Running CPU Stress Test + GPU Stress Test Will Trigger A Crash??

audaciousishan

Reputable
Jul 10, 2015
43
0
4,530
HI There!

MY Specifications :

CPU : Intel Core i5 4590@3.3 GHZ
Memory : 8+2 GB Corsair 1600 Mhz
Motherboard : MSI B85-G41 PC Mate
HDD : Samsung HD502HI (500GB)
GPU : MSI GTX 760
PSU : Corsair VS650
Sound Card : Asus Xonar 5.1 DG
OS : Windows 7 Ultimate 64 Bit


So actually I'm going through a crashing issue of my system since I bought it, but earlier the crash was into BIOS. Later carried onto with Random crashes and now even crashes when the system is just idle. I've made a separate thread for that and haven't yet a solution to it but soon gonna update it as I'm gonna get some update.

Right so that thread is separate but related so I just mentioned.

Regarding this thread when I perform the CPU stress test with prime95 it kept going on and I left it overnight with the test running on. I woke up in the morning and the stress was still going on. I thought just to add a little more stress on my GPU side by side so I ran the GPU stress as well with Furmark. It started up very slow and found laggy distorted images in the GPU test. Soon within 2-4 minutes my system just got crashed.

Also I'm running the CPU stress test and the main window of CPU stress (where it displays all the passes and all) is disappeared itself but the stress test is still going on as my CPU usage is still 100%. I guess its going on in the background but is it also normal? I can't see that.

Then I ran the GPU stress separately for 1 hour no problems at all. Temperature maximum reached to 82 degrees. But when I combine both the stress test i.e. CPU + GPU my system gets crashed.

Question : Is it normal for my system (as the configuration stated above) to get crash when I try to run both the stress tests at the same time?? As separately both stress tests worked/performed with no errors.

Thank You! :)
 
Solution
OCCT has an option for a PSU stress test which combines the CPU and GPU stress tests.... they give a string warning that this test can literally smoke your PSU. I suspect a PSU issues but without a detailed components list (make and specific model numbers) , it's just about impossible to start diagnosing your problem.

Also Prime 95 is not recommended for Haswell and newer CPUs. You can run the older non AVX version but it proves nothing as it's not testing the CPU's newer instruction sets
OCCT has an option for a PSU stress test which combines the CPU and GPU stress tests.... they give a string warning that this test can literally smoke your PSU. I suspect a PSU issues but without a detailed components list (make and specific model numbers) , it's just about impossible to start diagnosing your problem.

Also Prime 95 is not recommended for Haswell and newer CPUs. You can run the older non AVX version but it proves nothing as it's not testing the CPU's newer instruction sets
 
Solution


Sorry forgot to update on my system specs 😀 I've updated the system specs now.
Prime95 v28.5
 


Well the smoking warning doesn't sound good 😀 As I'don't have money to buy new hardware if anything gets SMOKED 😛
But I just wanna know that doing the CPU + GPU stress tests the system getting crashed. Normal for my system specs?
 
The old Corsair sounds like it's having issues. As the CPU isn't overclocked, it's not too bad but running OCCT CPU stress test will result in test shutdown as tempos should easily exceed the default 85C shutdown trigger.

It is strongly suggested that you use an application based stress test such as RoG Real Bench.
 


Umm alright I'll check that out if I find it somewhere. That ROG Real Bench will stress test my CPU? And I can add the GPU stress test with it as well to check my system stability? Cause with Prime95 + Furmark = Crash. I've to run both stress tests separately.
 
You'll need to monitor both voltage and temperature levels (HWiNFO64) while running to see what is the problem. That's one think OCCT does all by itself. I suspect the PSU is being stressed so much that voltage stability (and or noise / ripple) ranges outside the ATX spec...
 


I ran the OCCT and my system just crashed after few minutes. I chose the PSU test tab and ran and system got crashed after few minutes. What it indicates? I'll again run the OCCT to see what happens now.
 
I can run Prime95 on my i5-4590 and it throttles when it reaches 100°C, but it obviously doesn't shut down. Does OCCT force the system to shut down when it detects the CPU temperature reaches 85°C?
 
Yes, by default the test is programmed to stop the test at 85C. You can change the limit.

But again, don't see the point of P95 on any Haswell or newer CPU. I mean it;'s good if you want to get your name on web site leaderboards and say my system is stable under P95 at xx C temperature. But as you can be P95 stable and fail under RoG Real Bench, what's the point ?

P95 puts a heavy synthetic load on the CPU but doesn't test the CPU multitasking various different instruction sets at the same time.

as an analogy ....

P95 is like testing your 6,000 pound tow rated truck pulling 8,000 pounds at seal level on a cool day on an open road.

RB is like testing your 6,000 pound tow rated truck pulling 6,000 pounds at at sea level, in the mountains, at various temperatures and traffic conditions
 


Well after all this I ended up with my faulty PSU. I mean my PSU was faulty and sent it for RMA and they found it faulty and dead. So my related issues are all solved. Thanks to you who suggested me for OCCT. I was getting crash when I was using that in the PSU tab, which made me think that my PSU is faulty. And the doubt was assured by the service center guys. Now I got the replacement and everything going smooth! :)

Thanks Bro :)
 


Nopes. Earlier also my system was having some random crashes. That time I never used these stress tests. These stress tests I used and performed later when I was trying to figure out that what is wrong with my system.
 
Well glad it worked out well. It's best to avoid the manufacturer's "builder series"....er vendors (Corsair doesn't make PSUs, they buy from OEM and resell) ... Corsair HX (most sizes) and AX are all very good, the TX is more than serviceable. The rest I'd avoid in future purchases.

But again, testing a CPU with Prime 95 I find a wasted effort:

1. When the current version is used, it puts an extreme load on the PSU .... one that could never be duplicated using any combination of loadings for "real world" programs. As a result, you unnecessarily limit your OC because your are cutting it down so it can handle a thermal load that will never again be created .... unless of course you want to search for the next mersenne prime :)

2. If you use the old version, well than you have proved the system is stable under conditions when AVX SSE2 and other instruction sets are not present. Kinda like testing you car without making any left turns or going into 3rd gear. Those instruction sets are part of some programs you might use so what is the point of testing with it ?

3. You can be Prime 95 stable and still fail in RoG Bench. I still use Prime 95 (new version) for the purposes on setting TIM in a new build. I put on a moderate OC, bring temps up to 85 or so and let cool down, rinse and repeat sets the TIM nicely. After doing so.... I started looking at stability and found that my P95 stable "shot in the dark" OC was not RoG RB stable. Needed a bit more on VCCIN, Core and cache voltages. Can't be sure why but my guess is that P95 puts a large but very specific single type of load on each core whereas RB is a multitasking type benchmark throwing numerous different types of load which are likely having their affinity set to different cores. Think of it like standing at the end of a tennis court while I fire balls at you with a ball machine at 140 mph and you gotta block them w/ your racket .... now stand in the middle of the court while I put 4 machines in each corner of the court and fire 4 balls at you at 100 miles an hour. I think you'll get hit more in the latter case.