12900k Cinebench score extremely low

Lankoss · Monday at 10:00

PC Specs:
CPU -I9 12900k
MOBO- Rog Maximus Z690 Formula
GPU - Zotac rtx 3090
RAM - 64GB Dominator Platinum (5600mhz)
Cooling - CPU and GPU water cooled with Corsair Hydro

Context:
I bought a PC yesterday second hand. On face value everything looks okay but after running Cinebench when I got home I've found that the CPU score is only 1400 (it should be 28,000), so something is definitely off. Below is some context and what I've tried so far.

When I went to view the PC, it would turn on no problem but wouldn't give a video out from either the GPU or the MOBO, I was told that it had no issues turning on days prior as the HDD's were wiped and a fresh install of Windows installed by a local IT guy at his shop and brought back. I did speak to the guy on the phone and all parties did seem genuine so I figured maybe the MOBO got damaged in transit so I decided to buy the PC anyway as it was a good price. After getting it home and trying to boot it I found the MOBO had a screen and showed a Memory error, I checked and sure enough one of the DIMMS wasn't full seated, after doing that it booted no problem and it was indeed a fresh install of windows.

First thing I did was install the latest Nvidia driver direct from the website then did all the Windows updates, once those were all done I updated the bios to the latest version and restored the MOBO to factory settings after, which I believe is the correct thing to do. Once that was all done I did the Cinebench and got the score of 1400. After that score I went back to the bios and enabled XMP on the RAM as I was planning on doing a before/after anyway, but got pretty much the same result score wise.

Things I've checked:

CPU temps look okay, the highest temp I see after a 10min run in Cinebench is 83C, it was mainly around 80c or below so from what I could tell, it didn't appear to be thermal throttling, i've no idea what else to look for when benchmarking so not sure how else to tell if the CPU is throttling besides its temperature.
Its a fresh install, so I dont appear to have any programs running in the background hogging CPU bandwith, the only things I have open are CPUID and Hardware Monitor, I get the same score regardless of whether they are open or not.
I installed intel software to check the CPU was genuine, which it passed without issue, so i'm reasonably sure its a genuine 12900k.
Ran Cinebench on the 3090 gpu and it was 28k+, so that seems to be fine.

Can someone help point me in the right direction with what to look for and how best to look for it? My main ideas for what to check next is seeing what wattage the CPU draws and the voltage, but I'm not sure how best to do that and what I should expect to see. My other main concern is what to check CPU wise in the bios settings, my experience with bios's is pretty much limited to going to EzTuner and enabling XMP. I have my other PC still with Discord on it, so there is an option for someone knowledgeable to talk me through what to look for in real time if someone can spare the time, or I can just reply on here as and when people suggest things to check, it may be slow going though. I'll update the list above as I do more checks as I have nothing on today, so i'm going to plug at it.

Edit: I found an option to run MemTest86 within the bios (weird as I thought it had to be run via a USB) so i've been running MemTest86 while typing this up and it's come back with errors, none on the first pass but on pass 2 it got to 10 errors and I cancelled it, below is the prudent info from it if someone can help interpret it.

Tests passed 16/17 (94%)
Lowest error address: 0x1FF21D1AC (8178MB)
Highest error address: 0xC3E25D1AC (50146MB)
Bits in error mask: 0000000001400000
Bits in error: Total: 2 Min: 0 Max: 1 AVG: 1
Max Contiguous errors: 1
CPUS that detected memory errors: 0, 1, 4

I'm going to try a few different RAM variations and try Cinebench again.

Lankoss · Monday at 11:11

I've ran different variations and below are the results in a single pass of Cinebench (Multi Core), RAM sticks are numbered 1,2,3,4.

RAM 1+2 in slots 1+3 = 1458
RAM 1+2 in slots 2+4 = 1481
RAM 3+4 in slots 1+3 = 1461
RAM 3+4 in slots 2+4 = 1478

Rather than carry on I decided to try PassMark Performance test instead, ran that on the CPU and got a score of 42,404 which is right inline with the 12900k so I'm perplexed to be honest. The PC "seems" normal, its just the Cinebench score that is throwing me off. I am going to download Steam and put Wukong benchmark on and try that with the 2 Ram Dimms.

