Can i damage 4790k using Prime 95 ?

L

lpiotrowski

n00b
Joined
Oct 17, 2014
Messages
56
Hi. My cpu is 4790K stock + stock cooling. . Also yesterday i tested cpu with Prime 95 Blend 10 hours, without crash. But today i read that testing with Prime 95 Haswell cpus ,can overvolt cpus.
Can i damage my cpu with Prime 95 Blend tests running many hours? I am little scaried now after reading that running Haswell cpus in Prime 95 are dangerous. ( ,can overvolt cpu )

I am just only ask. Can i damage 4790k using Prime 95 Blend? I used that software to test stability of cpu. But today i read that using Prime 95 with Haswell is dangerous due to overvolting.
I used Prime 95 Blend test for 10 hours , no crashes. But is there any chance that i damaged cpu by any way ?
 
The issue with Prime 95 on Haswell is that the new P95 versions use AVX2 instructions. When Haswell receives AVX2 instructions, it boosts the CPU voltage over what you had currently set it to. You will never run into an issue with a stock setup running P95, since it it a relatively mild voltage boost to a low voltage.

Issues could occur when a user is *right* on a bleeding edge of a high overclock and already boosting the CPU voltage to get that overclock. Then P95 starts and the CPU kicks in even more voltage, which can push the CPU over the edge. Prime 95 doesn't overvolt the CPU, people push their CPUs too hard and then get bothered when Prime 95 causes the CPU to do what it is designed to do and get pushed even further.
 
So there is no chance to damage cpu ( all stock settings + stock cooling ) using Prime 95 blend for many hours? I left once for 10 hours. I never repeat that.
 
ocellaris said:
The issue with Prime 95 on Haswell is that the new P95 versions use AVX2 instructions. When Haswell receives AVX2 instructions, it boosts the CPU voltage over what you had currently set it to. You will never run into an issue with a stock setup running P95, since it it a relatively mild voltage boost to a low voltage.

Issues could occur when a user is *right* on a bleeding edge of a high overclock and already boosting the CPU voltage to get that overclock. Then P95 starts and the CPU kicks in even more voltage, which can push the CPU over the edge. Prime 95 doesn't overvolt the CPU, people push their CPUs too hard and then get bothered when Prime 95 causes the CPU to do what it is designed to do and get pushed even further.
Click to expand...

The problem is that the integrated voltage regulator (IVR) will often supply more voltage than is required when synthetic benchmarks are used. This is an issue with adaptive voltage. If you use manual voltage settings you can use Prime95 but the problem you run into with that is heat soak. Lets say your trying for 4.7GHz at 1.30v. What you often run into is a perfectly stable system for 20, 30 or even 60 minutes. But if you watch the CPU temperature you'll see the temps creeping up ever so slowly. You end up building up more heat than your cooling system can actually dissipate.

The general consensus is that adaptive voltage is the best way to go for actual overclocking and use. If you are concerned about Prime95 and OCCT, etc. then you have to understand that you won't find a maximum stable 24/7 overclock with those synthetic workloads. You'll fall short of what you can get using adaptive voltage. Granted if you can only hit 4.6GHz with Prime95 then that same overclock should be stable for everything else.
 
In addition to the Haswell-specific issue with Prime95 you could concievably damage the chip from the heat. The stock cooling solution is not designed to keep the chip cool 24/7 pegged at 100% usage so if you have some reason to run your CPU at 100% constantly I recommend upgrading the cooling or you risk premature failure. With Intel's high-end CPUs the stock cooler is barely capable of cooling the chip.
 
I am using stock clock speeds in cpu and stock voltage settings.


OK then. How can i be sure now that something didnt happened with cpu now? My system is stable but i want to be sure now that nothing bad happened with cpu after 10h torture test in Prime 95 blend. Thx
 
Last edited:
lpiotrowski said:
I am using stock clock speeds in cpu and stock voltage settings.


OK then. How can i be sure now that something didnt happened with cpu now? My system is stable but i want to be sure now that nothing bad happened with cpu after 10h torture test in Prime 95 blend. Thx
Click to expand...

