PDA

View Full Version : Two CPUs running at different voltages



george12
04-07-2009, 04:03 PM
I have a computer here which crashes under load. It's a ridiculous machine with 2x QX9775s, so 8 cores total.

In the BIOS voltage monitor it shows Proc 0 vcore to be about 1.24 V, while Proc 1's is only 1.12.

Is this even remotely normal?

I suspect that's not the only problem though as the CPUs appear to run at about 60-62 degrees at idle in the BIOS which is very close to the maximum allowable temperature. However it may well maintain that under load which would probably be OK...

Any thoughts?

qazwsxokmijn
04-07-2009, 07:00 PM
Holy son of a.....why are you so lucky as to get your hands on these babies??:stare::stare:

The two QX9775s may be two different steppings. The older one may need more voltage and the newer less. Just like my C1 Q9450 can overclock to 3.4GHz at 1.28v whereas the C0 ones I've seen needs 1.3v or more. I've even seen other C1 Q9450s needing more voltage than another C1 (such as mine).

As for the temp - may be another case of the 45nm faulty DTS. My Q9450 as with so many other 45nm CPUs are plagued with faulty DTS; or DTS that cannot be properly read by programs available today. My Q9450's cores #0 and #1 cannot show temps lower than 38c; core #3 is the only properly functional one; and core #4 is permanently stuck at 45c - even if ambient case temp is 5c!

EDIT: oh, it crashes? Have you checked the heatsinks are properly mounted and if they are applied with a proper amount of thermal paste? Try running Prime95 for a minute and monitor the temp using realtemp. Are the heatsinks hot to the touch?

george12
04-07-2009, 08:07 PM
Accidentally posted this in chat.

I have a computer here which crashes (freezes, never a BSOD) under load. It's a ridiculous machine with 2x QX9775s, so 8 cores total. He says it happens mostly during use of CS4, but I can replicate it using Prime95 (on the mode that uses lots of memory - haven't tried the max-heat flavour yet). I hope for his sake it's the RAM, but it sounds unlikely.

In the BIOS voltage monitor it shows Proc 0 vcore to be 1.218V, while Proc 1's is only 1.136V. That's quite a difference between identical CPUs.

Is this even remotely normal? Could 1.136V be too low for stability? The voltages are all on auto in the BIOS, and it says that VID is 1.25V.

I suspect that's not the only problem though as the CPUs appear to run at about 60-62 degrees at idle in the BIOS which is very close to the maximum allowable temperature. However it may well maintain that under load which would probably be OK...

However under load, CoreTemps shows a temperature of around 50 degrees max on all cores. I'm about to do a memory test tonight, but it seems like heat. The memory is fully buffered, 4x 1GB. An initial run of memtest showed nothing, so I will run it overnight.

Once its warm it will freeze in Prime95 in under 30 minutes despite never claiming temps above 50 degrees. Rebooting immediately after the freeze and checking in the BIOS shows the same 60 degree temps as at idle.

The memory controller temp is also around 60 degrees. Power supply rail voltages seem OK.

Any thoughts? Will update with results of memory test.

george12
04-07-2009, 08:09 PM
Holy son of a.....why are you so lucky as to get your hands on these babies??:stare::stare:

The two QX9775s may be two different steppings. The older one may need more voltage and the newer less. Just like my C1 Q9450 can overclock to 3.4GHz at 1.28v whereas the C0 ones I've seen needs 1.3v or more. I've even seen other C1 Q9450s needing more voltage than another C1 (such as mine).

As for the temp - may be another case of the 45nm faulty DTS. My Q9450 as with so many other 45nm CPUs are plagued with faulty DTS; or DTS that cannot be properly read by programs available today. My Q9450's cores #0 and #1 cannot show temps lower than 38c; core #3 is the only properly functional one; and core #4 is permanently stuck at 45c - even if ambient case temp is 5c!

EDIT: oh, it crashes? Have you checked the heatsinks are properly mounted and if they are applied with a proper amount of thermal paste? Try running Prime95 for a minute and monitor the temp using realtemp. Are the heatsinks hot to the touch?

I may be unlucky as it could be tricky to diagnose when every part costs $600+. Haven't removed the HSF yet, don't want to until I've eliminated some other things. I have reposted this in PF1 - chat was a mistake!!

Jen
04-07-2009, 11:35 PM
Accidentally posted this in chat.I've moved and merged your threads into this one.

george12
05-07-2009, 04:43 PM
Hmm I was happy when I saw all those posts until I saw it's just been merged :p

Ran memtest for about 16 hours, no errors, did run to 103% though... presumably that's just a programming error in memtest86 though?

I think it's got to be either heat or insufficient vcore on one CPU.

qazwsxokmijn
05-07-2009, 06:13 PM
From what I can remember memtest86 will keep testing the RAM until you stop it. So once it hits 100% it will start all over again.

george12
05-07-2009, 07:06 PM
From what I can remember memtest86 will keep testing the RAM until you stop it. So once it hits 100% it will start all over again.

Yes, but I mean when it hits 100% it keeps going until 103% before moving on to the next pass.