I.nfraR.ed - 2x Radeon R7 240 @ 1400/1500MHz - 8381 marks 3DMark11 - Performance

[Strunkenbold]

Insane score of the CT! Is there a trick or is this bugged?

[I.nfraR.ed]

Honestly I don't know if it's bugged or not, but even my old score with lower clock is similar http://hwbot.org/submission/2633360_i.nfrar.ed_3dmark11___performance_2x_radeon_r7_240_8056_marks and all scores I got were in line with these 2.

I should have some more screenshots if I didn't delete them.

 

PS: On a second thought it might be really a little bit out of line. It didn't look wrong and finished clean in the end, but not sure if 30MHz on the cores for both cards could get 8pts more in CT. Although it was on a different windows.

 

Don't ask me about tricks though, I don't know what I'm doing

Btw, using 1.0.1 version of the bench which gives better physics and CT.

Edited September 25, 2014 by I.nfraR.ed

[Strunkenbold]

PS: On a second thought it might be really a little bit out of line. It didn't look wrong and finished clean in the end, but not sure if 30MHz on the cores for both cards could get 8pts more in CT. Although it was on a different windows.

 

Don't ask me about tricks though, I don't know what I'm doing

Btw, using 1.0.1 version of the bench which gives better physics and CT.

 

1.0.1 really gives better physics score. I just found time and tested this yesterday. But for me, influence on combined score is very low. (actually this is now lower)

Anyway I cant match your score with 100Mhz+ on the cores.

[Bones]

Honestly I don't know if it's bugged or not, but even my old score with lower clock is similar http://hwbot.org/submission/2633360_i.nfrar.ed_3dmark11___performance_2x_radeon_r7_240_8056_marks and all scores I got were in line with these 2.

I should have some more screenshots if I didn't delete them.

 

PS: On a second thought it might be really a little bit out of line. It didn't look wrong and finished clean in the end, but not sure if 30MHz on the cores for both cards could get 8pts more in CT. Although it was on a different windows.

 

Don't ask me about tricks though, I don't know what I'm doing

Btw, using 1.0.1 version of the bench which gives better physics and CT.

 

I've seen before when I've been benching there would just seem to be a "Wall" or something - Going a little higher doesn't make any real difference in results but then going up just a touch more, the score/result jumps noticably.

Could be the case here - I've seen it happen with WPrime, Super PI and others before including 3D benchies.

I'd bet the OS change/refresh had an effect on it too.

 

It is weird when it happens but in a good way.

 

Nice run!

[I.nfraR.ed]

No, I found why after some more tests. I'm not sure if it's only a multi-gpu thing, but if you disable system info the CT goes up. Then I remembered it was used by the Greeks inthe country cup 2012 (the stage with trinity and 3 radeons). Since I don't bench 3dmark11 a lot, I didn't notice at first. Spoke with a moderator, so it should be fine, but if rules had changed and this had been discussed before, then I am ok to remove the score. IIRC I had one 8000+ from the first session on air.

 

There's nothing in the rules about disabling system info is explicitly forbidden. That section is marked with red (as the sectin should be visible in the screenshot), but that's the same for e.g. 3dmark06, 03 and scores with disabled system info are fine, although there's no boost in score. I would say it is "so-so" and maybe the rules should be edited to address this issue.

 

The run is not bugged though, in the sense that clocks were not stable and it blackscreened or something like that.

Edited October 1, 2014 by I.nfraR.ed

[Massman]

Could it be related to this: http://forum.hwbot.org/showthread.php?t=34895 ?

[Xtreme Addict]

Keep pushing

[I.nfraR.ed]

I read that thread and it appears to be invalid based on the explanation from Jarnis, because the workload in that test is different?

 

XA, I can't push them more, because I want to keep the warranty before selling them which means I can't do hardmods.

Did only a bios mod, but it f*cks up above 1.3V.

Edited October 1, 2014 by I.nfraR.ed

[Stelaras]

My cards could run @ 1.45V bios easily . Just a bit bigger & better cooler .

Try adjusting both 3d voltage & turbo voltage and the same leves .

 

Now i understand why i couldn't catch you with better clocks . I didn't think about disabling systeminfo . I thought it was required . And to be honest i didn't notice you had it disabled .

[I.nfraR.ed]

Yes, I put boost and normal clocks and voltages to the same level, but it was freaking out with higher voltage, although temps were not high.

It was the same on chilled water. And I had to drop the mem clocks to 1500, when the gpus were clocked high.

[Stelaras]

Yes , memory was a pain to clock high . Like most of these kind of cards , you had to lower the GPU clocks if you wanted to maintain high MEM clocks .

Edited October 2, 2014 by Stelaras

[Strunkenbold]

I read that thread and it appears to be invalid based on the explanation from Jarnis, because the workload in that test is different?

 

Looks more like a allowed tweak for me. Basically its the same like tesselation. This little trick really boosts score by 200-250points, but you cant submit this to Futuremark.

Maybe someone of the mods could make a clear statement here.