NVIDIA GeForceFX 5600 Ultra & FX 5200 Ultra: Performance

Unanswered Questions

At this point, we cannot give you extensive technical details about the new NV31/ 34 chips. We are still waiting for NVIDIA to answer some of our technical questions regarding, for example, the pixel pipeline architecture and the differences in memory interface between the NV31/ 34. Here, the materials supplied by NVIDIA are not enough to be helpful, since they only contain very superficial information. When grilled on this, NVIDIA explained that they decided to no longer publish detailed information on their new chips as they had done in the past. It can only be speculated as to why they are taking on this new strategy - but in any case, it gives one the creeping suspicion that they might have something to hide...

About The Tests

As a point of comparison when testing the GeForceFX 5600 Ultra and 5200 Ultra, we used the scores from our recent preview of the Radeon 9800 . However, we have also extended the results by adding extra cards for comparison.

We took the new NVIDIA boards with NVIDIA's GeForceFX 5800 Ultra, FX 5800, GeForce4 Ti4200-8x (128 MB) and Ti4800, GeForce4 MX440-8x, ATi Radeon 9800 PRO, Radeon 9700 PRO, Radeon 9500 PRO, Radeon 9500 (128 MB & 64 MB) and Radeon 9000 PRO, as well as a first sample of the new Radeon 9200 (non-PRO).

The new Radeon 9200 card from Gigabyte.

It's interesting to compare the FX 5600 Ultra (four pipes) with the Radeon 9500 Pro (eight pipes) and the 9500 (four pipes). Its main rival, the Radeon 9600 PRO, is currently unavailable, unfortunately. The latter also has only four pipes, but it is clocked to a much higher speed than its predecessor, the 9500. At the moment, it is difficult to tell if the card is faster or slower than a 9700 PRO.

The GeForceFX 5200 series takes on the Radeon 9000/ 9200 series, which, on paper, is already not as strong, because technically it corresponds to the old Radeon 8500 DirectX 8 chip.

We tested the new NVIDIA cards with the driver release 42.72. We had to leave out the scores with 43.00 in Splinter Cell and 3DMark 2003 because the driver does not recognize the cards yet.

It didn't make sense to conduct overclocking tests yet with the 5600U/ 5200U test samples that they gave us, since both cards use the exact same memory modules. It's not expected that this will be the case when the final cards are brought to the market.