Skylake CPU: potentional damage caused by Prime95

[AnonymousSovereign]

Hello everyone, I bought a new PC (not in one piece) just a few weeks ago and therefore I ran some benchmarks as well as CPU stress test in order to determine whether everything is working properly and particularly whether I installed my CPU heatsink correctly. I used the newest version of Prime95 and I ran Small FFTS. After 3 hours of this test I got the maximum temperature of 62°C, which is AFAIK not bad at all eventhough I was surprised because idle temperatures (28°C) almost immediately changed to 60°C after I started the test... However, a few days later I read some articles saying that running the newest versions of Prime95 (the ones supporting AVX) can damage your Hasswell CPU (And I suppose Skylake is the same) because of high voltage or something like that (basically because of AVX and 100% utilization of CPU, which change the vcore). I also found some user reports asking why the temperature raises so quickly on ther CPU during the Small FFTS test - the same case as mine. + I HWMonitor also showed max vcore 1,632 during the test which is insane (On the other hand it shows similar results - 1,580 V even during the same test using older version of Prime95 - 26.6). Isn't skylake CPU's vcore supposed to run on maximally 1,2 V??? I'm quite concerned... :-(

Could I somehow damage my CPU (i5 6600 - non K) with my 3-hour stress test?? I'm going to use this CPU 24/7 as my main processor for many years (at least five), so that would be a problem...

Thanks in advance

 

[GhislainG]

Why are you concerned if it didn't even reach 100°C? Even if it did for a few minutes it still would run fine. What do you think happens when a cooler fails?

 

[synphul]

I don't think you damaged your cpu though I would check in the bios to see what the vcore is set to. 1.4v is around the upper safe limit for skylake, 1.35v or lower would be more preferred for 24/7 use. Voltage can raise temps but both voltage and temps are something to watch. You can't just keep applying any voltage so long as temps are ok. Temps could be fine and voltage still high enough to damage the cpu.

It's not unusual for temps to jump straight up, it's an instant load and cpu's produce almost instant heat. It's not like waiting for an oil radiator style heater to slowly warm up, apply voltage and load to it and within a second or two the temps are up. Think of a flame on a candle, you don't have to slowly wait for your finger to burn if you touch it.

If you set the core voltage manually in the bios then it should stay where you set it and not overvolt.

 

[AnonymousSovereign]

I don't think you damaged your cpu though I would check in the bios to see what the vcore is set to. 1.4v is around the upper safe limit for skylake, 1.35v or lower would be more preferred for 24/7 use. Voltage can raise temps but both voltage and temps are something to watch. You can't just keep applying any voltage so long as temps are ok. Temps could be fine and voltage still high enough to damage the cpu.

It's not unusual for temps to jump straight up, it's an instant load and cpu's produce almost instant heat. It's not like waiting for an oil radiator style heater to slowly warm up, apply voltage and load to it and within a second or two the temps are up. Think of a flame on a candle, you don't have to slowly wait for your finger to burn if you touch it.

If you set the core voltage manually in the bios then it should stay where you set it and not overvolt.

Thanks for your comment, however, the strange thing is that I've never had to change vcore in the past - I just let it on auto and everything was fine. But with this CPU (i5 6600) and Gigabyte gaming K3 mainboard I have this problem, eventhough vcore is set on auto... And if I set it manually (lets say 1,2 V) I would get static voltage (now my vcore move from 0,5 V to 1,6 V) and that would cause higher temps and energy consumption, wouldn't it? Anyway, I should probably try other programs for finding out my vcore voltage and if it doesn't help I will probably update my BIOS (not sure there's a newer version for my MB though) since I found a person with similar problem (using different mainboard and CPU) and he solved it by updating BIOS... ;-)

 

[synphul]

I haven't used that motherboard though so long as you leave eist (speed step) alone and don't turn it off, the vcore should be allowed to fluctuate as needed. I'm using a gigabyte board (z97) and it works as it should. Maybe being so new it needs a bios update on that board. My cpu isn't under load and it drops around .78-.8v and rises to 1.28v under load.

Leaving vcore on auto allows the motherboard to give the cpu as much power as it wants and if the motherboard chooses wrong your vcore could be too high. Setting vcore shouldn't set static voltage it should only set the upper limit. Unless of course you also turn off eist and don't allow the cpu to idle down at all. If it were stuck at a fixed 1.4v it would use more power and generate more heat yes. 1.6v is too high, if it were my system I'd shut it down immediately.

Skylake can handle slightly higher vcore than haswell refresh/devil's canyon, but only by around .5v. Safe 24/7 for 4th gen was considered to be around 1.3v and for skylake it's 1.35v - many agree nothing over 1.4v unless you don't mind a gamble to hit a world record or something. A bsod or shut down would occur if you went overtemp or didn't have enough voltage but not sure it would do so with too much voltage. Just because the system isn't shutting down doesn't mean it's not harmful. It may not kill the chip today or tomorrow but I would think 1.6v would eventually shave noticeable lifespan off that cpu. If it were my cpu I wouldn't try it to find out.

I've had pretty good luck with hwinfo64 myself, though I can't speak to how accurate it is for skylake. Only because skylake is still somewhat new(er) and sometimes it can take software a little bit to catch up with newer tech as problems are found. Same goes for the bios.

I can't imagine why it would be applying so much vcore to a cpu that isn't overclocked. Auto should be fine for a locked 6500 but nothing is perfect which is why bios are able to be updated, to fix any potential quirks or bugs. It can also be on a board by board basis, one model to another. A gaming 3 board may need the update while a z170 gaming 7 may not, that sort of thing.

Update: This is similar to another user's problem and I believe this helped them.
http://www.tomshardware.com/answers/id-2826394/6500-turbo-frequency.html

 

[AnonymousSovereign]

So, I installed CPU-Z and AIDA 64 and they show (fortunately) much more safe and common voltage of my processor - around 1,2 V at most.... That's an significant difference compared to HWMonitor vcore result. Surprisingly, result of CPU-Z and AIDA 64 looks almost the same as CPU VID in HWMonitor - but as far as I know CPU VID is not the voltage measured by sensors but just an expected voltage of CPU, which can be pretty different compared to real voltage. Therefore I hope these results from AIDA64 and CPU-Z aren't just CPU VID but real vcore - thus I can rely on them and my CPU won't damage, right???

Screenshot: http://

Thanks