At stock speeds you'll be fine.
 
lpiotrowski said:
I am using stock clock speeds in cpu and stock voltage settings.


OK then. How can i be sure now that something didnt happened with cpu now? My system is stable but i want to be sure now that nothing bad happened with cpu after 10h torture test in Prime 95 blend. Thx
Click to expand...

Why did you run the test like that stock? I dont know of a way to check if it broke other than bench-marking with a different program to load it up for a bit.
 
I'd say if it never crashed during the test you're fine, especially if you were at stock voltages. Not that I'd do it again...

Haswell-E's are a lot easier to damage I would think. I've heard of OC'd 5960x's spiking to 400W running that test. I think it's much harder for the larger chips to safely regulate themselves at those wattages. 4790k at stock is probably 1/3 or 1/4th of that.
 
Last edited:
One of my burn-in procedures is to run Prime 95 on a 5960X at stock speed with an H110 for 72 hrs. Have done this over a dozen times with no problems. CPU temp has never exceeded 124 F!
 
Dayaks said:
I'd say if it never crashed during the test you're fine, especially if you were at stock voltages. Not that I'd do it again...

Haswell-E's are a lot easier to damage I would think. I've heard of OC'd 5960x's spiking to 400W running that test. I think it's much harder for the larger chips to safely regulate themselves at those wattages. 4790k at stock is probably 1/3 or 1/4th of that.
Click to expand...

Intel CPUs are harder to damage than you think. I've locked up many of them, caused BSODs, overvolted the piss out of them and they've definitely over heated. I've never had one die on me through all of that and I overclock with some questionable motherboards too. (And I have no regard for the CPUs health.)
 
Dan_D said:
Intel CPUs are harder to damage than you think. I've locked up many of them, caused BSODs, overvolted the piss out of them and they've definitely over heated. I've never had one die on me through all of that and I overclock with some questionable motherboards too. (And I have no regard for the CPUs health.)
Click to expand...

Good to know since I just bought a 5960x and was quite afraid of killing it! I'll probably avoid overvolting the piss out of it but would like a decent OC.
 
Last edited:
I have also heard and encountered this myself. Not the death of the chip in question, but the very hot temps from Prime95. There are specific versions of Prime95 that "dangerously" ups the temps to near 100C+. I was in another forum about this issue, and it seemed that AIDA64, which is what I use, realistically stresses the Haswell chips compared Prime95 (specific versions ofc). Seems highly recommended for stressing. So that would be another option to consider if you wish to stress test that chip while attempting to OC it. I was fine with AIDA64 stressing my system with no flaws for at least 12 hours, but all I do is game and stream. Even so, the "streaming" load is pretty much offloaded to my 2700k streaming rig, so even less stress.
 
Last edited:
Prime95 does get the chip heated...as long as the you have proper cooling and are monitoring the temps regularly I'm sure its safe to run.
 
Programs don't damage CPUs, too much voltage damages CPUs.
 
xorbe said:
Programs don't damage CPUs, too much voltage damages CPUs.
Click to expand...

Even then you really need to run excessive voltage over time or you need to overvolt the crap out of a CPU all at once to kill it.
 
Dan_D said:
Even then you really need to run excessive voltage over time or you need to overvolt the crap out of a CPU all at once to kill it.
Click to expand...

Thats good news for quite a few of us here...last time i figured out i was using unsafe voltage almost a year past bye lol:eek:
 
As a previous poster said there's no danger as long as you don't run it with adaptive voltage and set voltage manually but even then I have stability tested loads with IBT prime Asus realbench etc whilst forgetting all the time to swap from adaptive voltage to manual and have been okay but definitely only use it and stabilty test on manual voltage only.

As long as your temps are okay and you're on manually set voltage it's safe.
 
