Low Time Spy Scores - RTX 4090 STRIX OC - 30K Graphics

sk3tch

2[H]4U
Joined
Sep 5, 2008
Messages
3,401
Hi guys,

I have several similarly-configged rigs (others have 7800X3D - so basically the same although this guy should score way higher) that all "easy button" to a higher Time Spy score.

https://www.3dmark.com/spy/44597975

https://valid.x86.fr/16jfxx

I get only around 30K graphics score - config above and results above. Driver 546.33 (December 12, 2023). Other rigs get 37K graphics score so there is something seriously off here. Doing a compare - https://www.3dmark.com/compare/spy/44583772/spy/44638331 - everything looks like it should be OK the STRIX OC even clocks higher than the FE (as it should).

I have had this STRIX OC 4090 since the day of launch...can't imagine it's "tired" or something, lol...I run it with no OC and if anything I just do 120% power target.

sk3tch_4090_strix.gif


Things that I have tried:
  • Physical on-card BIOS switch verified on “P-Mode” (Performance)
  • Disabled G-SYNC
  • Disconnected 2nd monitor
  • Turned off Memory Integrity in Windows 11
  • Reboot
  • Messing with GPU Tweak III v1.7.0.7 (OC Mode, Default Mode, 120% power target, 100% power target)
  • Turned off "0dB Fan" and "Power Detect" in GPU Tweak III
  • Re-BAR: ON
  • XMP: ON (6000 MHz DDR5 - same as others)
  • Hardware-accelerated GPU scheduling: ON
  • NVIDIA Power management mode: Prefer maximum performance
  • Windows Power Setting: High Performance
Anything obvious that I am missing? It's a 7950X3D but I did a clean OS install and have kept up-to-date with the AMD driver updates - I also monitor everything using ParkControl Pro to make sure games are using the cache cores.

EDIT: just had this pop up (Windows Update replaced my AMD drivers for my onboard GPU) - maybe somehow this caused it? Trying the fix now - https://www.amd.com/en/support/kb/faq/pa-300.
EDIT2: nope - that wasn't it - equally crappy Time Spy score after installing and rebooting: https://www.3dmark.com/3dm/105526693
 
Last edited:
Only thing I can think of is that the benchmark is not running on correct CCD. Otherwise, not sure why your score would be that low. My TUF gets about 38K+ maxed out and overall Timespy score of 30K or something.
 
Only thing I can think of is that the benchmark is not running on correct CCD. Otherwise, not sure why your score would be that low. My TUF gets about 38K+ maxed out and overall Timespy score of 30K or something.
Thanks for the reply! Yeah I watch the CCDs in Park Control while 3DMark runs - all looks fine and the comparison shows about 15% more performance for CPU (over 7800X3D) so that jives.
 
That 120% power target probably means you're hitting the wall on the power limit. Set it back down to 100% and see what you get.
 
unless you have the exact same setup you will never get the same results. your power plan should be on balanced too, as per amd.


that also says 31k, not 37...
your within spitting distance, stop fretting and go game.
37K at the link. 30K with this STRIX. So it's a big difference. Nearly 20%.
 
OP do you have a max framerate set in nvpanel settings? This is usually recommended to help with VRR / Gsync but it will lower your score if enables for timespy
 
OP do you have a max framerate set in nvpanel settings? This is usually recommended to help with VRR / Gsync but it will lower your score if enables for timespy
Good suggestion! It is set to off. I am beginning to suspect ParkControl Pro. It changes the power plan settings on the fly...I have it disabled and I am gonna do a run.

EDIT: that wasn't it. https://www.3dmark.com/3dm/105575552

Also tried Armenius 's suggestion and left it at 100% power target - no change.

also did balanced power per pendragon1

I guess I will try a card swap this weekend. Very crazy. I can't imagine how I am going to pull off an RMA with ASUS if it turns out the card is just a 20% slower dog at this point. Ugh.
 
Last edited:
Good suggestion! It is set to off. I am beginning to suspect ParkControl Pro. It changes the power plan settings on the fly...I have it disabled and I am gonna do a run.

EDIT: that wasn't it. https://www.3dmark.com/3dm/105575552

Also tried Armenius 's suggestion and left it at 100% power target - no change.

also did balanced power per pendragon1

I guess I will try a card swap this weekend. Very crazy. I can't imagine how I am going to pull off an RMA with ASUS if it turns out the card is just a 20% slower dog at this point. Ugh.
Has the system and motherboard ever had a different AMD CPU in it, before the 7950x3D?
 
Has the system and motherboard ever had a different AMD CPU in it, before the 7950x3D?
Good call out - it did. 7950X - but I did a clean OS install for the 7950X3D and I monitor via ParkControl Pro that the core parking is working and games are hitting the cache cores.

