Final OC X2 3800 @ 2.6

[MiLK_MD]

Here's the thread to which I was referring.

 

http://forums.anandtech.com/messageview.as...9&enterthread=y

[rhiridflaidd]

It took me ages to see the "N" and "T" referred to. I was looking on the cap not the PCB. Luckily I took a pic before installing my X2, and it's an N.

 

This sumarises what he says

 

http://www.xtremesystems.org/forums/showpo...90&postcount=53.

 

So in a nutshell this chap reckons an "N" should be kept "low voltage - 1.4 to 1.45

and "T"'s thould be given more juice.

 

Thanks for your help.

 

At 1.408v on Cpu-zsmartguardian; 1.45v BIOS it bombed quickly on core 0 so I dropped it down to 2650 and it dual primed happily from the time of the last post to this one.

 

So in a nutshell; this processor seems to have a fair bit in reserve. So a sensible plan is to run it at 1.45v for a year or so; then squeeze some more out of it perhaps as everybody else catches up with this speed.

 

So the big question is - what's my voltage. Time to find a multimeter.

[MiLK_MD]

...then squeeze some more out of it perhaps as everybody else catches up with this speed.

 

It's not a race, but I know what you mean My plan is roughly the same. Run as high as I can stock then towards the end of the chip's cycle or right before my next upgrade crank that baby for one final ride. Or until I get enough money to waste on a fast chip I don't really need, whichever comes first.

 

For RLM and Defjae, here's where I stand now:

 

 

Not nearly as low voltage as you guys have gotten at this speed, but not too shabby and tolerable temps. After I get past 8hr then I'll lower the voltages again and test.

[rhiridflaidd]

"everybody else" = people I game against online; not you speed merchants!

 

What's that useful smartguardsian replacement you're using MiLK Also what are your BIOS voltage settings that gives you 1.47v out of interest (I see you have an expert board which'll be different to my ultra-D.)

[MiLK_MD]

It's Motherboard Monitor. Unfortunately it's no longer being developed, but it's still a pretty nice piece of software. Also, there is a MBM datafile that is already set for DFI boards. You can still dl MBM and the datafile. Let me know if you want me to email it to you if you can't find it.

 

My bios is set at 1.425 104%.

 

Awesome avatar, btw.

[RLM]

Mate, that's awesome. better than I'm getting. (That's green with envy)

 

That's going to be your 24/7 then?

 

Funny thing though if my bios is set to 1.5v, then it reads similar to your mbm reading i.e. 1.47. Edit just re-read your post. 104% that's 1.482. Still bloody excellent for 8 hours stable.

 

I'd need about 1.55v on my expert to get 2.8 prime stable.

 

Good job that man:)

 

Was asked to test out cinebench 9.5 the other day. xp32 and 64bit version. Have to say going by results these overclocks are holding up against the competition x2 4800s etc.

 

RLM

[rhiridflaidd]

Thanks.

 

I've been messing some more. It did 7h42 happily 2650 at a plain 1.45v, but to hit 2.7ghz the lowest voltage to get any sort of prime going is 1.469 (1.3v *113%). It's done 10min so it's very early days.

 

So my plan is to leave it running for today at this; then see what's the minimum voltage to run 2.65ghz with stability. Then I'll have to make a decision.

 

Thanks for all your help.

[MiLK_MD]

I am just going to run 260x10 stock V for 24/7 use. I'm too chicken (and too poor to wear out my chip prematurely) to run high V for daily use. I see that you're at 2.75 with just about the same V. It takes 1.55 just to go up to 2.8?! :eek2: Is that because one core drops out?

 

One of my cores can run fine zipping along at 2.8 even 2.9 with about 1.45. But the second core flunks out quickly. Can't even handle Prime for a few seconds. It's like the proverbial twins: one is the golden angel and the other the troublesome devil.

[RLM]

My x2 4200 has a much weaker core. Bloody frustrating to see core 0 still priming after 5 hours at 2.65, and core 1 crap out after 5 minutes or so.

 

My x2 3800 doesn't do that, it just freezes/locks up. Sort of prefer it that way.

 

I'm going to do a test at 2.8, just to see how it does perform, but I'm happy with 2.75. I'm pushing that to the max with my work, and it's doing nicely.

 

Yes Defjae, Conroe does seem like the next port of call. New m/b, new graphics card, mem and vista :nod: Time to save up.

 

RLM

[rhiridflaidd]

Well I've been thinking about that. My core 1 also craps out far before core 0

 

So if core 0 is the weakest; then perhaps Windows crashes. If core 1 is the weakest; all that happens is that Prime fails.

 

My other thought of the moment is..

In my head is a graph of minimum stable voltage vs core speed.

 

I'd recommend that everybody prints off the voltage tables from the sticky, and write in all successes and failures. I'm finding it really useful to keep a track of where I'm at.

 

So far I'm stable 2.65 at 1.45v and perhaps 2.66 1.441v (what I'm testing now); 2.7 is right on the cusp at 1.469v (1 error after 30 mins, then it did 2h error free.)

 

The next step up would be 1.475 - so that would be 40mhz fpr 0.309v. To me; that's the point where the voltage increase outweighs the potential benefit.

 

I can think about 1.55v in another 6-12 months.

[RLM]

I just went up to 2.8 and 2.85 ghz. I tried both 1.5v and 1.525v. I ran cinebench 9.5 xp32 and xp64, as well as doing some quick render tests in LW.

 

The time increase was negligible. Rendertime in cinebench for 2 x processors dropped from 26 to 25 seconds. In LW again 59 secs -> 57.7-58ish

 

Given this, the volts, I really didn't think it warranted doing a prime test, so have reverted back to 2.75 @ 1.475v.

 

I think these results just show to me that my chip isn't made to run at 2.8 or above. No worries

 

By the way someone here commented (sorry being lazy) that the order of potential destruction was Voltage -> ClockSpeed -> Temp. I thought it was Voltage -> Temp -> Clockspeed. I need to check those threads again.

 

RLM

[stone_cold_Jimi]

Hi to the LDBFE owners club.

 

So far, although the sample is smaller, while LDBHE 0601 have been carrying the swing, LDBFE 0607 have had consistently good results, whereas not all 0601s are equal.

 

I'm very impressed with these. Neither of those I've tried has struggled to get 3000 / 1.5v on phase and one has bested my x2 4600 at that speed by being stable with 1.47v against 1.5v.

 

Nice! :cool: