[SOLVED] Computer crashes from graphic intense games but not overheating

[Arthianne]

Computer crashes when i launch graphic intense games such as Black Desert Online, New World Alpha, ARK: Survival evolved etc. It's been happening for a long time now and I'm just deciding to fix it (Crucify me later) I've checked my drivers and they're all up to date, I've checked my temps and all of those are fine at 40 c I've checked the cords and theyre all fine as well. Seeing how its only an issue when i run graphic intense games I'm under the assumption it's a graphics card issue. Below are my specs, if you all want me to run tests with other programs im game as well, I just want this fixed, and I can't figure it out.


Operating System
Windows 10 Home 64-bit
CPU
AMD Ryzen 5 5600x
RAM
32.0GB Dual-Channel Unknown @ 1064MHz (15-15-15-36)
Motherboard
ASUSTeK COMPUTER INC. ROG STRIX B550-F GAMING (AM4) 35 °C
Graphics
VX2458-mhd (1920x1080@60Hz)
2269W (1920x1080@60Hz)
4095MB NVIDIA GeForce RTX 2070 (Gigabyte) 51 °C
Storage
1863GB Western Digital WDC WD2003FZEX-00SRLA0 (SATA ) 35 °C
465GB Hitachi HP SSD EX900 500GB (Unknown (SSD))
Optical Drives
Optiarc DVD RW AD-7240S
Audio
NVIDIA Virtual Audio Device (Wave Extensible) (WDM)

 

[Flayed]

You could try completely removing the drivers using Display Driver Uninstaller in safe mode. Then install the latest drivers.

It could be that your graphics card is faulty. Trying another graphics card would be best if you can borrow or source one from somewhere.

 

[scout_03]

could you post a picture of both ram spd in cpu-z page .

 

[geofelt]

When a graphics card is under heavy load, it will draw more power.
I would think the psu is the likely issue.
EVGA sells some very good units and some not so good.
I think yours is in the not so good category.
PSU testers are useless and can only confirm a dead psu. They say nothing about proper operation.
Your unit may be out of warranty which is 2 or 3 years.
Put in a RMA asap if you can.
The only way to confirm a psu issue is to replace with a known good replacement of sufficient power.
See if you can't borrow one to test with.
If all else fails, buy a quality replacement which will have a 7-10 year warranty.
Buy from a shop with a good return policy. Expect to pay a 10% restocking fee.

 

[Arthianne]

You do not need 32G of system memory for gaming especially with the graphics card you have, plus the fact that it is a mismatched set to boot. The memory speeds may very well match but all of the timings do not, which forces the CPU to juggle down to what it can run, resulting in a loss of performance.

Two different brands of memory is not a good idea, it is hard enough to manufacture memory module matched sets that run together problem free from the factory, much less get competing brands to work together never meant to be on the same motherboard.

Some motherboards do not like to have all 4 slots occupied which you may discover a little foot note about that in your motherboard manual, but whether it is frowned upon or not in the manual, if you are after performance don't do it.

Though your Graphics card is plenty capable it still has to communicate to the CPU through the system memory which is the communication bridge between the 2, if traffic is not flowing smooth across that bridge then you will have a problem.

I suggest dropping the Corsair memory and run the GSkill in the motherboards suggested slots for 2 modules.

Resolving gaming crashing issues like this is a process of elimination to find the problem and it is not just as simple as one size fits all.

If you do not have GPU-Z installed then install it and go to the sensor tab click on it, and at the bottom check Log To File, then you can run GPU-Z while gaming to have a recorded temperature file to see exactly where your Graphics cards temperatures go while heavy gaming right up to the crash occurs. You'll find the GPU-Z Sensor Log in My Documents and you can review your temperatures vs clock rate to be sure they do not get too high.

Try these suggestions and see what happens.

Thanks for the explanation! In college they told me that running different kits is fine as long as they're the same speed, getting an explanation like this helps me improve and understand so I appreciate it, I've removed the corsair

 

[Arthianne]

Download and run memtest!

