Having game freezes, cpu stress test FATAL ERRORS on my brand new modern AMD BUILD

CRYPT

CRYPT

CRYPT

Associate
Joined
22 Feb 2023
Posts
6
Location
London, Grays
Hello there folks, I decided to buy new desktop build, because since 2013 i was on laptops since i came to Englad to find life opportunities ^^ My Alienware totally or almost died at 2016 and since 2016 to nowadays i was working/playing etc on my MSI laptop GT73 VR... so long 7 years cpu dunno what, but 1080 gpu and i can still play all of games released nowadays ^^, but it's not that topic.... Because i was lucky i could buy my dreamy super tower desktop Cooler Master C700M (used one but with reasonable price) case and it was one of the main reasons why i decided to build some proper gaming build... I decided to hear answers to buy build from experienced folks and bought altogether this build...
Gigabyte B650 Gaming X AX (Socket AM5) DDR5 ATX Motherboard
https://www.overclockers.co.uk/giga...ocket-am5-ddr5-atx-motherboard-mb-5bf-gi.html
240 liber = 7100kc

AMD Ryzen 9 7900X
https://www.overclockers.co.uk/amd-...hz-socket-am5-processor-retail-cp-3dh-am.html
440 liber = 13000kc

Kingston KC3000 PCIe 4.0 NVMe M.2 SSD
170 liber = 5000kc

Seasonic Focus PX-850 850W 80+ Platinum Modular Power Supply
https://www.overclockers.co.uk/seas...-platinum-modular-power-supply-ca-06p-ss.html
180 liber = 5300kc - Tady je ale problem ze to je jako preorder, na amazonu je posledni kus ale o cca 600kcz drazsi

Sapphire Radeon RX 7900 XTX Nitro+ Vapor-X 24GB GDDR6 PCI-Express Graphics Card
https://www.overclockers.co.uk/sapp...ddr6-pci-express-graphics-card-gx-3a4-sp.html
1200 liber = 35 500kc

Kingston Fury Beast Black 32GB (2x16GB) DDR5 6000 CL36
173 liber = 5100kc

Silentium Fortis 5
60 liber = 1800kc

Thermal Grizzly Kryonaut
10 liber = 300kc

Sorry for this copy/paste post - i am just too lazy reedit, so posting it as i wrote it to my 'czech' forum

--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Now about my problem....

Since the very first start my system was booted to the bios which i had to reset, because there was some error due to decryption or dunno what (Windows wasn't installed yet), i googled it and there was written as a brand new desktop build, simply reset bios... So i did it... When i powered on my desktop it was booted normally to bios and i was there. Temperatures, hdw check was correct, everything set to default but unfortunately from 30 secs to 2 minutes the signal to monitor went off and my keyboard switched off too... Cpu Cooler was spinning, same for GPU... Power Button on case doesn't work so what i had to do was turn off PSU and get it back... Friend of mine told me check BIOS version and update it if anything, so i did. It worked for one day.... I tested 4 games, but not for vely long - WoW which i play mostly, but only for about hour, Firewatch for about 30 minutes - these games seemed ok, but I installed God of War and after aprox 5 minutes game freezed during cutscene and i had to reset whole PC as it was unresponsible... Another game i tried it was Kingdom Come Deliverance - same problem, played for about hour and game freezed and had to switch PSU and get it back.... (all of those games were set on ULTRA - checked temperatures and clocks - everything great, CPU about 60 C, GPU 60-65 C). The worst is, that my build frozen while simply surfing and using Google Chrome... That was rly bad.... Besides that i noticed because my motherboard has 4 lights flashing when something ain't right (GPU,CPU,RAM,BOOT), that my CPU was flashing red... I decided to test a few things... First of all i removed CPU cooler, removed CPU from socket, everything cleaned it up, put cpu again in to MOBO, do litle movement just for proper sit cpu to MOBO, repaste, and.... It worked for aprox one day, meanwhile i tested my memory modeles separately and always in another slot - 100% succes, so ram modules seems very ok. Btw after i removed and get back cpu back my MOBO doesn't indicate anymore CPU LIGHT error... So what i did afterward... Did CPU stress test via Prime95 - FATAL ERROR on my 12th core (check image)



Then i got more ERRORS while testing another time on more cores etc etc etc....

p.s. forgot to write here my temperatures while stress testing cpu were stable at 95 C for more than 3,5 hours,m but i as have seen 90-95 should be ok for this new ryzen gen

Edit:// Just got a message from Repair Center - posting it here


It seems the PC is fine as it is a new AMD architecture platform.
Unfortunately, all users will have some problems due to DDR5, memory compatibility, or overheating CPU like these ryzen processors are designed to operate over 90 degrees plus constantly.
The GPU on its own draws over 420W plus CPU , all fans might exceed your 850 W rating. The Power supply you have may not be enough.
Our technician got a system with 7950x and uses a dual tower air cooler and his temperature is lower than the test we did on your PC.
We changed the thermal paste on your processor, but there was no improvement due to the single fan and single tower air cooling.
On ryzen 7000 CPU is recommended to undervolt the CPU socket so you get a lot lower temps and more.

