Question Anyone burn out Ryzen CPU's PCI-E lanes?

[VirtualLarry]

Trying to figure out how common that this is. Trying to diagnose whether I need to replace a mobo, or a CPU.

 

[VirtualLarry]

Did you hold down the power button to clear the caps?

Don't higher-end PSUs have bleeder resistors across the primary caps? Surely, they wouldn't have held a charge for 6hr, would they have?

I highly doubt it's the PSU. It's new, it's high-end (more or less), 10yr warranty, EVGA G1+ 80Plus Gold, 650W. Also, neither of the cards were dead, nor the CPU, just the slot. Which leads me to believe, that it was something on the board, maybe a fuse, maybe not, something died, in the power-delivery to the slot. Or, maybe the modular cable died, but I tried swapping PCI-E cables, and that didn't make a difference.

 

[DrMrLordX]

Don't higher-end PSUs have bleeder resistors across the primary caps? Surely, they wouldn't have held a charge for 6hr, would they have?

They should. But you also have to worry about the caps on the motherboard.

 

[VirtualLarry]

I gave it another shot. This time, I un-plugged EVERYTHING, and left the PSU switched ON in back, and then pressed the POWER button on the front of the case several times. Then I let it set for 10 minutes. Then I hooked it all back up. No dice. So, I removed the card that I had in the primary slot, that didn't appear to be powering on, just as a precaution not to damage the card. (Hopefully not.)

So, now I'm looking for a replacement mobo. Newegg has some Refurb MSI B350 Tomahawk ATX boards for $60 or so.

 

[Shmee]

I would go for B450 boards at this point though.

 

[pf100]

PCIE lanes died with my ryzen 5800x and I have to run it at PCIE3x4 on the GPU slot for it to work. I bought a used ryzen 7 5800x last year for $230 when they were going for $350 new. Gaming performance was horrible and much worse than my previous ryzen 5 3600. I was using an x570 motherboard, so it has pcie4 pcie lanes for the gpu. Turns out after a lot of troubleshooting and hair-pulling that the pcie slot for the gpu was running at pcie 1.1 speed. So I set the gpu slot (PCIE_E1) manually to pcie4x16 and it was still at PCIE 1.1. I then set PCIE_E1 to PCIE3x16 and that fixed my performance problems but then I was running at PCIE 3 but whatever. Fast forward a year and I recently bought a 5800x3d and gave the 5800x to a family member. So I put the 5800x in another system with a B450 motherboard which is PCIE 3 and it worked fine as expected. Then after several boots in the normal course of things it all of a sudden wouldn't show a screen. So I put the gpu in the PCIE 2x4 slot and it worked. Then I forced all the slots to PCIE3x4 and I could put the gpu back in the first gpu slot.

So now, the pc I built for my family member has an rx 5700 xt with a 5800x, but the gpu is running at PCIE3x4 and loses about 10 fps in games like that. I'm going to leave it like that because it's still 10 times better than their previous R9 290X with horrible AMD FX cpu. I hope the cpu doesn't die because I don't need to spend $200 on an 8 core cpu.

 

[Shmee]

Huh, that is odd. I haven't heard of a 5800X having that kind of issue. Seems like it would be rather rare. Now the first 5900X wasn't quite stable, I think that was a core issue. I hadn't heard about PCIe lanes dying in a long time.

 

[MadRat]

Since when did motherboards become exempt from the 90 day warranty mandated by federal rules? Send it back to the manufacturer and demand replacement. They advertise these functions. They imply working PCIE lanes when they do as a bisic function.

Businessperson's Guide to Federal Warranty Law

Table of Contents State law (the Uniform Commercial Code). Sections 2-314 & 2-315. Section 2-313. Introduction For the full legal texts listed below, consult the supplement to this manual. Understanding the Magnuson-Moss Warranty Act

www.ftc.gov

I realize the thread is from 2019. But never hold on to defective hardware like that. You should have done an RMA.

 

[DAPUNISHER]

Huh, that is odd. I haven't heard of a 5800X having that kind of issue. Seems like it would be rather rare. Now the first 5900X wasn't quite stable, I think that was a core issue. I hadn't heard about PCIe lanes dying in a long time.

It was bought used, so who knows what abuse it was subjected to.

@pf100

I would not feel the need to buy them an 8 core CPU if it dies either. A Ryzen 5600 is plenty for a 5700XT.

