Question Come to mock me, stay to give me advice... New Question, w/Pic

blckgrffn

Diamond Member
May 1, 2003
9,127
3,066
136
www.teamjuchems.com
I built this premium 5820k machine for my Dad back in ~2014 when DDR4 was new and premium. It started life with a 7950 and he had that matched with a 2560*1600 30" LCD. He was gaming less and I asked him about it, he said it was getting slow and sometimes it rebooted in new games. So we popped in a 5700 he got for $300 shipped from Best Buy right after launch (remember those days of wretched excess?!?) and realized we needed a new monitor because no DVI on the GPU and we weren't paying for an active adapter. Cue an awesome new Dell 32" and everything was great. When it ran. I was pretty certain it was not the GPU because I test drove it for weeks in my rig with zero issues.

Finally, after a power supply replacement, a new UPS, bios updates (a new bios for this thing dropped in 2018!), retuning, a NVME drive fresh Windows 10 install, this video card we finally just pulled the trigger on a whole new everything but the NVME drive and GPU. And it's worked great for a week. Yay.

I've brought the old rig home, blew out the dust bunnies and installed a GTX 970. If I can figure out why it wasn't stable, I need to sell it now while the market is wonky. So I reset all the power connections, hooked up the supposedly optional 4 pin molex by the CPU socket, zeroed out the modest OC, installed W10 on old SSD and checked out HW info to see if there was any weird readings. Turns out there was.

I HAD BEEN USING THE PCIE 2.0 4X ELECTRICAL/16x PHYSICAL SLOT FOR THE GPU THE ENTIRE TIME. ALMOST 7 YEARS OF DUMB.

I moved the GTX 970 down one slot, which was a claimed PCEI 3 16x/16x slot. HW info is telling me it is a x8 slot. A even more careful review of the manual reveals an asterisk that only with a richer, better, and handsomer 40 PCIe lane CPU does this really become a x16 slot. I am loathe to use the actual x16 slot because it covers the NVME drive slot and makes that 4 pin power connector a really tight fit. I figure the x8 PCIE3 slot is a lot better than a x4 PCIE 2 slot and I am done with that for now. I'll let you do the math.

So it's up and running again. W10 activated, drivers installed and Novabench stable (ha, that's a joke.)

TL;DR: What stress tests do you think are reasonable for it to pass to feel OK about reselling it on FB/Craigslist?

Mainly it would have issues gaming, but also at times I guess it would buckle under running Chrome. It did have 1 & 2 core turbo set to 4.0, all core to 3.6 for the last few months, but that was after I had it for a weekend, flashed the bios, ran the MS memory test utility overnight and felt that other tuning tweaks left it in a good state. It didn't have any issues while I had it of course, or during the just-before-covid BL3 LAN weekend we had. I thought I had cured it then.

I want to feel confident that the reset and minor changes made have squashed whatever gremlins plagued this thing. I don't want to have to put a big asterisk on it "probably a POS!" If it comes to that, I'll just sell the CPU alone (eBay maybe $100) and use that to get a big portion of some Microcenter deal.
 
Last edited:

lobz

Platinum Member
Feb 10, 2017
2,057
2,856
136
Detecting user errors when their implications are not very obvious is really hard most of the time, so I don't think anyone should mock you, kudos for finding it out :)

I ran my 1070 in a Z97 mobo with the x16 slot set to to x8 for years, so I could have a 970 pro as a boot SSD in a PCIe adapter. To use a slot that is directly linked to the CPU, I had to set the main slot to x8, and it was a real pain in the brain to set an nvme boot drive up, but it worked and the GPU was perfectly fine with x8, it shouldn't hold a 5700 too severerly back either.
 
  • Like
Reactions: blckgrffn

blckgrffn

Diamond Member
May 1, 2003
9,127
3,066
136
www.teamjuchems.com
Honestly, you might just be better of parting the darn thing out. I usually see greater profit that way.

I think you are right.

Selling locally is one cash transaction you know? That's worth a bit.

Plus I like the idea of it going back out there and being useful. Sentimentality getting in the way here. I know you aren't supposed to treat your systems not like pets but instead like livestock. But I *did* raise cattle and pigs and sheep and I never liked dealing with their end on the farm either :p Much better they leave intact on a trailer.
 

Shmee

Memory & Storage, Graphics Cards Mod Elite Member
Super Moderator
Sep 13, 2008
7,405
2,440
146
What is the board? You could upgrade it to an 8 core xeon for cheap possibly. E5 1660 V3
 

blckgrffn

Diamond Member
May 1, 2003
9,127
3,066
136
www.teamjuchems.com
What is the board? You could upgrade it to an 8 core xeon for cheap possibly. E5 1660 V3

Asrock x99 extreme4 - there is one in the FS/FT forum :eyes: but I think I would be better served in the next couple of months trying to get something a little newer, platform wise. Still, supposedly a pretty solid board. Decisions.
 

blckgrffn

Diamond Member
May 1, 2003
9,127
3,066
136
www.teamjuchems.com
Well, ran three days of Rosetta without issue.

Then an unexpected WMI reboot event... then another about 12 hours later.

I've let Rosetta finish out (set to get no more tasks), but that was really disappointing.

I am a little bit unsure of how to proceed - I am going to get a bootable memtest going but if that doesn't find errors I'll probably toss the board and sell the CPU and memory. That seems like a shame but I can't sell something that can't be trusted to run a few days under load at least.

FWIW, it rebooted (not hardlocked) and just logged into the desktop because I hadn't set a PW and resumed crunching. So windows didn't freak out, there was just the WMI events in the logs and the shorter than expected uptime in task manager to clue me in.

HWinfo shows ~45c high temps on the CPUs when loaded, under that for an average.

Edit: Currently running Memtest x86+ 5.31b for a day or so...
 
Last edited:

blckgrffn

Diamond Member
May 1, 2003
9,127
3,066
136
www.teamjuchems.com
So, to my chagrin, when the eBay auction finally ended (7 days is way too long!) and I got ~$60 net after shipping and fees on the 5820k which I sold because I just could not get the setup get stable, I was greeted by this epic thermal paste job.

IMG_7512.jpg

What. The. Crap.

Of course, that's my doing and there was never really heat related issues - stock clocks and a Hyper 212 Evo and a fan at 100% it barely crept into the 60s under a load like Rosetta, yet would reboot every 1-4 days.

Is this spectacular fail of HS paste job a culprit? Of course I sold and shipped the CPU, leaving me with a nice x99 board, 16 GB of old slow DDR in 4 sticks a burning curiosity to know if this was the problem.

So... in the name of science I just bought on eBay (sigh) a e2623 v3 Xeon, a 3.0 ghz (boost to 3.5 ghz, which I am hoping like with previous Xeons in performance boards I can set the all core to, nbd really) 4C/8T CPU sporting the full 40 PCIe Lanes for... $15 shipped. I mean, compared to the 1240 v1 Sand Bridge 3.3/3.7ghz 4C/8T Xeon I just sold on eBay for $36 shipped, it seems like a solid little bargain for the purposes of my testing. There are cheap ($10 shipped) six core CPUs but they are all of 1.6 ghz and that's just not useful. This should be quick enough for general usage if it turns I out I am just an idiot.

I guess I netted $45 too? I mean, the 5820k wasn't really a gaming stunner without some clocking up, so I am thinking that really this is pretty low risk situation at the moment.

But... think that paste was my problem? TIA :)