Theres bad cooler contact on his vram.
Memory clock and Memory clock average should always match.

the_Scarlet_one
Can you post the actual benchmark scores, especially the links to the benchmarks, so that people can look at what is happening?

This is his https://www.3dmark.com/pr/777893
 
Here is mine https://www.3dmark.com/pr/777889
 

Yup. Actually i think he's using the xoc bios and I'm on stock, but the cards are the same model.

DDU drivers and reinstall fresh. Close PX1 during runs. Maybe reinstall Windows.

SeanDude05
DDU drivers and reinstall fresh. Close PX1 during runs. Maybe reinstall Windows.

Completely new windows install produced no change

bradsour
Definitely seems like his is having a memory issue. Are you using default stock settings on both or are they OC'd? Are you running stock BIOS or the 450 Watt BIOS?

I'm running stock, they haven't changed anything. Xoc and stock bios produced no difference.

vivagil
Yup. Actually i think he's using the xoc bios and I'm on stock, but the cards are the same model.

use the oc bios for better fan curve....your thermals are killing your performance

bmx045

vivagil
Yup. Actually i think he's using the xoc bios and I'm on stock, but the cards are the same model.

use the oc bios for better fan curve....your thermals are killing your performance

it's kind of strange because on equal settings in game and benchmarks mine is actually performing better even while hotter. The only difference in cooling is that he has an aio and I'm on good ole reliable cooler master hyper 212 evo.

My best estimation is that something isn't making proper contact with his heatsink/cooler on the gpu.  Best guess, memory (VRAM).  The great thing about having the ftw3 is you can check the icx3 in Precision X1.  It should give temperature sensors across the board.  See if one is WAY hotter than others. Or too cool even.  Basically, we don't want to see a large delta between temperature sensors on the same "pieces" (aka Memory, or Power, etc.).  A large delta would mean something is wrong (bad mount, bad chip, dead chip, bad power delivery for examples).  At the hottest, last I checked, my delta was around 3C between the three different memory sets and I think 5C on the two die sensors.  And, Ignoring the performance metric for a sec, you can run X1 during the benchmark with the ICX sensors showing to make sure.  I think X1 has an always on top setting, but if not, you can run the benchmark windowed in a custom run.  Feel free to report back the ICX temps.
 
That said, Neither one of you are running your RAM at 3200MHZ.  They appear to be running baseline for the motherboard.  Just putting that out there.

I wonder if his card is affected by the PCI-E load balancing power draw issue that an alarming large number of FTW3's seem to suffer from. This would explain the 58C operating temperature as well. I suggest monitoring power draw either with Killawatt from the socket or Hwinfo64. 

arestavo

vulcan1978
I wonder if his card is affected by the PCI-E load balancing power draw issue that an alarming large number of FTW3's seem to suffer from. This would explain the 58C operating temperature as well. I suggest monitoring power draw either with Killawatt from the socket or Hwinfo64. 

I think that's primarily a 3090 issue.

It's more likely a CPU/RAM bottleneck. SMT disabling and faster RAM will help. An easy test would be to just swap the cards around and see if the problem follows - or stays with that CPU/RAM combo.

Even bottlenecked he shouldn't be scoring 9500 in Port Royal with a 3080.  I mean, I'm running a 3900x for the moment, albeit with my ram at 3600.  And I'm scoring 11700 with all my background crap running.  9500 is way too low.  A CPU/RAM bottleneck should not be making a 2000 point difference in that benchmark, since it is so heavily GPU dependent and not so much CPU/RAM.  I would still like to see an ICX screenshot with a continuous benchmark run/stress test going.  
 
That said, never a bad idea to swap gpu's in this instance.  Easy way to eliminate any outside factors other than gpu.  I mean, it could be something wrong on the motherboard itself even. A swap would settle the issue of GPU or not either way.

arestavo

vulcan1978
I wonder if his card is affected by the PCI-E load balancing power draw issue that an alarming large number of FTW3's seem to suffer from. This would explain the 58C operating temperature as well. I suggest monitoring power draw either with Killawatt from the socket or Hwinfo64. 

I think that's primarily a 3090 issue.

It's more likely a CPU/RAM bottleneck. SMT disabling and faster RAM will help. An easy test would be to just swap the cards around and see if the problem follows - or stays with that CPU/RAM combo.

It's a possibility, I didn't realize PR was CPU or memory dependent: 
 
5950x slower In Port royal Benchmark then 9900k | Overclock.net

arestavo

vulcan1978

arestavo

vulcan1978
I wonder if his card is affected by the PCI-E load balancing power draw issue that an alarming large number of FTW3's seem to suffer from. This would explain the 58C operating temperature as well. I suggest monitoring power draw either with Killawatt from the socket or Hwinfo64. 

I think that's primarily a 3090 issue.

It's more likely a CPU/RAM bottleneck. SMT disabling and faster RAM will help. An easy test would be to just swap the cards around and see if the problem follows - or stays with that CPU/RAM combo.

It's a possibility, I didn't realize PR was CPU or memory dependent: 
 

It is up to a certain point, then it doesn't really scale better. Well, maybe a couple of points (literally) with a higher OC/RAM speed - but there is a definite, hard drop off point.

All this talk about it being the cpu and ram. For giggles, I turned my ram down to 2133 this morning. As a reminder, my machine is a 3900x with smt still on, ram now at 2133, and a 3080 running stock for this instance. I ran port royal. I got 11463. So I dropped 350ish points from moving my ram way down from 3600. But it's not 2000 points like OP brothers Port Royal score.

Just switch the cards in the machines, let's confirm its the gpu. Then probably just get an RMA going.

maybe the difference in cpu