I wanted to throw out some information to clear up some incorrect info I have read here (no disrespect intended), as well and many other places. Haswell's don't just up their voltage when using synthetic benchmarks. They up their voltage whenever the advanced instruction sets are being put into use by the running program. AVX, AVX2, FMA3 do this for sure, I have even seen it with SSE3 in use. If any program you run uses these instruction sets, your CPU voltage will increase over your set voltage (when using Adaptive or Stock).

I know this as fact because I run a number of different Boinc projects and many of them use these instruction sets. For example, at PrimeGrid, their LLR tasks use AVX. When I run these on my current CPU, my set voltage of 1.182 goes up to 1.216 every time. All of my Haswells do this (see sig), just at different voltages. I use a version of P95 that uses the FMA instruction set and it does the exact same thing to my voltages: 1.182-->1.216. I have also seen this behavior with tasks that use SSE3, but not always. AVX and FMA are the killers.

Anyway, as I wrote earlier, I just wanted to clarify this. If it is redundant info, I apologize.

Update: I also wanted to add that the CPU does this voltage jump on a per core basis. Meaning, if you are running the AVX instruction set on only one core, only that core ups its voltage. The other cores will run at the set voltage. Likewise, if you run AVX instructions on all four cores, all four will up their voltage. I have observed this behavior many, many times.
 
Last edited:
Great information, this explains why Prime95 and IBT run so much hotter than any game.
 
And that is the reason why you DO want to test your system with the new Prime95. Ya, most games don't use those instructions... for now. That may very well change in the future. Things like FMA3 have a good bit of utility in physics simulation and so you could find games start to use them. Particularly if they use Havok, which is Intel and thus Intel's people will optimize it for their tech.

If your CPU is stable with Prime95's intense test, then it is going to be stable with anything. If it has issues with Prime, well then it might work for other things, but might have problems in the future.
 
Sycraft said:
And that is the reason why you DO want to test your system with the new Prime95. Ya, most games don't use those instructions... for now. That may very well change in the future. Things like FMA3 have a good bit of utility in physics simulation and so you could find games start to use them. Particularly if they use Havok, which is Intel and thus Intel's people will optimize it for their tech.

If your CPU is stable with Prime95's intense test, then it is going to be stable with anything. If it has issues with Prime, well then it might work for other things, but might have problems in the future.
Click to expand...

I am with Sycraft. Back in the Q6600 days, a lot of people were hitting 3.6 GHz and then had to slow down their overclock during summer months because the CPU lost stability with higher ambient temps. But those people would ensure everyone their overclock was stable. BF2 and BF3 have had a tendency of causing stability issues as well, on systems that were "100% stable" otherwise.

I would much rather push a chip while stress testing so I am more certain that something won't pop up in the future that would introduce instability. I don't want to go in after the fact and start making adjustments. I try using the best available tools at the time and those are whatever pushes the hardware the most. Just trying some games and apps means little for future stability.
 
Sycraft said:
And that is the reason why you DO want to test your system with the new Prime95. Ya, most games don't use those instructions... for now. That may very well change in the future. Things like FMA3 have a good bit of utility in physics simulation and so you could find games start to use them. Particularly if they use Havok, which is Intel and thus Intel's people will optimize it for their tech.

If your CPU is stable with Prime95's intense test, then it is going to be stable with anything. If it has issues with Prime, well then it might work for other things, but might have problems in the future.
Click to expand...

this is not true... prime95 isn't anymore what it was in the past and actually you can pass days of prime95 and still fail/crash/bsod while gaming CPU intensive games firstly because while gaming the voltage are varying according load across the cores specially with today intel overclocks as most people tend to use adaptive/offset voltages.. which are constantly increasing/decreasing according load... second you can pass hours of prime95 and then fail apps like handbrake.. actually nothing its better for stability test than real world usage, benchmark programs based in real world usage like Frybench, Cinebench, RealBench will detect way faster a unstable overclock than prime95, IBT, Lynx, Aida64, etc.. I think in gaming Crysis 3 its the best game to test a overclock..

best way to test a overclock its a mix of several programs.. between synthetic and real world..
 
You must log in or register to reply here.
Back
Top