After I swap GPUs - if the issue still persists (with known good) - then I will definitely do a fresh install. Must be some kind of gremlin in there!!! Thank you for the reply.
 
Good call out - it did. 7950X - but I did a clean OS install for the 7950X3D and I monitor via ParkControl Pro that the core parking is working and games are hitting the cache cores.

After I swap GPUs - if the issue still persists (with known good) - then I will definitely do a fresh install. Must be some kind of gremlin in there!!! Thank you for the reply.
did you do a good bios clearing after swapping cpus?
 
did you do a good bios clearing after swapping cpus?
Yeah it resets automatically. All 4 of my rigs have had CPU swaps in the last year (all X3D now) and the other 3 don’t exhibit problems.

CPU scores are good across the board it’s just this STRIX 4090 - seems like a cluster of CUDA cores died or something it’s always a 20% lower graphics score.
 
I guess I will try a card swap this weekend. Very crazy. I can't imagine how I am going to pull off an RMA with ASUS if it turns out the card is just a 20% slower dog at this point. Ugh.
I'm interested to see how this will go if you end up having to do it.
 
Hi guys,

I have several similarly-configged rigs (others have 7800X3D - so basically the same although this guy should score way higher) that all "easy button" to a higher Time Spy score.

https://www.3dmark.com/spy/44597975

https://valid.x86.fr/16jfxx

I get only around 30K graphics score - config above and results above. Driver 546.33 (December 12, 2023). Other rigs get 37K graphics score so there is something seriously off here. Doing a compare - https://www.3dmark.com/compare/spy/44583772/spy/44638331 - everything looks like it should be OK the STRIX OC even clocks higher than the FE (as it should).

I have had this STRIX OC 4090 since the day of launch...can't imagine it's "tired" or something, lol...I run it with no OC and if anything I just do 120% power target.

View attachment 626342

Things that I have tried:
  • Physical on-card BIOS switch verified on “P-Mode” (Performance)
  • Disabled G-SYNC
  • Disconnected 2nd monitor
  • Turned off Memory Integrity in Windows 11
  • Reboot
  • Messing with GPU Tweak III v1.7.0.7 (OC Mode, Default Mode, 120% power target, 100% power target)
  • Turned off "0dB Fan" and "Power Detect" in GPU Tweak III
  • Re-BAR: ON
  • XMP: ON (6000 MHz DDR5 - same as others)
  • Hardware-accelerated GPU scheduling: ON
  • NVIDIA Power management mode: Prefer maximum performance
  • Windows Power Setting: High Performance
Anything obvious that I am missing? It's a 7950X3D but I did a clean OS install and have kept up-to-date with the AMD driver updates - I also monitor everything using ParkControl Pro to make sure games are using the cache cores.

EDIT: just had this pop up (Windows Update replaced my AMD drivers for my onboard GPU) - maybe somehow this caused it? Trying the fix now - https://www.amd.com/en/support/kb/faq/pa-300.
EDIT2: nope - that wasn't it - equally crappy Time Spy score after installing and rebooting: https://www.3dmark.com/3dm/105526693
So, if it helps, I score almost 40K graphics score with my 4090 and my 5950X. So, I can confirm your setup has something going on...

https://www.3dmark.com/spy/41712070

Your clocks on the 4090 seem pretty low for "boosting", you could try with an OC on top of the power limit to see how that effects your score. I run mine at +120 core / +1200 memory for benchmarking, for gaming I run +100/+1000 respectively.

However, your 7950X3D seems to have an issue (although your speeds do not indicate that). My CPU/Physics Score + FPS beats yours by a large margin, and I am on a 5950X, not a 7950X3D.

What are your memory timings by chance on your DDR5? Do you have an AIDA64 memory benchmark you can post as well? Perhaps something is not configured right on that side of things...
 
So, if it helps, I score almost 40K graphics score with my 4090 and my 5950X. So, I can confirm your setup has something going on...

https://www.3dmark.com/spy/41712070

Your clocks on the 4090 seem pretty low for "boosting", you could try with an OC on top of the power limit to see how that effects your score. I run mine at +120 core / +1200 memory for benchmarking, for gaming I run +100/+1000 respectively.

However, your 7950X3D seems to have an issue (although your speeds do not indicate that). My CPU/Physics Score + FPS beats yours by a large margin, and I am on a 5950X, not a 7950X3D.

