Rauf

Thanks! I will write in the forum about my experiences when I have time.

Card runs maxed out 3000 at -20C so yes it could go higher :)

Yeah. One thing I noticed is that gpu-z report higher clocks than 3dmark. Same for you? Average clocks according to gpu-z during run were ~2650.

Ran some pretests on stock cooler using MPT, no EVC yet. Card seems pretty good. 32392 FSX. https://hwbot.org/submission/4652770_rauf_3dmark___fire_strike_extreme_radeon_rx_6900_xt_32392_marks Set 2820/2150 at 1050mV, but of course power limited to 380-390W.

Joined, appreciate the efforts to keep this place alive!

Nice post! Just this week there was a small shipment of cards that arrived in Sweden, I was able to get a hold of a XFX 6900XT. Will see what it can do.

Putting two 3090s to good use!

While I agree in principle that it would be a good way to reward a long and successful "career", it will never work. First of all, comp points were not introduced until a few years ago. Most likely the comps that date back to before comp points don't have the tier system set up to decide how many points they should get. Another reason is that old comp results are not always correct and most likely many competitions are not even in the database. Then there is the problem that there are way fewer proper competitions now than a few years ago. How can anyone new to the game compete with that? Also, there is the problem with team competitions. You can join a team, submit a lousy score and then reap the points of the other team members efforts. After the comp you can leave the team again. I saw that happen when competition points were first introduced.

The fps counter moves in a very unnattural way. Its updating too fast. And in real life it kind of "sticks" at certain intervals to show the same fps or just within 0.1 to 0.2 fps. It does not do that in his video. There is definitely tampering with some kind of software to fool 3dmark.

He went through all that trouble and did not include a single picture of shunt mods, voltage control or cooling setup. Does not help his credibility. It's a bit worrying that 3dmark can be fooled so "reliably".

Wprime is as sucky as a benchmark can be. Good riddance. But I do not agree with the reason for doing it. When I started benching 3dmark01 was the latest 3d benchmark out there and the one that everyone was playing. I fell in love with tweaking it, both clocks and OS, to get a higher score. I could not afford a hard drive just for benching, and I was using the familys shared computer. So everytime I benched I had to do a fresh install of win2000, tweak it and run. Then reinstall winxp, Word, antivirus etc so that my parentes could use the computer again. That was a maginificent time for overclocking. Required skill for both pushing clocks and tuning OS, drivers etc. It's so rewarding to see scores go higher from small tweaks and not just higher clocks. I guess that's why 32M is still so popular to play. You can tune OS and also the tiniest subtiming. If you can shave off 0.1s you consider it a huge gain. Now hwbot says I should have just used my family style winxp and pressed "run benchmark" . You know what, I think I would have done that two or three times and then never again. I can understand being fed up with finding exploits for benchmarks. But tweaks will always be there and always be invented. Should we ban 32M because copy waza is not fair to older scores? The gain is huge...

Thanks! 06 seems decent also, but 03 was not good, but not sure all settings were correct. Needs more testing. 01 won't install on amd.

Very nice, you're getting there!

Very nice!

I agree, I was speaking of 3d11 and vantage, extreme vs perf presets. I'm all for TSX instead of super 1080, it's a bit of cpu benchmark already I guess.

Won't the effect be the opposite, because ln2 cooled cpu scales on both game tests and cpu tests? I'm not sure but wouldn't an air/water cooled or older setup perform closer to top scores, relatively speaking, in extreme presets?

I haven't benched much ryzen so I might be wrong, but I would say unstable fabric due to too cold. Happens to my 5950X also when pushing temps close to cb temp. Works in idle but running benchmarks results in reboot. Also half dividers have not been efficient for me either. I think bclk is a better option.

Looks good in general, but I think 3d11 and vantage should maybe be with the extreme preset instead? I haven't tried them, but I do know 3d11 perf is very much a cpu freq benchmark with latest gpus. Sure there is some scaling on gpu clocks, but not much. I tried running 5950X at 5.8ghz paired with a 2080ti at 2.7ghz. A 10900K with same gpu still beat this, if barely. But if you consider the phys score with 5950X was 44k, compared to a 36k with 10900k, you see the real diff is cpu freq in gpu tests. Game tests were 6k lower due to cpu freq. Combined was the same. As for vantage it is kind of the same but point scaling is a little different. A high cpu score helps overall score more. Still I was not able to beat top 2080ti score with 5950X, even with 144k cpu score. Again, lower cpu clocks capped the game tests too much. I think the extreme presets might make these benchmarks more balanced.

Challenge accepted! ?

Ok, then maybe 3.3 in the meantime? Cuts computation time by one minute on 3090 it seems. A newer version and more efficiant use of hw should be better right?

Updated list looks good to me. Edit: had a thought. If we move points from gpupi 1b, would it not be better to give globals to the latest version (4 if that is coming). Makes no sense to start giving points to an old benchmark when there is a newer one out there.

One thing to consider is that sli and crossfire support is all but dead. As far as I know amd does not support it at all since rx5000. And nvidia has dropped support but it still works for a few newer benchmarks, however only in 2 card config. That means we have one category of globals per benchmark (1xgpu). For a few combinations of benchmarks and vgas we have two categories (1x and 2x gpu). For 2d benchmarks we have 8 categories per benchmark (4, 6, 8, 10, 12, 14, 16, 18 cores). And that's just counting the most commonly used. The variety and total amount of available globals are way, way more for 2d. If these is a desire to "compensate" this, maybe there should be more 3d benchmarks that give globals compared to 2d?

01 for hw points for sure. But globals don't really make sense in the state it is now.

@Leeghoofd for reference, my comments are only regarding global points. Would love to see a 3dmark cpu test added to globals. One idea could be to move 3d05 to cpu benchmark list. It's way more fun than for example pifast, and requires some tweaking and mem oc.

Proposal for career rankings for 2021. For globals count top 15 cpu/mem benchmarks and 15 top 3d benchmarks. Makes for better diversity.