Question Is my Ryzen 3800XT faulty, or is it something else? (Crashes under load)

Oct 5, 2020
21
1
15
0
Hey all. So recently I built a new PC in anticipation for the RTX 3080 launch. I ended up getting an FE by being in the right place at the right time -- in the middle of the night restock just happened to be on their page. Below is something I posted on Reddit and everyone is blaming the card, but I realize the place I posted is is a bit biased. I'm looking for help here because I'm going nuts. I thought I had it working with Core Performance Boost on Auto, but it seems like the only way I can get my PC to run without crashing is to turn the boosts off and set a fixed ratio. I think it's at 42.25 right now, giving me ~4.2GHz while under load for gaming. I know "Fortnite" the Devil's game was listed in this write-up, but I was trying games that all have RTX. It now crashes even in games without RTX on (I tried Avengers). Anyways, here's what I wrote. Specs included at the bottom:

Summary of Issue
I am at my wit's end here. I just put together this system and finally finished it with the 3080 just this past week after using a 1080. I've been playing around with certain RTX features and had originally thought this was some sort of graphics card issue. Basically if I am playing any game with ray tracing, I have eventually reached a point where it would hard crash my entire PC. I'd even have to reset the time when my PC booted again.
Moving along, I started wondering if it was a heating issue, but looking at the temps in my Task Manager and Core Temp show well below crashing temps. I do know that AMD systems sometimes report the temp incorrectly in there, but I'll have you know my hand test, it feels like it's running cold. I am using an ARCTIC Liquid Freezer II on it and my CPU temps appear to not even be going above 50C on load and voltage. GPU is also hovering around 71C and normal voltage (for what it's listed as (200-350W).

A few weeks ago I was getting extreme frame stutter in some of my games and I ended up turning off PBO (from Auto) in the BIOS and things were fine. This time I was wondering what I could even do in my BIOS menu to try and get things to work correctly. What I ended up finding was that there was another booster in the menu that was ALSO set to Auto. I shut that off and then restarted.

My initial findings were perhaps that the outlet I was using I found to not have a ground terminal that was actually functioning, so I moved it to one that did (using my power strip). After tinkering with my BIOS and whatnot I tried this and things seemed stable. I then was noticing my speeds weren't going above 3.9GHz in Core Temp so I ended up rolling back my settings in the BIOS and letting that second boost setting in the Advanced options (I honestly forget the name, Core Performance Boost?, but it's not PBO -- I shut that off). I ended up restarting and booting up Fortnite and then somewhere after dying, I full on crashed and restarted. This is leading me to believe it's not an outlet problem, but this stupid random clock boost problem. Upon restart, I was able to change the ratio to 42 and run things no problem without crashing with an idle temp below 40C even when playing the game.

I guess my problem here is that now I have to run the clock consistently at that instead of using the boosts to regulate when I need those speeds because if I do, then I am most certainly going to crash. It happens so suddenly I can't tell if it's just one core performing poorly or some arbitrary "crash" threshold it's hitting.
I am a bit of a laymen when it comes to OC and AMD, but since the actual STOCK options on the MOBO and processor are causing crashes, I've had to jump into the BIOS and play around.

Does anyone have any ideas on what I can do here? I'd like to not have to run my processor into the ground by locking it at a speed. It's not overheating or even reaching dangerous temps, but I'd rather have the power when I need it running games on High->Ultra. Control does the same thing if I turn off DLSS. I just think it's really weird that a Graphics problem is being hit by my CPU. I had thought I'd got one of the faulty cards from NVidia, but since I am able to change my clock speeds/boosts and fix the issue, I don't think that's it.


System Configuration
Motherboard: Gigabyte X570 Aorus Elite
CPU: AMD Ryzen 7 3800XT
PSU: EVGA 850W GA Gold
Display(s) (Make, Model and Type):
Acer Predator G-Sync 1440p 144Hz
Extended Dell G-Sync Monitor (model is irrelevant) 1440p 144Hz
System Memory: CORSAIR VENGEANCELPX16GB (2X 8GB) DDR4 3600(PC4-28800) C141.45V AMD OptimizedMemory
GPU (VBIOS): NVIDIA RTX 3080 Founder's Edition
OS (Version): Windows 10 x64 (Version: 19041.264)
Driver: GeForce Game Ready Driver 456.71
Applications: Fortnite or Control and Direct X 12
Background Apps: Discord, Spotify, Chrome, Core Temp, NetLimiter
Game Software Settings: RTX On (High), DLSS On (Performance)

Steps to Reproduce
Enable RTX and Ultra Settings in either Control or Fortnite
Play for a while
Crash entire PC after a bit.
 

