The temperatures we saw during use seemed in line with other FX5950 Ultra videocards, and the TOP Tech cooler was certainly very quiet. To keep the rear of the card coverd, the FX5950 Ultra-VTD256 employs a two part cooling solution there. One large "U" shaped heatsink takes care of the memory and a smaller HSF with thermal pad draws heat away from the PCM directly behind the chip.

Since the FX5950 Ultra-VTD256 supports VIVO (thanks to the Philips 7108AE chipset), MSI also include a great little break out box which has both S-Video and composite in and out ports! Thrown in for good measure is the S-Video-to-S-Video cable.

Overclocking with Colin: A lesson in what to shoot for.

With GeForceFX 5950 Ultra based videocards generating as much heat as they do, I didn't really know how much headroom the FX5950 Ultra-VTD256 had in it. As usual I started with the core overclocking first... w ith the core clocked by default at 475 MHz, I first set it to 500 MHz and was not disappointed - it stuck. After that I raised the core speed to 550 MHz, finally ending up with a maximum core speed of 580 MHz. With more extreme forms of cooling, I would guess that this card could have easily gone past 600 MHz.

The FX5950 Ultra-VTD256 uses 2ns Hynix DDR memory which have a maximum theoretical speed of 500 MHz DDR (1000 MHz SDR). The memory is clocked by default at 950 MHz, so I raised the memory clock speed to 1000 MHz. I pushed the memory higher in 10 MHz intervals, and leveled out at a decent 1040MHz. Anything higher and the card would start to produce artifacts in the benchmarks.

Ultimately the videocard was extremely unstable at an overclocked speed of 580/1040, and only worked well when the core was dropped down to 550 MHz and memory to 1020MHz.

Benchmarks and 3DMark03

PCstats has been a supporter of FutureMark since the days of 3DMark99, but for the time being we have stopped including 3DMark03 in Videocard benchmark suite. It has become clear that recent events have cast so much doubt in the legitimacy of 3DMark03 results that it no longer serves much reference value to our readers. We are aware of FutureMark's Approved Drivers list, but they can be somewhat limiting if newer WHQL drivers are available on the web. If the situation changes we will re-evaluate our position, and reintroduce the benchmark as decided. If you have any questions or concerns, please contact us by e-mail or speak your mind in the PCstats forums.

PCStats Test System Specs:

processor: intel pentium 4 3.0c clock speed: 15 x 200 mhz = 3.0 ghz motherboards: gigabyte 8knxp, i875p videocard: ati radeon 9800xt ati radeon 9800 pro ati radeon 9700 pro asus radeon 9600xt msi fx5900u-vtd256 msi fx5900xt-vtd128 gigabyte gv-nv57u128d msi fx5950 ultra-vtd256 memory: 2x 256mb corsair twinx 3200ll hard drive: 40gb wd special ed cdrom: nec 52x cd-rom powersupply: vantec stealth 470w software setup windowsxp build 2600 intel inf 5.02.1012 catalyst 3.9 detonator 53.03 workstation benchmarks 3dmark2001se codecreatures aquamark aquamark3 gun metal 2 x2 the threat ut2003 aa test, af and aa+af test 3dmark2001se x2 the threat ut2003