General / Off-Topic Stress test of CPU and RAM

[lysan]

So I maybe got a problem with my new PC, constant blue screen and crashes, I suspect A) Faulty CPU B) Faulty RAM or AB combined.
Did a stress test with Prime95 and it failed, so the conclusion would be CPU right, well I also tested the RAM with memtest86 and it failed too.

Now I don't know is it both or just one of them?? I still got warranty so my first idea was to just RMA both?

I only get a few information's from device manager so it's really difficult to pin point where the culprit is.

question what is AMD PSP 11 device?? i get and error from that one?

 

[Northpin]

Could be the power supply too
But i would go trying a different set of ram sticks first then try another PSU before thinking the CPU or the MB are bad

 

[lysan]

Could be the power supply too
But i would go trying a different set of ram sticks first then try another PSU before thinking the CPU or the MB are bad

I have tested the PSU all values are good, and it's a EVGA 1200 platinum brand new out of the box, so I doubt that could be the issue.

 

[Robert Maynard]

Are the BIOS settings relating to RAM voltage correct?

I had issues with an AM5 board and getting it to play nicely with 6,000MHz DDR5 (that needs 1.35V).

@Morbad will likely have some good advice on the topic.

 

[nemolomen]

Are the BIOS settings relating to RAM voltage correct?

This was also my first thought.

I had issues with an AM5 board and getting it to play nicely with 6,000MHz DDR5 (that needs 1.35V).

@Morbad will likely have some good advice on the topic.

OP, I typically use IntelBurnTest (you can select how much ram to use for the test) for stress testing.

Windows 10 or 11?

Do you have all of the correct BIOS RAM timings as well as the "best" drivers/BIOS extensions for your CPU, motherboard chipset(s), devices, etc...?

* The "best" driver is not always the latest if there is a more stable version for your specific hardware.

Also, a long-shot but sometimes works, have you tried removing the power cord, trying to turn the computer on several times, then plugging it back in and turning it on? I have seen this fix issues like described on more than a few occasions over the last 30 years where things just weren't working properly for no apparent reason. It basically drains all of the residual power which is stored in capacitors, inductors and other components and/or creating electric fields which could have an affect other components, and thereby, how the computer starts up when powered on.

 

[lysan]

Are the BIOS settings relating to RAM voltage correct?

I had issues with an AM5 board and getting it to play nicely with 6,000MHz DDR5 (that needs 1.35V).

@Morbad will likely have some good advice on the topic.

don't know will check it's an AM5 with a 7600X cpu and 32gb kingston fury ram

update:

it was set to a different value, I'm testing again and will return with the result.

 

[Robert Maynard]

don't know will check it's an AM5 with a 7600X cpu and 32gb kingston fury ram

update:

it was set to a different value, I'm testing again and will return with the result.

Don't assume that 1.35V is OK for your Ram!

 

[lysan]

Don't assume that 1.35V is OK for your Ram!

I picked it from the XML profile so i guess that is ok? however did a cpu torture test with prime95 and no errors now so that's good, now I will do the memtest86 and see if the result changed.

 

[Robert Maynard]

I picked it from the XML profile so i guess that is ok? however did a cpu torture test with prime95 and no errors now so that's good, now I will do the memtest86 and see if the result changed.

That'll be fine. Good to hear that one stability test has passed.

 

[lysan]

That'll be fine. Good to hear that one stability test has passed.

Looks like this solved the problem, the memtest has been running for 3 hours now and no errors, 10% more and it should be completed, if no errors pop up I will do some graphic test al call it a success, thanks mate that saved my day 🫡

 

[Morbad]

So I maybe got a problem with my new PC, constant blue screen and crashes, I suspect A) Faulty CPU B) Faulty RAM or AB combined.
Did a stress test with Prime95 and it failed, so the conclusion would be CPU right, well I also tested the RAM with memtest86 and it failed too.

Now I don't know is it both or just one of them?? I still got warranty so my first idea was to just RMA both?

If it's faulty hardware, it's most likely to be the memory. However, I would check for configuration problems before assuming hardware.

Are you using the EXPO profile or did you manually set the memory frequency/timings/voltages?

What motherboard is this again?

I only get a few information's from device manager so it's really difficult to pin point where the culprit is.

You'll want to check Event Viewer (administrative events) for any errors that occur around the times of the crashes.

question what is AMD PSP 11 device??

That's the AMD security processor. It's almost certainly not the cause of your issues, but it should have drivers installed. These are part of the chipset driver package, or can be installed separately/manually.

I had issues with an AM5 board and getting it to play nicely with 6,000MHz DDR5 (that needs 1.35V).

Not using sufficient memory or SoC voltage could definitely be the cause, but the EXPO profile to even set the memory to 6000MT/s, without manually configuring it, should set things accordingly.

Looks like this solved the problem, the memtest has been running for 3 hours now and no errors, 10% more and it should be completed, if no errors pop up I will do some graphic test al call it a success, thanks mate that saved my day 🫡

I'd give y-cruncher's component stress tester an overnight run.

