Question Random Memory BSOD's with XMP Profile

[sze5003]

I'm using an Asus z370E board with corsair lpx
(CMK16GX4M2B3000C15W) memory. 4 sticks so 32Gb total.

I used to have 16gb and bought another 2 sticks 2 years ago.

With my 8700k enabling the XMP profile for 3000mhz has not given me any issues. In fact it has ran like this for 3-4 years now.

Recently I've swapped my 8700k for a 9900k (this past weekend actually). Enabling the XMP profile with the 9900k I thought everything was fine for a few hours, until I started getting random BSOD's like Memory_management and IRQL_Not_Less_or_Equal, etc.

So I disabled XMP and changed the frequency in bios to 2993 and that also would bsod after some time in windows. I have settled on 2666mhz and it ran fine for a day or two without any blue screens.

Is the cpu bad or just luck related to the memory controller?

 

[BonzaiDuck]

Is it possible to enter the BIOS, set the XMP profile to 3000, and check the monitored value of the VCCIO and SA voltages (second photo) after re-booting and re-entering BIOS?

The voltages for the 2666 speed seem to be what they should be. Also, we're only talking about a 333 Mhz increase to the spec OC voltage. I'm having my doubts that you have any problems with motherboard Auto-over-volting.

Somehow, I imagine you have a RAM stick on the fritz . . . . The only way to find out is to test 'em.

I've had Gigabyte boards and Intel boards. It's become like my choice of Seasonic PSUs: I don't bother looking at other brands, unless they're re-badged or a spin-off of the same company (like AsRock spawned of ASUS). I never find anything coming up short on ASUS boards, and yet I don't much incline to shelling out for "top-end". I may be tempted to spring for a Maximus or something in the future, but I'm happy as a pig in s*** with my Sabertooth, and expect the same from my WS workstation board.

Their RMA reliability is also stellar. Paradoxically, I blame myself for damaged boards, but they STILL honor the warranties, and they just replace the entire board. [I just avoided telling them the whole story -- or any of it -- about the windy, dry day and the vaping-pen in the USB port.]

 

[sze5003]

Is it possible to enter the BIOS, set the XMP profile to 3000, and check the monitored value of the VCCIO and SA voltages (second photo) after re-booting and re-entering BIOS?

Here's the voltages when I set the xmp profile for the 4 sticks. Shortly after rebooting I got a bsod but it happened so quick it didn't actually collect a dump file for viewing in blue screen viewer.

I guess I will need to find time to test the xmp profile for each ram stick or set. Sounds like a job for a weekend whenever I'll get a chance.

 

[BonzaiDuck]

1.23 and 1.24V are not excessive. They might have been, some several generations ago in the Intel Ark files, but then as I recall the threshold was 1.25..

Yeah -- that's it -- test 'em a kit at a time -- I would say. You KNOW which is which. That way, you could find the faulty module in as little as two tests, if you discover the faulty kit right away and then get the luck of the draw as to which module it is.

Filling all four slots of DDR can require more VCCIO and VCCSA voltage, but less likely at stock CPU clock and voltage settings.

If you nail it down to which kit it is, I would RMA the kit of both modules, or see if you can do that by inquiring through the RMA process or with tech-support prior to completing the RMA preparation.

I suspect your motherboard does a better job of automatically setting those voltages than mine does, for being a few generations ahead.

Oh! A little later here . . . . I read your comments with the pictures more closely. It may be that those voltages reflect the 2666 speed, if those are what you saw before you rebooted. But it looks as though you could be in a quandary if you can't at least boot into the BIOS after exiting with the changes, just to have a look. Your board could be doing what mine did. Even so, I didn't experience your symptoms and I quickly found the sweeet spot for the VCC -IO and SA. .

But one thing at a time.

Hopefully you can find out which kit it is. We'll suspend any suspicion that your used CPU has a damaged IMC for the moment. IF it's the RAM, everybody will actually be relieved. You should be able to RMA the defective module or kit.

 

[sze5003]

Would running HCI Memtest 64 on two sticks be enough, and how long would that take?

I think after work I can do some preliminary testing today. Pull 2 sticks, enable xmp, boot to windows and wait for the bsod since it seems to happen within minutes.

Maybe it's a combination of running 4 sticks and the processor's IMC that doesn't play well.

Hopefully I can figure out what it is sooner than later.

If not, I'm running them at 2666 until I upgrade to 12th gen.

 

[BonzaiDuck]

Would running HCI Memtest 64 on two sticks be enough, and how long would that take?

I think after work I can do some preliminary testing today. Pull 2 sticks, enable xmp, boot to windows and wait for the bsod since it seems to happen within minutes.

Maybe it's a combination of running 4 sticks and the processor's IMC that doesn't play well.

Hopefully I can figure out what it is sooner than later.

If not, I'm running them at 2666 until I upgrade to 12th gen.

That all makes sense. If the errors come hard and fast with BSODs, I'm only guessing it would show up with a 100% first pass with HCI-Memtest-64. Download the utility on another computer and burn the CD version to run on the system that BSODs. I assume that BSODs occur after booting to windows, so the program, bootable from CD, should run.

For those young enough that they have to work, I sympathize. I remember spending a weekend of my time troubleshooting an audio card, fitting in my personal hours to make my systems work right, and juggling all the balls. Being a computer enthusiast takes time. Most of the mainstreamers I know don't want to visit those prospects, so for them it's shopping at COSTCO and Best Buy, or calling Geek Squad.

