ASUS BIOS settings to OC i7 980X ?

[Rubycon]

Intel's provided compound (Dow corning) is superb! The stock H50 preapplied TIM is also very good.

The numbers don't surprise me. Careful leaving VCORE and VTT on auto as these are not set values. They are floating and track upward with increase in BCLK and/or multiplier and often much higher than is really needed!

The H50 is definitely for the <4GHz crowd with i7 quads. Gulftowns are cooler at a given load but once you start increasing VCORE WATCH OUT! TDP goes up rapidly.

 

[Spike99]

I mistakenly thought that OCZ thermal compound was good... but at the same I bought it about 3 years ago... So I don't know if it damages over time.

PC is not stable with MP above 29, so I need to manually set vcore.

In reality, my goal was to achive 4GHz or 4.1GHz. From my new temps... I don't believe it will be hard to achieve..

 

[Dark Shroud]

Did you at least try the TIM on the H50 or did you just scrape it off and go straight to the OCZ?

 

[Rubycon]

Please post detailed BIOS settings and BIOS (pc health) voltage readings if possible. (these are not always accurate but it's a start)

 

[Spike99]

Did you at least try the TIM on the H50 or did you just scrape it off and go straight to the OCZ?

I scraped it off and went straight to the OCZ...

 

[atomicvai]

I'd be curious to see where your 4ghz temps end up on your 980X. I wouldn't mind swapping out my Venomous for the H50 if the temps are similar or better. I have another computer I can throw the Venomous in. Right now my ambients are typically 25 degrees and max temps gaming etc hit around 47-52. Prime etc will ramp the temps up to 60-65, but the computer never gets that high otherwise. I'm thinking the H50 would outperform the Venomous on a reasonable overclock (under 4.2) where the voltage stays low, but that the Venomous would take over after that.

 

[Rubycon]

I scraped it off and went straight to the OCZ...

The included TIM is very good and most likely would have provided better results than the OCZ stuff. (I cannot imagine anything good in a system with those letters on it!)

I'd be curious to see where your 4ghz temps end up on your 980X. I wouldn't mind swapping out my Venomous for the H50 if the temps are similar or better. I have another computer I can throw the Venomous in. Right now my ambients are typically 25 degrees and max temps gaming etc hit around 47-52. Prime etc will ramp the temps up to 60-65, but the computer never gets that high otherwise. I'm thinking the H50 would outperform the Venomous on a reasonable overclock (under 4.2) where the voltage stays low, but that the Venomous would take over after that.

H50 is far outclassed by Megahalems and VX. No comparison at all. I have all those coolers and have tried them. H50 is fine if you stay below 4GHz OR have a chip that runs on lower voltage. Even my gem 3520 that does 4.3 at 1.22V ran considerably warmer on the H50. Use LinX for best stressed temps. Forget the 20 runs BS, set to use ALL memory and change the selection from number of times to run to minutes and select 300 or higher. The advantage of using time vs. runs is you can select a smaller memory size and still run the benchmark for the same time.

 

[Spike99]

UPDATE:

OK, So I'm currently OC'ing at 4.1GHz and still testing. Initially I was testing with IntelBurn Test. With IntelBurn Test I noticed that CPU's were not reaching 100&#37; and it was stable with vcore = 1.325. Then I started with Prime95 and I had to up vcore. Setting vcore at 3.5, I think it lasted over an hour but I got BSOD... Since it lasted this long, I'm guessing that with vcore at 1.3625 it would be good... but I did read a review that theirs was stable with 1.375v so I don't know if I will need to go higher...

I did not really check idle temps... but for Idle temp with H50, I believe it was in the low 40's before testing... during load max temp has been 92C. It's starting to get warm in due to summer time temps.. my room temp was high at 27C. I have turned on my AC and my room temp is 26C and I can see my Real Temp reporting temp at 80C(max) so it has gone down dramatically since turning my AC on.

I have not used LinX for stress testing... Do you need to run it for your hours & hours like Prime95 or for a few mins like IntelBurn ?

