Strange problem with vcore and stability

aml

Distinguished
Jul 2, 2009
2
0
18,510
Hello, and sorry for my bad english. I think I need some help trying to get my OC stable. Been struggling with my setup for a week now without much advance.

My setup is following:

- i7-920 D0
- Asus P6T-deluxe V2
- 3x2gb 1600mhz Team Xtreem tri-channel kit

And the settings I've come into after some tweaking:

- BCLK 180 (x20)
- Mem. freq. 1400-something mhz
- Mem timings set at 9-9-9-24 (default 8-8-8-24)

- 1,30v QPI/Dram voltage
- 1,86v PLL voltage
- 1,12v IOH voltage
- 1,64v Dram bus voltage

- Speedstep, turbo mode and both spread spectrums disabled
- Loadline calibration enabled
- Basically all other things at "auto" or some other default setting

And with those settings, I can get the CPU stable (running prime95 small ffts - test without problems) with about 1,22vcore voltage. That sounds pretty average and so on, no problems at all. I have been dropping the vcore with 0,01v steps and that looks like the lowest stable value.

But, the problem is here, I can't get my memory stable at those settings or even by raising Dram bus voltage and/or QPI voltage or basically any other voltage besides the vcore. I only get my RAM stable by raising the vcore to almost 1,30v. Then I can run prime blend test without errors. But that vcore value seems to be too high for the chip itself, being stable at only 1,22vcore when doing CPU stress testing instead of memory stress tests.

And to make things even more complicated, memtest86 run for several hours without any errors with those settings and 1,22vcore, which is extremely unstable when stressing memory. Blend test crashes immediately and I get random BSoDs even when not stress testing anything. The bsod message says the following:

IRQL_NOT_LESS_OR_EQUAL
STOP: 0x0000000A

Now can you help me troubleshooting, does this look like problem caused by bad RAM or could it has something to do with memory controller (part of the chip itself)? When the memory seems to have hunger for much higher vcore to run stable than the chip itself.

I can't test with another memory kit before next week, propably that makes things more clear but now I need some help. :??:
 

aml

Distinguished
Jul 2, 2009
2
0
18,510
Yeah, it propably should take that without a problem in 24/7 use but that big difference in vcore between CPU stable value (~1,20v) and RAM stable value (~1,30v) still bugs me. I don't want to keep up too high vcore voltage in case the chip itself doesn't need that much, only the faulty memory.

I believe it's just bad memory but can't be sure before testing with some other memory kit which I should be able to do in a couple of days.