Not the most likely solution, but try to blow out the memory slots with an air duster / can of compressed air. Sometimes dust gets in the memory slot and prevents it from communicating correctly.

Check the manual on how your 8x 23GB memory modules should be slotted.

https://www.dell.com/support/manuals/en-nl/oth-xlt7920/precision_7920_om_pub/memory-configuration?guid=guid-ce1561c8-88a9-45e3-bc3b-762cbb37e073&lang=en-us

Remove OEM modules and test with new memory modules only to see if the system will accept them.  Then OEM can be added back.

Memory configuration with DIMM slot locations for Precision 7920 tower.

https://www.dell.com/support/manuals/en-us/oth-xlt7920/precision_7920_om_pub/memory-configuration?guid=guid-ce1561c8-88a9-45e3-bc3b-762cbb37e073&lang=en-us

@TheQuickFox​ Thanks for quick reply. I don't think it is dust. Always blow before install. The "wiggeling technique" doesn't help either. Also, I've put them in exactly like the manual. That is 1, 2, 3 and 4 on each CPU. Double checked and tripple checked.

Tried now booting with five modules, which not particularly surprising doesn't work.

Btw, how long does the first boot with memory check supposed to take after installing 256 GB? After how many minutes can I call it a failiure? It is very stupid design with no feedback.

@Chino de Oro​ I've done that. Tried all the modules pairwise. There is nothing wrong with the new memories and the system accepts them individually. They all have the same part number.

I've now managed to get up to seven modules installed on the mobo and boot into windows (after a lot of trail and fails), but after adding the eight it won't boot. Period. I also tried adding seven, boot, shutdown, replace one module, boot, shutdown and then use the module that the system just booted with in the eight slot, only to fail.

I reckon either the socket or the CPU is deffective!?

Otherwise, this system is superwired. Did a full diagnostics with memory test (6 modules) with a successfull result and no errors. Awesome, except when I close the test, the computer refused to boot up again. Had to cut the power and do a couple of power cycles , before it came back up. It seems to be totally random.

Also in CPU-Z don't report timings table for the modules. Is this normal?

I have the exact same system and a similar problem. I am able to populate 6x16GB in all of the white slots (total 192GB) and the system boots up and works. The minute I start populating the black slots (first to 256GB, then to 384GB), as indicated in the manual. It does not boot past POST. Weird things happen. The power LED was white for a while without POST, and then suddenly all the fans came on like a blower. I let it sit for a bit, but am wondering what the deal is?

Update. Took out all the memory modules and re-seated them one-by-one, carefully. Rebooted. It eventually showed the DELL Logo. I was able to confirm that 24 16GB modules are all working and 12 channels are enabled. Has to be my sloppy hands seating some of the DIMMs incorrectly.

The way this system deals with memory problems is absolutely abysmal and perhaps the single most terrible way any manufacturer has implemented dealing with such issues. I have a Dual 6140 which has given me incessant, repeated but varied grief when it comes to starting up, how long the boot might take THIS time, whether the LED will indicate a problem... or not, etc. etc. Here is a list of issues that may or may not come to pass as you merely just want to start USING your $5000+ computer:

- Ages to get a screen with DELL logo showing
- Ages for the blue bar to "finish finishing" and letting OS boot
- Stuck/frozen not accepting KB inputs ad aeternam
- Delayed switching on or off of the NumLock key (10-15 seconds after pressing)
- Not indicating F12 was pressed, or being "too late" to do so, even though the message shows
- Just not showing you anything. Sittin' there, chillin'...
- Booting into some test instead (fan test is NOT ram related!)
- Showing some led codes (x White + y Amber)
- Giving NO data on what DIMM is the offender (yes, I enabled all the Bios logging!)
- Having NO LED's on the motherboard to indicate issues
- WHEA errors in the Event Log (needless to say, not telling you module number)
- The DIMMs working fine in a HPE DL Gen9 server (and that one tests well!)

A spring loaded sausage with a Peruvian bee inside would give me more data on what the problem might be than this thing does. But fear not! It DOES have at least a dozen sensors to annoy the bejeezus out of you when one fan is missing! Coz THAT'S important!

