Hi. My second post. I bought a Asrock Fatal1ty gaming x399 threadripper motherboard as “opened, but unused” from ebay… along with a threadripper 1950x, which was all I could find at the time. Anycase, shock and horror when I saw what appeared to be rows of bent pins. (see pictures).
First I took closeups with my phone. I studied the pics and I couldn’t really make out exactly how the pins were bent, but they looked more or less evenly spaced. Keen to use my new torque screwdriver, I proceeded to clamp down the cpu and fitted a noctua 12cm threadripper cooler. Then I paused and decided to go read more about bent threadripper pins. I ordered a magnifying glass, ifixit kit, tweezers etc. I have bent back occasional intel pins in the past, buy never on this scale and never amd. I prepared for hours of work.
But just before I opened up everything again I thought I ‘d check what happens when I switch on, just in case the clamping broke some pins. Lo and behold the system posted. I installed windows 10, cinebench, etc. The system’ s been running 3 days. I did updates, ran cinebench stabilty tests twice and updated the bios as well.
My question is:
Do I consider this solved? Will it work like this indefinitely, or should I expect problems down the line? Was I just lucky?
a lot of those pins will probably be i/o pins. I’d expect random i/o on motherboard not function properly (eg. some pci-e slots be non-functional, link to chipset could be degraded etc)
You can see in some of those closeup pics that some pins are bent outward and not in the right place. As was said, it works right now, but you may find some IO device non functional or giving errors when in use, or even bluescreens every time you populate a certain dimm slot, or something of the sort.
Temperatures are holding. Power is stable. Cores are hitting right frequencies. Put in a beefier gpu, two lsi 9260-8i, a samsing 980 pro nvme. Need to still check final pcie slot, reserving that for asus hyper m2 card. That should be able to fully test a x16 slot.
All sata ports work. Will test all m2 slots. I appreciate the advice. Right now I’m curious to see where it goes. It’s a kinda lab pc. Not the main pc. All up for a journey. Thinking if it’s not broke, don’t fix it. Maybe claiming down to 1.5Nm helped. Maybe they were alligned close enough when that happened. Maybe some crucial pins are damged. Let’s see.
I will report here on findings. If at the end of the day it’s done for, then I can draw some conclusions, like me being a fool for going ahead.
uploaded photo doesn’t reliably show if damaged section is only that one “scar” at the bottom or if there are some bent pins in top section as well. Like it’s been mentioned it seems that this bottom scar goes across unused pins and some RAM related ones. Try to run full memtest with all 4 dimms populated.
I overlaid your photo over scheme linked by @EniGmA1987 it’s very rough of course since this scheme doesn’t seem to keep TR CPU proportions correctly but overall it seems that you might have been very lucky. Assuming that part marked in yellow is just photo artiffact and pins are not in fact significantly bent there, then red section indeed mostly goes right across unused pins xD If yellow is also bent then it seem that it might go across memory channel B and some probably not very essential I2C/SPI etc.
I would think some of them are probably bent, but not enough to “miss their mark” in terms of hitting a pad on the CPU.
First off – who knew these things had so many reserved and unusued pins? I sure didn’t.
Second – I imagine (and it looks like this is in fact the case) that a lot of these are for voltage and ground. I’d run a load test, if it doesn’t crash under load, call it good. The advice to check PCIe slots and all that is also good. I honestly think you would have found a problem already if you had one and you’ve “dodged the bullet” as it were. If you find a PCIe or RAM slot is acting up? Just label it or something so you aren’t tempted to use it in the future, and call it good. I had a mobo at one place I work blow a cap for just once PCI slot (yes this was many years ago)… totally asymptomatic, there was nothing in that slot so it didn’t affect anything (why a cap would blow when there was basically 0 current running through it? I have no idea.)
I’ve had one board where a few pins were bent – didn’t affect anything. And one (Athlon XP Era) where the corner actually chipped off the freakin’ die, that was one of those ones where you REALLY had to horse the clips down on the heatsink, and no heat spreaders back then. I’ve read (going back to the 1970s) that they’d intentionally leave nothing important at the corners so they wouldn’t have losses if this happened cutting the wafer into dies. Whatever the case, no problems on that CPU either.
