3990x and 256gb ram unstable aida64

hey all. Having issues with aida64 memory test being unstable with my 3990x. All stock except pbo and xmp. P95 blend passes, as does all aida64 tests except the memory test. As soon as i load the memory test my pc freezes up. Even tried setting memory to 2933 and lower timings. It freezes up in 5 seconds.

Pc is otherwise seemigly stable, i do 24/7 heavy number crunching for work with no issues so far.

2933Mhz is high for 3990X with 256GB. I am currently running 2800Mhz.

I advice you to start at 2666Mhz and then increase.

What is your mobo?

1 Like

Yes, what was said here.

See the gskill video I did on their 256gb kit. It is an insane accomplishment that it’s stable @3600 with 8 sticks.

2933 is pushing it. Try 2933. What’s the ram voltage a d soc voltage? Mobo?

1 Like

About that video, is there a list of all the motherboards you tested and the maximum speed you achieved with 256GB?

I had really bad luck with XMP / DOCP on my 3900X. It seemed to have different problems depending on which BIOS update I was on. That was an ASUS x570 board.

I would try manual memory settings.

Its the gigabyte trx40 aorus master.

I actually found out that it is stable, the computer just occasionally lags to hell for a few minutes (looks like its frozen/crashed) but will recover after a few minutes. No errors detected by aida64, the test continues to run in the background, the screen/mouse just becomes periodically unresponsive.

My ram kit is the 256gb tridentz neo kit by g skill (hynix die) which supposedly was tested to run on a 3990x at xmp settings and marketed as xmp ready for ryzen.

My ram voltage is 1.35v, soc voltage 1.036v

Happy to see my 3990x is seemingly stable afterall with 256gb and 3200mhz.

Ive actually heard that the gigabyte trx40 aorus master is an excellent board for pushing ram, that combined with one of the g skill 256gb 3200mhz qvl kits maybe whats allowing me to have a stable setup.

1 Like

I have this exact same bahaviour in Aida64 with my 3950x and 2x32gb corsair kit, freezes for a minute or two and then continues the memtest. I think the system freezes because Aida64 is loading up all the ram with data for its stability test.

Just what was in the video. And keep in mind it’s board and CPU dependent.

1 Like

Hey Wendell thanks for responding, unfortunately my celebration may have been premature… I was curious what ram temps were you getting on 256gb of ram, because I’m having a hard time reigning in the temps during heavy use. All my sticks get up to 65-71c, especially with RGB leds on. I have to turn them off just to get them around 62-67c, because I noticed once the heat gets up I lose stability and the system powers off. note* I also get intermittent system power off during heavy workloads if I increase the ram voltage above 1.37.

On another note, I’d appreciate if you could let me know what asus set the xmp settings to? I have the gigabyte aorus master trx 40, and as you said in the video, only the main timings got changed, not the other voltages like soc, vddg, proc odt, etc. Also, can you confirm the 256gb kit is samsung b die? My kit which was originally rated for 3200mhz is hynix. I may have to upgrade to a TRX 40 Zenith EA and the Gskill 256gb 3600 if these issues persist…

Sorry for all the questions, we’re on the bleeding edge of tech here, so any help is appreciated!

It’s so odd, my setup was working just fine as a 3970x and 256gb of memory. Why would dropping in a 3990x affect the stability :frowning:

Does upping the soc voltage to like 1.2V improves stability for you?
keep the memory voltage at 1.35V.

256GB memory at 3600mhz on a Hynix kit is unlikely to happen on TR.

Hey all, my trx40 aorus master was rocksolid with my gskill 256gb 3200mhz kit when I had my 3970x. With how smooth things were, I decided to upgrade to a 3990x for work. Unfortunately, after the first week, I started noticing the PC would shutdown during heavy ram intensive loads at 3200mhz. I decided to downclock to 2666mhz and the pc will no longer shutdown, however, HCI memtest will detect errors.

What’s going on?

I would re-seat the CPU, you might not be making good contact on on all the pads. You still might not hit 3200 since you changed the CPU and thats part of the equation for if the memory can hit X clock.

1 Like

No real experience on my part here, but from what I’ve read and watched I’ve heard the same. The STRX40 socket isn’t very friendly and you have to make sure you aren’t over tightening when using the torx screwdriver included. You shouldn’t go past the first click. I also believe there is an order to which the screws should be tightened. You could also try resesting the RAM.

Okay I will try reseating the cpu… I was thinking that couldn’t be it since there’s no shutdown when I drop the ram frequency to 2666 or when I remove half the sticks. But will give it a shot

Do not assume that OC on RAM will work when switching CPUs, especially if the TDP or architecture changes. I know we are not getting to the 5GHz mark with DDR4, but anything above 2667 is going to be factory overclock.

Now to your errors - please take a note that frequency is less likely to cause errors in this range and it is probably the timings.

tl;dr Yes your patch from the controller in the CPU package to the sticks is not running well.
Re-seat the CPU, carefully tighten it enough on all sides and clear the CMOS to force relearning the modules.

EDIT: If you did not run memtest with the old CPU, do not assume you had no errors there. It could be the sticks or the MB. Issue could have just gone worse instead of appearing.

Thank you for responding. It is possible I had errors before with my old setup as well, but now it is way more noticeable when I slam the mobo/ram with 128 threads instead of 64.

I just put the 256gb ram into my second rig which has my old 3970x and a msi trx40 pro wifi mobo. All stock settings except xmp ram and 1.35v ddr. Ran memtest and got 4 errors on test 7.

I then tested my 3990x rig (haven’t yet reseated cpu) with aorus master trx40 and a 8x16 128gb kit I have laying around. So far no errors.

Seems like the 256gb ram may a culprit, however that doesn’t quite explain why it causes my 3990x rig to power off right away during memtest, while my 3970x can run memtest with only 4 errors. I plan to rma the 256gb, but Maybe I should still do the reseat on my 3990x?

What power supply are you using?

Some motherboards are better at providing clean power to ram than others, and sometimes you can compensate by using a higher quality power supply.

Evga 1200 p2

I’m starting to think reseating my 3990x won’t help. If it was improperly seated it shouldn’t have been able to pass memtest with my 128gb kit at 3200mhz.

The errors on my 256gb kit could be because the msi trx40 pro wifi and aorus trx40 are lower end trx40 boards, unlike the zenith extreme, which is what g skill qvl tested their kits on. Wendells videos also seem to confirm the zenith extreme being able to run higher clocked ram compared to other, cheaper trx40 boards.

I placed an order for a zenith extreme alpha and will be conducting tests with my 256gb kit and will post the results here…

Did you do a reset CMOS and removed the CMOS battery, waited a bit and put it back in? That will force re-training of the memory on the next bootup, maybe some timings are not as well trained with the other memory controller(in the 3990X).