Prime95 one core failure
Prime95 one core failure
My configuration;
ASUS Crosshair V Formula
AMD Phenom II X6 1090t
G.Skill Ripjaw series DDR3 1600 1.5v
ASUS Radeon HD 6950 DCII/2gb
Corsair 600w Gamer series
Corsair H60 (with added fan for a push-pull radiator system)
The current fully stable configuration;
[email protected] (with a .05+/- load variance)
2000mhz HT
[email protected]
Core reaches 54c after 24h
I can achieve 3.9 with these settings, and pass a 10m prime95 6 Test run. Still not fully stable until the 24h test completes.
Before I proceed with the 24h test, I aim to hit 4.0 for testing. If I can't reach that during the 2-hour evaluation, I'll revert to 3.9 for the extended period.
The challenge lies here;
I apply the 4.0 settings. All monitoring data confirms it's functioning as intended within the specified limits. However, when I launch Prime95 for a blend test, around 30-45 seconds in, core04 (the fifth core) fails Test1. The remaining four cores succeed in Test1 and move on to Test2. After roughly two minutes, Test2 starts on random cores, occasionally failing core03 (fourth core), while others pass. All other cores complete up to Test10 before I halt the test.
Is this an issue? I've seen reports indicating that boosting the vcore resolves the problem. Yet, at 4.98v in the BIOS, the .05 variance can push it close to 1.55v, which is outside safe limits. This raises doubts. Some users have managed to reach stable 4.3 with this chip. I acknowledge there are variations, and I'm not among the fortunate ones who consistently hit the minimum.
Just a reminder; the RAM is set to default values, manually adjusted to guarantee stability.
Is there a method to evaluate a single core? It seems unclear how to do this in Prime95 configurations. [/strike]
I managed to launch Prime95 on designated cores by changing process affinity in the task manager and setting the cores to run at 1 during the blend test. The two cores involved passed Test10, which gives me confidence the hardware is correct. This suggests the issue might be related to voltage settings.
Someone has information about performance trends. At idle the temperatures are low, around 27°C in the core. Under load, even at [email protected] it stays below 54°C. However, one core consistently fails when others are active. Another test showed success with first two cores (core0 and core1), but failure occurred when switching to core3 and core4 (4th and 5th). The reason is unclear.
It looks like some of the CPU cores aren't performing equally, especially on my 3930K model (core 10 tends to fail first). Increasing voltage might help, but it's clear you're already running at a high Vcore. I think this issue also occurs with other stress tests such as OCCT?
Could it be that my RAM requires higher voltage? The 1.55v might be too low, and both my CPU/CPU voltage could also be insufficient. My motherboard is set to auto, which keeps it at 2000mhz since my core clock remains the same. Would adjusting to a lower voltage ratio with a higher core clock yield better performance? Should I reduce my RAM and CPU speeds instead?
I understand on some models I can change the multiplier without altering clocks, but others need clock adjustments to achieve good results. It seems a mix might work best. It’s possible my power supply isn’t adequate either. The 600w PSU worked fine with my old NVIDIA 260, but the new 6950 consumes much more power.
If your RAM is rated at 1600Mhz @ 1.5V then 1.55V is sufficient, just ensure your timings are configured properly. Your HT frequency is acceptable. Are you experiencing any BSODs? If yes, what is the corresponding bluescreen code? OCCT serves as another stress testing tool similar to prime95.
I don't see a BSoD occurring. I've only used Prime95 (and now OCCT) and normal browser/music use while letting the system run at 4.01 for a short time. Prime95 occasionally fails one core at a time. I've tried single affinity and dual affinity tests. When testing any two cores except those that fail, they pass Test15 without issue. However, when I include the two problematic cores, one of them fails.
OCCT ran for 10 minutes and passed without problems. The voltage readings in the program were inaccurate, which is normal since each program behaves differently. CPU-Z's rog flavor for my motherboard works correctly, matching BIOS readings.
I'm still puzzled about what's happening. It seems like a rounding error—like getting .5 when you expect .4 or less.
I've been adjusting my memory settings and often encounter that error when it's unstable. Perhaps trying smaller FFTs in prime numbers could help. Also, verify your memory timings in the BIOS are accurate.
Finding the most accurate method to read every configuration the RAM should hold is essential. Your BIOS and CPU-Z data are reliable sources, though a few settings might be missing. The goal is to ensure the RAM matches the programmed values.