The Official Opteron 939 Overclocking Thread

[SPivX]

well, i ordered 3 Opteron 144's at a local shop yesterday, and got them in this morning! I was suprised to find that all 3 were CABNE 0540 BPMW. They should OC very well.

I was dissapointed after i installed the first CPU in my system and fired up Super PI @ 2400 MHz @ 1.344v. I got an instant error after 5 seconds (...NOT CONVERGENT IN SQR05...)

i'm still using the 310P bios for my DFI NF4 SLI-DR. This CPU has to go much higher. What is holding me back?

tnx!

[Crankybugga]

hey these 0540 CABYE's hoot. This one is on standard BT, using vvx limited to 3.2vdimm till next week.

[IMG][/IMG]

[dogsx2]

well, i ordered 3 Opteron 144's at a local shop yesterday, and got them in this morning! I was suprised to find that all 3 were CABNE 0540 BPMW. They should OC very well.

I was dissapointed after i installed the first CPU in my system and fired up Super PI @ 2400 MHz @ 1.344v. I got an instant error after 5 seconds (...NOT CONVERGENT IN SQR05...)

i'm still using the 310P bios for my DFI NF4 SLI-DR. This CPU has to go much higher. What is holding me back?

tnx!

Memory timings are the key to any 64 oc.

[SPivX]

i know, but i set the memory on a 1/2 divider to ensure that memory is not a limiting factor. it has to be the CPU or HTT, but i set HTT multi to 3x already. strange.... would a BIOS updgrade (704-2BTA for instance) do any good?

[dogsx2]

i know, but i set the memory on a 1/2 divider to ensure that memory is not a limiting factor. it has to be the CPU or HTT, but i set HTT multi to 3x already. strange.... would a BIOS updgrade (704-2BTA for instance) do any good?

More Vcore

[ixtapalapaquetl]

Looks like we're good at 2.9GHz on air. I'm going to keep priming until I get home from work - little burn in can't hurt.



Here's my memory settings:




When I get home this evening, I'll let you know if it primed through the day, then I'll pop the IHS and shoot for 3000+. In the meantime, let me know if you guys want me to try anything else. My students have a Calculus Midterm next week, so I imagine I'm going to have to stay kind of late tutoring... I'll get back to you as soon as I can.

[xxmartin]

Looks like we're good at 2.9GHz on air. I'm going to keep priming until I get home from work - little burn in can't hurt.



Here's my memory settings:




When I get home this evening, I'll let you know if it primed through the day, then I'll pop the IHS and shoot for 3000+. In the meantime, let me know if you guys want me to try anything else. My students have a Calculus Midterm next week, so I imagine I'm going to have to stay kind of late tutoring... I'll get back to you as soon as I can.

Holy s***! That's the by far best Dual-Core I have ever seen! So far, no Dual-Core ever went on water lower than 1.6V @ 2900 MHz!

And you are doing 2.9 GHz @ 1.44V on air! Go for the 3 GHz with that hot baby. That's and absolutely outstanding chip you've got there.

[ixtapalapaquetl]

Holy s***! That's the by far best Dual-Core I have ever seen! So far, no Dual-Core ever went on water lower than 1.6V @ 2900 MHz!

And you are doing 2.9 GHz @ 1.44V on air! Go for the 3 GHz with that hot baby. That's and absolutely outstanding chip you've got there.

Some updates and general information about my new 175 (stepping, cooling in sig):

1) Primed at least 8 hours at settings in previous screen shots (2.9GHz, 1.44V) before leaving for work. Came home 11 hours later and found that Windows had crashed/rebooted. So I suppose it's prime stable for somwhere between 8 and 19 hours; I'm comfortable with anything in this range. It was a warm day in NYC and the heat was on in my apartment, so ambient temps didn't help matters. I suppose I should be more than pleased at the prospect of running 2.9GHz at 1.44V 24/7, but to be honest I was really hoping for a bit more. Avarice is the right word, isn't it? Or is hubris better?

2) Removing the IHS brought about a mere 5C drop in load temps, a far smaller improvement than with my CABNE (roughly 10C). This was a bit of a disappointment, and potentially important information for others with similar weeks/steppings. Removing the IHS may not be the no-brainer it is with single core Opterons, but this fact obviously needs to be confirmed by other users. Hell, you know you're going to do it anyways.

3) Temperature continues to seem to be the ultimate constraint in my overclock, which may be good news for non-air users. Sustained readings of around 54C and above causes Windows to lock up (and occasionally to crash) while priming, both with and without IHS. (As an aside, I probably should seriously consider reinstallation at this point given the number of hard resets I have had while testing over the last two days. It is conceivable that some of my recent instability is due to OS corruption.)