Recommendations: a power supply of 1000W+ and getting a 360mm.aio or at least dual tower air cooling with a dual fan for a push-pull configuration. Right now the PC got a single fan and the temperatures are exceeding 95 degrees result of the tests.
 
Last edited by a moderator:
Hi and welcome to the forums. :)

First off a bit of friendly advice. Please do not post links to other sites that are competitors of OCUK. This forum is owned and supported by the shop and links to competitors are not allowed for obvious reasons.

I would tend to agree with the repair shop. Your cooler is rather optimistic for such a high end cpu and you would have been much better off with a AIO water cooler or at least a oversize dual tower heatsink. I wouldn't be happy with those temps and if it's hitting 95 degrees now it certainly isn't going to cope as the weather warms up. I feel that your psu while it is a very high quality one could be borderline for such power hungry high end components. Personally I would have gone with at least a 1000w unit for that build.

When you flashed the bios (I presume you are on version F3 now?) did you clear cmos afterwards and re-entered your settings in the bios? Is XMP set and running correctly with the correct voltage? Sometimes it can be better to set the speed/timings/voltage of your memory manually for stability. Also mentioned by the repair shop is that you could try undervolting the cpu and especially the gpu. Quite a few on here have had to do this on the new platforms to achieve manageable temps and keep power consumption in check. You can achieve big power savings by doing this while minimising any performance loss. Because the gpu will be drawing less power it will run cooler and may boost higher and for longer periods. My RTX3070 for example is undervolted and runs cooler as a result. It now boosts higher than it did before and also has a 600mhz overclock on the memory so is actually better than it was at stock settings.

One last tip, if you add all your items to the OCUK basket you will see a bit on the bottom of the basket that looks like this < >BBCODE. If you click on that and copy the text that pops up you can paste it on here and it will look like this:-

My basket at OcUK:

Total: £2,277.94 (includes delivery: £8.00)​

​
​
​
 
Last edited:
Hello and tx for reply

I have MOBO with version F3 which was released as a the lates one. Anyway, due to my calculation it's not physicaly possible that it would do 'weak' PSU as when i did stress test of cpu it cause only CPU to be very loaded, not rams, not gpu etc... Besides that testing showed like energy consumption of CPU about 175W and even if u put everything at the same time together, it's not possible by my opinion, that it would exceed PSU 850W.... 420W GPU + 180W CPU + something on ram, coolers mobo etc and it's not possible to get over 850W.... Not to mention i would never be propably able to test every hdw at the same time on 100% usage!

Aio Cooler - yes i bet it could be better choice than tower, but honestly as we all know thanks to folks like Gamer Nexus and others check out what they say
https://www.youtube.com/watch?v=nRaJXZMOMPU or https://www.youtube.com/watch?v=SIv3CpYVNo0&t=1262s
so for me as i have stable 95 C temperature doesn't matter what cooler i would change it for by my opinion

After BIOS update it just asked me if i want to reset bios after boot to default so i did, i haven't cleared cmos

------------------------------------------------------------------

on reddit some folks described something similiar

Reddit - Dive into anything

www.reddit.com www.reddit.com

Helped to him cpu replacement

-------------------------------------------------------------------

Right now i am highly thinking about support contact and send MOBO and CPU to OC UK for further testing and if faulty, would ask for replacement....
Tomorrow i am planning to undervolt GPU, put RAM modules to default 4800MHz and run test, but i highly doubt it will be success
 
Be careful with power ratings. They refer to continuous power. Both CPUs and GPUs can consume far more over very short periods due to surge current. The issue arises as to whether or not the PSU can cope with the surges. Unfortunately in the consumer space it is very difficult to find details on this. One brand of PSU might cope whereas another might not. The safe option is to use a higher continuous rating than you need in order to ensure that that the PSU can cope with the surges. That's why experienced gamers will generally advise a 1kW+ PSU for high end gaming systems - it is from experience.

Having said that if the OP's system has crashes or hangs in low power states then it is not a power or thermal issue. Based on the previous mobo CPU error LED and the error messages then it would seem that investigating the CPU further is a good idea. Are there any interesting entries on the Windows event log? Particularly WHEA errors.
 
My priority in your situation would just be to make it not crash, once you can achieve that, then you can identify the problem. To start with, assuming you have sorted out XMP/EXPO & DRAM voltage, then I'd force the CPU to run 'underclocked', fixed at the base clock (or with a power limit) and then cap the 7900 XTX's clock at the base clock. You could even try running just from the onboard graphics.

