need help with Prime errors

leebeloola · 7 Apr 2016 at 20:39

I'm trying to figure out which hardware is responsible for the Prime errors I'm getting. I've gone back to default settings and I'm still getting the same errors. I was going upgrade the RAM and get an SSD, but now I'm wondering if I need to replace the CPU and motherboard.

It passes the small FTT test, but I am getting “rounding was 0.5, expected less than 0.4” hardware errors within a few minutes of running the large FFT or Blend tests. I've run Memtest86+ several times overnight for 13 passes and I didn't get any errors. I've run tests on the HDDs and used intel diagnostic tool for the CPU, but haven't found anything. I've also been checking the CPU temp while testing and it's well below 40*C.

System specs:
MB: Gigabyte GA-H55M-UD2H Intel H55 (Socket 1156)
CPU: Intel Core i3 530 @ 2.93GHz (bought it @4.2MHz, but lost settings after bios flash)
RAM: Corsair DDR3 PC3-12800C9 1600MHz Dual Channel Kit (2x2GB) CMX4GX3M2A1600C9
Graphics: AMD Radeon HD 5770 CuCore 1024MB GDDR5
PSU: Corsair GS 600W Power supply (CMPSU-600GUK)
Cooler: Akasa AK-CCX-4002HP Venom CPU Cooler

zipp0r · 8 Apr 2016 at 05:07

if you pass small ftt for hours and hours but error on large/blend within minutes, its likely to be a problem with the memory/chipset/memory controller.
could be not enough voltage for the ram or memory controller or too tight memory timings e.t.c e.t.c so look to that area.

TheMadDutchDude · 8 Apr 2016 at 05:48

As above, those errors are related to your memory, your memory controller or the chipset, the first two are more likely.

Things to check:

1) Make sure your RAM voltage is correct,
2) Check your IMC voltage (it may need increasing) and,
3) If all else fails, reset to factory defaults and try it fresh stock settings.

Let us know how you get on. It could be that the parts are starting to fail. They are fairly old after all.

Aretak · 8 Apr 2016 at 14:17

I wouldn't be 100% sure that the issue is memory-related, even if the symptoms suggest that it is. I spent days trying to get a 1090T overclock stable recently, with all the symptoms suggesting memory instability. It'd pass small FFTs AND Memtest86 just fine, yet fail with blend, IBT and during games. I was absolutely certain that the core itself was stable due to it passing small FFTs, yet absolutely nothing I tried in terms of adjusting NB voltage or memory speed/timings worked. After much frustration I eventually worked out that the core was the thing causing instability, but only when the memory was in use as well. Something about the combination of it all resulted in crashes, despite each individual piece passing stress tests.

I ended up dropping the core a mere 125MHz and everything became rock solid. It was perfectly stable at 3875MHz and 1.4V, yet not at 4GHz and 1.45V. I could even drop the NB voltage by 0.1V AND bump my memory up to 1666MHz CAS8/CR1, whereas with the CPU at 4GHz it would fail Memtest86 instantly even at 1333 and with extra voltage if I set CR1. It was the core causing the problems all along, despite the memory being the thing showing the symptoms. It's amazing that such a tiny 125MHz difference could cause so many problems, and that an extra 0.05V wouldn't make it stable. I guess somewhere around 3875MHz is just a hard wall for my chip.

Not really related to the thread tbh, but I thought it worth noting that problems can lie in unexpected places. In terms of this specifically, I'd definetly try bumping voltage a little. If it's not stable even at stock now, it suggests that the chip might have degraded. Perhaps the previous overclock had a fair bit of extra voltage applied if you bought it that way, and has degraded the chip over time. If the memory is passing Memtest86 for 13 hours, it's almost certainly not the RAM itself that's faulty IMO. The one thing Memtest86 does really well is detect faulty sticks, even if its usefulness for stability is somewhat limited due to it only really stressing the memory and not everything together, which leads to it not triggering errors such as the ones I was experiencing under real loads.

leebeloola · 2 May 2016 at 18:29

Sorry for the long silence, I had to put the project on hold for a while. I ran the small FTT test again and it ran for 17h (I couldn't remember how long it ran for last time).

I tried increasing the vCore to 1.21875V has resulted in the Blend test running for 1h 1m instead of the previous 3m. The max temperature is only 1 degree higher. I increased the vCore again and the test ran for 1h 13m with +1 degree, the next bumb up in V resulted in 1h 20m and again +1 degree higher. Seeing as the time difference isn't that big I gave up on this.

As for RAM, I've tried the following:
comand rate 1T to 2T = errors after 3m
tRC 34 to 41 = 22m
enabled XMP profile = 2m
I did load optimised default when I made changes to a different setting.

leebeloola · 2 May 2016 at 18:57

I really don't think it's the RAM. I replaced it & tested it repeatedly 2 years ago after adding it to my pre-existing RAM didn't work. It was the same product number but each set had a different density. I thought one of the sticks was bad, so I tested them all. I had to update the BIOS but unfortunately all the update description said was "improved memory compatibility".

I'd like to figure out if a RAM/MOBO compatibility problem or an old overclocked CPU is more likely. I would really like to understand how to diagnose this problem further. I haven't found anything on testing the MOBO and I'm not sure how to interpret the Blend test errors versus the passed small FTT & memtest.