Question PC wont boot with more than two cores (Ryzen 7 3700x)

Sep 6, 2019
9
0
10
Heya Forum (first time posting, so bare with me, im not very tech-savvy)

Roughly a month ago i completed my new gaming rig, and its been running OK since.
Now, in the start i had some problems restarting the pc, it would turn off but get stuck booting up, with fans blasting at full force. And i would have to kill the power completely to get it running again.
That eventually stopped, and it ran well for a couple of weeks, then suddenly one morning last week, it switched off, and now its booting up till BIOS, but as soon as it has to launch windows it crashes - unless i enable it to only run the lowest amount of cores i can, which seems to be two in the case of my CPU.
(BUILD
CPU: Ryzen 7 3700x
GPU: Nvidia GTX 2060 FE
RAM: 2x8 GB G.Skill Ripjaws 3200Mhz
MOBO: MSI B450 A Pro
SSD: Corsair MP510 240Gb (Boot drive)
HD: 1tb WD
PSU: Corsair CX600 600W)

I ran Memtest multiple times, with different configurations, like 1 stick of RAM and then switched them around, and with both of them in.
Running the test with all cores active created some problems, in the start it showed: {UEFI Firmware Error] Could not start CPU 3 (sometimes CPU 2, as it wrote the sentence many times)
I then switched off SMT within the BIOS and ran Memtest once again. This time it ran smoothly until Test 7, where it "crashed" due to too many error (Test: 7 Addr: 105CB8 Expected: FF7FFFF Actual: 000080000 CPU: 2) which it also wrote multiple times with differing numbers and characters.
I then went into BIOS once again, and enabled it to utilise the two cores mentioned above, this is where i discovered it could boot with this config.

Like i said, i can boot into windows just fine, with the two cores active(MOBO shows it as: Two-cores 1-1). As soons a i configure that to something like two cores with (2-0) or above, it crashes and restart as soon as it tries to start windows.

I humbly hope you guys can/want to help me resolve this.
 
With windows 10 and these types of problems, I have learned that doing these things first is always a good idea to make sure there are no problems with the Windows install:

1. Open Start.

2. Search for Command Prompt, right-click the top result, and select the Run as administrator option.

3. Type the following command and press Enter, let run until its complete: SFC /scannow

4. Type the following command and press Enter, let run until its complete: DISM /Online /Cleanup-Image /CheckHealth

5. Type the following command and press Enter, let run until its complete: DISM /Online /Cleanup-Image /ScanHealth

6. Type the following command and press Enter, let run until its complete: DISM /Online /Cleanup-Image /RestoreHealth

7. If step 3 failed but steps 4-6 found corruption and fixed it repeat step 3 again.

Report back when this finishes (5-10 min)
 

Infikiran

Distinguished
Oct 11, 2008
107
4
18,715
Another thing to check is make sure your CPU is not overheating. When you disable cores it runs at a lower thermal temp, which would explain why it is running more stable with less cores. If it is indeed the CPU overheating, check and make sure your heatsink and fan is securely fastened and making contact with the CPU. Also make sure the appropriate amount of thermal paste is used. If you used the stock heatsink it should have already had some on it.
 
Sep 6, 2019
9
0
10
Completed the SFC/scannow and it yielded: Windows Resource Protection found corrupted files ect.

Ran the other tests without realising i had to reboot, so the other tests didnt go well.
I restarted and ran the other tests and they completed successfully, ran the SFC again, and it too completed successfully.
 
If you can't successfully pass memory tests with your CPU fully enabled at it's default, stock settings, the last thing I would be having Windows do is verify and possibly rewrite a bunch of system files. How do you know it's doing this correctly if you don't have reasonable assurance of the values passing through system memory?

Sort out your memory tests first.

What is your BIOS / UEFI reporting for CPU temperature (we're assuming you used the stock cooler that came with the CPU and that it is installed correctly.)

Things you might try:
  • Verify that you have installed a single memory module into DIMMA2 or DIMMA2 and DIMMB2 when running a matched pair according to MSI's recommendation; counting from the CPU socket, DIMMA2 is the second memory slot from the CPU while DIMMB2 is the fourth (may also be silk screened on the PCB)
  • Run your memory at the lowest available settings, perhaps 1066 MHz or 2133 MT/s
  • Run your memory using an XMP profile selected in BIOS if one is available
  • Adjust the voltage to the memory modules; many modules run at 1.2 V for standard speeds but require 1.35 V or higher to achieve stable results when running at advertised speeds; verify the voltage after setting it as some BIOS do not honor manual memory voltage settings if other settings have been modified
  • Verify you are running an up-to-date version of BIOS that supports your CPU; there have been updates to the AGESA version that boots Ryzen 3000 model CPUs and MSI lists several revisions for the MSI B450-A Pro that "Improve memory compatibility"
  • Reset your BIOS to it's default settings, especially if you've previously used that motherboard for a different CPU such as a Ryzen or Ryzen+ model
  • Disable any overclocking, PBO, special, or performance features in BIOS that will cause the CPU to run outside of it's stock parameters
 
Sep 6, 2019
9
0
10
Once again thanks for taking the time to review my problem!

