Overclock on x2 3800+

[Doclife]

Originally posted by: jpbelauskas

Originally posted by: Doclife
I overclocked my socket 939 X2 3800+ to 2.9GHz with these settings:

CPU multiplier = 10X
Mobo bus speed = 290MHz
Vcore = 1.575V
HTT = 3X
Memory Divider = 133MHz (10:15). Thus, 2GB (2 X 1GB) DDR400 is running at DDR387

It has been running stably for about a year now.

I'll try these out. did play around with your memory timings? or are they running 2-2-2-5?

My memory is running at 2.5-3-3-8. Memory timings doesn't increase performance all that much so I relaxed it a bit to increase stability.

 

[geokilla]

If you can't pass Prime/Orthos 24hours, then try raising your vcore up a notch until Orthos/Prime passes. If it still doesn't pass after raising the vcore by say .05v, then I would assume that you've hit your max OC.

 

[rasczak]

seems i've hit the wall at 2.8ghz. i loosened my memory timings to 2.5-3-3-8 and upped my vcore to 1.5v, i haven't run prime95 yet, but i was checking my temps with sppedfan and noticed temp2 is running 54c.

can anyone tell me what temp2 indicates?

 

[DerwenArtos12]

Originally posted by: MarcVenice
How can I measure my chipsets temperature ? I'm really really curious, because the stock fan that was on it was well, a piece of crap, no heatsink, and a really tiny little fan inside a metal housing, maybe aluminium, with a glob of thermale paste between in. I replaced it with a 'heatsink' cooler, and when I touch it, it feels pretty warm, not to hot to touch, but warm nonetheless. What has me baffled is, if this new heatsink cooler gets this warm, how the hell could the old crappy cooler keep things cool enough? I'm using speedfan atm btw, what temperature is for the chipset? I also only get a Core temp, not both cores, which kinda annoys me. But Coretemp isn't vista 64x certified, other options perhaps ?

Depends on your chipset but, should be available through a program like smartfan or the monitoring program that your motherboard manufacturer slapped together. If you post your motherboard I'll be happy to look into your best options.

 

[DerwenArtos12]

Originally posted by: jpbelauskas
seems i've hit the wall at 2.8ghz. i loosened my memory timings to 2.5-3-3-8 and upped my vcore to 1.5v, i haven't run prime95 yet, but i was checking my temps with sppedfan and noticed temp2 is running 54c.

can anyone tell me what temp2 indicates?

in speedfan that is typically chipset temperatures. I'd reccomend getting on DFI's website and getting their temperature monitoring software, it's a direct port from the hardware moitoring chip manufacturers software and is surprisingly reliable. Besides that check out DIY-street, they have all the files necessary to make MBM5 work for our boards. I've tried speedfan twice now on my NF4UD and I really don't like it.

 

[DerwenArtos12]

Originally posted by: geokilla

Originally posted by: DerwenArtos12

Originally posted by: geokilla
Whoa!! Temps above 69C? You're going to burn out your X2 quickly at this rate. What cooler are you currently using right now? Please don't tell me it's the stock cooler.

What stepping is your X2? Is it F2, F3, or GO? I read a couple of months ago that the new X2 3800+ can get up to 3Ghz with ease. Too bad my F2 tops out at only 2.4Ghz.

Like others say, keep your HTT at below 1000Mhz. The HTT doesn't even affect your computer's performance, and even if it does, it's a minor performance hit. Trying giving your X2 a bit more vcore as well. Sometimes, just that little bit will make all the difference.

Finally, what's your system specs?

Please read an entire thread, or for that matter an entire post before posting. He stated quite clearly that it was his chipset running at 69C which is hot but, not out of line for an NF4 with stock cooling and a signifigant overclock.

Sorry bout that then =.=. But 69C is still high for a chipset is it not? I have no idea what the max is for chipset temps, so yea...

It is high but, chipsets are much simpler and made on a larger manufacturing process than modern cpu's so they can take higher voltages and higher temperatures, similarly to how graphics cards take higher temps very well. I believe, if memory serves me correctly that nf4 is on 130nm and nf5/6/7 are all on 90nm, 965 and 975 were 130nm and the new 3 series and 4 series are on 90nm.

 

[MarcVenice]

A8N5X is the motherboard I've got. The temp monitor application that came with it, can't be installed, since the whole CD that came with the mobo won't even run in Vista 64x. I'm going to try smartfan, but if you have any suggestions they would be greatly appreciated

 

[mindless1]

Originally posted by: jpbelauskas
hrrm,

interesting i remember running prime95 yesterday and forgot to mention this,

i received an error

hardware failure error,

"Rounding was 0.5, was expecting less than 0.4. Hardware failure detected consult stress.txt file."

Ignore the "hardware failure" part, just know your system isn't stable yet.

Isolate variables. Run memtest86+ for several hours, ideally at an additional 5% higher memory clock speed than what you intend to use regularly. If that passes for hours of running, put the memory clock back at the -5% intended o'c speed and proceed to run Prime95 again. If you get errors right away, your voltage is too low or speed too high. If it takes awhile for errors, the temp is getting too high but ultimately if your 'sink is on good and you haven't set the board to throttle the fan too much then it's the same need, to lower speed and or voltage to get the temp down.

Frankly, a minor performance difference isn't worth the hassle, in regular use you will never notice a single digit percentage performance difference.

 

[rasczak]

Originally posted by: mindless1

Originally posted by: jpbelauskas
hrrm,

interesting i remember running prime95 yesterday and forgot to mention this,

i received an error

hardware failure error,

"Rounding was 0.5, was expecting less than 0.4. Hardware failure detected consult stress.txt file."

Ignore the "hardware failure" part, just know your system isn't stable yet.

Isolate variables. Run memtest86+ for several hours, ideally at an additional 5% higher memory clock speed than what you intend to use regularly. If that passes for hours of running, put the memory clock back at the -5% intended o'c speed and proceed to run Prime95 again. If you get errors right away, your voltage is too low or speed too high. If it takes awhile for errors, the temp is getting too high but ultimately if your 'sink is on good and you haven't set the board to throttle the fan too much then it's the same need, to lower speed and or voltage to get the temp down.

Frankly, a minor performance difference isn't worth the hassle, in regular use you will never notice a single digit percentage performance difference.

Yea, I kinda felt the same way about it after a while. I ended up bringing my OC down to 2.7ghz. seems to run reliably fast for what i want to do.

Thanks for all the help guys. I'll continue to troubleshoot the prime95 errors. if I find anything I'll post.

Cheers!
Joe

 

[nerp]

Yeah, prime stable @ 2.7 is way more desirable than a prime-failing 2.8 to 3ghz chip, if you ask me. Like others have said, you won't see or feel any difference and stability should never be a tradeoff for speed.

 

[DerwenArtos12]

Originally posted by: nerp
Yeah, prime stable @ 2.7 is way more desirable than a prime-failing 2.8 to 3ghz chip, if you ask me. Like others have said, you won't see or feel any difference and stability should never be a tradeoff for speed.

QFT. I can do 3ghz prime stable for a week but, it more than doubles my heat output from 2.75 and stresses the chip a lot more(not that I can actually manage to kill this damn thing) thats why I run at 250x11 most of the time. I don't play games very much anymore(except monopoly) so there is really no point in trying to squeeze that last little bit out of the system.