I'm not completely clued up on how Ryzen performs with Prime (overclockers would give you better advice), but I don't think core errors are supposed to happen (even if the CPU is overheating), so if it still errors after reducing the clock (or running with a power limit), I think I'd consider RMAing the CPU directly to AMD.

The most common problem I've seen so far with AM5 is memory configuration and voltages.
 
Could be a bad draw on the silicon lottery.

I’d try putting the chip into eco mode and seeing how it fares, did you mess with any voltage already?
I’d be tempted if eco doesn’t work to increase the voltage in very tiny increments.
 
therealdippy said:
Be careful with power ratings. They refer to continuous power. Both CPUs and GPUs can consume far more over very short periods due to surge current. The issue arises as to whether or not the PSU can cope with the surges. Unfortunately in the consumer space it is very difficult to find details on this. One brand of PSU might cope whereas another might not. The safe option is to use a higher continuous rating than you need in order to ensure that that the PSU can cope with the surges. That's why experienced gamers will generally advise a 1kW+ PSU for high end gaming systems - it is from experience.

Having said that if the OP's system has crashes or hangs in low power states then it is not a power or thermal issue. Based on the previous mobo CPU error LED and the error messages then it would seem that investigating the CPU further is a good idea. Are there any interesting entries on the Windows event log? Particularly WHEA errors.
Click to expand...
Just checked my event log and not a single WHEA error SINCE my first power on... Had only BSODs

and just today i tested for 2 minutes CPU via Prime95 blend test and i got FATAL errors on 7 cores....
 
Last edited:
I admit I don't know Prime 95 but my expectations are that it should not generate errors if the hardware is healthy and operating within spec (i.e. enough power and cooling). If you are running default BIOS settings, so no overclock or undervolting and you are experiencing BSOD then each needs to be investigated to source. Based on the info it is pointing to the CPU.

If that is a pcie gen 5 motherboard then it might be worth forcing it to gen 4 or gen 3 if possible to see if that helps. The signal integrity needed to achieve gen 5 without uncorrectable errors is much higher than for gen 4 and you can't guarantee that all components are actually as good as the manufacturers claim, especially in the consumer domain.
 
If your memory has been configured correctly and memtest86 (or similar) checks out (you do need to manually check the dram voltage has been set to what the manufacturer recommends), then I'd RMA the CPU to AMD after verifying with one more benching program that it is producing errors at stock settings.

Underclocking like suggested would help you to check what's happening.
 
Last edited:
First thoughts (and they're in no particular order)
  • That CPU cooler isn't good enough imo... it might be sitting at 95C but it could also be thermally throttling at that point... so ideally we need clock speeds from hwmonitor as well.
  • Is the cpu actually mounted in the socket properly, I've seen reports (and video's) where it's just slightly off in the socket. Are any of the pins in the socket actually bent or not 100% connecting...
  • I'd have gone bigger than 850W for the psu with your setup
  • Have you done a smart test on the ssd, while rare it's possible that could be dodgy... it seems to lock up when trying to read new data.

Having said that, I've seen so many issues with ddr5 compatibility and am5, it's actually put me off first gen am5....so I'd try removing any xmp/expo applied to the ram and see if it's stable at 'stock' ram speeds.

@Tetras - I've never used prime95 on AMD because it (iirc) hit something that wasn't great for the cpu, there is supposedly an AMD specific version now but no idea if it's been updated to AM5 though. Personally I test with real world stuff that I might actually end up doing like encoding/rendering (ie cinebench/handbrake etc). Ryzen master is another option if you can find the stress test bit.. it really needs a simple 'stress test' button lol
 
lsg1r said:
@Tetras - I've never used prime95 on AMD because it (iirc) hit something that wasn't great for the cpu, there is supposedly an AMD specific version now but no idea if it's been updated to AM5 though. Personally I test with real world stuff that I might actually end up doing like encoding/rendering (ie cinebench/handbrake etc). Ryzen master is another option if you can find the stress test bit.. it really needs a simple 'stress test' button lol
Click to expand...

I see. Do they report back if the CPU is producing errors? I know Intel has a diagnostic software, but I didn't think there was anything like that for AMD, unless Ryzen master has this function?

CRYPT said:
Currently testing some settings for last 2 days and it seems it's working. I will keep info updated later
Click to expand...

Which settings did you change?
 
Tetras said:
I see. Do they report back if the CPU is producing errors? I know Intel has a diagnostic software, but I didn't think there was anything like that for AMD, unless Ryzen master has this function?
Click to expand...
Honestly no idea on Prime95 reporting stuff.

Ryzen master is pretty limited in what it does, it can give 'real time' info about core speeds/boosts etc and useful for pbo/curves but it doesn't look like it does errors. It might show errors if there is something to report but (touch wood) I don't appear to have any obvious issues on my setup, but then mines still pretty much running stock.
 
You must log in or register to reply here.
Back
Top Bottom