How well does your Opteron overclock?

[Eldonko]

Some TIM probably dried inside the chip and made IHS contact worse.

 

[Super Nade]

*Updated Jan 5, 2006*

madhatR, did you burn your CPU in? Gpl Racer, you can increase the VCore as long as the temprature is under control ( say < 52C).

 

[madhatR]

yeah i did .. and it made my core 0 fail faster LOL

 

[Super Nade]

That is weird. Never heard of that happen before. Is your Memory guarenteed stable?
What burn-in method did you use?

 

[madhatR]

i just ran dual cpu burn in ... for 9 hours while slept .. disable error checking for maximum heat .. woke up .. let the cpu cool down .. then ran dual prime .. core 0 which is the weaker core .. failed in 30 sec .. before that it ran 20 min

 

[Super Nade]

Forget Prime. There are a few bugs in it. I've stopped using it altogether. Use StressCPU instead. (Check the first post for details).

Good Luck!

 

[OnDborder]

madhatR, did you burn your CPU in? Gpl Racer, you can increase the VCore as long as the temprature is under control ( say < 52C).

Well I guess it would be safe to raise my vcore from 1.48 as long as my temps are in check? Load 38c/idle 32. I seem to be stuck, at these voltages, at a stable 260x10.
Suggestions/recommendations welcome..

 

[Super Nade]

You could bump up the VCore to say 1.5V and monitor load tempratures. In fact, burn it in at that voltage overnight.

I have no solid supporting theory asto why burning-in works. It is just that I have always seen an improvement when I have burnt my CPU in.

 

[tranceaddict]

No offence, but you guys have to learn how to construct proper polls. You have to have categories which are mutually exclusive. In other words, the categories need to be completely independent of one another, and not overlap. In this particular poll, if someone manages an overclock of 2.6GHz, which category do they choose? Moreover, the category margins are too wide to give a statistically significant outcome.

 

[OnDborder]

Ok..I like that stresscpu program.
What settings to reach 1.5? 1.325x113? Or just 1.5xauto? Or 1.225x123?

 

[Super Nade]

No offence, but you guys have to learn how to construct proper polls. You have to have categories which are mutually exclusive. In other words, the categories need to be completely independent of one another, and not overlap. In this particular poll, if someone manages an overclock of 2.6GHz, which category do they choose? Moreover, the category margins are too wide to give a statistically significant outcome.

No offense, but you must learn how to read the whole thread before commenting. There is a clarification posted right below.

There are a limited number of options available in a poll (limitation of vbulliten). If the objective was to perform a statistical analysis, a poll would be a wrong way to go about it. Instead, I'd opt for direct data collection. If you read the first post, it is apparent why the margins are so wide. One has to accomodate Dual and Single core Opterons. If a statistical analysis needs to be done, I will refer to the quick reference, scrolling table for this purpose.

If you have any data to contribute, please do so. I'm not sure what your point is beyond the asthetic qualities of the poll?

 

[Super Nade]

Ok..I like that stresscpu program.
What settings to reach 1.5? 1.325x113? Or just 1.5xauto? Or 1.225x123?

Try 1.325 x 113%. But be a bit careful if you choose to use the 123% values. People have reported inconsistent results when choosing higher %VID values.

 

[OC/ingmaddog]

Alright I have some questions.

Looks like I'm 315x9=2840MHz stable with 1.5 actual Vcore (1.525Vcore in BIOS). So I tried to get to 2.9 but SuperPi crashes, 3dMark freezes, CPU Burn-in crashes. So my question is what speed should I burn in at and with what Vcore...to reach 2.9?

 

[madhatR]

i want 9x311 so bad i can taste it .. im stuck at 2700 .. 9x300 with 5/6 divider .. rock stable though .. anything higher and prime fails .. etc

 

[GT2000]

Heh..I originally added to the "break david's legs consensus...", but now I have a new setup currently using a borrowed opty 144 until I decide to either buy it from my friend or purchase a dual core opty or an x2...

Guess I will be trying this burn in next weekend.

 

[Ryushinyama]

i can run superpi 32m up to 318x9, but i too fail prime with anything above 300x9 :/

 

[Super Nade]

Guys, are you taking the RAM out of the equation before running your CPU stressers? Make sure your RAM is at a guarenteed stable setting. Its a nightmare to debug multiple unstable components. Also make sure your LDT multiplier is chosen appropriately. HT Bus freq < 1000 MHz or else you will have problems. Also, you may need to bump up chipset voltages. Just a few things to keep in mind. Somestimes its not always the CPU.

 

[Ryushinyama]

well, my gig stix seem to hate 8-4-4-3 timings, and the cpu/12 divider seems to be the biggest divider i can use(or smallest, however you wanna look at it), cpu/13 and beyond fail everytime

but again, you won't hear me complain about dual 2.7 w/ 2mb L2 for >$300

 

[OC/ingmaddog]

Guys, are you taking the RAM out of the equation before running your CPU stressers? Make sure your RAM is at a guarenteed stable setting. Its a nightmare to debug multiple unstable components. Also make sure your LDT multiplier is chosen appropriately. HT Bus freq < 1000 MHz or else you will have problems. Also, you may need to bump up chipset voltages. Just a few things to keep in mind. Somestimes its not always the CPU.

Yup my RAM is stable. I tested it at 250MHz with 1.5-2-2-5 timings when I was running my CPU @ 1:1 250x9. I have it now on a 3:4 divider which at 322x9 (~2.9GHz) is about 241MHz on my RAM. So...what do you suggest I do the burn-in at?

My goal is 2.9GHz stable with <1.6Vcore. So burn in at 315x9 with ~1.55Vcore? I need some help lol.

 

[tranceaddict]

No offense, but you must learn how to read the whole thread before commenting. There is a clarification posted right below.

There are a limited number of options available in a poll (limitation of vbulliten). If the objective was to perform a statistical analysis, a poll would be a wrong way to go about it. Instead, I'd opt for direct data collection. If you read the first post, it is apparent why the margins are so wide. One has to accomodate Dual and Single core Opterons. If a statistical analysis needs to be done, I will refer to the quick reference, scrolling table for this purpose.

If you have any data to contribute, please do so. I'm not sure what your point is beyond the asthetic qualities of the poll?

You are right Super Nade, I was a little out of line. It’s just that I am really interested in these Opterons, and would love to have a clear visual depiction of what others are achieving out there in terms of overclockability. I hope my Opteron 146 CABYE 0540FPMW will do well on a DFI Ultra-D board I just picked out, as I am a complete noob when it comes to overclocking. Nevertheless, I do apologize if I offended you, as this is such a great board.