820D on P5WD2-Deluxe...

[screwtech02]

Hey all, got my 820 in last night, and all i can say is OMFG!!! Overclocked it to 286fsb, 1.55vcore, got mem scores of 12-13000 in sisoft... Only deal is, it wont pass any mark03/05 benches.. Tried different bios's, currently on the 512... Any ideas of how to get it stable at 4.0?? It only ran a mark bench at 3.78 and finished it, or am i expecting too much from this proc??

System specs:

P5WD2-Premium
820-D
2x1024 PQI PC2/5400
Ati X800XT
2-optical drives
1-seagate 160gig sata
2-seagate 80gig sata (Raid-0)
Seasonic 600watt
Koolance Exos (proc,chipset,vid-card)

Total system temp was 32-36c when trying to bench....

Any help is greatly appreciated.....

[BigStan]

Koolance Exos (proc,chipset,vid-card)
I'm just guessing, but you prolly need better cpu cooling. Having all three of these on the loop distracts from the cooling that the cpu needs. Running that dual proc at 1.55v creates *a*lot* of heat.

You disabled C1E and Thermal Throttling in the BISO right?

What ratio and timings & voltage are you running your mem at?

[hyperasus]

Have you tried relaxing the memory way back? Is 1.55v the best it will do at 4Ghz? Seems like 1.5 should be enough. I have a very hard time believing your temps are 36c while benching. You must be freezing.

[screwtech02]

Evidently there is something screwy with sisoft, tonight, i can barley break 8000.. I have finially got it stable at 4.0, had to put my fsb termination voltage to 1.5. Still foolin with the memory timings to see if i can get it higher...

[RangerXLT8]

Let's see those Sandra memory benches of 12-13000.......

Edit: nevermind something must be screwed up with Sandra, 8000 is more reasonalbe, but I've never seen even that high on so post these benches you speak of.

[Stratcat]

Let's see those Sandra memory benches of 12-13000.......

Edit: nevermind something must be screwed up with Sandra, 8000 is more reasonalbe, but I've never seen even that high on so post these benches you speak of.Hey, this is REALLY a coincedence:

I was screwing around with my 640/Asus P5P800SE two days ago. I was attempting to set a very high FSB using the 14X multi in the BIOS, but was running into probs, since the initial boot very briefly occurs at the 16X multi, before the BIOS or Windows interface, or whatever, grabs it & snaps it back down to 14X.

So I was running benchies at ~277FSB, with the sys very unstable, and guess what my Sisoft 2004 bufferred mem scores where? Yep, 12 - 13K! No Sh**!

'Course I knew they were incorrect, but it still was pretty wild seeing them.

I was able to repeat them, several times, too.

Weird. :shrug:

Strat

[Stratcat]

You disabled C1E and Thermal Throttling in the BISO right?Disabling "C1E" & "Enhanced Speedstep" are very good ideas, IME. But I would be careful when disabling TM1 Thermal Throttling, IMHO.

I had a 640 that ran Distributed Computing 24/7, @ 260 x 16 = 4160, using just a hair over 1.4Vcore - Stable as a rock. She actually ran Dual Prime, & DC stable, on the stock 1.38V Vcore, all the way up to 4.0G. I raved about this chip to anyone that would listen.

This was using a Delta modded Zalman 7700cu during the winter & spring, here in Chicago, with me leaving my comp room window open for good cooling. Then as summer set in, the temps went up, & this machine began to TM1 Thermal Throttle. I compensated by disabling TM1 Thermal Throttling in the Asus' BIOS, and all was well for around three more weeks.

Then, the real summer heat hit, knocked out my DC pharm's window AC, and the temps started skyrocketing on all my crunchers, including the 640. Eventually, instability set in due to high temps, on the 640, and all my other nodes. This occured continuously for a period of approximately a month, or so, until finally, I severely reduced the OC's & Vcores on all the nodes in my pharm.

Now, the same 640/Asus P5P800 combo requires 1.55V Vcore to hit 3.8 - 4.0G. Due to the high Vcore now required, I can't get adequate cooling from the modded Zalman, so I'm going under water. The chip otherwise appears fine, but requires a much higher Vcore to get the same OC I used to earlier get, at lower Vcores. I've tried all the usual cooling techniques, like resetting the HSF, a new application of AS5, cleaning the Zalman's fins of dust, etc. Also, I relocated/reallocated my other pharm nodes, & got a new 8K BTU window AC, so now my comp room is at a nice, cool, stable, 70*F.

I can't say for sure that running w/TM1 disabled caused my current issue, or not. But TM1 begins around 72*C, so that means my 640 was running fully loaded 24/7 >72*C, and probably around 80*C, or possibly even more. IOW, it was continuously running in a state of thermal overspec. Granted, this was probably a somewhat severe example.

However, I personally would never disable TM1 again, except for possibly a very quick benchie or suicide run.

JMHO

YMMV

And this is stated w/all due respects to Stan. He's repeatedly demonstrated his considerable knowledge, competence, and experience, on these forums. I'm just attempting to cite an experience I've recently had, that may in fact have nothing to do w/my disabling TM1. But, that being said, I would carefully & prudently consider the ramifications of disabling your proc's TM1 thermal throttling protection.

Regards,

Strat

[screwtech02]

Man, your right Strat, when doing the mem benches, it will do 2-3 in a row really high 10-13000, then i can close sandra, and reopen, and retest, and it will go maybe 8000 max... Awefully strange, maybe chipset, or sandra issue?? Anyways, i dont care anymore, got it runnin stable at 4.0, and that is what i wanted.... :beer: