Question Possible HyperX RAM issue with AMD/X470-F mobo rig

[30Forte]

Hello everyone! I've been following threads and reading guides on Tom's for some time now, but this is the first thread I post.
I finished building my PC about 4 days ago - I followed all the online tutorials I could and went through all the installation guides to make sure I did it right. The specs are as follow:

AMD Ryzen 7 2700X CPU
Asus ROG Strix X470-F motherboard
16GB (2x8GB) HyperX 3466GHz CL19 RAM (purchased in the same blister)
Gigabyte RTX 2070 8GB Gaming OC VGA
Corsair TX750M PSU (80-plus gold)
Crucial BX500 SSD

After assembling the set I updated the mobo BIOS to the latest version, run the BIOS/UEFI Wizard to set the mobo parameters for gaming, then installed the OS (Win 10 64bit) using a bootable USB drive withou any issues at all.

I've been using it on a daily basis, 3-4 hours per day since then, and for three times it crashed on me:

1st time: while playing Assetto Corsa the video went off (monitor couldn't find the video source) but apparently the computer didn't reboot or shut down - the mobo and stock CPU cooler LEDs remained alight. Power or reset button were useless, so I switched the PSU off for a restart. There were no error messages on post screen or in the OS. I should have checked the LEDs in the mobo that indicate where the issue could have happened but I didn't...

I thought it could be the video board as it was running quite hot (70-80 degC) so I changed the profile in the utility OC software to something more conservative (changed the fan RPM mapping to higher speeds and set the target temp to 70), which kept it under 70 degC.

Then, when starting Assetto Corsa again the next day I noticed the mobo amber LED was on (indicating DRAM fault) and went on to restart the system, which made the LED turn off and don't come back again.

2nd time: it happened just as the 1st time, one day after, except there was no game running. This time I looked at the mobo issue-indicating LEDs and again the amber one was on. Turned off the rig again by using the PSU switch and entered the BIOS/UEFI, then decided to check the RAM settings and enable the DOCP profile to set the RAM to its advertised speed of 3466MHz (until then the BIOS and the monitoring software indicated it as working @ 2933 MHz, CAS latency of 16).

3rd time: it happened not much time after the 2nd crash. When playing another game the computer rebooted out of nowhere, which made me think the DOCP setting had some influence on it. I changed it to the DOCP conservative 2933MHz profile and had no crash since then. But I used the PC for only 4-5 hours after this, so I'm still wary of its stability.

The RAM is in the mobo QVL, which I used as reference for the purchase, but oddly it's not among those Kingston recommends for AMD-based rigs (https://media.kingston.com/pdfs/hx-product-memory-ddr4-amd-ryzen-and-hx-compatibility-us.pdf). It's probably not a B-Die type either.

So the question would be if you guys think this could somehow be related to something else than the RAM and if there would be something I could do to try to get it solved once and for all. I read the troubleshooting sticky post, from which I think maybe a one-step up in RAM voltage could help stabilize it (followed by the recommended testing), but I'm also considering resetting the BIOSand updating it again to make sure it's not corrupted and to get everything back to default. These RAM modules weren't my first choice, but where I live it's become impossible to find a G.Skill Trident-Z 3200MHz CL14 set of modules, which would have been my first option (however, if these HyperX modules prove to be somehow incompatible I will try to get the G.Skill ones in a trip to Canada next month).

 

[Roland Of Gilead]

Test the mem running with memtest86+ and run it for 4 passes. If it fails it's the ram. Also try testing with just one stick at a time. See if it's just one faulty DIMM. You can also try setting the timings manually instead of XMP/DOCP. You can set the timings and voltage from within the bios. Do these things first, before ordering new DIMMS.

 

[TechyInAZ]

It could very well be your memory. 3466MHz is not easy to run on a lot of Ryzen chips. Which is why Ryzen is better off with 3200MHz at tighter timings like CL14 and CL16. (And 3466mhz at CL19 I can bet you is slower than 3200MHz at CL14, cause timings are super important aswell.)

