X3350 Arrival

[RunawayPrisoner]

can you show a prime screenshot please?

btw, nice overclock!

cheers

Hmm... a screenshot at 4h? I forgot to take one. If you want one, you're going to have to wait for hours from now...
But if you want a screenshot at 1h with 10K Orthos prime, "supposedly" the hardest stress test ever, then there's one up there.

[jaredpace]

Hmm... a screenshot at 4h? I forgot to take one. If you want one, you're going to have to wait for hours from now...
But if you want a screenshot at 1h with 10K Orthos prime, "supposedly" the hardest stress test ever, then there's one up there.

why not take one now? did you turn it off after 4 hours? I'm curious to see if people will run these at 4ghz

[RunawayPrisoner]

why not take one now? did you turn it off after 4 hours?

After 4 and a half. I have to turn it off so some critical websites I'm going to (mostly business and shopping ones) won't lag and give me a whole bunch of troubles. To post that 1h screenshot, I had to resubmit the post twice because 10K just stresses the system way too hard. It rose my temps from 65/62/62/64 last time to 68/66/66/67. Now I got a way better fan... a bit on the loud side, but it's acceptable.

Edit: Here, restarted the old Prime but this time, under Vista. Seems like under Vista, stress testing is more stressful?



Maybe hours from now, I'll give you another shot. And by the way, it's "--" for me in Real Temp since even SpeedFan reports the same temp. It's actually 2 or 3 degrees cooler than that in reality (measured) but I'll take it as it is reported by those programs.

[Pete]

What FPO is the daddy then? I can get these cheap so will get one to try out i think

[RunawayPrisoner]

Okay, as promised, here's another screenshot:


I changed the theme, benched some Crysis, browsed some websites... etc... and she's still going rock solid with 1.304v.
And oh yeah, in case you ask... yes, I DID bench Crysis with Prime running.

Edit: FPO/Batch#: L803B196

[cbutters]

Are you guys testing GTL values? Im playing with some, but having a hard time both understanding how they work (especially with 45nm chips) and what values to set. other than that what are your NB volts at 3.6-4.0?

[RunawayPrisoner]

Another two hours, another screenshot (Vista theme color changed on purpose):



Seems really REALLY solid at 1.304v today. I'll start playing with GTL tomorrow to see if I can get to 4GHz... or if I can lower vCore even more. It's been set at "Auto" for a while now... and I thought that if something is not broken, don't fix it, right? But hey, experiments are good. :p

[HKPolice]

Another two hours, another screenshot (Vista theme color changed on purpose):



Seems really REALLY solid at 1.304v today. I'll start playing with GTL tomorrow to see if I can get to 4GHz... or if I can lower vCore even more. It's been set at "Auto" for a while now... and I thought that if something is not broken, don't fix it, right? But hey, experiments are good. :p

Seriously, you need to stop using 4 instances this old version of prime. It is NOT an accurate test for stability stop misleading readers. Use 2 instances of Orthos, 10K FFTs IN PLACE, Priority 5 or higher and run it WITHOUT any other programs in the background, including RealTemp and CPU-Z.

[karl_eller]

Seriously, you need to stop using 4 instances this old version of prime. It is NOT an accurate test for stability stop misleading readers. Use 2 instances of Orthos, 10K FFTs IN PLACE, Priority 5 or higher and run it WITHOUT any other programs in the background, including RealTemp and CPU-Z.

Or, you know, he could run a single instance of Prime95 v25.x, which will automatically multi-thread to all 4 cores when stress testing

Eller

[RunawayPrisoner]

Okay then. Right now, I'm trying to get 4.00GHz stable. Once I'm done with that, I'll start the 10K again. (And FYI, it did pass over 4 and a half hour)

Edit: Okay, started it. 4.00GHz seems like it'll need a lot of trials and errors for me, and I'm not up to that tonight. Let's start orthosing.

Oh, and how many hours have to pass for me to know that the CPU is completely stable?

[sofarfrome]

Seriously, you need to stop using 4 instances this old version of prime. It is NOT an accurate test for stability stop misleading readers. Use 2 instances of Orthos, 10K FFTs IN PLACE, Priority 5 or higher and run it WITHOUT any other programs in the background, including RealTemp and CPU-Z.

I don't believe he is misleading anyone. His version of stable just differs from yours. Frankly, I wouldn't run it for as long as he has.

[aussie-revhead]

The longer the better

What mobo is that mate ? Im going to go order one ... my P5K is priming at 460 but thats about it .

[aussie-revhead]

I don't believe he is misleading anyone. His version of stable just differs from yours. Frankly, I wouldn't run it for as long as he has.

Agreed , there are many different definitions of "stable" , I like to prime and game at the same time , if it survives that then its pretty good .

[RunawayPrisoner]

Thanks guys. And it's okay. I'd like to see just how much this chip has improved. Over a week, the voltage requirement for stable 3.8GHz has dropped from 1.35v to 1.296v now, and Orthos 10K, as mentioned above, has gone over 3 hours at 1.296v. I'll keep going for another 5 hours just for the heck of it.