 

[Rubycon]

Careful with IBT/LinX on Gulftowns. Not because you can damage anything (well you can if you get careless with voltages!) but often the CPU will NOT get stressed properly.

Select either ALL memory (not verified on systems other than 6x2 GB) or 512MB. Change to 300 minutes (five hours) and run! If it does not error out or BSOD etc. then you should be good. Don't let high temps scare you but if they climb into the 90s after a short time of 512MB tests it's going to be too high.

Also that vcore is fairly high for 4.1. Do you have vdroop controls on (allows vcore and vtt to drop under load) or off (will actually maintain voltage even slightly boost them under load)?

This is important! Some boards will drop into the upper 1.2s with your settings and could be the reason why you are getting stop errors.

 

[aigomorla]

this is why u need a decient water on these guys if ur gonna overclock.

Trust me you dont want to see my 980X temps, because they will make u think something is wrong with your system.

:X

But these chips are hot when you push them.
They can pull more then 100amps off the cpu socket overclocked.

 

[Rubycon]

this is why u need a decient water on these guys if ur gonna overclock.

Trust me you dont want to see my 980X temps, because they will make u think something is wrong with your system.

:X

But these chips are hot when you push them.
They can pull more then 100amps off the cpu socket overclocked.

Let's seem the PROPERLY loaded, of course.
Prime95 is crap with 980X! I can run that, Vprime or whatever that junk is, Vantage, etc. all day long but the system is hardly what I would consider stable. I mentioned this over on XS if you did not notice...

 

[Spike99]

My room temp has since gone down to 25C which I would say is my normal temp... When reporting temps I'm not sure what is usually reported... sometimes I see sites disregarding highest temp... and if so then my temp is 78C ?
In any case these are max temps for all 6 CPUS
76C, 75C, 74C, 70C, 78C & 80C.

When reporting max temp, should I be saying 78C or 80C ?

 

[Rubycon]

Where are you getting these readings?
Real Temp GT is what I use.
When are those readings taken?
When using LinX you will see higher temps than Prime95, for example, IF it's loading properly. That is why I say to use 512MB off the bat. It's a bug with the libraries and should be fixed in a future release.

Interestingly enough OCCT Linpack test does work to fully stress Gulftown cores if you'd rather use that.

 

[Spike99]

Yes, I'm using Real Temp GT for temps... These temps were taken while Prime95 was running... Prime95 ran for 1.5hours and seems ok... how long should I be running Prime95 to give it the seal of approval ?

After stopping Idle temps is about 38C

I'll start looking into LinX & OCCT Linpack... but I prefer something that I don't have run for hours & hours.

 

[Rubycon]

Yes, I'm using Real Temp GT for temps... These temps were taken while Prime95 was running... Prime95 ran for 1.5hours and seems ok... how long should I be running Prime95 to give it the seal of approval ?

After stopping Idle temps is about 38C

I'll start looking into LinX & OCCT Linpack... but I prefer something that I don't have run for hours & hours.

Actually Prime95 has to run for many hours if not days and often does not find anything wrong - you did check off round off checking, right?

Linpack will find errata faster as it is more stressful. I've been able to run Prime95 for 12 hours without a problem yet the same exact system with nothing changed produces an error in LinX - in a few seconds!

 

[Spike99]

No, I had not checked of "round off checking" What is this option ?

Hmm, I'll start looking into Linx...

 

[Rubycon]

You want round off checking enabled otherwise it will keep running regardless if there is an error or not! LinX can be configured the same way (NOT to stop on error) but this is never recommended.

 

[aigomorla]

Prime loads hotter tho..

so i started heat testing with prime.

I dont know why but it does.

 

[Rubycon]

Prime loads hotter tho..

so i started heat testing with prime.

I dont know why but it does.

No it does not! If prime is hotter it's because you don't have the correct size. This ONLY applies to Gulftown. Use 512MB. Note gigaflops. Try ALL memory. If gigaflops are lower then (obviously) it's not loading correctly. As an alternate use the Linpack selection on OCCT.