PaulM · Monday at 11:24

Are you using the latest version of Cinebench (2024)? If so it looks like ~1500 is to be expected in multi core score for a 12900k, the ~28k score would be the older version Cinebench R23.

https://nanoreview.net/en/cpu/intel-core-i9-12900k

Lankoss · Monday at 11:35

PaulM said:
Are you using the latest version of Cinebench (2024)? If so it looks like ~1500 is to be expected in multi core score for a 12900k, the ~28k score would be the older version Cinebench R23.

https://nanoreview.net/en/cpu/intel-core-i9-12900k

I am using 2024 yeah. Mannnn, that exact thought did occur to me yesterday that the score could be fine and that the version could make a difference, but everytime I googled it the only info I could find on expected score was the 28k one. You're a legend for letting me know that Paul, that explains it. I have just run BM Wukong Benchmark and its come back suggesting Very High settings which is also in line as it suggests medium settings with my older rig. I am going to reinstall all 4 dimms, try again, then try again with XMP enabled as I am dubious whether there is anything wrong with the RAM either, I haven't notice any stability issues with the RAM installed with XMP enabled prior to running the MemTest and i've read there are reasons for errors being thrown up besides actual RAM issues so it may be a false positive.

Threepwood · Monday at 12:36

whack all the memory sticks in, and do a mem test, then you will know 100% without having to mess around with the memory any more too!

ED209 · Monday at 12:51

Check Windows and make sure all the cores are enabled\detected.

If not check the bios or clear the bios completely and restore to defaults in case of anything the previous owner had done

Lankoss · Monday at 14:08

Threepwood said:
whack all the memory sticks in, and do a mem test, then you will know 100% without having to mess around with the memory any more too!

I did run a Mem test in the Bios earlier this morning and it was throwing up errors, everything seems stable when I run in game benchmarks however. After some googling it seems my board has issues with 4x16gb DIMMS when in XMP mode and causes boot loops and stability issues, which I did witness. I have slightly increased the voltages and was able to boot back into windows with XMP enabled after that. Once my longer kettle lead arrives I can do some proper gaming and see if the problem persists or not.

Tetras · Monday at 16:05

Lankoss said:
I did run a Mem test in the Bios earlier this morning and it was throwing up errors, everything seems stable when I run in game benchmarks however. After some googling it seems my board has issues with 4x16gb DIMMS when in XMP mode and causes boot loops and stability issues, which I did witness. I have slightly increased the voltages and was able to boot back into windows with XMP enabled after that. Once my longer kettle lead arrives I can do some proper gaming and see if the problem persists or not.

That's very unlikely to be board specific, running 4 sticks of DDR5 has been problematic since it was released, though later boards/BIOS have helped somewhat.

Increasing the voltages is a standard fix (even on DDR4, that was sometimes necessary), but you could step down the frequency a little (still with XMP enabled).

The expected problems are reflected in Intel's official spec, which only rates the memory controller to 4000 with single rank sticks or 3600 with dual rank sticks.

Lankoss said:
Edit: I found an option to run MemTest86 within the bios (weird as I thought it had to be run via a USB) so i've been running MemTest86 while typing this up and it's come back with errors, none on the first pass but on pass 2 it got to 10 errors and I cancelled it, below is the prudent info from it if someone can help interpret it.

A small number of infrequent errors is usually a config issue (or OC instability) rather than faulty memory.

Lankoss said:
when I got home I've found that the CPU score is only 1400 (it should be 28,000), so something is definitely off.

As mentioned above, scores depend on version.

Lankoss · Monday at 19:43

Tetras said:
That's very unlikely to be board specific, running 4 sticks of DDR5 has been problematic since it was released, though later boards/BIOS have helped somewhat.

Increasing the voltages is a standard fix (even on DDR4, that was sometimes necessary), but you could step down the frequency a little (still with XMP enabled).

The expected problems are reflected in Intel's official spec, which only rates the memory controller to 4000 with single rank sticks or 3600 with dual rank sticks.

A small number of infrequent errors is usually a config issue (or OC instability) rather than faulty memory.

As mentioned above, scores depend on version.

Thanks, I will leave it at 5600mhz for now and see how I get on, if I still have stability issues I will step it down in stages and see if that sorts the problem. I only increased the voltage very slightly, so I have more headroom there also but i'm more comfortable with decreasing the frequency slightly then I am increasing the voltage.