about Prime95

[BTRY B 529th FA BN]

Lately, i've been testing the stability of my FX60 system and I keep getting errors at what ever clock i set the cpu at. The blend test says it tests some of everything 'lots of ram tested' I keep getting errors so this leads me to believe there are two possibilities, maybe a 3rd like the MB. Right now I am running Memtest86 and Prime95 25.6 build 6. How reliable are these test tools and with Prime 95, how can i isolate the testings to just the cpu to leave out the variable of the memory?

thanks,

troubled

 

[DarkRogue]

Is it failing due to small FFTs or large FFTs?
Also, does it fail at stock settings too?
The common advice for fixing Prime fails is to increase the vcore of the CPU though.

 

[BTRY B 529th FA BN]

Just got done running memtest86 and it passed (just need to check the other pair or sticks 2x512's)

I've underclocked it to 2.4 and running 'In-place large FFT's' then will try the small FFT's: will get back with results.

 

[bryanW1995]

if you're not failing memtest but you are getting errors in prime 95 then it's almost certainly the cpu. as darkrogue mentioned, try raising vcore in small increments until you regain stability.

 

[DarkRogue]

Yeah, the likely culprit is probably Vcore, as I learned the hard way.
My current rig was also my first "taste" of overclocking and I tried everything possible to avoid raising the vcore, but it just would not play ball.

My RAM would pass 5 passes of memtest, but immediately fail Prime/Orthos as soon as I hit large FFTs. I thought it was a RAM problem, or maybe the MCH/FSB not getting enouch juice, etc. Ran those tests over and over with different settings but in the end, I caved and upped vcore a few notches and that worked like a charm.

Rather than underclock though, I'd suggest you return EVERYTHING to completely stock/factory settings and run it again, just to rule out the possibility of bad hardware. If it's not stable at stock settings, it's not going to be stable at any settings.

 

[BTRY B 529th FA BN]

Caved is right. After this bit of OCing experience, i don't think upping the juice in any of the area's is a good idea but at any rate, the 1st pair of 512's tested ok, haven't tested the second pair. It was underclocked it to 2.4 and in Prime95 it (blend) ran stable for the 1st 1024k section; i usually find out within the first minute but it passed it and started on the next number. I stopped it and tried the maximum heat, In-Place large FFT's and that seemed good. 2.4 with voltage set to 1.300 but in OS readings say between 1.31 & 1.34. The default setting of 1.35 (in OS) reads 1.39 also no over-voltage on the northbridge. It also seems more stable with just one pair of 512's using 1 memory channel. I'm gonna try the other 512 modules about 4 days when I get my new heatsink, otherwise i'll have to take out the MB just to fit on the new sink. It's out right now ...long story.

 

[BTRY B 529th FA BN]

Well, for some reason things are running stable. Even with the voltage set to 1.300 in bios and its clocked at 2.8 but I've been only running the small FFT's (maximum FPU stress...) Does that test only stress the CPU and not the memory? I did a memtest86 on the 512 pair i have in there now and it passed.

The story is I decided to buy a new platform and so basically I took out everything except my drives and PSU from my gaming case and put it into the primary machine's case. Did a fresh install (running 1 primary drive instead of a RAID 0 setup) using the fairly new Liberty 500... things are stable for some reason. I'm not unhappy, i'm just seriously wondering what was the problem. The other machines PSU is a fairly new; Infinity 720w also and ...never seemed to have any problems, hmm.

 

[DarkRogue]

Doing CPU-only (small FFTs) is not an accurate way to determine if your system is stable.
When I was having trouble overclocking my rig, I really thought it was my RAM because I could run small FFT's forever and not see an error, but as soon as I hit large FFTs (either from a custom large FFT setting or from blend) it would error out.

After raising the Vcore though, all the problems disappeared, so even though it initially appeared to be a RAM problem, it really wasn't.

 

[BTRY B 529th FA BN]

Thanks for the info. Gonna try that out