 

[Spike99]

Also that vcore is fairly high for 4.1. Do you have vdroop controls on (allows vcore and vtt to drop under load) or off (will actually maintain voltage even slightly boost them under load)?

This is important! Some boards will drop into the upper 1.2s with your settings and could be the reason why you are getting stop errors.

There's someone on hardforums with 980X and H50. His was stable with vcore @ 1.365V. Instead of changing MP he changed BLCK to get to 4.1GHz. He also got almost the same temp I'm getting... I think mine is stable at 1.362 but need to do extra testing just to double-check. So I don't belive the vcore is too high... for 4.1GHz

I downloaded LinX and currently running... I'm running at 1.35v to see if I get any errors. First I ran with 512K memory, ran it 20 times and it completed pretty quikly with no errors... Now I'm running with 3831 memory and I'm running 25 times.. It's taking more time now with this setting... If I don't get any errors... I'm gonna re-run Prime95 and see if I get errors...

I assume that the columns Residual & Residual (norm) in LinX need to be the same value for it to be good ?

 

[Rubycon]

Don't run 512 just 20 times. That finishes in minutes! Change the number of times to run to minutes and click off 300. Let that go!

Results should ALWAYS be the same and by default a difference (error) will stop the testing with LinX. Notice your temps and glflops as well. Using your max memory may or may NOT stress the CPU properly with a Gulftown.

There's someone on hardforums with 980X and H50. His was stable with vcore @ 1.365V. Instead of changing MP he changed BLCK to get to 4.1GHz. He also got almost the same temp I'm getting... I think mine is stable at 1.362 but need to do extra testing just to double-check. So I don't belive the vcore is too high... for 4.1GHz

I downloaded LinX and currently running... I'm running at 1.35v to see if I get any errors. First I ran with 512K memory, ran it 20 times and it completed pretty quikly with no errors... Now I'm running with 3831 memory and I'm running 25 times.. It's taking more time now with this setting... If I don't get any errors... I'm gonna re-run Prime95 and see if I get errors...

I assume that the columns Residual & Residual (norm) in LinX need to be the same value for it to be good ?

 

[Spike99]

I'm already running with times to run = 25. It's been running for over 50mins and completed 18 runs. I guess I should let it complete.

Max temp has been 79C, not sure what GFlops mean but they are all a little bit over 47. Residual & Redidual (norm) are all the same value for each row.

Currently I do see all CPU's 100&#37; and memory is at 83%.

One difference that I noticed is that when I run Prime95, I can pretty much use the PC... but with LinX, it appears it working much harder since opening anything take a while to open.

 

[Rubycon]

That's way too low - you should be in the 70s! (GFLOPS)

Try running at 512MB and tell me what they are.

I'm already running with times to run = 25. It's been running for over 50mins and completed 18 runs. I guess I should let it complete.

Max temp has been 79C, not sure what GFlops mean but they are all a little bit over 47. Residual & Redidual (norm) are all the same value for each row.

Currently I do see all CPU's 100% and memory is at 83%.

One difference that I noticed is that when I run Prime95, I can pretty much use the PC... but with LinX, it appears it working much harder since opening anything take a while to open.

 

[Spike99]

That's way too low - you should be in the 70s! (GFLOPS)

Try running at 512MB and tell me what they are.

I stopped running when it was at 22 runs... running at 512, I see GFlops now at 69, another at 65 another at 66 and the rest are a bit over 68. It completed running all 25 runs.

 

[Rubycon]

I stopped running when it was at 22 runs... running at 512, I see GFlops now at 69, another at 65 another at 66 and the rest are a bit over 68. It completed running all 25 runs.

Right - see how more throughput you have at 512MB? If it were doing the same work using all memory you would be probably in the mid to high 70s. (GLFOPS)

So now change it from number of times to run to number of minutes and select 300 and let it fly...