1. My tempts are just fine, in the BIOS as well as when running Memtest. Didnt go above 60 degrees C. (I dont use stock cooler, got a vertical mounted Noctua fan running, with a beefy heatsink)

2. I have verified that the RAM sticks are indeed in the right slots.

3. I did see a new BIOS update roll out, which im goning to try and update now.

My question: If the RAMsis indeed unstable, wouldnt booting into windows with the two cores be just as unstable as with all cores, which it is not?
 
Wouldn't booting into Windows with two cores enabled vs more be just as unstable? Maybe.

There are several sources for power floating around your CPU, and yet another for your memory. If any one of those supplied powers is on the verge of being insufficient, such as because you could have ended up with poor quality silicon or have parts with over ambitiously specified capabilities, running only two cores may barely be keeping things under the threshold that is causing your issue, and there may be some CPU loading cases where you could pass that threshold even on two cores.

No PC hardware is perfect, so the slight variations in hardware are expected. This is usually accounted for by setting the bar lower for all parts so more meet their rated specs. You may have equipment that doesn't meet even the lowest common denominator for what remains stable at stock settings. You could also simply have a defective part or parts.

Tinkering with voltages, within reason, is usually safe, and sometimes is all you need to stabilize a borderline part.

If the CPU is defective, I would imagine that any DDR4 you stuff in the motherboard will give the same or at least very similar results.

If the RAM modules are defective, you should be able to see the same errors when tested in another system.

If one or more of the motherboard memory channels is defective, using the other channel should be sufficient to avoid the issue, but if both were defective I'm not sure the best way to suss that out without using a different board.

I would go so far as to wonder if the CPU was fully seated into the socket and no pins were bent or missing before locking the socket. Really wouldn't give a satisfactory explanation to the system seeming to work with only two cores enabled though. I know it's not a fun task, but removing and re-seating the CPU is always an option.

There comes a point in working through issues like this that either dedicated test equipment, or at least replacement parts is where you end up at, short of an RMA for the suspect parts. At this point, I wouldn't be comfortable blaming either the CPU or the memory modules with any assurance.

I would start by trying to eliminate the memory as a possible culprit, testing every possible safe configuration I could. Again, low clock speeds, conservative timings, and increased voltage. If all variations of memory settings fail to cause an improvement to the situation, I would look for an alternative Ryzen CPU to test with. If the 3000 series CPU you are currently using was an upgrade, I would fall back to the previous CPU and retest.

Temps near 60°C sounds a bit high for a Noctua aftermarket cooler, unless you are loading the CPU, neither of which the BIOS or a memory test should be doing. Are you sure you have it properly mounted? Are you using a fan speed reducer?
 
Sep 6, 2019
9
0
10
Just checked the temp again, read it wrong, idles around 30 degrees C.

I just tried tinkering with the volts and speeds of the ram, with no results.
I updated to the new MSI BIOS patch, and now only one stick of my two sticks of RAM are showing... It did say its a BETA version.

I did suspect bend pins and such, so i took out yesterday, inspected it - nothing, and put i back in, no results.

When the problem first arose, i was quick to suspect the mobo being the culprit (i dont know shit about faulty PC parts)
So i was quick to order a X570 board which coming monday, but it doesn't seem like you guys suspect the mobo being the problem...
Would it be worth trying the board out? I mean, its an ok upgrade from the B450 anyway.
 
30°C sounds better.

It shouldn't hurt to wait and see if your components work in the X570 board (unless you miss a hardware return window at a vendor you purchased equipment from), but there will be little benefit from the newer board if the B450 board is not to blame. The only significant upgrade is the PCIe 4 support, but returns on that are mostly limited to some NVMe devices. It's also been talked about before by reviewers that, unless you spend a significantly greater amount, you generally don't get better VRM designs between B450 and X470, and the same seems true of the X570 boards.

I don't have high suspicions the motherboard is the problem, but after trying an X570 board (does not need BIOS updates for Ryzen 3000 support out of the box), if the problem persists, I would return and replace or RMA the CPU.
 
Sep 6, 2019
9
0
10
Now, i did actually hear that the VRM on the midrange X570 boards actually being pretty beefy, giving you are buying the right one... Which i lack the knowledge to determine.

Ill wait till monday, and see if the mobo is at fault.

Thanks a ton for the replies and help, i really do appreciate it!
 
Sep 6, 2019
9
0
10
I ended up getting a new CPU after the place i bought it checked it through, RMA ofc. If you are having the exact same problem as me, which is that one or more cores are faulty, then there is no other way around, than getting a new CPU.