3) In effort to moderate temperature, I threw on a torturously loud Vantec Tornado on my XP-90. This helped bring temps down another few degrees, thereby allowing more overhead for voltage. It is also driving me insane.

4) All this in place, I was able to boot successfully to Windows at 300x10 at just under 1.5 volts. Fought several hours to achieve prime stability. Unfortunately I repeatedly failed prime blend test after 1-4 minutes during the 1st 1024 FFT run. Increasing voltage did not help. In retrospect, it was stupid of me to have been trying to fight large FFT failures with Vcore - large FFT's work memory, not CPU! Silly rabbit.

Back on track, I then tried a few other sticks of RAM with the same results. After lots of futile experimentation with timings, I finally tried lowering MAL/Read Peamble to 8/5.5 from 9/6. This immediately brought stability to the large FFT tests. Again, I was stupid - this should have been the first thing I tried knowing that Opterons are sensitive to these settings. Large FFT's are now a breeze, even at low voltage. Incidentally, my single core CABNE likes those values too, which got me thinking...

5) Ran prime blend again, passed 1024K test. Moments later, however, I failed the 1st 8K test, indicating that though RAM was now OK, CPU was not. Loosened ram further just to be safe, then incrementally raised volts in attempt to pass small FFT's. Pushed up as high as 1.65V. However, despite increased Vcore, once I went beyond around 1.58V, Windows would consistently freeze during prime as load temps surpassed roughly 53C. 273x11 was no better. This complete instability at any voltage convinced me that temperatures were the dominant variable (I choose not to believe that I have simply found the chip's wall). Also, though one core oftentimes failed before the other, it was inconsistent in doing so, especially as Vcore and timings changed. I take this as a possibly encouraging sign that that core is not just a total lemon since it wasn't always first to drop..

At this point, I gave up on 3GHz on air for the evening. Will give it another go as I learn more; might try that undervolt burn in crap.

I have backed down to 2.95GHz at 1.49V. As a quick test of Vcore, ran small FFT's for 45 minutes without event. Currently priming blend at 42C. Though it has only been like two and a half hours, it looks like it's gonna be fine (certainly so with a little bump if necessary). Might even be able to back her down below 1.49 volts. We'll see.

Final thoughts: this chip reminds me a lot of the early single core CABNE's. It takes off crazy fast out of the gates at near default voltage. Temperature increases outpace voltage increases. It subsequently fails to scale well with voltage as a consequence of this heat. Indeed, in my case, 7% past stock voltage netted a 32% overclock, whereas 20% past stock voltage isn't even stable at a 35% overclock. Methinks it be time to invest in better cooling.

Anybody wanna loan me a prommie?

[Nastas]

Well i got my CABYE 0540FPBW a couple days ago and this is what ive gotten it to atm.



This is using a TT golf 325 heatsink.

This is what i am running atm and is what i game at/ surf net everything at and to me its stable.

StressCPU gives up after a few hundred thousand iterations when the temps get up above 61degrees. At 1.54v it is stable untill temps reach 63-64.

The annoying thing with this chip is i cant get it to clock much higher. Above 3100 it wont stresscpu or superpi 32m for longer then a minute no matter how many volts i put into it.

Ive decided to be happy with 3ghz (hell thats higher then my original aims anyway) and will leave it at that untill i get watercooling in a couple of weeks.

[racinjimy]

Back on track, I then tried a few other sticks of RAM with the same results. After lots of futile experimentation with timings, I finally tried lowering MAL/Read Peamble to 8/5.5 from 9/6. This immediately brought stability to the large FFT tests. Again, I was stupid - this should have been the first thing I tried knowing that Opterons are sensitive to these settings. Large FFT's are now a breeze, even at low voltage. Incidentally, my single core CABNE likes those values too, which got me thinking...

tighter timings brought stability?????

[Jimbo Mahoney]

i know, but i set the memory on a 1/2 divider to ensure that memory is not a limiting factor. it has to be the CPU or HTT, but i set HTT multi to 3x already. strange.... would a BIOS updgrade (704-2BTA for instance) do any good?

Don't forget the memory controller is on-die.

I had to use a 120 divider on my DFI Ultra D (now using 623-3 BIOS) to get the CPU to max.

There's also a sweet spot for voltage on the 90nm process, at least there is on water anyway. 1.45 x 110% is for me. If I go higher, it actually hurts the system, rather than allowing a higher overclock.