Hardware Hell and Me

Moderator: Jere

Post Reply
User avatar
I am nobody
Moderator
Posts: 13289
Joined: Tue Aug 07, 2007 7:26 pm
Location: -89.97814998,-42.2333493
Has thanked: 6 times
Been thanked: 174 times

Hardware Hell and Me

#1

Post by I am nobody » Sat Nov 02, 2019 3:18 pm

Posted about this in the Discord a few times, but now that it's (finally) completely resolved (I think), I need to vent the whole 3 day story.

Spoiler because long.
Spoiler.
So I got a new GPU and two new RAM sticks on Wednesday. The GPU needs a 650W PSU vs the 550W my current one asked for, but I already had 650, so didn't worry about it. It also needed two 8pin connectors instead of one, which meant changing my VGA power cable and is potentially significant in retrospect. I know now, but not at the time, that buying 2 8x2 kits (let alone over a year apart) instead of one 4x8 isn't safe, but that mistake turned out to be fortunate.

Anyway, I get everything plugged in and run some benchmarks and stress tests that give me pretty numbers across the board. Then I go back to playing The Outer Worlds and enjoy an hour of much smoother frames before it freezes for a half second and my computer restarts.

So I immediately assume it's a PSU issue and spend some time reconnecting everything and then verifying that my current wattage should be enough and that everything else I've got plugged in won't change that. I unplug some USB HDDs I'm not using just to drop the power draw a little and pull up another benchmark. Restart. I open it up again, unplug a SATA drive I'm not using, and then run the benchmarks for 20 minutes with no issue. So great, it works! Boot up TOW again, realize I didn't have the settings maxed out and fix that, start the game, and restart within a second of the game loading. Spend a bunch of time checking RAM, drivers, temperatures, CPU, etc, with no issues. Curse the universe and go to bed.

Talk about it with a work friend the next day and we agree that it'd be weird for a PSU or temperature problem to not have an obvious relationship to load or usage and he suggests I reseat everything. I open it again, reconnect every single power cable and redo the GPU seating, then boot up a more demanding benchmark and immediately reset. Given that the GPU has been absurdly quiet other than a slight high pitched noise and I got it open box, I assume it's at fault and put the old one back. Everything seems fine. Yay! I fill out a return form, pack up the box to take to UPS the next day, and go to bed.

I don't have time to go to UPS after work, but did have an hour for TOW, so I start that up and get an hour out of it before, sadly, an abrupt restart. Then it turns out I'm not busy after all and I decide to use my newfound free time to trudge three miles round trip to Microcenter and buy both something to test my PSU and a new one so that I don't have to walk back if, as expected, the old one is bust.

Get back from a surprisingly cold walk and unplug the whole PSU again to test it. The tester says it'll stop beeping once the 24, 8, and 6 pin connections are in if everything is normal. All voltages are normal according to the display, but it doesn't stop beeping. I decide it's probably a weird bug and that it must be a RAM issue after all, so I take out the new stuff and go back to benchmarks. Restart. I do a fancy graphics driver uninstall after that and discover that, even though Windows couldn't see it, there actually were two different drivers installed, so surely that must have been the problem. Install the right ones again, benchmark, restart.

I bite the bullet and open the new PSU, voiding the return. Plug it into the tester and the beeping actually does stop this time. Unplug the old one, put the new one in, turn it on, benchmark, and it's fine! Yay!

This morning I run lots more tests to confirm it's fine and don't have any issues. I eventually put the new GPU back, more tests, still fine. Yay! But my Windows marked half my RAM as hardware reserved and it's running at half speed. No! The usual fixes don't help. I now learn mixing sets even of the same model isn't safe, so I unplug the new stuff and restart. Now I only have half of that RAM, and it's still at half speed. Benchmarks confirm that it's running like absolute trash, at one point achieving the 0th percentile for that model. Swapping the slots each remaining stick is in doesn't change anything.

I put everything back since half of 32 is at least better than half of 8 and notice that BIOS sees all 4 sticks and the intended speed, but halves both figures in the total display anyway. Changing settings manually doesn't help, so I update BIOS, which also doesn't help. I then remove three of the four sticks and restart, which works. I swap the old one that was left for the other old one and restart, to nothing. So one old stick is dead. One old stick + one new stick boots, but crashes when Windows tries to start. Two new sticks and no old ones boots and shows the correct total memory and speed. Problem solved!
tl;dr: PSU that was, in retrospect, showing signs of problems already crapped out when asked to run a more demanding graphics card. Somewhere in the cycle of restarts, it also killed one of my old RAM sticks and caused the other one to run at half speed. My motherboard resolved the conflict of one dead, one half speed, and two fine sticks by refusing to acknowledge anything other than the presence of the dead stick and one of the fine sticks and slowing the other fine stick down to match the half speed one. That configuration apparently looks okay to all the RAM diagnostics I tried.

Also, Windows shows hardware reserved RAM in terms of PB now, which is kinda weird, but I'm going to hope I can ignore that for now.

User avatar
Apollo the Just
Moderator
Posts: 15515
Joined: Mon Jun 19, 2006 3:37 pm
Location: Piccolo is Gohan's Real Dad
Has thanked: 334 times
Been thanked: 460 times
Contact:

Re: Hardware Hell and Me

#2

Post by Apollo the Just » Sat Nov 02, 2019 11:08 pm

Remind me again why I decided to change careers to IT

User avatar
I am nobody
Moderator
Posts: 13289
Joined: Tue Aug 07, 2007 7:26 pm
Location: -89.97814998,-42.2333493
Has thanked: 6 times
Been thanked: 174 times

Re: Hardware Hell and Me

#3

Post by I am nobody » Sun Nov 03, 2019 7:57 am

This particular problem probably wouldn't have been as bad in a tech desk scenario because there would've been already-purchased backup parts to test with and, worst case, I imagine most companies would recognize that an open-ended amount of employee time spent on a solution that may or may not work very quickly becomes more expensive than just buying a new machine or trying to claim warranty. Most of the reason my experience was so miserable was that I made not spending additional money the priority, which meant a lot of wasted time trying to test the broken old parts in-place.

Speaking of warranty, that PSU was only like 18 months old. I'll have to see if I can claim that.

User avatar
S1x
Posts: 9948
Joined: Mon Jul 23, 2001 1:00 am
Location: Loading......
Been thanked: 27 times

Re: Hardware Hell and Me

#4

Post by S1x » Sun Nov 03, 2019 9:43 am

I've been troubleshooting my own desktop for a while now and I really wish I had spare components around just for tests on which of the current parts are at fault.

I'm going to be getting a new GPU soon because I'm somewhat sure it's the source of weird occurrences I've been having in games, but if the problems continue afterwards I'm going to be really frustrated

Post Reply

Return to “Computer Gaming & Tech”