Darkbreeze

Retired Mod
Corsair only lists TWO memory kits that are 2 x8GB for kits between the speeds of 3333mhz and 3800mhz on their compatibility validation tool the Corsair memory finder, for your motherboard model. Neither of them is 3600mhz and neither would seem to be your kit. What is the EXACT model of your memory kit so I can check it against the motherboard QVL. If they are not listed on the QVL and they are not listed as compatible by Corsair on their memory finder utility, then they are probably simply not compatible with your motherboard and of course incompatible memory can mean anything from simply "not listed AS compatible" all the way up to "won't even POST with them installed" or anything in between.
 
Oct 5, 2020
21
1
15
0
Would it help if I posted the link to where I got it? Corsair Vengeance. This should be CL14 RAM. You can tell I basically threw my money away with this build and with the trouble I've been having I will probably not do it like this again. Thanks for your reply.
 

Darkbreeze

Retired Mod
SO, that memory, CMK16GX4M2Z3600C14, despite being a CL14 kit, does not appear on the motherboard QVL nor does Corsair list it as being compatible with that motherboard. They do list it as being compatible with some OTHER X570 and B550 motherboards, but not with that one. So again, that doesn't necessarily mean that it can't be used with it, but it is almost always a pretty surefire indicator that if you are having issues they are, good chance, related.
 
Reactions: Turtle Rig
Oct 5, 2020
21
1
15
0
Interesting catch! I wish I knew to look at that instead of hastily throwing together this build. I wonder if it's worth trying to order RAM that's on this list. Any recommendations from that QVL list? You don't have to search, you've been helpful enough. The return window actually closes today on it. Woops! I can still file, likely before the end of the day. I wonder if this is a good idea.
 
Oct 5, 2020
21
1
15
0
I'm in a bit of predicament now since if I send this RAM back it has to be tomorrow and I need this computer for work. I am wondering if this is all worth it. This RAM was unfortunately almost 200$ and now I probably won't be able to sell it since almost no one is looking to spend 180$ on RAM. :(
 

Darkbreeze

Retired Mod
My recommendation, send the RAM back.

Get a kit that's validated through G.Skill. I've seen nothing but problems with the use of Vengeance memory on Ryzen systems, especially the Vengeance LPX kits. I've probably seen fifty to a hundred LPX kits in various threads on here, both CL14 and CL16 kits from 2933 all the way up to 3600mhz, that simply were problematic on a variety of different Ryzen platforms. If you google it, you see that there are a lot of people out there who simply couldn't ever get those kits to work right with Ryzen systems and I am talking about long after most of the original Ryzen memory compatibility issues for all three generations were mostly cleared up through subsequent BIOS updates.

These, would be a good choice, and are validated for your board. There are three pages of compatible memory kits listed here. Flare-X kits are specifically made for Ryzen, and the listed Trident Z and Ripjaws kits all have a really good track record of compatibility with all Ryzen platforms.


X570 Aorus Elite compatible G.Skill memory kits
 
Reactions: Turtle Rig
Oct 5, 2020
21
1
15
0
Thanks, I truly appreciate it. I am sending these back to Amazon tomorrow, but unfortunately I'll be out of RAM for the time being until I can actually get new RAM. I might head to Best Buy and pick up a compatible Corsair option (it's all they have locally for me) and just return it when I get my new shipment.

I'm also running GPU tests and CPU tests now to verify that it's neither of those giving me issues either (since the RTX 3080s have had crashing issues of their own). I have until the end of tomorrow to send out the Amazon RAM so I'm going to be testing my PC until then. Thanks for your help.
 

Darkbreeze

Retired Mod
Yes, DO keep in mind that there might very well be something other than the memory causing your issues, but without having KNOWN compatible memory modules installed it REALLY makes it hard to want to target something else knowing that it might simply be a secondary or tertiary memory timing that the motherboard simply does not like but cannot find a comfortable setting for that plays nice. Or anything along those lines. Could just not like the IC's (memory chips) but that would be improbable given that they are B-die sticks. I believe it is something else in the makeup of the LPX modules that Ryzen dislikes.

