Realbench - Stress test instability issues

[ezy10z]

Hello,

My system is continually failing the Realbench V2 stress test. I'm running all of the tests at once. Sometimes the stability test will last for 15minutes and then sometime reach 1hr. I'm trying to complete the 8hr test.

Below are the details of my PC.

Asrock Z97 Extreme 4
i7-4770k
8 gig RAM
H100i with Noctua fans

With the following overclock it run 8hrs in Prime95 blend temps around 78-88.

Multi 44 (all cores)
CPU cache 39
Vcore 1.2425
Cp Cache voltage 1.15
Voltage setting to fixed on both cpu and cache
LLC auto
memory xmp 1 at 16000
memory voltage 1.5

Before trying Realbench I started the vcore at 1.2 an multi at 46 and tested in prime95. Over 3-4 days it was clear that 1.244 vcore was about the limit as my temps in prime95 were 85-88 max. So gradually reduced multi to 44. I tried increasing cpu cache but system seemed to be more stable with 39 so With those settings I run prime95 for 8hrs and stopped it there.

For the last two days I've been trying to pass the Realbench stress test but its impossible. With the settings described above it fails after 15 min when running all tests. I've gradually reduced cpu cache by 1pt and retest but continue to fail. I've tried cpu cache at 35 and it fails. I've gradually increased cpu cache voltage to 1.2 but still fails. I reduced multi to 43and kept vcore at 1.242 and still fails. I reduced multi 42 and kept vcore at 1.242 and managed 1hr and thought I was close so increase vcore to 1.25 and failed in 10 minutes!!. Seems no matter what I change the results of the stress test are usually worst even if appears to be a change to improve stability. I've reset bios and started from scratch but continues to fail. I tried reducing memory to 1333 and started with 44 core and same cache settings that that run for 1 hr but failed in 20 min. Increase cpu vcore 1.251 and failed in 9 min. This is doing my head in. Also I did run mem86 but only 1 pass and no memory errors occurred. Is there anything else I can try? I've never had such problems with a stress test. For example at 44 multi with settings described in first paragraph I can run benchmarks with no issues. Cinebench, 3dmark, Heaven etc. Memory on GPU is +200 but I used that on my AM FX 8350 setup for nearly 12 months and had no problems with it. I would appreciate anyones help. Thank you.

[Myk_SilentShado]

Do you get a BSOD? if so what code?

[ezy10z]

Myk SilentShadow wrote:
Do you get a BSOD? if so what code?

No BSOD occurs but the following message " instability detected " is displayed at the bottom of the Realbench window. Then the test stops.

[ezy10z]

This is the error that occurs . I googled it and seems it could be a glitch with Realbench?

Handbrake: Encoding: task 1 of 1, 91.69 % (7.44 fps, avg 1.26 fps, ETA
Handbrake: 00h04m40s)
Handbrake:

Handbrake: Encoding: task 1 of 1, 91.69 % (7.44 fps, avg 1.26 fps, ETA 00h04m40s)
Blender:BLF_lang_init: 'locale' data path for translations not found

[ezy10z]

I'm still trying to fix this issue. Can anyone assist?

I've searched google for the error shown in my previous message and the only results I find is it's a path error or glitch with the test.

Is it a stability issue or something else? I can complete the benchmark test ok.

I tried the stress test yesterday with 43 multi, 41 uncore, 1.28 vcore and got the same error instability message after 1hr 10 min. I increased the vcore to 1.29 and failed in 20 min. So increased vcore to 1.30 and failed in 15 min. Is there something else I should be looking at?

I would appreciate any feedback. Thanks.

[ezy10z]

I've done my last attempt at trying to resolve this as all my trouble shooting and testing and no response on here for a long time it's not worth the effort. At this stage I dropped to 41 multi and 35 uncore with vcore ranging from 1.18-1.3 and the stress could end at 10min or 1.5hrs with the same settings. To me this is a glitch. I even tried my 4.2ghz setting that passed a 24hr Prime 95 27.9 and same thing happened. Based on feedback from another site I'm going to stick with a modified version of x.264 to stress test haswell and works the cpu harder than realbench. It's easy to find any stability issues and you dont have to run it for a long time.

[Praz]

Hello

If you cannot successfully run the RealBench stress test it is because of instability or software/driver corruption or misconfiguration issues. Not a glitch with RealBench. However, if testing only with software that your system successfully runs provides piece of mind than that is what you should do.

[X-ROG]

Although not really aware of the settings available on your ASRock board, things to try:

Can you run the stress test successfully ...
> at default (stock speed) settings?
> with less memory capacity selected?

Can you pass the benchmark itself without issue?

What temps are you getting during benchmark load?

Prime95 doesn't test your whole subsystem, only CPU and memory. It does nothing to your chipset and it doesn't oscillate the loading between components. Is your PC generally stable in gaming/software?

[ezy10z]

Praz wrote:
Hello

If you cannot successfully run the RealBench stress test it is because of instability or software/driver corruption or misconfiguration issues. Not a glitch with RealBench. However, if testing only with software that your system successfully runs provides piece of mind than that is what you should do.

MarshallR@ASUS wrote:
Although not really aware of the settings available on your ASRock board, things to try:

Can you run the stress test successfully ...
> at default (stock speed) settings?
> with less memory capacity selected?

Can you pass the benchmark itself without issue?

What temps are you getting during benchmark load?

Prime95 doesn't test your whole subsystem, only CPU and memory. It does nothing to your chipset and it doesn't oscillate the loading between components. Is your PC generally stable in gaming/software?

Sorry for the late reply but unfortunatley I didnt receive the notification for an update to this thread.

As far as the stability in games my PC seemed ok except for Crysis 3. I wasnt able to complete the latest version of Prime95 as temps were to high but other tests seemed to be fine. I had also run memory stress tests, X.264 to get a guide if I was close to stability.

Two weeks ago I installed a second GTX 770 and the card wasnt detected. I replaced the Asrock MB with a Gigabyte Z97X-UD5H and the 2nd card was detected. Another thing I found was the cpu temps had dropped by almost 20-25 degrees when running Prime95 28.5. To me it sounds like the Asrock Extreme board had a faulty 2nd pcie slot and perhaps something wrong with the socket or power phase.

Last night I run Real Bench for the first time since installing the Gigabyte board and completed the 8hr stress test successfully 🙂 🙂 🙂 I can also run the Prime95 28.5 for 24hrs and temperatures didnt go over 78 degrees with a slightly higher overclock. My OC is 45 multi (+195 offset), 41 uncore (+140 offset), 2400 RAM, 1.875 vrinn.

Thanks for your response.