 

[aigomorla]

INSPECT PINS on CPU people.

Maybe one is bent, or is not making good contact.
Inspect Sockets as well.
Threadripper/EYPC has MASSIVE memory and PCI-e issues if the CPU is not even Torqued properly, and they make you use a special tool to torque it.

But i always resocket the CPU, inspect socket, for bent pins or poor contact.
And many times i have been cursing upset because i realized somehow i bent a stupid pin reseatting the cpu and ruined a 400 dollar + board, unless i can fix it with a magnify glass and high precision tweezers.

 

[Markfw]

INSPECT PINS on CPU people.

Maybe one is bent, or is not making good contact.
Inspect Sockets as well.
Threadripper/EYPC has MASSIVE memory and PCI-e issues if the CPU is not even Torqued properly, and they make you use a special tool to torque it.

But i always resocket the CPU, inspect socket, for bent pins or poor contact.
And many times i have been cursing upset because i realized somehow i bent a stupid pin reseatting the cpu and ruined a 400 dollar + board, unless i can fix it with a magnify glass and high precision tweezers.

Except, I never want to do that on my Genoa's. 6096 pins ????

 

[William Gaatjes]

PCi-express is very electrostatic discharge sensitive. ESD. To prevent ESD sparks damaging the board or the gpu : Please make sure you have proper ESD wristbands and crocodile beaks to connect to the ground plane of the motherboard before placing the gpu. Also complete disconnect the pc from any cable to prevent a ground voltage level elevation effect. You can preferably connect the chassis of the pc which is connected to the GROUND plane of the motherboard through proper ESD cable with (1MegaOhm) to real EARTH (Ground). And then connect your ESD wire connected to your wristband to the Grounded chassis of the pc which of course is also connected to the ground plane of the motherboard. To make sure that the metal parts of the chassis of the pc (the housing, the enclosure) is connected to the groundplane of the motherboard. Use a multimeter. It should be 0 Ohms. Otherwise, connect the ESD wristband and grounding to a ground connection on the motherboard. When placing the gpu , always touch the large metal parts of the gpu, the groundplane. Because that is where the charge will be stored the most as it has the highest surface area. And this way you will level the charge differences between the ground plane of the motherboard and the ground plane of the gpu card out. It takes a second or 2.

Never touch the PCI-express pins as these are ESD sensitive and the PCIexpress pins are impedance tuned during training of the PCI express links. Training is like testing and tuning for making the almost perfect electrical path for the data to pass over to. So, touching these pins may result in ESD damage and therefore reducing PCIexpress performance. PCI express is to a degree fault tolerant and will sent those PCIe datapackages again when CRC is wrong because of datacorruption. And every time a datapacket needs to be resend, is a reduction in useful databandwidth or PCIe bandwidth will happen because now the same packets needs to be send more often instead of only one time for each packet. Think, weird stuttering behavior or lower fps than possbile in almost worst case scenario. Worst case scenario is of course : no image.

One PCIExpress link is in a way a sort of, very fast RS485 line (A differential serial dataline (data+ & data-) and with an OSI layer on top of it. Similar to TCP/IP has CRC checks and USB has CRC checks. USB is also a differential serial link with an osi layer on top of it.

Also make sure the slot is clean an no copper swarfs are present on the pins as these can create short circuits. And burn out the motherboard and gpu.

CRC = "A cyclic redundancy check (CRC) is an error-detecting code commonly used in digital networks and storage devices to detect accidental changes to digital data. Blocks of data entering these systems get a short check value attached, based on the remainder of a polynomial division of their contents. On retrieval, the calculation is repeated and, in the event the check values do not match, corrective action can be taken against data corruption." Like sending the data again after not acknowledging the received corrupt data.

 

[q52]

Oh, and if you get a Ryzen CPU or APU, prefer one with SMT (simultanous multi-threading) Enabled on it. This 1200 has Four cores, and SMT disabled, and ... it bogs down. I'm mining on all four cores, and web browsing, and man, it is a little big sluggish. Whereas my Ryzen R5 1600 wouldn't bog down nearly as much, due to some SMT threads still having CPU time available.

you need to reduce the process priority (`nice` level in Linux) on the mining so that your other processes e.g. browsing dont get throttled by the mining. This will eliminate "bog down" on pretty much all CPU's regardless of their age or performance.