Graphics card might be my next target. It's also possible that for some motherboards, there might need to be a round of BIOS updates to fix compatibility or other issues on some boards. IDK, but it wouldn't FOR SURE be the first time something like that was necessary. We can investigate further if known compatible sticks don't cure the problem.
 
Oct 5, 2020
21
1
15
0
That's a fair assessment. I'm hoping this problem is temporary. I am happy I ended up getting G.Skill RAM as most reviews say it's really the best overall for a Ryzen system. I ended up going with +2CL on it, but I doubt I will notice much overall difference.

One of the more curious thing that has been happening in this build however has been that if I lock my CPU speeds, I've not been getting any problems. I know that PBO has caused issues in games like Destiny 2 where I would get some strange stuttering that would freeze for 2-10 seconds at a time and shutting that off fixed it, and then turning off Core Performance Boost and setting my ratio settings to what I have listed seemingly fixed things. Could it just be that my processor cannot handle the erratic jumps when it's boosting when I'm playing a game at some points? Is locking your speed via the ratio a bad thing so long as you're not over-working it? I guess this is one of those things that I'll unfortunately have to troubleshoot for some time. Thanks for taking the time out to sort of spell some of this out for me. I'm sure I'll be back in here asking more questions as this progresses.
 

jasonf2

Honorable
Oct 11, 2015
293
64
10,890
6
Setting a modern processor at a fixed rate is grossly inefficient at the very least and limiting performance. These things are designed to find thermal ceilings and balance single thread performance against multi by floating clocks automatically. Ryzen processors are great but RAM overclocks for me have been a nightmare. They don't like odd clock timings and from my personal experience SPD timings are about as good as you get from corsair vengeance. Anyone building today needs to pay not only attention to QVL, but listed clocks under QVL.
 
Last edited:
Oct 5, 2020
21
1
15
0
A bit of an update. I restored Core Performance Boost and the game that crashed randomly hasn't done it since. I have no idea what the hell is happening now. I just called for a return on the RAM and I probably don't need it (I had no other choice since they were past return date and they agreed to take them back). Now I have no idea what's happening.

Do any of you think that perhaps one of the cores on my processor is hitting a threshold for a millisecond that is triggering my PC to just shut off (no BSOD) to black screen (the "on" button my case is curiously on when this happens)? I just have no idea. I might end up sending this RAM regardless as I can deal with the hit of losing a CL14 kit and just taking a CL16 kit, but I really don't know where to go from here.
 

Darkbreeze

Retired Mod
Do any of you think that perhaps one of the cores on my processor is hitting a threshold for a millisecond that is triggering my PC to just shut off
No, because if that was inbuilt behavior based on a specific set of conditions, we'd have seen it from a lot of other Ryzen owners as well, and we haven't.

Anyone building today needs to pay not only attention to QVL, but listed clocks under QVL.
This is a 100% fact. There are a fair number of motherboards out there that traditionally list memory kits that do run on that board, but only at the JEDEC SPD configuration, not at the XMP or A-XMP profile settings. It is getting better, because most motherboard manufacturers are now specifically listing whether the kit is validated for XMP or not on the QVL, but another thing people rarely pay any attention to, especially with Ryzen, is the fact that that doesn't really apply once you move to a four DIMM configuration. Most Ryzen QVL validations are for two DIMM kits and the ones that are for four DIMM kits are generally going to be at lower speeds than for equivalant two DIMM kits. Really doesn't apply here though because it IS a two DIMM kit and it's NOT on the QVL or the memory manufacturers compatibility list for that board.
 
Oct 5, 2020
21
1
15
0
Thanks for all the help everyone. I ended up snagging a 3600MHz kit from G.Skill. I kind of wanted some illumination on my case anyways to counter the reflection. I saw some people OCing that RAM with some stellar results compared to the LPX sticks I have. For the meantime I'm able to move my old HyperX into the build just to hold it over, but I will resist playing anything on it until I get the new RAM since these, too, were not on the QVL.

I truly didn't know these existed until this thread.
 
Oct 5, 2020
21
1
15
0
UPDATE: So I got the new RAM today. Ran a few games. Fortnite, Baldur's Gate 3, Avengers (heavily CPU intensive), Destiny 2. All seemed to be fine. Went into Control, ran around and noticed some stuttering. Then I ran into a room with bad guys and shot my gun once and BOOM, full crash. I am so unbelievable defeated right now. I don't know what to blame and it's extremely frustrating.
 

