Brand new 9950x3d crashing after a few minutes of AIDA64 CPU + FPU stress test

[redzo]

Hello Everyone,

Could a 9950x3D owner confirm if aida64 stress test with only CPU+FPU checked freezes the entire system after a few minutes of runtime ? This is on stock. No OC.

I have Aida64 at the latest non beta version. I have never enabled PBO on this system. Everything is on stock, except for the RAM. I had the RAM initially at the expo profile, but I reverted both modules to the standard jedec settings and redone testing with only one module installed at a time: same results. Aida crashes when I run the cpu+fpu stress test and freezes the system. Didn't manage to run memtest86 so far, I will do this next, but I don't think that both memory modules could be defective, small chance.

So far I have ruled out:
- the OS install.
- BIOS version: I have downgraded to the first release that supports the new x3d chips and got the same results.
- I uninstalled the AMD drivers and got the same results with them uninstalled.
- CPU cooling / high temperatures.

This is a brand new retail box 9950x3d, with new motherboard and RAM. I found other people reporting exactly the same issue on reddit. They have different motherboards and memory than me. Some got replacement CPUs with brand new ones, but no luck: same issue. Unfortunately, I do not own any other AM5 CPU to test.

I can't tell for sure if this is a firmware agesa bug with the new x3d cpus, AMD driver bug or defective CPU.

Could someone with a known working 9950x3d(PBO set on auto, everything on stock), W11(with the latest updates) with latest AMD drivers installed confirm if aida is suppose to crash like this at stock settings on a 9950x3d?

Thank you very much !

 

[thesmokingman]

What are your memory settings? Default, or other?

 

[ir_cow]

Disable XMP/EXPO and try again. I was having issues with the 9800X3D until the newest BIOS that came out last week. Mostly related to failure to boot, but also memory problems.

 

[redzo]

I am currently on EXPO profile which is 6000mt, c30, 1.4v as I will test the ram and I want to increase the chance for errors.

But it doesn't seem to matter. I have ran them individualy(one at atime in single channel) at the jedec settings and I encountered the same issue.

What are your memory settings? Default, or other?

 

[GerKNG]

runs fine on my side with 64GB of RAM (DDR5 6000 CL30)

 

[redzo]

Disable XMP/EXPO and try again. I was having issues with the 9800X3D until the newest BIOS that came out last week. Mostly related to failure to boot, but also memory problems.

The other people that I found reporting this on reddit are only for the 9950x3d
I am afraid that we may have a defective batch, but then again, this cpu is too new and driver / firmware bugs can't be ruled out

runs fine on my side with 64GB of RAM (DDR5 6000 CL30)
View attachment 391505

Thank you ! After the memtest completes I will probably end up returning the CPU

 

[LabRat 891]

Looks like others are already on top of the RAM angle.

Longshot, probably unrelated, but...
if it were any other new max core'd high-TDP CPU, I'd look @ power stability.
-Pretty much everything these days can suffer from brief heavy-current transients, and consequently brief periods of severe vDroop if the circuit cannot supply that current.

Are all the CPU power leads connected to the motherboard? Most boards these days have at least an 8/4+4 *and* a 4-pin or 8-pin.

 

[redzo]

Looks like others are already on top of the RAM angle.

Longshot, probably unrelated, but...
if it were any other new max core'd high-TDP CPU, I'd look @ power stability.
-Pretty much everything these days can suffer from brief heavy-current transients, and consequently brief periods of severe vDroop if the circuit cannot supply that current.

Are all the CPU power leads connected to the motherboard? Most boards these days have at least an 8/4+4 *and* a 4-pin or 8-pin.

Besides the 2 cpu power connectors which I plugged in, the motherboard has an additonal 3rd power connector(pcie) located at the bottom of the board. As far as I know, that one is for power to other pcie extension boards, but I do not have any installed. I did not plug in any pcie connector to that one. All I have is 2 x m.2 ssds plugged into the pcie 5.0 m.2 slots.

