I finally got around to getting current gen hardware for my home pc… All the Ryzen 2 goodness was too much to keep away from and combined with a timely trip to the US where I ended up at Microcenter meant that the parts were in the bag:
Ryzen 5 3600
16 Gb DDR4 3200 - Corsair LPX Vengeance
Ryzen 3k support on Linux is still a little iffy AFAICT. There have been issues where systemd won’t boot due to a bug on AMD (fixed in an AGESA update etc). So I downloaded ubuntu mainline kernel 5.1 and 5.3 and installed them along with nvidia proprietary drivers and made sure that both those were working before racking up the hardware.
Hooking up a PC isn’t hard at all - but even then, first boots trigger a few butterflies nevertheless, where you hope that you didn’t end up with a part that’s DOA. In any case, I got Microcenter to do a POST test on the parts, so that wasn’t an issue - but still.
Hooked up the CPU to the board and connected the stock cooler and put in the RAM sticks. Powered the board from my current SMPS and it booted up. Did a bios update to the latest on the manufacturer’s site and checked again if the system booted up. Next, I hooked up my hard disks and got to the SDDM login screen. Given that all this went pretty boringly, I was just not prepared for what happened next!
At this point, there’s only clean up ops - basically shut down the PC, hook up front panel button/IOs etc, tie up cables to the case and close the side panels. Again connected up the power to the SMPS and switched on the SMPS and I hear a POP! I thought I’d misheard - and powered on the system and it came right up. Still, that nagging doubt wasn’t good - so I shut down again, removed the power to the SMPS and repeated. This time there was definitely a POP! - no mistaking that.. and if that wasn’t sufficient, I also thought I saw a spark out of the PSU. Gingerly, hit the reset button (NOTE: this is not a good idea - if you have failing electronics, sending current to expensive parts isn’t advised. But then what can i say - curiosity killed the cat and all that).
Nothing - system’s dead - no fans, no CPU lights etc. I’m panicking on all cylinders and remove power first of all. Frying a SMPS would be bad - but frying a new CPU/MOBO would be worse. What’s more - there’s no way to tell if the latter had happened as I don’t have a set of spare PSUs lying around.
I did open up the SMPS but did not find anything obviously wrong visually (typically, either blown fuses or blown capacitors etc) and I wasn’t going to risk plugging it in again. So off I went to the market and bought a new PSU - an Antec 650 modular PSU the next day. Plugged it in and thankfully, system boots up - my relief was palpable.
BIOS - only RAM at 3200Mhz - everything else is at stock.
OS - so I got rid of both the custom kernels (5.1 and 5.3.1) that I had installed from .deb files. Instead, I added the Ubuntu Kernel team’s proposed ppa and installed 5.3.0 along with its
linux-modules-extra. This is required since the kernels installed via deb files did not have the
extramodules and so no
aufsfilesystem and so docker daemon was failing to start.
Temp monitor - thankfully my board apparently uses something like a
nct6775module for thermal monitoring which is supported. With that installed and loaded, I could see CPU and Motherboard temps.
Right after a system build, the first thing is to do a stress test to smoke out
issues if any. Prime95 is great for this and you can run a stress test that
will peg all your cores to their max over an extended period. If the program
crashes or your computer shows wonkiness (abrupt restarts, runaway thermals) -
that’s a good indicator that something’s off. So I ran
prime95 for a 20 min
period and things didn’t go haywire… temps were high at 84C but apparently
that’s par for the course. The Ryzen 3600 is supposed to boost up to 4.2Ghz -
but I never went above 3.9Ghz. That’s not alarming but something to look into.
So other than a few hair raising moments, things went well - new hardware works. I had a 10 year old system that I could probably load to 40% sustained load when I was doing a lot of work on it, now I have a brand spanking new system that I can’t even load 15% with a similar workload (1 Windows VM under QEMU/KVM, a bunch of containers on the host, Firefox with about 20 odd tabs and miscellaneous other programs running .