Came across this thread after many hours of troubleshooting. I'm currently trying some of the suggestions here that I have not currently attempted. I've been trying to get my system to come back up for 2 days now after adding a second CPU and a small amount of RAM. The one bright side I see is I didn't spend 5k+ on the 7920, and only wasted 1k on the massive paperweight. 

Wish I would have known this before purchase.

@MrServer  I think you summed up everything I would want to say. If anyone comes across these BEFORE buying, save yourself the trouble, and do not buy these. Nothing is worth this hassle. 

@jdhensley Cheers to that - let me add a bit more nuance to my struggles so it is an accurate reflection of the facts. It just so happens that in the last month I moved around a lot and so the computer had to be packed up and rearranged a few times. Every single time after setting my gear up at the new place there were problems. To be clear - I am extremely careful with my gear; I've flown EU-AUS 4 times with a camera bag full of HDD drives, never had a single failure (but a lot of questions and tests for explosive residue at customs each time...) so, rest assured the issues were not due to rough handling...

Mitigating factors beyond DELL's control
If you look at eBay, there are quite a few socket 3647 CPUs that are sold as half-defective with 1 or 2 non-functional memory channels. I suspect this series of processors is particularly sensitive when it comes to the memory used. The super fidgety and fragile CPU socket is not helping things either - so far I've done about a dozen CPU replacements (on SuperMicro, HPE and Dell) and though for me they all went fine it is easy to imagine the smallest mistake like having unsteady hands, inserting the CPU the wrong way around (WHY is pin 1 so vaguely indicated!?) or the flimsy plastic clip not fitting smugly resulting in terminal damage to either the socket or the chip. Frankly, it's a miracle it even works at all.

Guesswork: Hopeless Diagnostics
After 3-4 instances of me nearly tossing the damn thing out of the window, it was revealed that I had a single 32GB PC4-2933 Hynix RAM module that exhibited correctable ECC errors. This crucial bit of information became known when, at some point, after the 4th move and it again giving me led codes (with nothing changed after power down at the previous place!) the 7920 spontaneously booted into a RAM test. Hallelujah! CPU0_DIMM2 it said. Well I'll be lost...

Nothing in the DELL manual or documentation suggests, indicates or recommends doing a RAM test when having to deal with the problems mentioned in my previous post... you just gotta get lucky.

The Supreme Source of Confusion is that, apparently, WHEA/single bit correctable ECC errors only start to happen after a while; I had seen WHEA errors in Windows Event Log and swapped 4 out of 8 of the DIMMs and for the first day or two no new WHEA errors were logged.  All is well now, I thought. Alas, it was the "wrong" 50% of DIMMs. So I replaced the other set of 4 Hynix modules with 4x 32GB Samsung PC4-2666 and voila, Calculus, the God of Computation hath blessed this cursed device and allowed it to finally boot like a normal PC. No more WHEA errors, it's been 2 weeks now.

Extremely limited RAM options
The list of permitted memory configurations for this system (and the 7820, don't even get me started on THAT thing with it additional CPU board!) is very small. Forget using DIMMs of more than one size. Ye shall limit thyself to 4x, 8x or 12x or 16x modules.  You can't add 48GB per channel, nor replace 1x 32 with 2x16GB for one socket/side. I'm almost amazed it now works with 4x PC4-2666 and 4x PC4-2933 32GB from two (!) different manufacturers! It's a Miracle, I tell ya!

When I had the audacity to add another 4x 32GB (matching one of the other sets of 4) showing a screen was no longer a feature that was available. I strongly recommend you just stick with 32GB modules though - other humans have had assorted problems with installing more than 12 16GB sticks (and once you HAVE 12/16 DIMMS, the only option to upgrade is to either replace ALL modules by bigger ones, or installing another 12/16, which will duly give you plenty of junk!)

Once the damn thing actually works... it works well. It's just that the diagnostics are WOEFULLY inadequate. Pesky. Petty. Temperamental. Vague. Uninformative. Prone to one making random guesses, or contemplating ritual disembowelment.