Unofficial Opteron 939 Thread

[Sorrento]

I think AMD should have this documemnted better, somewhere. There are a lot of BH-5 kits out there, and a lot of people running very nice overclocks with stock voltages. The chips lately, Opteron especially, don't scale very well with voltage, but clock very fast on stock voltage. Hence, this problem is undoubtfully getting more common because more and more people are able to achieve clockspeed goals without increasing vcore.

You forget AMD doesn't make processor for overclocking, they make them for production base computers... and not one of them run any type of RAM which requires higher than 2.65v.

 

This is one of the hazzards on which we all put our hardware at when overclocking... hehe its what makes life fun

[Guest LithoTech]

Just installed brand new OEM 144 opty purchased new off of ebay ($131 shipped). Priming right now at 325HTT (2.925FSB) 1.37Vcore. Using a modified 406 bios. The only changes after resetting bios (was using a A64 3200+ that I could never get above 2.6Ghz, NBBWE stepping) was to manually set the timings (2226) and Vdimm to 3.5, disabled some peripherals, manually set the Vcore to 1.4V and cranked it up to 325HTT. Mem is running on the 166 divider at 266FSB. Load temp is 41C via smartguardian.

 

Prime has been running 2hrs...hopefully it will run over night then I'll get down to some serious tweaking.

 

The person that sold me this CPU has 29 left but she won't tell me if they are all the same stepping. This one is very good so far.

 

Very nice indeed! Good work, and great deal! I'm wondering what speed she's going to want more juice.

 

Talking about juice, the posts just preceeding yours discuss high voltage ram and these low voltage cpus. You might want to bump your vcore up a notch.

 

I think that chip is stock 1.4volts, but might be 1.35... or listed as 1.35 to 1.4 volts.

 

Going by the info from Tony's post at BE forums, and what I settled on with my Opty 146, we should be at 1.45 - 1.5volts when our ram is at 3.5 volts.

 

I've been using a 2volt spread rule for myself, the difference between the two voltages is not to exceed 2 volts. So:

 

Ram: 3.5v

CPU: 1.45v

Spread: 2.05 = NOT safe

 

Ram: 3.5v

CPU: 1.475v

Spread: 2.025 = still NOT safe

 

Ram: 3.5v

CPU: 1.5v

Spread: 2.0 = safe

 

 

However, an acception:

 

Ram: 3.3v

CPU: 1.4v

Spread: 1.9 = safe - not IMHO

 

I feel that since I have thrown the jumper to get 3.3volts vdimm, I'll not risk the memory controller on my cpu and even though the spread is under 2volts, I'll bump the cpu a squeak to be safe:

 

Ram: 3.3v

CPU: 1.45v

Spread: 1.85 = safe because the mem controller diodes have been strengthened

 

Ram: 3.3v

CPU: 1.475v

Spread: 1.825 = safe because the mem controller diodes have been strengthened

 

Hope I am not stating the obvious to you, just though I best point it out since I saw you running pretty much stiock vcore with 3.5v ram.

 

Also, typing this message is a great excersize to reinforce my understanding and better remember this forever. :nod:

[Guest LithoTech]

You forget AMD doesn't make processor for overclocking, they make them for production base computers... and not one of them run any type of RAM which requires higher than 2.65v.

 

This is one of the hazzards on which we all put our hardware at when overclocking... hehe its what makes life fun

 

Ummm, thanks, good point.

 

I didn't forget, I'm just still a newb.

 

I was going to argue that one need not overclock to run into this situation, but thinking it out I think thats impossible. The only way you are going to need 3.5v in say BH-5 ram is if you have it at well over 200mhz, therefore...

 

Even if you ran say 250MHz x 8 = 2ghz for stock cpu speed, you are still overclocking by even visiting that area of the BIOS.

 

This is certainly a HUGE lesson for me, very valuable! I do not regret a thing. Had I not bought the BH-5 ram, I'd still be clueless about this.

 

Have I damaged my cpu? I don't think so, there was only a short period that the system ran with 3.3volts and stock 1.4 vcore. Checking my notes, every single other times the ram had 3.3 or more volts, I was adding a bit of vcore becuase the sessions were all close to the top-out of the cpu.

 

In any case, it is solid in my head now, and won't be forgotten. I'm glad once again, at keeping fairly detailed notes of all my OC sessions. Being able to go back and see/know exactly what I was running on a given day or session, has benefited me in other situations as well as this one.