Don't assume that 1.35V is OK for your Ram!

There are very few DDR5 modules/ICs that can't handle at least 1.35v. A lot of the tighter latency Hynix M/A-die stuff wants 1.35-1.45v at 6000MT/s.

What's a more concerning possibility is what SoC voltage EXPO will set, especially if the board's firmware hasn't been updated.

 

[lysan]

I'd give y-cruncher's component stress tester an overnight run.

Great thanks! is that one better than OCCT?

That's the AMD security processor. It's almost certainly not the cause of your issues, but it should have drivers installed. These are part of the chipset driver package, or can be installed separately/manually.

I've installed the drivers but it won't run! A hardware defect?

UPDATE: I updated my bios and that fixed it

 

[Morbad]

Great thanks! is that one better than OCCT?

It's different. No single test does everything, though some try. I usually wind up using about a dozen different tests and test combinations to cover as many bases as possible, so that I run into as few surprises as possible under real-world use.

y-cruncher tends to hit the Fabric and memory (the FFT test in particular is extremely good at finding bad memory settings) a bit more reliably than OCCT, but OCCT is easier to use for per-core CPU testing. Tests that cycle through individual cores are important because modern CPUs will essentially never reach maximum clock speeds under all-core loads, and some errors will only crop up at those peak clocks.

I've installed the drivers but it won't run! A hardware defect?

View attachment 392045
View attachment 392047

UPDATE: I updated my bios and that fixed it

First time I've seen the PSP device cause an error like that, but old firmware could certainly do it. With as many security and stability issues as are found post-release, it's important to check the firmware versions on any new board and update as needed.

For example, AM5 had a pretty critical issue with board makers setting extreme SoC voltages to make sure borderline EXPO profiles were stable. It was later discovered that SoC voltages past 1.3v could damage the IOD to the point of catastrophic failure, where degraded traces/vias would get hot enough to blister the CPU substrate and reflow the solder under the heatspreader, often killing the CPU and the motherboard in the process. AMD had to step in with a voltage cap built into their AGESA to (mostly) assure sane values for EXPO memory.

And, evidently, the security processor part of the SMU firmware needs to be updated to even work with the current drivers.

 

[lysan]

Need to undervolt my cpu, because 91 C is a bit too hot for my taste during the stress test if I need to run it for a very long time, however i ran it for 2 hours and no issues, and some of the problems I had before is now gone.

 

[Morbad]

Modern CPUs will boost until they hit some kind of limiter, usually thermal, and just sit at that temperature while varying clocks and voltage to keep from exceeding it.

On Zen 3 and 4 you can use Curve Optimizer to adjust the frequency/voltage curves, which tends to boost performance first, then as max boost clocks are reached, start reducing temperatures as larger negative CO offsets are applied. Just targeting voltage directly (with a negative vcore offset, for example) will almost always harm performance, as will simply reducing the thermal limit.

These parts are expected to be able to run indefinitely at 95C and won't be harmed by such temps, as long as all other relevant limiters are in place. Of course, if one wants to use more voltage elsewhere or increase current limits, temp limit needs to come down to ensure the same reliability/longevity.

 

[lysan]

Modern CPUs will boost until they hit some kind of limiter, usually thermal, and just sit at that temperature while varying clocks and voltage to keep from exceeding it.

On Zen 3 and 4 you can use Curve Optimizer to adjust the frequency/voltage curves, which tends to boost performance first, then as max boost clocks are reached, start reducing temperatures as larger negative CO offsets are applied. Just targeting voltage directly (with a negative vcore offset, for example) will almost always harm performance, as will simply reducing the thermal limit.

These parts are expected to be able to run indefinitely at 95C and won't be harmed by such temps, as long as all other relevant limiters are in place. Of course, if one wants to use more voltage elsewhere or increase current limits, temp limit needs to come down to ensure the same reliability/longevity.

More like I already live in a hot place, having my PC acting like a heater is not fun unless it's a windy day and i can open the window

 

[Morbad]

More like I already live in a hot place, having my PC acting like a heater is not fun unless it's a windy day and i can open the window

The 7600X is not a power hungry part; there is maybe 10 watts to save through tuning.. These parts reach high temperatures because of their thermal density (each core is less than four square millimeters), not because they produce a ton of heat.

If you want to reduce the amount of heat the system is dumping into the room, tuning the GPU will be the main place to do that.

 

[lysan]

The 7600X is not a power hungry part; there is maybe 10 watts to save through tuning.. These parts reach high temperatures because of their thermal density (each core is less than four square millimeters), not because they produce a ton of heat.

If you want to reduce the amount of heat the system is dumping into the room, tuning the GPU will be the main place to do that.

Yeah I just did that, it went from 90C to 70 by setting the voltage to 1150 mv

 

[Navigare Necesse Est]

The first thing I would check is power issues.