What are your memory timings by chance on your DDR5? Do you have an AIDA64 memory benchmark you can post as well? Perhaps something is not configured right on that side of things...
Thank you for your post. Maybe we are onto something! I run the same memory kit on all 4 rigs - https://www.gskill.com/product/165/393/1665020366/F5-6000J3040G32GX2-TZ5N.


aida_7950x3d.png


Didn't get that pop-up in my 7800X3D / RTX 4090 rig:

aida64_7800X3D.png


Also check out that North Bridge Clock! HALF!

Checked BIOS, disabled Hyper-V in Windows, hard-set North Bridge clock - same values and results. Strange. Probably a bad reading from AIDA64.

Gonna swap cards that will tell the tale for sure.
 
Last edited:
Good call out - it did. 7950X - but I did a clean OS install for the 7950X3D and I monitor via ParkControl Pro that the core parking is working and games are hitting the cache cores.

After I swap GPUs - if the issue still persists (with known good) - then I will definitely do a fresh install. Must be some kind of gremlin in there!!! Thank you for the reply.
Resetting the CMOS is always good. But, I mainly asked, because AMD says you can't swap from an X3D to a regular Ryzen, without an OS reset/reinstall (something about the X3D chipset driver can't be removed and it messes up regular Ryzens). And I thiiiink it may also be the same if you swap between single CCD X3D and dual CCD X3D. And you said you have some 7800X3D. So I was wondering if you'd done some musical chairs between CPUs and Mobos, without OS resets.
 
Resetting the CMOS is always good. But, I mainly asked, because AMD says you can't swap from an X3D to a regular Ryzen, without an OS reset/reinstall (something about the X3D chipset driver can't be removed and it messes up regular Ryzens). And I thiiiink it may also be the same if you swap between single CCD X3D and dual CCD X3D. And you said you have some 7800X3D. So I was wondering if you'd done some musical chairs between CPUs and Mobos, without OS resets.
Great call out - I did make sure for the 7950X3D to do a clean install. For 7950X to 7800X3D (move I made recently) - I did not - since the CCDs all have cache.
 
Swapped out an MSI Gaming Trio RTX 4090 from a “known good” rig (37K graphics in Time Spy) - I feel like it’s good news that the results were the same - 30K graphics:

https://www.3dmark.com/3dm/105630182

That means no hardware issues but some kind of software gremlin which means I don’t have to deal with ASUS RMA. I guess a clean Win11 install is in my future…

Pics of the surgery since you’ve read this far. Thanks!



IMG_4844.jpeg

IMG_4846.jpeg
 
Swapped out an MSI Gaming Trio RTX 4090 from a “known good” rig (37K graphics in Time Spy) - I feel like it’s good news that the results were the same - 30K graphics:

https://www.3dmark.com/3dm/105630182

That means no hardware issues but some kind of software gremlin which means I don’t have to deal with ASUS RMA. I guess a clean Win11 install is in my future…

Pics of the surgery since you’ve read this far. Thanks!



View attachment 626750

View attachment 626751
Well, the hardware issues COULD be within the system that is holding those GPUs back... it's just not a potential GPU issue... lol.
 
This fixed or what? I ran timespy for first time on my 7800X3D and scored over 30K with omgrafix score of 38K+. FYI only.
 
This fixed or what? I ran timespy for first time on my 7800X3D and scored over 30K with omgrafix score of 38K+. FYI only.
Nice, thank you!

Not yet - did a backup today with Macrium and will reinstall Win11 tomorrow or something. Frankly I'm relieved it's not the graphics card. I am wondering if it's the latest ASUS BIOS (I am super militant about being on the cutting edge BIOS) that has a patch for vulnerabilities:

https://rog.asus.com/motherboards/rog-crosshair/rog-crosshair-x670e-gene-model/helpdesk_bios/

ROG CROSSHAIR X670E GENE BIOS 1807

SUPPORTTABS_VERSION 1807
11.5 MB 2023/12/15
"1.Update AGESA version to Combo AM5 PI 1.1.0.1
2.Improve system stability
3.This update includes the patch for the LogoFAIL vulnerabilities.
4.Support graphics card with M.2 storage
 
People always suggest reinstalling windows but typically it's just a waste of time and these kinds of issues are hardware or bios related.
Yeah. I mess around with virtualization quite a bit so I figured a fresh start may get rid of some old BS that I'd need to find with a fine-toothed comb otherwise.

Now I'm like...mobo issue? It's crazy. I literally flash my BIOS to the latest version (did it with my X670E ProArt too - so "same" BIOS with mitigations and I'm getting a 37K graphics for that 4090 in Time Spy) and flip on EXPO I and start Windows. Everything seems to work just fine except a 20% slower GPU...one that I have tested other GPUs on and they exhibit the same behavior inside of this rig.