 

[deustroop]

Would running HCI Memtest 64 on two sticks be enough, and how long would that take?

I think after work I can do some preliminary testing today. Pull 2 sticks, enable xmp, boot to windows and wait for the bsod since it seems to happen within minutes.

Maybe it's a combination of running 4 sticks and the processor's IMC that doesn't play well.

Hopefully I can figure out what it is sooner than later.

If not, I'm running them at 2666 until I upgrade to 12th gen.

I'm not clear on whether you have the latest bios ? Attempting to run the stick pairs separately at xmp is the next step as you say.If one stick is ID'd as the problem , will 3 run at 3000MHz ?
This may ultimately be a board problem and 2666 will have to suffice. A s I mentioned the difference is not noticeable except to us enthusiasts who want the product we was promised and nothing less. On that point recall that any speed over 2666 is over clocking .

 

[sze5003]

I'm not clear on whether you have the latest bios ? Attempting to run the stick pairs separately at xmp is the next step as you say.If one stick is ID'd as the problem , will 3 run at 3000MHz ?
This may ultimately be a board problem and 2666 will have to suffice. A s I mentioned the difference is not noticeable except to us enthusiasts who want the product we was promised and nothing less. On that point recall that any speed over 2666 is over clocking .

Yea I have the latest bios. I had to update it when I swapped out my 8700k for the 9900k. The 4 modules ran fine at XMP with the 8700k.

It's when I attempt to use xmp with the 9900k that causes the blue screens.

 

[solidsnake1298]

I'm a little late to this thread. Just in case you haven't fully solved this problem yet, I wanted to add my two cents.

I had a similar problem with my parent's computer. It had a 6700K and a 2x8GB 3000MT CL16 gskill kit. I tried increasing SA voltage, DRAM voltage, etc. But even at slower DRAM speeds it still had stability issues. Even 2133MT. It would pass RAM tests with flying colors with the XMP profile. For hours and hours the RAM tests would run without issue. And then randomly once or twice a week my parents would call or text me that the computer crashed while doing nothing in particular. There was no software or task commonality to the crashes, leading me to believe that it was hardware and not software.

Eventually I found that the problem was the command rate. For whatever reason, even though the kit was rated for 3000MT CL16 at CR1, I had to manually set the kit to CR2. Regardless of speed, it could not run at CR1 and be 100% stable.

 

[BonzaiDuck]

I'm a little late to this thread. Just in case you haven't fully solved this problem yet, I wanted to add my two cents.

I had a similar problem with my parent's computer. It had a 6700K and a 2x8GB 3000MT CL16 gskill kit. I tried increasing SA voltage, DRAM voltage, etc. But even at slower DRAM speeds it still had stability issues. Even 2133MT. It would pass RAM tests with flying colors with the XMP profile. For hours and hours the RAM tests would run without issue. And then randomly once or twice a week my parents would call or text me that the computer crashed while doing nothing in particular. There was no software or task commonality to the crashes, leading me to believe that it was hardware and not software.

Eventually I found that the problem was the command rate. For whatever reason, even though the kit was rated for 3000MT CL16 at CR1, I had to manually set the kit to CR2. Regardless of speed, it could not run at CR1 and be 100% stable.

Interesting . . . . My ASUS motherboards all default to CR = 2 when the CR setting is configured as "Auto". Generally, I've found that tweaking the command rate is easier than fiddling with the latency timings, but it still requires bumping up VCCIO a bit to make it work for stable operation. And you can only do this with a 2x RAM kit as opposed to use of all four slots.

I recently replaced my 4x8=32GB with a 2x16=32GB kit, but ran into trouble with CR=1. It may also have something to do with the density of RAM on a memory module as well as the number of slots used.

 

[sze5003]

I filled out a support ticket with Corsair. Wanted to see what they say. I still have not had time to test each kit separately so I have been running them at 2666mhz. I'm having issues with my seasonic PSU right now and the new GPU I got.

Turns out a bunch of people on the EVGA forums have the same issue where the older model PSU's have issues with power spikes and transient load spikes of these 30 series cards where it trips the OCP protection of the PSU.

Ordered a 1000w PSU and then I'll individually test the Ram once I get everything stable again.

 

[sze5003]

This is the reply I got from Corsair:

"We do not recommend mixing separate memory kits as it can cause stability issues. Memory with the same speed, size (GB per stick), timings, and voltage would net your best chance of not having a stability issue. Unfortunately, we cannot guarantee mixing any two kits, even if they have the same part number, for guaranteed stability.

We simply cannot guarantee stability when mixing kits. The memory controller on your older CPU handled it fine, but obviously, your 9900k can't adjust properly. Unfortunately, we would not be able to assist or replace either of your kits due to a compatibility issue."

So it seems because I bought the ram at different times even though they are the same model, they are not guaranteed to play nice together.

I would need to get a full 32gb kit it seems.

 

[sze5003]

Update: I ended up buying a pair of G skill ripjaws Ram- 32gb kit and can run them at 3200mhz using xmp profile.

I ended up selling 1 of the Corsair kits and will probably sell the other kit too. Turns out just because a kit is on the motherboard QVL list doesn't mean you can buy it twice, use it together, and have it work at advertised speed.