revtothelimit - motorbike forum

For the last month I've been chasing my tail trying to understand a fault on a PC system I'm building for Mrs Mango. By process of elimination I'm looking at the CPU (a Ryzen 2400G) as the culprit. Memory sails through Memtest, PSU checks out with all voltages within spec and I've thrown 3 motherboards at it; the first one just hung randomly during boot and the last two didn't even POST.

So, is there a relatively simple way of testing a CPU, or do I just buy another one and try again? The suspect CPU is s/h and of unknown provenance (i.e. ebay), so no warranty.

Prime95 used to be the go to for stress testing, but it's a good 10 years since I last used it

At present it's moot, as I can't even get to the initial BIOS screen. I think it's time to buy another cheap CPU.

In my long and distinguished IT career, the only CPU's I ever had fail were AMD.

This is the first faulty CPU of any manufacturer I've ever encountered. As it's an ebay purchase that shouldn't be too surprising I guess.

It would be more likely that the pins on the CPU are damaged in some way. Under a microscope you might be able to see some damage and as you have tried more than one motherboard I don't think you have any other choice. Unfortunately as individuals testing by substitution is relatively expensive but it is typically the easy choice for professionals. A final desperate option would be to try a contact cleaner spray followed by a gentle blast from a compressed air can.

I've ordered another (s/h) one and we'll see how that fares. I've inspected the pins using a Lupe, and nothing seems amiss, but that's purely an external inspection.

The second CPU arrived today, and was plugged in to the 3rd motherboard, Predictably enough, it didn't work in exactly the same way as the previous CPU. This wasn't entirely unexpected but it did make a me just a tiny bit cross(er).

So I decided it was time to swap out the only major component not yet substituted - the memory. I purloined a pair of 8gb sticks of DDR4 from a PC that is bound for France, and inserted them into the memory slots with just the right amount of force.

You can probably guess what happened next.

The bloody thing booted up perfectly. Of course it did. So then I tried the memory that hadn't worked in 3 previous motherboards in the PC bound for France.

You can probably guess what happened next.

The bloody thing booted up perfectly. Of course it did. So it appears I have a pair of memory sticks that will only work when the motherboard suits it. It also means I have a motherboard and CPU going spare. But at least (failing further dramas) I have the right number of working PCs.

Perhaps there is an issue with BIOS incompatibility somewhere. Not all manufacturers will update BIOS for older boards indefinitely and it might not even be possible. Some boards are also very fussy about which memory slots are used if they're not all populated. I once had to send back a brand new AMD socket board because it needed a BIOS update to be compatible with a new CPU but there was no way of updating it because it couldn't even get to BIOS without the update, catch22!

It's been very odd. The first board I tried (a Biostar AMD AM4 socket with the B450 chipset) would boot and load the OS, but then hang once the BIOS had handed over to the OS. I then tried 2 x ASRock AMD AM4/B450 boards and neither would even POST with the original RAM. Switch to different RAM and it works flawlessly. And then just to really confuse matters, the 'suspect' (Samsung) RAM works perfectly in an older Biostar AMD AM4 A320 board.

Has to be a RAM timing/compatibility issue. All the boards only have 2 memory slots and both were populated on all boards, with identical sticks of RAM.

Might be interesting to run CPUID on the one with the Samsung memory and see how it reports memory and timing.

Good call. I'll do that in the next day or so - should be instructive.