RodroX

Estimable
Besides what Darkbreeze asked about the OS, I will add three question that may sound silly:

1. The last time it crashed (runing Control, inside the room with the bad guys), when you say full crash, is it like the PC get restarted, or that it crash to the desktop, or that it freeze and need to use the case reset/power button ?

2. How are you checking your Ryzen 7 3800XT and RTX 3080 temps in game while playing, are you using MSI Afterburner, or do you go back to the desktop and look at some monitoring software (if this is the case, what are you using to monitor temps) ?

3. Have you tried to play games with only 1 monitor plugged into the RTX 3080 ?
 
Oct 5, 2020
21
1
15
0
What graphics card were you running in this system prior to the RTX 3080?

Is your operating system a clean install or has it been in use for a while?
I originally had 1080 Founder's Edition in there and I never really pushed it. Only now I have been pushing it with RTX, but it's crashed in a non-RTX game like Avengers.

My OS was a clean install and then a completely new clean install after this first started happening and it didn't fix it. I picked that option from Windows 10 that let me completely reinstall Windows. I don't know if that's truly a "clean" install.

To further answer your question, I've only had this PC mostly built for a few weeks and only recently did I put the 3080 in.

Besides what Darkbreeze asked about the OS, I will add three question that may sound silly:

1. The last time it crashed (runing Control, inside the room with the bad guys), when you say full crash, is it like the PC get restarted, or that it crash to the desktop, or that it freeze and need to use the case reset/power button ?

2. How are you checking your Ryzen 7 3800XT and RTX 3080 temps in game while playing, are you using MSI Afterburner, or do you go back to the desktop and look at some monitoring software (if this is the case, what are you using to monitor temps) ?

3. Have you tried to play games with only 1 monitor plugged into the RTX 3080 ?
1.) It literally just shut off. My case is the h710 from NZXT. The PC shut off and the light ring around the power button was still on. I waited to see if the system would restart itself because I was assuming it was going to since that light was on. It didn't. It powered on when I hit the case button again.

2.) I am using several different things. HWInfo, Core Temp, Ryzen Master, even the built-in monitor for Windows (mostly to monitor activity on CPU/memory/GPU/disk drives). I just installed Afterburner after a friend's suggestion, but I'm not so clear on what I'm doing with it just yet. I don't necessarily go back to desktop to look as I know a lot of times leaving the window reduces the load on the readouts until I tab back in. I have a second monitor that I am checking these on.

3.) I actually haven't tried that yet and it honestly hadn't occurred to me.


One thing to note is that currently while levels are loading I sometimes experience stuttering in the loading screens. As in, my mouse will stutter around the screen if I'm dragging it around. I don't remember experiencing this kind of thing on my old PC at least on this consistent of a basis. I wonder if this is a symptom of something else or just a red herring. Kind of desperate for clues.

Some have also suggested it might be a connection issue, bad MOBO, and even a dud PSU on arrival. I often also think that the old wiring in the apartment that I'm in might be a contributing factor. I still can't rule out the video card, but I've tried video benchmarks and nothing gets triggered there and I don't so much as see a spike in anything.
 
Reactions: RodroX

RodroX

Estimable
mmmm, Im not sure the temps are the problem here, but is something to try out if you feel like it and if Ok forget about it. Heres a nice tutorial of MSIAfterburner:
View: https://www.youtube.com/watch?v=uIlMRRlKQGg
(you can add or remove thing from the OSD, no need to have the same the OP have).

HWInfo 64 - Sensors Only option is great to test temp cause you can always see how high it got by looking at the MAX value. But if the PC crash and shutdown you may be missing something. (remember to always keep only 1 monitor tool open at all times. More than one and you can mess up the sensor readings).

If the PC shutdown I hardly doubt is the monitor thing I mention, but Ive seenssome witchcraft stuff in my many years of experience, so why not try it, its free, and pretty simple to test (besides unplugging the monitor be sure to also turn it off).

This could be a PSU issue, and/or a house wiring issue.
 
Oct 5, 2020
21
1
15
0
That was another thing I'd been wondering. I was wondering if maybe running all of these sensors, maybe a certain one or a few at the same time was causing something to spike and just outright shut down. I'm also considering that I should try a UPS to see if maybe some sort of power issue in the house is causing it and if I can somehow trigger the normal crash but save it because of the UPS. That's an expensive solution though as I don't want to spend 200 bucks just for a test.
 

ASK THE COMMUNITY