T7910 won’t post - possible memory issues?

Question

I’m having issues getting my Precision T7910 with dual E5-2697 V3 and 128gb memory to post, show any display output or power to usb keyboard. Power supply tests green and when I turn on the system I get a solid white power button, just no bios or any screen display even if I hit f2 or f12 immediately after powering on.

some history - ever since I purchased this system as refurbished back in 2019, it would occasionally exhibit this issue on initial power up but if I cycled the power it would always boot fine on the 2nd try. Fast forward to a few weeks ago and see suddenly one day it would just not post at all. At that point I gave it a good interior dust cleaning and checked all the connections and then fired it back up and this time it posted! I ran the full suite of diagnostic and everything passed including all 128gb of memory. I noticed the following repeated in the most recent bios event logs…

-Invalid configuration information - please run SETUP program

-Time-of-day not set - please run SETUP program

“Ah Ha! It’s probably just a dead cmos battery” I thought, so I took out the old coin cell, waited about 15 min and popped in a fresh one. Ever since then I’ve not been able to get it to post again. At this point I’m suspecting a memory issue and upon closer inspection I realized that there are two different brands modules, 8x8gb of Samsung (white banks) and 8x8gb of Hynix (black banks). I removed all of the Hynix sticks, leaving banks 1-4 for both CPUs populated with the Samsungs. When I tried to boot in this configuration I got the Amber 2,7 power LED which supposedly means a memory configuration error. Next I removed the memory from banks 3 & 4 for both CPUs and it went back to the solid white power LED with no video signal or power to the keyboard.

Next I’m planning to try 1 stick of the Hynix in bank 1 only for both CPUs just to rule out bad Samsung memory, but beyond that I’m not sure what else to try. I’ve also tried another video card that’s known to be working in my T7610. I supposed I could get two approved 4gb crucial modules to test but not sure it’s with the money and effort for such and old workstation. It’s still a great render box when I need one so I’d love to get another year or two of life out of it.

mazzinia_ · Answer

Hello, how much time have you waited ( after booting with the new cmos battery ) with the solid white led, before turning it off ? It may be 'training' the ram

spiderhole · Answer

@mazzinia_​ Very interesting theory!  Is it typical that there would be zero display during this memory training process?  I don't think I've left it siting more than a few minutes after powering on with the white solid light but on Monday I'll try to leave it alone for a half hour or so and see what happens.  I know testing of 128gb takes a while in the bios diagnostics tools so perhaps you are right.  Thanks for the suggestion.

mazzinia_ · Answer

@spiderhole​ to be honest I've not had memory training yet, so I don't know ... but i suspect it would not have a display

spiderhole · Answer

Waiting an hour for possible 'memory training' and also using the jumper to reset the cmos have not resolved my issues.  I've ordered a Quadro K600 card off ebay to see if the higher end GPUs I've tried might be the culprit (I've tried known working GTX 1070ti and Tesla GP100 cards).  Stay tuned!

mazzinia_ · Answer

@spiderhole​ well, not the result i was hoping to hear :( Do you remember if you had secure boot on ?

spiderhole · Answer

@mazzinia_​ I had it set to UEFI with 'Always, except internal hhd' the last time it posted successfully.  Keep in mind I have the boot hdd removed for now and simply getting it to post to the bios at all.

spiderhole · Answer

UPDATE!

While waiting on the K600 card to arrive I decided to swap out all four of the remaining 8gb DIMMs with different modules, put both side panels back on and turn fire up the system. It’s ALIVE!!! I’m not sure if it’s bad memory or maybe the bios won’t post if the case covers are not in place but I’ll soon find out as I do more testing. More to follow.

mazzinia_ · Answer

@spiderhole​ There's an intrusion switch that prevents the computer to work with the side cover removed. If you want to operate while open, you have to tape down the switch.

spiderhole · Answer

@mazzinia_​ What's interesting is that it ran before I changed the CMOS battery with the left side panel and front bezel removed.  I even have photos of it in that state with the bios screen active.  Maybe it was a fluke, or perhaps only the RIGHT cover panel is preventing boot when it's off.

spiderhole · Answer

@mazzinia_​ What I'm also wondering is if there should would typically onscreen message in the event of missing cover panel vs just a blank screen, which is what I was experiencing.

mazzinia_ · Answer

@spiderhole​ The message I don't know, but you would find a message in the bios log. This yes. And the intrusion switch is only on the side panel that's normally removed to access ram/cpu

spiderhole · Answer

@mazzinia_​ That's the panel I had off and was able to post before with it removed.  I wonder if the switch could be faulty and that might be the source of all my issues with it not booting randomly over the years.  Now THAT would be funny!

mazzinia_ · Answer

@spiderhole​ that could be, everything mechanical can have issues. It's a simple cable ( 2 pins on the mb ) going to the switch ( mechanical pressure switch ). You can even bypass it by unplugging the cable and bridging the 2 pins it connects to

jas3245 · Answer

@spiderhole did you happen to figure this out? I recently bought a T7910 as a relatively inexpensive vessel to hold a couple of GPUs so that I could tinker with local AI. It came with 32GB of RAM and accepted the two NVIDIA 3060s just fine.

I decided to beef it up a little and add it to my Proxmox cluster to run a few VMs on it when I'm not in an AI mood.

It is running dual E5-2695 V4 cpus, and the 32GB is comprised of four 8GB sticks in
CPU0 Ch0 DIMM1
CPU0 Ch2 DIMM2
CPU1 Ch0 DIMM1
CPU1 Ch2 DIMM2

(as per the chart HERE)

All of that works great.

I picked up four 32GB sticks and did a 1 for 1 swap in all four of the slots above, removing the 8GBs. Doesn't POST.

This new batch are Dell branded SNPC7GC which I believe checks all of the boxes that the T7910 requires (RDIMM and ECC) and they are 2400's which I understand is still within the limits of the T7910 (the 8GBs were 2133s).

I was sometimes getting the 2/7 amber code and sometimes just the solid white LED, but never any video. I've tried a second set (by chance also Kingstons but without the Dell label) and had similar results. The latter were used, the former were new.

I have tried every trick I can find.... The BIOS came with the latest A34, but I tried downgrading to A33.

I've tried pulling the CMOS battery and removing the power cord for 20 mins.

I've tried adding just 1 stick to slot CPU0 Ch0 DIMM1 (works if I do that with one of the 8GB, but never with any of the 32's).

I've tried just letting it sit there for 20-30 minutes when it is in one of it's "solid white" modes.

I've tried pulling out both of the 3060 GPUs in case they were pushing the power too close to the edge.

It came with an NVIDIA Quadro that I've been using for the actual display. I tried swapping that out with an old ATI Radeon that I dug up.

I've tried Secure Boot and not Secure Boot and a few other totally random BIOS options.

I've tried cussing at it. A lot.

I have NOT tried messing with the panel switch as mentioned above... My GPUs wound up being a little too tall and I can't get the cover back on, ha! But it has always booted fine with the cover off and the original memory installed as well as during the times when I was trying the single 8GB (it would raise the "memory amount changed" flag and I'd go into the BIOS to acknowledge it as normal when the cover was off). So I don't think that would be it, but easy enough to mark off the list, so I'll do that next chance I get.

Precision Fixed Workstations

T7910 won’t post - possible memory issues?

Was this post helpful?