I'll do that now

 

[Arthianne]

Download and run memtest!

0 Errors on memtest 4 passes

 

[Arthianne]

could you post a picture of both ram spd in cpu-z page .

https://imgur.com/a/5VWU8rH

View: https://imgur.com/a/5VWU8rH

 

[Arthianne]

When a graphics card is under heavy load, it will draw more power.
I would think the psu is the likely issue.
EVGA sells some very good units and some not so good.
I think yours is in the not so good category.
PSU testers are useless and can only confirm a dead psu. They say nothing about proper operation.
Your unit may be out of warranty which is 2 or 3 years.
Put in a RMA asap if you can.
The only way to confirm a psu issue is to replace with a known good replacement of sufficient power.
See if you can't borrow one to test with.
If all else fails, buy a quality replacement which will have a 7-10 year warranty.
Buy from a shop with a good return policy. Expect to pay a 10% restocking fee.

My unit is within warranty, however I don't have an invoice of my purchase since it was bought in an emergency from best buy, so if I RMA it, my system will be down for weeks

 

[TommyTwoTone66]

You do not need 32G of system memory for gaming especially with the graphics card you have, plus the fact that it is a mismatched set to boot. The memory speeds may very well match but all of the timings do not, which forces the CPU to juggle down to what it can run, resulting in a loss of performance.

Sorry but that's nonsense. The only thing you lose with mismatched memory is performance, there's nothing about having mismatched memory that would cause memory errors or crashes, the only thing you would see is higher overall latency perhaps or a small reduction in peak data rate. Why would it cause crashes??

Besides which, it's only games that are causing the crashes. If it was triggered by high memory usage, than a few minutes on Chrome using 10-20 tabs would also cause a crash.

But sure, put the issue to bed, run a "full" Memtest with all 32gb of mismatched memory installed. I bet anyone $5 that you get zero errors.

It must be a GPU issue, and I'm thinking either power to the GPU is insufficient, or it's a cooling issue. Cooling is easiest to check. You can measure temps, and you can clean out the gunk in the fan and take the whole card apart and replace the thermal pads with new ones. You can't measure temps at idle, you need to start a graphics intensive game and look at what happens to the temp. If it shoots up instantly and goes above 100C, then there's your issue.

If your temps genuinely are fine, then have you got the right power cables connected to the GPU? is it a dedicated cable with only one endpoint, or is it a daisy chain type connector with multiple devices running off it? Try all the different ways of hooking up your PSU to your GPU. Try a different PSU (this can be difficult). If it does the same thing with another PSU, then almost certainly the GPU is faulty.

 

[carocuore]

Install a new HXI or P2, T2 power supply 850-1200 watts and try again. 9 times out of 10 these issues are power supply related.

Just... what?
He could also go ahead and build his own fission reactor as well /s

A fatal hardware error has occurred.

Reported by component: Processor Core

Is that CPU overclocked? make sure no auto-oc options are enabled on the BIOS Setup. WHEA uncorrectable errors are related to the CPU, not the graphics card.

I've read somewhere up those walls of text you're going to get a PSU tester, well, don't, all those things do is short 2 wires and check for the presence of any voltage, you can pretty much do that with a multimeter and it won't really tell you if the power supply is faulty because there's only a tiny current load on it. A borked PSU can work at the bench but not when you plug it to the rest of the components.
Yours isn't a cheap china unit though so it should be fine unless it came from a bad batch or something, in that case I think you can honor the warranty and get a replacement if you live in the US, Canada, Japan or the EU.

 

[TommyTwoTone66]

I've read somewhere up those walls of text you're going to get a PSU tester, well, don't, all those things do is short 2 wires and check for the presence of any voltage, you can pretty much do that with a multimeter and it won't really tell you if the power supply is faulty because there's only a tiny current load on it. A borked PSU can work at the bench but not when you plug it to the rest of the components.
Yours isn't a cheap china unit though so it should be fine unless it came from a bad batch or something

I am leaning toward PSU issue as well. The CPU might not have enough power to run any more if the GPU takes it all, hence the crash log showing a CPU failure.

