[SOLVED] New Ryzen build, Win 10 1903 nightmare

[dejectedpanda]

Hey,
I picked up a R5 1600 + MSI B450 Pro-Vdh Plus + 2x8GB HyperX RAM a few days back to replace my old 2500K system. Plugged it in and it booted right into Windows (1809), no reinstall needed. Ran my games and some benchmarks w/o any issues.

Then I decided to install a fresh copy of 1903 and its been nothing but freezes, reboots and BSODs ever since. So far I've got BSODs for rt640x64.sys, dpc watchdog violation and kmode exception not handed. I've installed all the latest drivers from my board's page but no luck.

Also, when I try creating a bootable USB with any previous version of Windows, Media Creation Tool fails at around 50% whereas Rufus outright crashes the PC or gives me this error. I've also got a bunch of USB device not recognized errors even though the drive in question works fine at other times.

Running sfc /scannow throws the "Windows Resource Protection found corrupt files but was unable to fix some of them" error. Even ran the DISM command but it didn't help.

Best part is I managed to install Steam and get my previously installed copy of Just Cause 3 working. Ran it for about half an hour, no problems. But use Chrome/Edge or other apps like Photos, and it randomly freezes/restarts when its not stuttering and lagging like hell.

What driver or fix do I have to use to solve this? Will going back to 1809 or even 1803 help? I am at my wits end trying to find a solution and I could really use some help.

PS: Even though its tied to my MS account, my Pro retail key doesn't work anymore when I try reactivating it using the Troubleshooter.

 

[Colif]

I not blaming you for what you got, I know it wasn't intentional

 

[Darkbreeze]

1.344 is NOT the same as 1.35v. If it shows 1.344 in the BIOS, on the ACTUAL setting, not the stock information listing you generally see on the right panel or in default views, then that's what it's at. I'd run it at 1.35v because that is the stock voltage setting for most DDR4 over 2666mhz. There are a few that might be higher. Anything up to 1.4v is technically within DDR4 spec but I'd try to stay as close to 1.35-1.36v as possible. Obviously, lower is better, but lower also ALWAYS means less stability, until you get to the other end of the spectrum and begin exceeding spec for that memory platform of course.

One, two, or three passes of Memtest86 mean nothing. Literally. It's four passes, period, with ZERO errors, or else it's not stable. And you have to make SURE that it is actually running at that speed before you run Memtest because if the system detects instability it will reset the actual RUNNING configuration in some cases, without changing what the settings look like when you go back and visit the BIOS again. I have seen this happen many times when testing with Memtest and thought it was all good, only to discover in HWinfo that the system had made changes to the primary or secondary timings that were WAYYYYYYY loose, to the point where any gains you might have seen from the frequency increase were long since lost.

So after changing the memory configuration, at any point, you want to run HWinfo (And ONLY HWinfo) and scroll down to the reporting section for the memory (Run "Sensors only" when you start up HWinfo. Disable the "Summary" option) and verify that the speed and timings which you believe it to be running at are actually what it is running at. It does you no good to run the memory at 3000mhz if the system has altered the timings so that you are running at 18-18-18-42, just as an example. Those won't be accurate numbers probably but it gives you an idea because you would probably want to see something more like 16-16-16-36 or 14-14-14-35 or something along those lines depending on the actual specs of the memory, which for the sticks you've indicated should actually be

Default:
DDR4-2133 CL14-14-14 @1.2V

XMP Profile #1:
DDR4-2400 CL15-15-15 @1.2V

And there is no way those sticks are able to do 3000mhz with timings that make it worth even trying because to start with they are not high end sticks using high end ICs. I'd bet money that the system is grossly loosening the timings in order to get it to run at that speed, which negates any actual gains because the latency penalty will likely be enormous. You can verify using HWinfo. Take a screenshot of the memory section and post it here.

The fact that you have a low end motherboard and CPU, compared to other offerings on that platform, makes it that much more unlikely that you are able to actually do what you have indicated is being done. No possible way on that motherboard, with that 1st gen Ryzen CPU, and those sticks, and that voltage, that you are running with an overclock that is 800mhz over the XMP configuration. Even further is the fact that the XMP profile for those sticks is a fully odd numbered primary timings which Ryzen particularly hates. Check the timings, post your screenshot, and in the end I'd probably still return them but let's see what's actually going on first because even if you have to drop them back to the XMP configuration (Which by the way, makes it a lot less likely they'd run without major penalties at the "just try it" speeds you indicated but WON'T run at a much lower XMP profile speed) and make a few changes to the timings, it might be worth it.