Damaged PCI-E connection? The GPU-Z above shows I'm at the fastest connection speed. Not even sure how that would play out. I don't get any graphical glitches in games or in 3DMark which I guess would indicate a bad connection?
 
Yeah. I mess around with virtualization quite a bit so I figured a fresh start may get rid of some old BS that I'd need to find with a fine-toothed comb otherwise.

Now I'm like...mobo issue? It's crazy. I literally flash my BIOS to the latest version (did it with my X670E ProArt too - so "same" BIOS with mitigations and I'm getting a 37K graphics for that 4090 in Time Spy) and flip on EXPO I and start Windows. Everything seems to work just fine except a 20% slower GPU...one that I have tested other GPUs on and they exhibit the same behavior inside of this rig.

Damaged PCI-E connection? The GPU-Z above shows I'm at the fastest connection speed. Not even sure how that would play out. I don't get any graphical glitches in games or in 3DMark which I guess would indicate a bad connection?
Definitely could be a mobo issue. Have you done some in-game benches to see if it's ~20% or have you solely been using timespy?
 
Definitely could be a mobo issue. Have you done some in-game benches to see if it's ~20% or have you solely been using timespy?
I haven't - any ideas? I just eyeball Fortnite in-game FPS - it looks around OK but hard to tell since I game 1080p/360Hz on my other rig.
 
I haven't - any ideas? I just eyeball Fortnite in-game FPS - it looks around OK but hard to tell since I game 1080p/360Hz on my other rig.
Cyberpunk, tomb raider are two I recall off the top of my head, I'm sure other people might be able to chime in.
 
For what it's worth, my system is also a 7950X3D and a 4090 and my normal "vanilla" (aka. the free one) TimeSpy graphics score is only a hair under 31K. That's with straight out of the box settings for the graphics card and no CPU overclocking at all beyond turning on EXPO for the RAM.
 
Here's what mine pulled off 2 fresh runs after a reboot. I reset as many of the Nvidia settings as I could (turned off gsync, reset 3D settings, etc.) within reason.
The first one was entirely at stock speeds and the second one is with the Gigabyte GCC basic overclocking enabled.

https://www.3dmark.com/3dm/105911824
https://www.3dmark.com/3dm/105912184

That's with the free stand-alone (not via Steam) 3DMark installed and no settings changed. Everything finishes fine. Wild guess, it's probably a setting somewhere. I don't think it's a hardware thing.
 
Last edited:
Could be some of these things: https://www.reddit.com/r/buildapc/comments/109ec6z/4090_low_timespy_graphics_score/?rdt=41086
Granted, it's Reddit, but there are people claiming differences between Windows 10 and 11, plus several settings that seem to affect scores in a major way.

EDIT: Uninstalling virtualization made my graphics score jump up to 36.9K immediately with no other optimizations at all, so that's probably the culprit for most everyone who seems to have otherwise bulletproof hardware. It kills off your ability to run Android software in Windows 11, so keep that in mind. That probably only affects a few people.
 
Last edited:
Could be some of these things: https://www.reddit.com/r/buildapc/comments/109ec6z/4090_low_timespy_graphics_score/?rdt=41086
Granted, it's Reddit, but there are people claiming differences between Windows 10 and 11, plus several settings that seem to affect scores in a major way.

EDIT: Uninstalling virtualization made my graphics score jump up to 36.9K immediately with no other optimizations at all, so that's probably the culprit for most everyone who seems to have otherwise bulletproof hardware. It kills off your ability to run Android software in Windows 11, so keep that in mind. That probably only affects a few people.
Thanks! I have studied that thread extensively. :) You may be able to tell by my earlier posts. I do appreciate everyone chiming in. Very helpful. I have more things to try now - I am traveling but back home tonight to futz with it. I am going to sell my 4090 STRIX OC since I picked up a 4070S FE yesterday. Gonna cut down on my 4090 investments, lol.
 
A few things:

Try 100% power target if you haven't yet already.

Low Timespy scores can be CPU related, even on the GPU test. So anything that can affect the CPU performance could affect the score:
Ensure Ram is configured for dual-channel or quad channel, whatever mobo supports/how many modules you have.
This could be a case of a bios issue where the settings appear correct, but are not actually applied. And this is settings specifically concerning RAM. So load default values on the Bios and then Save the changes. Then reconfigure everything, including the RAM XMP settings, Rebar.
If that doesn't help, turn off XMP and try manual DDR settings at 1 step slower clocks. So if it's 6000 Mhz now for that XMP profile, Try 5667Mhz If that's what the next slowest DDR5 modules would run at.
If that doesn't help, I would roll back the Bios to the previous version you were running.

Did you try yet with cpu virtualization optimizations turned off?
 
Back
Top