Iirc it’s the same form factor of CPU as EPYC CPUs which had waaaaay more i/o than consumer TR, more RAM channels etc. I’d assume TR PRO / EPYC CPUs have fewer unused pins.
unless this was disclosed during sale get a refund from ebay - goods not as described.
with those photos i’m pretty sure you’d be fine to get a refund, ebay support is generally pretty good with this as far as being a buyer goes.
it may be fine, it may be fine until you decide to put more RAM in the system, some of the PCIe slots might not work, some USB lanes might not work properly and get errors (maybe only even at particular speeds/USB standards), etc.
fuck that. get rid of it.
otherwise - every single “wierd” problem you have on that board - you’ll never know if its just a “threadripper on OS whatever” thing, a chipset bug - or electrical noise/errors on your IO pins on the CPU socket. don’t give future @Rob1971 headaches to deal with.
Unless you like dealing with this sort of hassle? Personally i do plenty enough PC doctor/janitor shit at work.
there’s no bad products, only bad price. If he got that for llike 100$ with mobo I’d keep it XD I mean it’s Threadripper after all. i/o doesn’t grow on trees.
i wouldn’t be worried about time for troubleshooting, if there’s anything I’d be worried about is that one fine day it could decide you know what? f*ck that nvme drive in particular. xD Data integrity is the only real problem that you may face in such situation, with real catastrophic consequences.
But TR platform is TR platform. You never know reason for any single “weird” problem and with platforms like those there’s A LOT of weird problems. Few days ago I connected one more M.2 drive and one, random of my GPUs stopped working in BIOS all of sudden. And only in BIOS, in OS works just fine. Because… potato. I suspect it’s due to too many connected pci-e devices because I have 154 pci-e devices connected as of now and maybe ASRock never predicted someone will stuff this thing so much. when I connect too many USB hubs, keyboard from built-in IPMI doesn’t work (also chipset usb controller doesn’t work) in BIOS. But that’s just TR platform being TR platform - a lot of things can go wrong if you have… a lot of things there…
answering this question precisely - cost of new mid-tier TR platform (like 24 core+mobo) is my 1 full monthly salary so I’d have to struggle for 1 entire month to make it not worth it XD
This is great. i was looking for a map and found a list of pin assignments for this socket, but it is hard to visualize as there are many many pins. A large number are as you say unused, ground or mem pins. I will do a mem test. The machine has been running weeks now. I will be away for a while as it’s holidays, but will report back.
Yes this is old tech and as you can imagine, it’s hard to find locally sourced replacements short of importing from China. But even then it’s hard to find the Fatal1ty X399 motherboard, which. is identical to the Taichi, but including 10GB Ethernet, which is why I went for this. It cost around £350 for the motherboard and cpu. I wanted to replace my old fatal1ty Z68 motherboard in what is my hobby pc. Messing around, testing stuff and hooking up tech. Sort of like a workstation. I like that being separate from my main pc, where software is carefully curated. Also where the consumer Intel/amd platforms are super fast, they lack io and that became increasingly apparent. They are limiting,
So I’ve been watching a lot of youtube videos about homelabs and home setups and level1techs of course. These forums are brilliant. I set up a homelab. server with a Datto motherboard and this workstation with threadripper to experience more platforms. I fully expected switching on, that the machine would not post, or something worse would happen. I had tweezers, magnifying glass, lights, etc. ready for the long slog, but when it posted and all apparently seemed to work as it should, I decided to go along for the ride and will post what I find, here. I will try the various suggestions and maybe there will be something that’s weird.
As it is i filled 3 of the 4 pcie slots - tried 3 video cards. Will get a gtx1080 for this machine next month. Two LSI 9160 I had in the previous machine - one has 8 hard drives. The other has 4 ssds.
Planning to fill the 4th one with an Asus Hyper m2 down the line
Also have 4 x 32gb ddr4 and a samsung 980 pro nvme ssd installed. All ethernet (2x 1gb and 1 x 10gb working. All usb ports working,