 

[dejectedpanda]

@Darkbreeze
Yes, I did run 4 passes of Memtest86. Got the message saying it completed successfully and the option to save the report to the USB drive.

Here's the HWInfo64 screenshot - https://i.postimg.cc/GbTLWvms/HyperX.jpg. As far as I can tell, its 3200 16-18-18 like it says in the BIOS, unless I'm missing something obvious. And if the system is indeed loosening timings without my knowledge, where else should I look to determine whether that's the case?

Is the odd numbered primary timings the reason the 2400 profile is giving me trouble whereas the even numbered 3200 preset is seemingly working so far? Also, like I asked before, do you think my board and CPU are an issue anymore? I'd like to know if you think its necessary to get them replaced as well.

Btw, thanks for all the help you've given. I appreciate all the detailed info you've provided.

 

[Darkbreeze]

Looks to me like the system is making a HUGE change in terms of secondary timings with a 75 trc and over 500 tRFC secondary timing, which is enormous and likely why they look to be running at a a reasonable speed and primary timings.

Hard to say without seeing what those timings look like at the stock XMP configuration, but I think those numbers should be more like 40-60 and 250-350 range for that frequency. We could definitely try to reset the XMP and then tweak the timings, but I think you are MUCH better off returning those sticks and perhaps simply buying some sticks that are supposed to run at 3200mhz, then you are trusting them to be stable at that kind of an overclock. Consider, that is a HUGE amount of difference. Average overclocking margins for memory for most sticks are usually only in the region of 200-400mhz and then cranking down a little tighter on the timings if you're lucky. Or just a bump in clock speed at the same timings, or just the XMP clock speed and tighter timings.

Something certainly seems off in terms of you being able to see that kind of gain and there not be a problem with stability.

Run the custom memory stability test using Prime95 as outlined in the guide to see what happens then. Use the latest version of Prime and be sure to disable both AVX and AVX2 in the options you see in the first window that pops up. When you disable AVX2, AVX option will stop being grayed out so you can disable that as well. Set the configuration for the test as mentioned in the test procedure I suggested above.

Memtest is not an accurate assessment of full stability. It is merely a good PRE-assessment to see if there are any glaring configuration problems. If you pass the Prime stability procedure I would highly recommend also running the extended Windows memory test as well.

 

[dejectedpanda]

Looks to me like the system is making a HUGE change in terms of secondary timings with a 75 trc and over 500 tRFC secondary timing, which is enormous and likely why they look to be running at a a reasonable speed and primary timings.

Hard to say without seeing what those timings look like at the stock XMP configuration, but I think those numbers should be more like 40-60 and 250-350 range for that frequency. We could definitely try to reset the XMP and then tweak the timings, but I think you are MUCH better off returning those sticks and perhaps simply buying some sticks that are supposed to run at 3200mhz, then you are trusting them to be stable at that kind of an overclock. Consider, that is a HUGE amount of difference. Average overclocking margins for memory for most sticks are usually only in the region of 200-400mhz and then cranking down a little tighter on the timings if you're lucky. Or just a bump in clock speed at the same timings, or just the XMP clock speed and tighter timings.

Something certainly seems off in terms of you being able to see that kind of gain and there not be a problem with stability.

Run the custom memory stability test using Prime95 as outlined in the guide to see what happens then. Use the latest version of Prime and be sure to disable both AVX and AVX2 in the options you see in the first window that pops up. When you disable AVX2, AVX option will stop being grayed out so you can disable that as well. Set the configuration for the test as mentioned in the test procedure I suggested above.

Memtest is not an accurate assessment of full stability. It is merely a good PRE-assessment to see if there are any glaring configuration problems. If you pass the Prime stability procedure I would highly recommend also running the extended Windows memory test as well.

The secondary timings with the 2400 XMP profile are 55 Trc and 421 Trfc. How much does the increase in secondary timings impact the performance gained from going to 3200 C16?