Though if the memory is on the QVL list, it should be fine.

I think your RAM is unstable, download memtest86, put it on a thumb drive, boot to it and run the default config. Takes 4 hours. (run this at your rated 3466mhz CL19 DOCP speeds)

If that fails, yeah you got a memory problem for sure.

Also make sure you are on the latest bios before running memtest86. BIOS updates for Ryzen can mean the difference between stable and non stable memory.

 

[30Forte]

Thank you very much for the fast replies! keith12 and TechyInAZ, I will be resetting the BIOS and updating it again to the latest version and I will leave the RAM test running overnight on current DOCP 2933 MHz setting. If it all goes well (even then, could there still be a problem with the RAM that went unchecked?) I will try DOCP 3466 MHz and perhaps a slight increase in RAM voltage, then test again.

TechyInAZ, I also felt the same way about getting this set of 3466mhz at CL19, as I mentioned I wish I could have gotten the GSkill CL14 B-Dies. I actually might purchase them in the near future and sell the HyperX, if they're ok and the performance boost is worth it.

 

[Calvin7]

You changed it to the DOCP conservative 2933MHz profile and had no crash since then. Higher memory frequency speeds over 2933MHz are for overclocking. Having crashes from failed overclocking means you need to work on your overclocking. You can also have problems by not overclocking the CPU while setting exotic memory speeds.

 

[30Forte]

The third crash, the one that came quick after a change to settings, happened with the RAM at 3466 MHz DOCP, but before that the memory was always at 2933 MHz. If I'm trying the 3466 MHz again, I will probably also raise the RAM voltage to try to gain some stability.

 

[TechyInAZ]

Ok, so now we know it is your memory controller that is not happy.

Your RAM should be perfectly fine, it is rated FOR that DOCP profile. (so leave RAM at it's rated voltage which is either 1.35v or 1.4v)

Try increasing your SOC voltage to 1.18v and test again.

if that doesn't work, try going back to that 2933 CAS16 and running it.

 

[30Forte]

When I turned on the PC yesterday, as soon as I got into the OS it crashed again, same way as before. Then I reset the BIOS and updated it again to the latest version. I've had no issues since (about 5 hours of use).

I did perform the MemTest86 check from a thumb drive with default settings and 4 passes, no errors were found. I'm inclined to believe the crashes resulted from the tweaking done by my mobo OC Wizard and then through the DOCP profiles.

This RAM is actually rated for 1.20 V: https://www.tweaktown.com/reviews/8...6gb-dual-channel-memory-kit-review/index.html

I'm thinking of keeping it in these standard 2933 MHz settings for a bit longer to make sure it won't crash on me again (does that testing method make sense or the memtest86 already said it all?)... I'd like to make sure it is stable in the standard setting given by the mobo before trying to run the more aggressive 3466 MHz DOCP again (and when I do that, I might go ahead and try the new SOC voltage, though 1.20v seems not enough for them to achieve 3466 MHz).

 

[TechyInAZ]

1.25v? Woah that is weird. I've never seen anything beyond 3200mhz at such a low voltage.

Yes good idea 30Forte, run memtest86 at that 2933mhz. If that fails, then you definitely should try running the memory at default JDEC spec (2133mhz) and see if it works.

1.25v at 3466, man that is a really weird kit. I'd personally try getting some 3200mhz 14 or 3200mhz 16.

 

[30Forte]

Yeah, I got future plans for a G.Skill 3200 CL14. I've been running until now the modules at 2933, default BIOS setting, no more crashes. Maybe because of this odd 1.25v setting it is not the most adequate kit to equip my rig. It seems that even though it figures on the mobo's QVL, it's not the best option for it's hard to remain stable with DOCP profiles.

 

[Roland Of Gilead]

Agree with TechinAZ. 1.25 is very low for a high clocked DIMM. Normally for 3200 or above you are looking at 1.35v for the mem. Also, SOC at 1.18 is pretty low too. At default my SOC is 1.2, and I run it at 1.21 for my DIMMS OC.