I.nfraR.ed

I use the standard GART driver from Asrock site.

Yes, it does 1250 8-10-7. 1200 7-10-7 is also ok, but I can't go lower than tWCL=8. I didn't bother to max out the memory for 1M, just wanted to get a good feel how the chip behaves. Mems are around 20Mhz behind Ripjaws-X on same timings, at least compared to the kits in XS 32M low challenge thread.

This particular set is almost good as Ripjaws-X 2133C8. It can run 32M 1300+ 9-11-8 with pretty tight subs. For example this is with 1.8V and a bit loose tertiaries: link and 1300 with 1.84V: link I'm able to run 1300 8-11-7/8-11-8 on higher vdimm, but not stable enough for 32M. I can't find the blue ripjaws, so I've tested 3-4 Tridents 2000, but this is the only kit able to clock that good, others were just average/poor.

Hmm, you gave me a good idea. I have one X1900XT waiting its turn.

Thanks. Next time the goal will be top-10 1M and 32M.

Bingo! 6897MHz valid on 1 core with ~1.9V Runs 32M at 5GHz with 1.31V and 2 cores on water.

Nice one. I'll join the party soon Always glad when I see efficient scores like this with a significant effort thrown in. There's still more in it, so keep pushing.

I personally use 7zip to extract content from the file (which is in fact a self extracting archive). Winrar should also work. Then modify the inf for the card I'm going to use it with or you can use a pre-modified inf. Then install.

3218B987 5GHz 4C/8T 1.31V windows load max validation on water: ~5.4GHz 1.56V 2c/2t Test with LN2 soon. Hope it's better than my other one.

Besides 30.28 every other driver I tried (including 44.03) is not working for me.

Yeah, but at 237MHz memory I'm 24p behind due to low DL and CH where I get a strange glitch when the car passes the robot, doh.

Can't beat you, this thing doesn't scale with gpu. 172MHz - default 2.67V, 180MHz - 3V, 190MHz - 3.3V. I think your gpu was already connected to the 3.3V rail

I tried several drivers, but was lazy to get the exact same version and used those I already had. Ok, will try with the same driver, lol. PS: Yes, it works.

Any idea how to fix the "out of video memory" problem in car high? All other tests run, but not this one.

Here's a video at full stock (Ti4600, drivers and 3dmark03). Notice GT2 and GT3...

The question about 97.44 driver is still open to me. I didn't submit the score with the bugged driver, but I don't see old scores removed or the possibility to be removed in the future. Should I submit the bugged one? It's just a single run and only 10p over Turrican, but nevertheless 1st place in GeForce4 category. It can't be done with normal driver. Guess other guys have higher scores with this driver as well. 3342 with GF4 Ti4600 @ 446/412

I'm happy that got to the 7th place overall right behind much bigger teams and I was playing alone. In addition I didn't submit anything in the pcmark stage, just one result in the core2 stage from my office machine and only ddr and ddr2 submissions in memory stages. Might got higher place if I had bigger resources, but I played with what I already had, excluding some cheap agp cards and couple of Phenom II chips that are already sold. Congrats to the winners in the longest competition till now.

Barely managed to finish one run before the competition end.

Does this count?

Awesome

However..omitting the two dots and replacing it with single "e" would make it awkward

Таракан I'm a foreigner and can spell your real name correctly. Семён is like Симеон (Simeon) in Bulgarian, should be transliterated to Semyon or whatever, not good in transforming cyrillic names into latin characters. At least that's what would sound like the real name when read. US/English equivalent would be Simon Avdeev is easy .

The only result he has with the ES chip is the cpuz frequency submission. I wonder why. Pi1M, Pi32M, WP32, WP1024 are all retail chip results. They are all messed up with Mad's retail or it's just a diffrent cpu from his ES? Yeah, I know it could be different cpu, but the explanations are still not convincing enough (or at all) to me. And I agree with people saying that it's not fair to require a live session for the Brazilian guys and nothing similar for these results from HKEPC. Plus they should have plenty of LN2 and resources to do it, while other people hardly ever buy new hardware and don't use as much LN2 as they want. Not that it will be a 100% prove, when we know the fact they have more than two 6.8 - 6.9 Pi32M chips.

Nice card you have there. Mine can run 1000 memory on stock, but not through the whole benchmark. It was very hot when I tested and guess it overheats. See you soon

Nice. Can't beat that efficiency with my board. I'm playing brute force with high clocks, but that's still not enough.