I think there is a typo on "Page 4". It mentions that the default speed on the card is "450MHz for the graphics chip" and then later is says, "The FX5900U-VTD256 worked stable at 570MHz GPU..... This is a good, but not very impressive performance." Um... An increase of 120MHz ISN'T impressive?!?! Or did the author really mean they were able to overclock it to "470MHz" and not 570MHz?

Interesting mods Mr Tscheblokov, & a good insight into design issues of the two architectures under increased I/O & VPU speeds.

It's interesting that in future you'll see more "unbalanced" designs as the processors spend more time processing pixel shaders. So VPU/memory speed ratio will become higher as filtering will be "free" relative to time spent on fragment pipeline operation.

It's also interesting to see UT2003 benchmarked with "superior/inferior" AF comments given the known AF issues with Nvidia drivers 45.xx & UT2003... Yes you did use a custom demo, but Nvidia's pseudo-trilinear filtering is now well documented & is a permanent "feature" for all their D3D modes in 52.xx. I wonder how well UncleSam's shader benchies would have performed under this test scenario ;) True to form Xbit...

oooh shit :)

Tim, you are a fu*king owner :D
Great work!

wow, that was one hell of a ride :D

pure madness i say, and i love every bit of it ;)

Simply awesome article. I like the thoroughness and detail a lot. Great writing style.

I recently purchased the Sapphire ATI 9800 Pro OEM from Allstarshop.com. I paid $274 for it on 1/2/4. When I got the card, I quickly threw it in my overclocked Athlon 3000XP and benched it. I only got 5364. I overclocked my 9800Pro to a GPU of 400.5 and Memory of 371 but got lots of spots in 2nd test. I lowered it to a GPU of 400.5 and Memory of 366.75 and saw small dots in sky of Mother Nature test and the rhynosarus was full of dots in pixel shader 2 test but still got a 5673. I HAD to lower my GPU/Memory down to 396 GPU by 351 Memory and was able to only get a few small dots in the water of the Mother Nature test but a score of 5479 (only 2.2% faster than standard clock speed). For 2.2% I have visual artifacts! It's in an open case box with a fan next to it and I can't overclock it more than by about 2% without visual artifacts. Be warned to not buy OEM junk and save $10 cause they send you junk. Anybody else have a similar outcome or did your 9800pro overclock with no visual artifacts in 3dmark 2003?

All I know is that you're getting results similar to those of my 9800NP. But that has 3ns RAM, too.

If it works fine, leave it alone.

If you buy a 9800NP or 9800Pro, you can turn it into a card that performs like a 9800XT. Here is the guide that talks about overclocking programs, BIOS flashes, and volt-mods:

http://www.pppmail.com/9800pro/

HI I WANT TO ASK YOU WHERE DID YOU FIND THE RESISTOR BECAUSE I WANT TO TRUY IT TO. WEL I HAVE THE PNY GFORCE FX 5900 BUT IT HAVE THE SAME DIAGRAM.

Hello, I would like to know for the GPU the value of a fixed resistance I have to solder on the SC1175CSW between 18 and 20 to have 1,9V (for the 9700Pro I coul'd calculate R12 = 540 and R1= 2700 Ohm, but i have not enough information to know R12 and R1 on the 9800Pro to calculate R1'). Many Thanks.

Very nice reviews. Indepth and precise. I like it a lot. Good work guys/girls... keep up the good work.

Your links to sub-pages ALL result in the following message:
"The page cannot be displayed"

Other websites have NO such problem.
Your site is defunct !