I'll run Prime95 as mentioned in your guide and get back to you with the results. As for the Windows Memory Diagnostic Tool, I ran the standard test but not the extended one. I'll run that as well.

 

[Darkbreeze]

You can learn a great deal about timings here at this link below. This is where I started out, and from there proceeded to ask questions of several of our own very knowledgeable members like Invalid error and Computronix. Both are disgustingly smart individuals, and I'm pretty sure they are both engineers with extensive architectural knowledge, so at least I was able to pick up some additional things from folks who really rather know their stuff pretty well.

This guide however is written by somebody who is pretty well versed as well, so reading it several times (I still pick up little things when I re-read it that I've missed previously) is a good idea. Bookmarking it and using it as a basic reference is almost a mandatory practice for anybody who wants a solid foundation for reference.

https://www.overclock.net/forum/18051-memory/1630388-comprehensive-memory-overclocking-guide.html

 

[Aeacus]

Looks to me like the system is making a HUGE change in terms of secondary timings with a 75 trc and over 500 tRFC secondary timing, which is enormous and likely why they look to be running at a a reasonable speed and primary timings.

Hard to say without seeing what those timings look like at the stock XMP configuration, but I think those numbers should be more like 40-60 and 250-350 range for that frequency.

Just for comparison, here are my HX430C15SB2K2/8 timings, running XMP 3000 Mhz profile (no manual tweaks),
image:

 

[Darkbreeze]

Major difference in Tras and Trfc, and of course, the odd numbered timings that Ryzen doesn't seem to like.

 

[ryzengamer]

There is nothing wrong in your hardware when updating windows makes your system unstable. Debian testing/Sid Xfce is the best OS, updates have been stable since year 2000.

 

[dejectedpanda]

@Darkbreeze
Thanks for the link. I've bookmarked it and will read it properly later. Anyway, I bumped Trc down to 60 and Trfc down to 450, and ran Prime95 with the mentioned settings for about 2 1/2 hours (power went out temporarily so had to shut it down, will run full 8 hours later) w/o any errors. Also, in your reply to Aeacus, you mentioned major difference in Tras but isn't lower better? Mine's at 36 while his is at 39.

@Aeacus
If its at all possible, could you tell me your secondary timings? TIA.

@ryzengamer
Is that Linux? I have no experience whatsoever with Linux unfortunately.

 

[Aeacus]

@Aeacus
If its at all possible, could you tell me your secondary timings? TIA.

Here's full data my BIOS has about advanced DRAM configuration. Didn't include primary timings since i shared those above.
Note: all values are set to "Auto",
image:

There is nothing wrong in your hardware when updating windows makes your system unstable. Debian testing/Sid Xfce is the best OS, updates have been stable since year 2000.

And on how your reply helps OP with his RAM issue?
Btw, choosing GNU/Linux distro is a personal choice. I, for example, don't like Debian at all since it's so barebones. For me, Linux Mint is best GNU/Linux distro. And on older hardware, Lubuntu. Besides the three mentioned, i've also used Ubuntu and Xandros.

 

[dejectedpanda]

Here's full data my BIOS has about advanced DRAM configuration. Didn't include primary timings since i shared those above.
Note: all values are set to "Auto",
image:

And on how your reply helps OP with his RAM issue?
Btw, choosing GNU/Linux distro is a personal choice. I, for example, don't like Debian at all since it's so barebones. For me, Linux Mint is best GNU/Linux distro. And on older hardware, Lubuntu. Besides the three mentioned, i've also used Ubuntu and Xandros.

Thank you very much.

 

[Darkbreeze]

You're right, I mentioned the wrong timing there. What we really need is somebody with a similar set of sticks, on the same platform, to chime in. In truth I'd like to see you set the XMP settings again and then work your way up from there.

I'd try setting the XMP profile with the primary timings set to 16-16-16-36-422 and don't touch any other timings except those five and then set the voltage to 1.35v. Then work your way up from there.

Truthfully, any memory that won't run at it's XMP values, especially after bumping the voltage up a bit, that were not a matched set to begin with, would go back right away while they still can. But if you're happy with the idea that they're running right, then not much I can argue against that.

Those Intel settings don't help much really, because the platforms and the memory configurations are just way too different.