Prime95

Associate
Joined
14 Apr 2004
Posts
69
I was trying to baseline a new build before playing around with oc. Ran prime95 on tourture test (Blend) and it results in an application error (prime95) between 45mins to 1 hour. I then thought I should properly check the RAM thoroughly, ran memtest86 for 12 hours and no issues.

I am not entirely sure if prime95 torturetest (Blend) will run for a sustained period of time? mixed reports from google from 20mins to 24 hours.

*cpu idle temp 30 under prime95 torture, doesn't break 70

Intel Core i7 4770K
Gigabyte Z87X-UD3H Motherboard
Patriot 32GB DDR3 Viper 3 Mamba Quad Channel 4 x 8GB 1600MHz
Corsair H80i
 
Last edited:
Just leave it running overnight or while you go to work, there is no perfect time frame just use your own judgment, you'll get a failure sooner or later if things are unstable if you test often enough. Also, don't just use blend which is more of a memory test, small and large FFT's both provide additional stress to the CPU.

The idea of running it for hours and hours is that it gives your system time to warm up, it's usually best to do your stress testing in the summer when conditions are the harshest, there's no point overclocking to the limit in winter because it's likely to fall flat on its face come the summer.
 
Last edited:
Are you ever going to actually use your system for hours on end in the way that Prime does

Yeah, I need to start using this new box for Adobe AE renders. They max the cores to 100%.

I installed CoreTemp. Found temps were at 90 degrees after 15mins so I stopped the test. I then thought I might have the H80i configured wrong. Switched from default fan mode to balanced (fans much louder) ran test again and it eventually hit 90.

Something quite wrong here. Going to run again with fans of performance mode.
 
With that cooler and CPU on stock you shouldn't be reaching 90c unless you have installed the cooler wrong or your case has zero fans.

Check temps with Hwmonitor and check that your BIOS "Auto OC" functions are not enabled. They overvolt your CPU like heck.
 
Maybe the pump isn't sitting correctly, will try that next. I need to check if the difference between the cpu temp and h80i temp is correct.
ze7rjlA.png
 
Last edited:
29c coolant temp.. sounds like it's taking quite a pounding there. Your volts are not overly high either.. it's quite strange. Maybe reseat it yeah.

How's your case airflow and are the coolers mounted the right way on the H80?
 
How's your case airflow and are the coolers mounted the right way on the H80?

Its good, both h80 fans configured to intake from the back of the case.

Just stuck the intel stock cooler on the machine, idle's about 37 and after about 10min prime takes it into the 90's. Does this sound normal for the stock cooler?
 
Yes the stock cooler will even cause throttling in prime in worst cases.

I just did a test on H100i with mine, after 20min of prime my coolant temp is at 30c+ on silent mode, and temps under yours..

So maybe you had a bad fit on the H80 if the core temps are 90's but coolant temp under 30.
 
I think I have a bad fit, there is some wiggle with just the backplate fitted. Found a forum thread that talks about having to fit extra washers to the backplate to make it fit better (link) Going to have rummage around for some washers now...
 
Washers fitted, results of Prime after 30mins, fans on quiet mode. Better for cpu temp but coolant is higher. I could do with buying some thermal paste because of the refit. Make, whats your thoughts on the on these temps now, bearing in mind the cpu is still at stock.
AYl5Ksy.png
 
I think with that cooler and prime95 it's a fine temp. Prime is an unrealistic stress test, so your gaming temp will most likely be around 40 at stock. The coolant temp seems more realistic now; it's supposed to heat up over time, that's how watercooling works.

When you OC, prime temp will be 80-85 but gaming temps at less than 60.

Sometimes OC actually drops temps, as you really wouldn't need 1.2v for 3.4ghz.. you'll probably get 4.1ghz with that. (Alright, it wouldn't drop temps, but it will give more performance for same).

Great job :)
 
If your coolant temps are going up, try increasing your fan speed to try to keep them down and reduce the delta-T between the chip and plate. At stock, those are acceptable temps, but they aren't great. The fact that the temps have improved after a refit lean towards the fact that you could potentially improve further with better paste and paste application. How much paste did you use out of curiosity, and what method did you use to apply it?
 
You can call it unrealistic but it's not doing anything that a processor isn't designed to do.

Intel will have highly advanced specialist tools that test processors on an electrical level to make sure they can run at maximum efficiency without faltering, the best we end users have is software tools like Prime95 which use the processor as efficiently as possible and reports back if it gets a sum wrong.

The problem with Ivy/Haswell is simply that they are glorified mobile parts and aren't really built with overclocking in mind, IB-E runs considerably cooler than standard IB and that has two more cores and a much higher TDP.
 
Last edited:
Originally fitted with the standard paste that came preapplied to the h80i block. Since then its been a bit messy with the stock cooler I put on as part of the troubleshooting, then after that I refitted the h80i (with washers). I don't have any paste so thats next on the list.
 
Make sure you have all of your memory timings at the right settings, 8GB DIMMS seem to be a bit high maintenance compared to lesser ones. I doubt a CPU would fail Prime95 at stock (unless it was faulty) as they are built to do that sort of thing.

Try clocking your memory at 1333mhz as well and see if that helps, I've seen it in the past where modules will not run at their rated speeds. Memtest isn't as stressful as Prime95 it's more suited to finding faulty RAM (as in areas that are completely error prone regardless of anything) rather than unstable RAM.
 
Last edited:
Back
Top Bottom