[SOLVED] Help -- i7 9700k showing weird voltage behaviour ?

Mahbub1

Distinguished
Mar 23, 2015
435
0
18,780
Ok so i have i7 9700k (not overclocked) being cooled by noctua nh u12s in push pull.. when i set voltage in bios to auto and run prime 95 tests. Without avx. temps go to like 90 and i checked the voltage being used by cpu in hwinfo.. it shows a max voltage of 1.224 ... But when i go into bios and put put the cpu volts from auto to 1.230v or 1.240v.. and then run p95 without avx.. it reaches 80 degree but after running for sometime it crashes and pc restarts... My question is why? How is it that in auto voltage mode in p95 without avx processor is using a max of 1.224v and reaching 90.. but when i clearly set a manual voltage more than 1.224, which is 1.230v to 1.240v and run p95.. it runs cooler and crashes.. whats the mystery/logic here? Can someone help me understand.
 
Solution
First, Prime95 is quite artificial. I will not run for extended periods. Get the machine booted, run p95, launch Task Manager and let the cpu bar go all the way once, then I'll shut P95 off. that's a decent 2min burn.

185°F or 85°C is HOT, I killed a CPU running it hot. It was overclocked and overvoltage and ran 6 years, but when the radiator starved for water it hurt the CPU and had to be replaced. So, back to the hard burn on p95, 2min is plenty to verify it's stable.

As to why higher voltages are giving cooler temps and crashes. I suspect the thermostats are not reading in realtime and it's getting hotter w/ more voltage, but it's throttling and dies before it can tell you.

At the end of the day, I think you need to look at...

MonsterMaxx

Distinguished
Jan 23, 2015
110
15
18,615
First, Prime95 is quite artificial. I will not run for extended periods. Get the machine booted, run p95, launch Task Manager and let the cpu bar go all the way once, then I'll shut P95 off. that's a decent 2min burn.

185°F or 85°C is HOT, I killed a CPU running it hot. It was overclocked and overvoltage and ran 6 years, but when the radiator starved for water it hurt the CPU and had to be replaced. So, back to the hard burn on p95, 2min is plenty to verify it's stable.

As to why higher voltages are giving cooler temps and crashes. I suspect the thermostats are not reading in realtime and it's getting hotter w/ more voltage, but it's throttling and dies before it can tell you.

At the end of the day, I think you need to look at your cooler situation. These modern CPUs are generating a ton of heat. I have a 9900k and if I p95/torture it, it'll hit 170°F with the H100i watercooler. I feel that's safe. They cost $150, but if your case will support it, this would probably solve your problem.
 
Solution
I've never heard of a non-OC'd CPU crashing in P95, but, I'd speculate the manual voltage setting is just not enough for some of the peaks...; voltage needs can peak and subside many times in 1/4 second, perhaps all of which might not be shown in assorted monitoring tools.

In the P95 test, are all of your cores at 4.6 GHz (semi-normal) or 4.9 GHz (indicating MCE mode is clearly enabled)?