I'm using an ASUS P5E-VM HDMI board without any modification at all. Or maybe I just added too many fans and coolers but eh...

P.S.: And I'll post a screenshot of priming over 5000 hours tomorrow. Quite frankly... didn't expect that to show after just 9 hours of priming, but I guess that's what I have to expect from an old version of Prime95.

[RunawayPrisoner]

BUMP. It went on for 8 hours before my dad went down and turned the computer off... because he thought I forgot to turn it off! So anyway, that means I couldn't take a screenshot at 8 hours or anything. I'm starting another Prime session with 10K but at 1.288v today to see if the voltage requirement really is dropping day after day. Seems rock solid so far, and will report back in a matter of hours.

[mu||et]

RunawayPrisoner, Thanks for all of your work dude.

[RunawayPrisoner]

No prob. Sincerely, I didn't do much... just about 54000 hours of priming.



^= What in His Name is that? It was only a nine-hour priming session? Oh well...

And here's priming with the new settings and of course, new Prime95, also at 10K like the previous 2 tests that I could never take any screenshot of:
idle 1.304v
load 1.288v



Heavy droops, I know. But I am living with droops and drops rather than avoiding them, as Intel's specsheets even state that droops and drops are somewhat necessary. If it's solid with droops and drops, it's good enough, right?

[jaredpace]

i did 10 minutes of priming on my e8400 on air at 500mhz FSB x 9. does that count too?

[RunawayPrisoner]

Mmm... yeah. Pretty much? It seems you were only stressing one core, though? And I'd run Prime for at least 4 hours or beyond before I conclude that a clock range is stable. But yeah, each has his own definition of stability, right? I dare say that sometimes, I'd even consider 30-minute priming sessions a success... and continue to drop the volt until it gets really really low.

[jaredpace]

Mmm... yeah. Pretty much? It seems you were only stressing one core, though? And I'd run Prime for at least 4 hours or beyond before I conclude that a clock range is stable. But yeah, each has his own definition of stability, right? I dare say that sometimes, I'd even consider 30-minute priming sessions a success... and continue to drop the volt until it gets really really low.

I hear people say prime doesnt matter, or prime REALLY matters and you should pass 8 hours.

I dont bother with it, I overclock to the point where games and pc usage doesnt freeze/crash.

for crysis it was about 4400mhz.

Btw, nice job on your new yorkfield!

[RedHOT]

All I have to say is WOW.

http://valid.x86-secret.com/show_oc.php?id=336593

For the past few days I couldn't get anything over 425fsb now I'm doing 450 with no issues... Its all about the VTT and GTL settings.

Obviously going to push it more tonight... vcore is only @ 1.3v

[TEDY]

jaredpace what cooling and which fpo batch...i don't believe you :P

[RunawayPrisoner]

All I have to say is WOW.

http://valid.x86-secret.com/show_oc.php?id=336593

For the past few days I couldn't get anything over 425fsb now I'm doing 450 with no issues... Its all about the VTT and GTL settings.

Obviously going to push it more tonight... vcore is only @ 1.3v

Yes, that's the spirit! And I want to see this on that X48 motherboard at least!

And... thus far, 3.80GHz stable:

vCore: 1.325v (1.304v idle real, and 1.288v load real)
CPU volt ref: 0.61x
PLL volt: 1.54v (or maybe 1.58v something)
FSB term volt: 1.28v (I think that's it? Or maybe it's 1.36v)
NB volt: 1.47v
NB volt ref: 0.61x

Try that for 475 FSB, guys?

[aussie-revhead]

Runnaway - can you tell me what the cpuv ref and nb v ref mean and do ? The voltage settings in bios are self explanatory but I have never touched these multi's . I will try your settings now (P5K).

Just reading this too

[RunawayPrisoner]

Mmm... let's just say they control the way some voltage signals are handled. Using your PLL as the base to multiply, the value after the multiplication should be quite low, but don't let it be too low or too high compared to 2/3 of your FSB Termination voltage. It's trial and error, though, because the real voltage may not be what you set in the BIOS. You can try upping your vCore to compensate for the error in real voltage, though. Also try to consider (vCore x vCore ref)/(vNB x vNB ref). It seems the closer you are to getting 1 and as long as (vCore x vCore ref) is less than (vNB x vNB ref), the system is more stable. It still locks up on me quite a lot at 500MHz FSB though.

Edit: And this may explain why upping the vCore and vNB alone will make the system more stable in some cases... just because as the numbers get larger, they get closer to 1. Disregarding vFSB and vPLL, which are handled automatically by the motherboard (which, I think, also tries to get them to match as close as possible) we have (really high vCore) / (really high vNB) = somewhat... closer to 1. This might also be why on some motherboards, when vCore is settable to 1.90v, vNB can also be set to 1.90v.

I'm still not so sure... but at least that's how I understand it after trying to digest a whole bunch of articles about VTT and stuffs...