Weird ram issue, 2 stick not showing at all, then they were.

andrew breckill · 17 Jul 2020 at 19:02

A bit like asking for clairvoyance with this one. I know ram is a law unto itself at times. I added 256gb ddr4 ram (3200) to my shiny new asrock trx40 taichi, went into bios, bios showed 8 sticks, but two had no ram amount next to them, bios showed the speed, but capacity was showing as zero for these two sticks. So I quit bios and let it boot, got a message memory training failed for the two sticks, window booted OK, and 6 sticks showed in task managers screens. turned off power moved the ram to next two slots along, same issue, the sticks, not the slot seemed the guilty party. so rang OC and got an rma and ordered some new ram for use whilst it got sorted, whilst waiting for new ram Did a bit more reading, cleaned all contacts, moved all the ram around and checked everything was seated right, did a bios reset, and boom, the system booted with all 8 sticks, then next day they went off again, so reset bios, rebooted, they were there again but only at 2133mhz, sso usedd system like that, then today tried again to run at 3200mhz, and they all are, and seem to be working properly, (did loads of rendering tests, using maya, did some photogrammetry using reality capture, basically given it a good work out, so what do i do? keep memory as this is parr for the course or should i send it back still?

Tetras · 17 Jul 2020 at 19:34

Don't think you'll get much help with that round here, there's hardly anyone with a threadripper, let alone 256GB of memory. My impression is that putting max memory into any board is always going to be an iffy trial and error experience, because it's just not done and tested widely.

andrew breckill · 17 Jul 2020 at 20:05

thanks for the heads up. i used 128 gb on previous build and that was hassle free. so was a bit surprised at what happened. i will prolly give it a good workout over weekend and maybe give them another call on monday if it falls over. i have noticed the ram sockets for ddr4 seem a bit shallow compared to ddr3, on last board one stick spontaneously popped out once.

Tetras · 18 Jul 2020 at 02:32

Was the previous build an Intel by happenstance? They've typically been less fussy with memory than Ryzen based builds. Not sure what happened there, have you run anything that reports if there are errors, like Prime or memtest?

You might find this an interesting read, can think of a few possible implications from it:
- The large amount of memory gives it a problem detecting and verifying it before the system is required to boot.
- There might have been a dirty contact or a seating problem with one DIMM which disabled the channel.
- One or both of the DIMMs are failing.

https://blog.asset-intertech.com/test_data_out/2014/11/memory-training-testing-and-margining.html

andrew breckill · 25 Jul 2020 at 11:10

Hey all, No previous build was an x399 with 1950x threadripper.
I did run windows memtest once all chips were recognised at 2133mhz, and no errors.
As for present situation, I removed them all again, cleaned all contacts with rubbing alcohol and double checked all seated properly and system been running spot on all week. Which is great, as at first I was convinced two were faulty. All running at 3200mhz and system is kicking it. I often have 4 copies of a photogrammetry program running with up to 1000 photos to process in each instance and this amount of memory handles it perfectly. Also can use much higher particle counts in Houdini which is massively helping look dev.

hominid · 25 Jul 2020 at 17:41

The only thing I'd add is that it could be a problem with the seating of the CPU in the socket. I've seen it with intel server chips that sometimes you get a problem like one or more channels not working and re-doing the heatsink mostly sorts the issue - with a few needing multiple tries to get it to work on the LGA array.

Check with the torque driver you have done them tight enough, also remember the correct order also!

andrew breckill · 25 Jul 2020 at 18:17

i am going to redo the thermal paste as only had some cheap and cheerful akasa paste left when cpu arrived, got a tube of kryonaugt arriving next week, will double check cpu, when aio is off. i have fitted a few if these cpu's now so pretty confident its ok. like i mentioned been crash free and issue free all week and i have been doing some intense computations.