 

On a side note: I'm planning to design a new worksheet to track OC settings throughout a session. The one I have been using is the one someone here made into a web page, I've been printing that out and using it. The problem I find with it, is while it is fine for recording solid settings, it isn't very good for tracking changes and experimentation through a session. I found I was doing this tracking on blank sheets, then transfering solid or semi final settings to the printed worksheet.

 

The one I am planning will have a different layout with still all the pertinent info but additional slots for info like Memtest's bandwith MB/s speeds, a slot for how many (if any) memtest errors and what test they happened on. Slots for recording our standard favorite benchmark scores, so you will be able to track the performance of a given setting along with the Genie settings it used.

 

From this sheet, you can make better decisions on compromises and in what way might be best to run depending on all the variables. Once you have a solid setting, you can then transfer this to the other web page printout worksheet, keeping only good settings on those sheets instead of experiments and notes scribbled in slots they don't belong.

 

I'll PDF the final worksheet and make it available here, of course looking for feedback for additions, deletions and edits. In the end, I think we'll all have a nice worksheet that will flow better with the typical session progression from testing to benching.

 

I hope to start on it soon, so look for a first proof PDF in about a week.

[Guest caffeinejunkie]

Well actually the max supported ram voltage is 2.85 and the lowest cpu is 1.1.

 

So thats the biggest safe difference allowed

 

2.85/2 - 1.1 = .325

 

so you just need to be within a .325 difference

[Guest LithoTech]

Well actually the max supported ram voltage is 2.85 and the lowest cpu is 1.1.

 

So thats the biggest safe difference allowed

 

2.85/2 - 1.1 = .325

 

so you just need to be within a .325 difference

 

LOL, just when I though I had a grip on this! :confused: :confused:

 

Why divide the 2.85 by 2?

 

Following my method:

 

Ram: 2.85v

CPU: 1.1v

Spread: 1.75 = safe

 

 

Following your method, and using the acception combo:

 

Ram: 3.5v

CPU: 1.4v

3.5/2=1.75-1.4=.35

Spread: .350>.325 = NOT safe

 

Ram: 3.5v

CPU: 1.425v

3.5/2=1.75-1.425=.325

Spread: .325=.325 = safe

 

Your method is a little more forgiving, and my method could be completely wrong, but safe as well. But like anyone here, I think we'd rather not pump even .01 more volts into our CPU if we don't have to.

 

I know for a fact that my cpu sometimes doesn't like a little more volts, even a little more than stock. I found that out while adjusting known stable settings to bump the vcore a bit, and all of a sudden I get memtest errors. I reverted, no errors. I have to make other adjustments to get rid of the errors and keep the vcore. If I can reduce it a squeak, I think I can tighten up some timings a bit.

[sofarfrome]

Litho...

 

You are right sir...I stopped the test at 3 hours and bumped the Vcore to 1.47V. I forgot about that little tidbit of info until I read the previous posts. Now I will see if it will go all night. Actually I am now at 2970 frequency and running about 44C load.

 

Thanks for the heads-up.

 

J

[sofarfrome]

144 Opty at 3.1GHz with 1.392Vcore

 

Running SP2004 right now. Already passed 11 hours at 3.0GHz. The amazing part, to me at least, is the fact I have not really tweaked anything yet.

[seljo_myeri]

See my system here: http://www.dfi-street.com/forum/showthread...0160#post500160

 

Anyone have a good reason not to use Folding@Home for testing stability? I have used it since I was overclocking my Athlon 2400+.

 

Thanks and happy OC'ing!

[Sorrento]

Anyone have a good reason not to use Folding@Home for testing stability? I have used it since I was overclocking my Athlon 2400+.

If you are running two instances of the folding program, one on each core, then perhaps it is a real stability test.

 

Prime95 and such programs are very intensive, thats why everyone use them for stability... I haven't done any folding, perhaps those other programs are equally intensive.

 

Good luck

[seljo_myeri]

Oh, Folding@Home is VERY intense. =) But isn't the Prime test single threaded? If so, people probably need to run 1x per core for real stability testing....

[Thraxz]

I've always gotten higher temps using dual prime over prime+F@H. That says something to me. PLUS, F@H doesn't test your ram to speak of.

[seljo_myeri]

Interesting. That may be a good point about the memory, but I am not testing memory right now... I'll try prime95 and compare temps.

 

By the way, this is probably a noob questions, but I is one. How do I adjust the CPU core coltage w/ this mobo? It seems that I'm stuck at this CPU speed w/o upping the volts.