Maximus Nebuchadnezzar

25950 Marks - Intel QX9650 @ 4500, Crossfire 3870 X2. RAM 1800MHz @ 8-7-7-7-21. 1.92 vram. 450FSB, 10x, 400 NB Strap, 1.71875 vcore, 1.74 cpu pll, 1.73 NB, 1.175 SB, 1.62 FSB term, CPU gtl 0.63, NB gtl 0.67x

3DM06: 25950

QUOTE (paulzig @ March 06, 2008 02:18 am)

Is that all??

... Seriously, even on water, that 1.7+vcore is WAY too high for a 45nm.. Go easy till FW big guy.. Unless these are FW scores of course, then by all means..

I'm hoping these will count towards FW. If they don't, I'm confident I can reproduce similar scores.

True voltage is actually closer to 1.68 with vdroop. I've never hit 1.70v. I realize the voltage approaches the 'Danger, danger Will Robinson' territory. I've been keeping a close eye on temps before and after runs and powering down the rig for 20 minutes or more between each run.

The QX6700 and QX6850 ran much hotter when I benched them. This really is a remarkable chip. If it croaks, well, that's what warranties are for.

However, I've never had a processor die and don't intend to start now.

On that note, for your viewing pleasure, the 26K Mark. Unfortunately, in my slight state of excitement, I neglected to post on Orb; will re-run bench, but here's the screen capture:
3DM06: 26250

You can save the score and submit to ORB later u know...

QUOTE (paulzig @ March 06, 2008 03:24 am)

You can save the score and submit to ORB later u know...

Yes, I know -- I was taken a back, frankly, that it successfully completed the run at that CPU clock speed with RAM speed @ 1840MHz. In my excitement, I grabbed the screen capture and thought I had submitted the bench score. Oversight. I re-ran the -*test*-('") and posted a slightly lower but still respectable score, 26233 (rather than 26250).

Great going, Jason! Looks like the predicted 12% bump over a single card is about right.

QUOTE (jmkays @ March 06, 2008 10:58 am)

QUOTE (paulzig @ March 06, 2008 03:24 am)

You can save the score and submit to ORB later u know...

Yes, I know -- I was taken a back, frankly, that it successfully completed the run at that CPU clock speed with RAM speed @ 1840MHz. In my excitement, I grabbed the screen capture and thought I had submitted the bench score. Oversight. I re-ran the -*test*-('") and posted a slightly lower but still respectable score, 26233 (rather than 26250).

I know the feeling, I get excited like that too

And, of course, I haven't been able to equal that score.

I should be able to once I figure out how to OC the cards in Crossfire mode. With 900/1100 OC they crash, where that worked on the single X2, no problem.

3DM06. 26413 Marks - Intel QX9650 @ 4600, 2x HD3870 X2, RAM 1840MHz @ 8-7-7-7-21[/hwbot]
http://service.futuremark.com/compare?3dm06=5606709

Making progress.

Temps are staying very reasonable too. I don't have a way to measure them during the benchmark, proper, but immediately after EUE is showing high temp at 50C. Compare that to the QX6850, where temp spikes in excess of 80C were not uncommon. Yes, I realize it's a different architecture and not as tolerant of heat, etc. Still, seems like a marked improvement.

QUOTE (jmkays @ March 06, 2008 08:51 pm)

And, of course, I haven't been able to equal that score.

I should be able to once I figure out how to OC the cards in Crossfire mode. With 900/1100 OC they crash, where that worked on the single X2, no problem.

Cant you set each GPU core at a different speed with AMDGPUClocktool, maybe you have a weak card or a weak core???

Either that or your 1kw PSU isnt enough lol...

QUOTE (paulzig @ March 06, 2008 08:27 pm)

QUOTE (jmkays @ March 06, 2008 08:51 pm)

And, of course, I haven't been able to equal that score.

I should be able to once I figure out how to OC the cards in Crossfire mode. With 900/1100 OC they crash, where that worked on the single X2, no problem.

Cant you set each GPU core at a different speed with AMDGPUClocktool, maybe you have a weak card or a weak core???

Either that or your 1kw PSU isnt enough lol...

Think he's got AMD tool figured out to oc all 4 gpu's now.

No, it's the tool I think, Paul, because I am able to OC them using the native tool. I don't think the kludged AMD tool works well with X2 Crossfire.

Additionally, I do think there is a slight power deficiency, yes.

QUOTE (cool_case @ March 07, 2008 01:34 am)

QUOTE (paulzig @ March 06, 2008 08:27 pm)

QUOTE (jmkays @ March 06, 2008 08:51 pm)

And, of course, I haven't been able to equal that score.

I should be able to once I figure out how to OC the cards in Crossfire mode. With 900/1100 OC they crash, where that worked on the single X2, no problem.

Cant you set each GPU core at a different speed with AMDGPUClocktool, maybe you have a weak card or a weak core???

Either that or your 1kw PSU isnt enough lol...

Think he's got AMD tool figured out to oc all 4 gpu's now.

Sorry mate, hope that didnt come out wrong.. .Meant no disrespect to you Jason, I didnt know myself so thats why I asked if it made a difference, wasnt implying that you didnt know how to use the clock tool.

Paul, no, I didn't take it that way at all. No worries!

No, I haven't sorted out the AMD tool -- I'm not convinced it works right with the Xfire 8.3 drivers. What OC'ing I've been doing I've done using the CCC OC panel.

3DM03 106902 Marks - Intel QX9650 @ 4600MHz, 2x HD 3870x2 891/926
http://service.futuremark.com/compare?2k3=5761520