Failures do happen, and a PSU is working all the time to smooth out imperfections, spikes and dips in the power feed that it is getting. If OP lives in an area where the power fluctuates a lot it might stress the PSU and cause it to fail early. Or yeah, it might be from a "bad batch", or be from a good batch and just so happened to have a manufacturing defect that slowly developed over time.

 

[TommyTwoTone66]

My unit is within warranty, however I don't have an invoice of my purchase since it was bought in an emergency from best buy, so if I RMA it, my system will be down for weeks

This is interesting. What emergency caused you to need a PSU in a hurry? The last one failed? Maybe there is an issue with the supply into your PC, maybe some other equipment in your home is causing power spikes or a neighbour is using some crazy high powered equipment that is dumping volts into the grid at random times?

Try a known working PSU in it. Ideally borrowed from another machine. If that works fine, then I recommend a new PSU and maybe one of those power strips that has a surge protector built in, or even better a UPS unit.

 

[Unolocogringo]

Sorry but that's nonsense. The only thing you lose with mismatched memory is performance, there's nothing about having mismatched memory that would cause memory errors or crashes, the only thing you would see is higher overall latency perhaps or a small reduction in peak data rate. Why would it cause crashes??

With matching speed and CAS.
With EDO ram mixing was basically accepted, always worked.
With SDRAM that was true 99% of the time.
With DDR probably about 95 98% of the time it worked.
With DDR2 92-95% of the time it worked.
With DDR3 90- 92% of the time it worked.
With DDR4 it seems to be even more common again.
With worked I mean the computer booted and ran without errors.
BUT many times the memory speeds and latencies would be terrible and the computer actually slowe than before. BUT HEY it had more memory
I've been building ,upgrading, tinkering with computers for many years. And here at Toms hardware since 1996
This is my experience over the years and many others here. Exact numbers are unknown. but each generation this seems to be more a common experience.

 

[TommyTwoTone66]

With matching speed and CAS.
With EDO ram mixing was basically accepted, always worked.
With SDRAM that was true 99% of the time.
With DDR probably about 95 98% of the time it worked.
With DDR2 92-95% of the time it worked.
With DDR3 90- 92% of the time it worked.
With DDR4 it seems to be even more common again.
With worked I mean the computer booted and ran with errors.

Where do these percentages come from? Were you mixing ram in hundreds of systems personally to come up with those numbers? Or is that just like a "guesstimate" of how often mixing RAM works perfectly well, based on... gut feel?

Come on. How would it cause a crash? Memory modules are not complicated. If you have different modules from different manufacturers running at slightly different speeds and timings, the memory controller just automatically selects the best timing and highest speed which all modules can achieve. Why would that not work in more than 10% of cases?

 

[Arthianne]

I figured it out, My BIOS was out of date and causing issues with my CPU's communication. Sometimes the complicated issues have simple answers, thanks for your responses everyone

 

[Unolocogringo]

How To - Troubleshooting problems with PC memory (RAM) and XMP profile configurations

In the beginning, there was......BIOS Before you go ANY further, go to the manufacturer product page for your specific motherboard model AND revision. Revision is an important aspect because for any given motherboard there may be more than one version of that model which will be identified...

forums.tomshardware.com

This is a good read on memory.

 

[TommyTwoTone66]

I figured it out, My BIOS was out of date and causing issues with my CPU's communication. Sometimes the complicated issues have simple answers, thanks for your responses everyone

How lucky!

 

[Unolocogringo]

Where do these percentages come from? Were you mixing ram in hundreds of systems personally to come up with those numbers? Or is that just like a "guesstimate" of how often mixing RAM works perfectly well, based on... gut feel?

Come on. How would it cause a crash? Memory modules are not complicated. If you have different modules from different manufacturers running at slightly different speeds and timings, the memory controller just automatically selects the best timing and highest speed which all modules can achieve. Why would that not work in more than 10% of cases?

