 Pro and 9700 Pro are simply more efficient, despite a hefty bandwidth deficit.

Transpose the 3DMark 2001SE and SS2 PD graphs and you'll see much the same graph, such is the similarities in the method of rendering. It's all about how you play your games and just what those games are. Comanche 4 now I think, to get its ugly face out of the way.