System Crashes when playing 3D Games.

Hi all, having some sort of mysterious problem with my build.

CPU AMD A10-5700 3.4GHz Quad-Core Processor
Motherboard Asus A88XM-A Micro ATX FM2+ Motherboard
Memory PNY Optima 16GB (4 x 4GB) DDR3-1333 Memory
Video Card XFX Radeon R7 360 2GB Core Edition Video Card
Power Supply CoolMax 700W Semi-Modular ATX Power Supply
CPU Cooler Enermax ETS-T40F-RF 65.6 CFM CPU Cooler

This is an old build that I recently added a dGPU to and swapped out an old HDD. I have NOT overclocked this system at all.


When launching a game, the menu system and all work fine, when 3D gameplay is about to start however, the whole system shuts down. The ASUS mobo warns that it shut down the system because it detected a power surge (the power cord is run through a surge protector). Turning off the surge protection in the BIOS makes no difference. The system shuts down at the same point.

I suspected a CPU overheat, so I installed a new cooler. Temps are idling ~51C. This seems high to me, but others have told me that it is OK for this AMD CPU. The temps rise to ~65C right before the computer crashes (using HWMonitor). I have also wiped all the GPU drivers with the DDU uninstall tool and reinstalled them. Took apart the build, airdusted the fans and vents, and vacuumed the empty case. It is a lot cleaner now and airflow seems to be ok. The exhaust air is not super hot, only a little warmer than room temp.

The system still crashes at the same point. I have run Prime95 and the results are as follows:

SmallFFTs: did not fail after a few hours running.
LargeFFTs: one worker failed on one of the first few tests. Others continued for a while.
Blend: all workers failed within the first few tests.
In some docs I have heard that a voltage drop can indicate PSU problems, and I have observed that the 12V rail voltage does drop a bit (~10.5V) when starting Prime95.

I have also run AIDA64 tests:
Stress CPU + Stress FPU + Stress cache: run for a while with no failures
Stress system memory: fails almost immediately and indicates "Warning: Hardware failure detected! Test stopped"

I have run Memtest86 and let it go overnight. It passed through 3 iterations without any errors. This took ~9 hrs.

I am sort of out of my depth at this point. I really have little idea on what to do next to isolate the problem. I do not really have any spare components to swap in and out, but I have a Frys close by that I can run to. But I'd like not to spend any additional money if not necessary to identify the problem.

For now, I will try to swap around the DIMMs and see if that might help identify something, but I am not hopeful.

If anyone could offer some help, I would really appreciate it. Thanks in advance.





For future reference: The issue appears to be resolved with a replacement PSU.

mhermher said:

Hi all, having some sort of mysterious problem with my build.

CPU AMD A10-5700 3.4GHz Quad-Core Processor
Motherboard Asus A88XM-A Micro ATX FM2+ Motherboard
Memory PNY Optima 16GB (4 x 4GB) DDR3-1333 Memory
Video Card XFX Radeon R7 360 2GB Core Edition Video Card
Power Supply CoolMax 700W Semi-Modular ATX Power Supply
CPU Cooler Enermax ETS-T40F-RF 65.6 CFM CPU Cooler

This is an old build that I recently added a dGPU to and swapped out an old HDD. I have NOT overclocked this system at all.

When launching a game, the menu system and all work fine, when 3D gameplay is about to start however, the whole system shuts down. The ASUS mobo warns that it shut down the system because it detected a power surge (the power cord is run through a surge protector). Turning off the surge protection in the BIOS makes no difference. The system shuts down at the same point.

I suspected a CPU overheat, so I installed a new cooler. Temps are idling ~51C. This seems high to me, but others have told me that it is OK for this AMD CPU. The temps rise to ~65C right before the computer crashes (using HWMonitor). I have also wiped all the GPU drivers with the DDU uninstall tool and reinstalled them. Took apart the build, airdusted the fans and vents, and vacuumed the empty case. It is a lot cleaner now and airflow seems to be ok. The exhaust air is not super hot, only a little warmer than room temp.

The system still crashes at the same point. I have run Prime95 and the results are as follows:

SmallFFTs: did not fail after a few hours running.
LargeFFTs: one worker failed on one of the first few tests. Others continued for a while.
Blend: all workers failed within the first few tests.
In some docs I have heard that a voltage drop can indicate PSU problems, and I have observed that the 12V rail voltage does drop a bit (~10.5V) when starting Prime95.

I have also run AIDA64 tests:
Stress CPU + Stress FPU + Stress cache: run for a while with no failures
Stress system memory: fails almost immediately and indicates "Warning: Hardware failure detected! Test stopped"

I have run Memtest86 and let it go overnight. It passed through 3 iterations without any errors. This took ~9 hrs.

I am sort of out of my depth at this point. I really have little idea on what to do next to isolate the problem. I do not really have any spare components to swap in and out, but I have a Frys close by that I can run to. But I'd like not to spend any additional money if not necessary to identify the problem.

