unstable 555BE, brand new

[konakona]

Full system specs:
phII 555 BE (no unlockable cores, unfortunately )
sunbeam core contact freezer + 1850rpm gentle typhoon + IC Diamond 7 Carat
biostar TA790 A3+
ocz ddr3 gold 2gb x2
corsair 400cx
sapphire 4850 + s1 rev2 + 120mm yate loon
antec 300
tricools that come with the case + a yate loon for intake, another yate loon for vga exhaust
wd caviar black 1tb
hitachi 7k1000.b 1tb
toshiba sd-h802a

The story:
The system has had 965BE not too long ago, until it just died one day and had to be replaced. It was doing 3.8/2.8 all day with ok temps, occt stable for hours. The replacement was this 555BE, which was never stable at any speed (even at stock) in games without overclocking, which prompted me to subject it to some OCCT fun.

Here are some of the attempts I have made thusfar (I just highlited changes I have made on each successive run)

core: 3200
cpu-nb: 2600
vcore: 1.400 + auto
vcpu-nb: 1.2 + 0.20
crashed in the first minute

vcpu-nb: 1.2 + 0.22
crashed in 2-3 minutes

vcpu-nb: 1.2 + 0.24
crashed in 4-5 min

vcpu-nb: 1.2 + 0.26
crashed immediately

cpu-nb: 2000
crashed immediately

vdimm: 1.72
timing: 7-7-20-27 (1066)
crashed immediately

vcore: 1.4125 + auto
vdimm: 1.70
timing: auto (1066)
lasted a bit longer

vcore: 1.4250 + auto
about the same

vcpu-nb: 1.2 + auto
crashed immediately

CnQ / C1E off
HT Link tristate / nb power saving off
crashed immediately

moved ram modules to diff slots
crashed immediately

I am only using OCCT atm to monitor my temps, and it's reporting a relatively cool <47C at load (I keep my room a bit hot). Perhaps I should try realtemp or coretemp for a change? A wild guess that I have on my mind is bulging/leaking caps on the mobo, prolly should take a closer look at it whenever I get a chance. Any thoughts?

 

[Rifter]

CPU's usually dont just die, i would suspect the motherboard.

 

[konakona]

well I did too and was ready to get a new mobo... but to my surprise the mobo was fine (as in booted up) with the new cpu, but the cpu was not with a new mobo. tried the cpu with another mobo, no go. just for the hell of it, I stuck the cpu in the freezer for a half hour (sealed in a pouch, of course), took it out and tried again, still no go. yep, its the cpu.

oh btw, by crashing I mean "errors detected on core X", not hard lockups or BSODs.

 

[Rifter]

RMA the CPU then. Also did you try the CPU in another mobo first? or after? the mobo could be defective and actually killing the CPU's, i have seen that happen before although it is rare.

 

[konakona]

I didn't have enough time to do full stability testing on a new mobo as the new system to be built around it lacked OS on it, but I did test to see it POST.

Could be ram for all I know, never was a big fan of OCZ stuff anyway

 

[Rifter]

yeah OCZ is not the best, do you have another system you can memtest the ram with?

 

[richierich1212]

What vdimm?

 

[konakona]

I have already done the memtest using the bios built-in feature. No probs there.

These are rated 1600 8-8-8-24 @1.65v. I have tried anything from stock all the way up to 1.76v for vdimm, still crashing in OCCT regardless. At first I was just letting the bios configure it automatically since it seemed to have work fine with the previous cpu (1066 7-7-16-?). Now I am just trying the stock settings to see if it fares any better.

 

[konakona]

seems like the combination of 1.26v on the nb (not the cpu-nb) and 1.72 on the vdimm was a bit better, was able to hold up for half hour or so. lower the nb voltage to 1.2v, immeadiate fail. needs a bit further testing on the vdimm I suppose, at least there is some hope now!

sadly I don't recall what I had for the nb back when I had it working with the 965BE. why should it need any more than stock for an unoverclocked system? faulty mobo or ram? subtle incompatibility between ram and the rest?

EDIT: nvm, tested again at the same setting again and it was flaky this time, crashing just few minutes into the test. oh well
nb might be getting insufficient cooling, time to whip out speedfan! <- been there done that, nb stayed constant at 49C and the core was at 42C just before it failed (about 6min or so)

 

[Rifter]

seems like the combination of 1.26v on the nb (not the cpu-nb) and 1.72 on the vdimm was a bit better, was able to hold up for half hour or so. lower the nb voltage to 1.2v, immeadiate fail. needs a bit further testing on the vdimm I suppose, at least there is some hope now!

sadly I don't recall what I had for the nb back when I had it working with the 965BE. why should it need any more than stock for an unoverclocked system? faulty mobo or ram? subtle incompatibility between ram and the rest?

EDIT: nvm, tested again at the same setting again and it was flaky this time, crashing in just few minutes into the test. oh well

This is making me think i was right and the mobo is indeed the problem. Do you have the latest bios?

 

[konakona]

This is making me think i was right and the mobo is indeed the problem. Do you have the latest bios?

yeah it should be, flashed it like 2 months ago. doubtful that they made any meaningful updates to the now aging 790gx since then.

actually, the dead cpu is a different issue altogether. you might be right, the mobo might be the problem at hand. but that 965BE is dead beyond repair, does not POST in a good mobo that is married to another cpu now.

 

[Vesku]

I'd follow Rifterut's advice and try to RMA the dead Phenom II and also test the 555 BE in a different motherboard.

 

[konakona]

The 965BE was bought from fry's at the beginning of the year, I don't think their warranty on cpus lasts all that long (like 30 day return only or something). Yeah, testing the new one in another machine might be an option.