Probably in the 125-150 system builds over the years at least.Very Low-Ball estimate. Never counted or kept track of all the systems I worked on/repaired/built/updated
Personal/ family/ friends/local businesses/local charities anyone who needed help.
I keep at least 2 folding computers running .Running 3 right now.
Have probably 20 of my favorite old computers in the basement now dating back to to my first Pentium 75mhz.
Most still boot and run but are useless relics. Would have more but wife decided she would "clean" the basement and threw away almost all of my older computers/parts.
Been learning and helping on toms hardware forums since 1998 (?) right after the "defphi forums fiasco.
Been a contributing member of toms hardware forums folding team since 2004.
Held the Mad Onion 3D mark record for 6 months with aP3 1.26 and a Ti 4200.
Just a few memorable moments.

As stated above these are from personal experience and may not translate to world wide numbers. But my data set is much higher than most people.
From personal experience and as forum member of Toms over the years those questimates of mine are probably not far from other peoples experiences with lots of upgrade experience.

 

[Karadjgne]

Come on. How would it cause a crash? Memory modules are not complicated.

Memory is very complicated, so much so that they require roughly 40ush different sets of timings and subtimings.

If you have different modules from different manufacturers running at slightly different speeds and timings, the memory controller just automatically selects the best timing and highest speed which all modules can achieve. Why would that not work in more than 10% of cases?

Ram is built by only a handful of OEMs. You'll find the exact same SkHynix chiplets in Corsair LPX, Kingston HyperX, Adata, Gskill, oloy, etc. So manufacture Branding is nothing more than a paint job on an aluminium heatsink and a sticker.

As to the memory controller, it automatically selects the loosest (worst) timings of the sticks, not the best. It also selects the slowest common default speed, not the highest all the sticks can achieve, that's the purpose of xmp or manual override.

Ram chiplets are unique. Every single one is punched from a sheet of spun silicon, and like a fingerprint, the chemical and mineral makeup of that particular punch out is microscopically different. It's all in the impurities, and those have a strong affect on the subtimings. Get 1 subtiming too far out of alignment and compatibility with another stick, you get no joy. Even if the same brand, same shelf, consecutive serial number, same size, speed, timings, color, heatsink, everything.

There's only 1 Guarantee when mixing ram, there are absolutely no guarantees.

 

[Arthianne]

Memory is very complicated, so much so that they require roughly 40ush different sets of timings and subtimings.

Ram is built by only a handful of OEMs. You'll find the exact same SkHynix chiplets in Corsair LPX, Kingston HyperX, Adata, Gskill, oloy, etc. So manufacture Branding is nothing more than a paint job on an aluminium heatsink and a sticker.

As to the memory controller, it automatically selects the loosest (worst) timings of the sticks, not the best. It also selects the slowest common default speed, not the highest all the sticks can achieve, that's the purpose of xmp or manual override.

Ram chiplets are unique. Every single one is punched from a sheet of spun silicon, and like a fingerprint, the chemical and mineral makeup of that particular punch out is microscopically different. It's all in the impurities, and those have a strong affect on the subtimings. Get 1 subtiming too far out of alignment and compatibility with another stick, you get no joy. Even if the same brand, same shelf, consecutive serial number, same size, speed, timings, color, heatsink, everything.

There's only 1 Guarantee when mixing ram, there are absolutely no guarantees.

So with this you're saying as long as /I use xmp there shouldnt be an issue with running all 32 gb? both sets passed the mem test

 

[Karadjgne]

Won't and shouldn't aren't synonymous. Hopefully you won't have an issue, but realistically it's a crap-shoot. The ram will either work great, work with some manual adjustments, or just plain not work. Xmp might work no worries, but just as often it requires you manually plug the values in and maybe add a little voltage to the ram or voltage to the memory controller in the cpu etc.

Only thing for sure is you will not find out until you try the sticks. Nobody can say for certain what Will happen, only what Could happen. It's the biggest reason why it's always recommended that you don't add ram, but purchase a full kit, in the size and speed you want. The factory has already tested the ram for compatibility and they Do have the choice of thousands of sticks to pair up. You get 1 shot, work - make it work - fail.