For now, I will try to swap around the DIMMs and see if that might help identify something, but I am not hopeful.

If anyone could offer some help, I would really appreciate it. Thanks in advance.

Click to expand...

As the guy says to the doctor "Doc, it hurts when I do this" Doc says "Well, don't do it"

I don't have your CPU but from what you have said, it may be the PSU, also, what are you using for cooling? Is it the HSF that came with the CPU? If so, look in to a NH-D14 or NH-D15 to cool your CPU.

WhitehawkEQ said:

As the guy says to the doctor "Doc, it hurts when I do this" Doc says "Well, don't do it"

I don't have your CPU but from what you have said, it may be the PSU, also, what are you using for cooling? Is it the HSF that came with the CPU? If so, look in to a NH-D14 or NH-D15 to cool your CPU.

Click to expand...

No. I replaced it with a Heat Sink and Fan (listed above in the main post). Replacing it did drop the temps decently. I played around with swapping the DIMMs too, and it gave some results that don't completely make sense right now. But dropping the memory to 8GB (from 16GB) stabilizes it.

I've at least figured out how to keep it from crashing but still don't totally understand why. Need to do more tests swapping things around but I ran out of time on Sunday. Might just eventually get a new PSU too just to see what difference that makes. This PSU is pretty old now.

65c sounds pretty darn hot did you use thermal paste between the cpu and heatsink?
if workers are failing in cpu-z any test its a cpu issue, either its getting too hot, or a failed overclock.
run memtest with all 16gb in after you get the cpu temps fixed. if dropping to 8gb fixes it then you probably have a bad stick of memory.

wagex said:

65c sounds pretty darn hot did you use thermal paste between the cpu and heatsink?
if workers are failing in cpu-z any test its a cpu issue, either its getting too hot, or a failed overclock.
run memtest with all 16gb in after you get the cpu temps fixed. if dropping to 8gb fixes it then you probably have a bad stick of memory.

Click to expand...

My initial reaction was that it was too hot, so I spent some time improving that (new HSF, un-dusting and rearranging some components to improve airflow). I got some feedback elsewhere that my temps weren't too bad (which surprised me, but I took their word for it). I'm not an expert in applying thermal paste (have only done it a couple times), but I am pretty sure I did it ok. Just dropped a BB sized amount on the mounted CPU head (?) and pressed the heat sink onto it. It did wiggle and slide a little as I secured it in, but nothing major.

My other thought is whether I may have stressed the CPU too much over the years (with the stock heatsink+fan, this is like a 3-4 year old CPU), and maybe permanently damaged it in some way so that it runs hot. I am not sure if this actually happens. Is it possible that I have weakened the CPUs ability to regulate its temps from overuse?

When I ran memtest (3 passes, no fails), it was with all 16GBs. That made me think that none of the sticks were bad. Should I let it run longer? It took ~9 hrs to get through 3 iterations though.

I will run cpu-z tonight when I get home. Thanks for the suggestion. I will also try some other memory configurations to isolate whether it is a bad stick (but I got some somewhat confusing results last time. I may have two bad sticks? Or maybe a bad RAM slot?). Or maybe a bad lane (is that a thing?) that connects the memory to the CPU? The signs do seem to indicate something having to do with memory.

Try each stick individually, one at a time, in each slot. This tells you if the stick is bad, and if the slot is bad. If the slot is failing though, seems to me it would be reading only 12GB from time to time as, in my experience anyways, bad slots can usually be identified as the stick you put in them not always being detected. The 8GB part makes me think it could indeed be bad slots or sticks. Here's to hoping it's the sticks, you might have to buy an FM2+ board to be able to find one at all for that APU.

So I thought I had identified the problem here. I reconfigured the sticks and kept running the memory stress test (with AIDA64) to isolate the stick/slot that was bad. I identified that one of the sticks was indeed causing a failure. It was not 100% reproducible, so it took a while to figure it out. I isolated it to a single stick and since removing it have not failed the stress test. Additionally, the game I was using to test was passing. I was able to get into the 3D gameplay no problem. All has been working fine since then.

However, I just installed a new game which is probably a little more graphically intensive. To my dismay, on the first launch, my system crashed again when it just got to loading something (I'm not sure what it was loading at that point, but I assume some graphically intensive portion after the menus).

I just checked Prime95 (I hadn't done so since stabilizing the system), and the Blend test again fails (all 4 workers fail on the first iteration). So, it seems not only did I have a bad DIMM, but something else is wrong with the system. Again, I am trying to salvage an old build, so I guess I shouldn't be surprised.

I am running cpu-z now. I have never run it before so I am not sure how to interpret this (I will read up on it). It is just a bar graph like graphic that keeps changing the number. Hovering around ~3000. Not sure what that number means. No indications of any sort of error though.

Any other ideas on what may be wrong are appreciated. Thank you.

PSU. That Coolmax thing was garbage when it was new.