 

[LabRat 891]

Besides the 2 cpu power connectors which I plugged in, the motherboard has an additonal 3rd power connector(pcie) located at the bottom of the board. As far as I know, that one is for power to other pcie extension boards, but I do not have any installed. I did not plug in any pcie connector to that one. All I have is 2 x m.2 ssds plugged into the pcie 5.0 m.2 slots.

Cool.
Then you've already addressed what I was directly-referring to.

 

[GerKNG]

what i would do with a problem like yours. (coincidentally i just did that with a PC last week with a freezing ryzen 5 system)
completely take your PC apart incl GPU, RAM and CPU out of the socket, wiping it with IPA, check the socket, unplug the 24pin and remove the cmos battery and drain it completely for a full CMOS reset and not just triggering a "go to default" with the clr rtc jumper as long as it has power through the 24 Pin.
put everything back together and just enable EXPO and save/exit.

 

[kckim]

Recently, I built a system with the 9950X3D, ASUS TUF B850 Plus WiFi, and G.Skill DDR6000 CL30 and windows 11 24h2. While there were no issues in system check programs such as Prime95 and OCCT, a problem occurred during the AIDA64 stress test.

As soon as the test starts, within about 5 seconds, sensor reporting stops. The CPU load does not properly reach 100% either. If the test continues, the system experiences freezing, becomes extremely slow, or reboots.(When EXPO is not applied, there are no issues. I requested a RAM replacement from the manufacturer, but the problem remained the same.)

I have found issues with ASUS motherboards and the 9950X3D on several communities. Although the motherboard models were different, the symptoms were the same. The solution is to use version 23H2 or lower.

 

[A Computer Guy]

What is your voltages? Run ZenTimings get screenshot with and without expo.

for example

 

[redzo]

Recently, I built a system with the 9950X3D, ASUS TUF B850 Plus WiFi, and G.Skill DDR6000 CL30 and windows 11 24h2. While there were no issues in system check programs such as Prime95 and OCCT, a problem occurred during the AIDA64 stress test.

As soon as the test starts, within about 5 seconds, sensor reporting stops. The CPU load does not properly reach 100% either. If the test continues, the system experiences freezing, becomes extremely slow, or reboots.(When EXPO is not applied, there are no issues. I requested a RAM replacement from the manufacturer, but the problem remained the same.)

I have found issues with ASUS motherboards and the 9950X3D on several communities. Although the motherboard models were different, the symptoms were the same. The solution is to use version 23H2 or lower.

I eventually concluded that there was an issue with aida running on Windows 11.

I am now on the same build and the issue went away by itself: no more aida locking the entire system when running that test. But to be fair: windows received updates and I am always on the latest firmware BIOS build. There is nothing that would point out to anything else wrong: the system is very stable and I've been using it ever since I've started this post here. This issue with aida was the only thing giving me the creepers about keeping this CPU.
Except for the screen turning on an off randomly which seems to have been fixed by the latest nvidia driver, everything else is fine.

What is your voltages? Run ZenTimings get screenshot with and without expo.

for example

View attachment 393046

Sorry for the late response.
I did not mess up with memory timings as I thought that there was no reason to do it.
I've let memtest run over one weekend for about 20 hours and no errors were detected. After that I've completely ruled out memory issues.

 

[overclockedamd]

Hmm i read about this a while back on overclock.net the overall consensus there was they have no idea why it does it other than to aggressive of a -CO which was stable 99.99% of the time other than AIDA64 CPU + FPU. Some were bone stock and had the same issue. So that doesn't make much sense. I would not be surprised if it is simply a fringe case issue where it is having issues with core parking and the 3d cache jumping around and AIDA64 not being coded properly. If they can run literally any test for days and have no issues BUT AIDA64 crashes almost instantly to a hard lock there is a common theme AIDA64. Was there a AIDA update since you retested this?

*edit* Sorry am at work didnt have time to read the whole post. I see you already drew that conclusion. Move along nothing to see here.

 

[alexbirdie]

Sorry to say, but I am happy with my 9950x3-build. I am using windows 10, and during testing, when I pushed the CO-values too hard( lower than -25), then I experienced crashes with AIDA, although cinebench ( different versions) succeeded. In my signature you can see my ( and my son's, which is identical to my build except the used AIO-cooler) build.