Overclock.net banner

Performance Limit Thermal/perfcap Reason thrm on load change RTX 3090

7493 Views 9 Replies 5 Participants Last post by  kiriakos
Greetings Enthusiasts,

i have a reproducible behavior with the Performance limit Thermal flag being triggered on load change of my GPU, maybe someone is experiencing the same or can reproduce the same with a simple testing

GPU: MSI RTX 3090 Suprim X 24GB, no oc, gaming Bios
CPU: Intel I9-9900k stock
Motherboard: Gigabyte Aorus Master Z390, F10 Bios
RAM: GSkill Trident Z 64GB(4x16) DDR4 3200MHz, XMP enabled, no overclock F4-3200C14D-16GTRS

PSU: BeQuiet Dark Power Pro 11 1200w, 4x12v Rails, 12V1/2(CPU/Mobo/disks etc) 35 Amps each Rail, 12V3/4 (pcie) 45 Amps each Rail

Operating System & Version: Win10 Pro 1909, clean install

GPU Drivers: using ddu when changing drivers, currently on 461.40

whenever there is a larger load change the Performance limit Thermal is flagged yes (hwinfo) or perfcap reason thrm in GpuZ, however neither core Temps nor VRAM temps are anywhere near the throttling threshold.

MSI Afterburner spikes on the temp limit graph, at the same time i saw a drop in Power % but non on core or mem clock graphs.

for example during Benchmarking (3Dmark, ROTTR/SOTTR, RDR2, Division2, Ghost Recon Breakpoint etc.) thermal flag is tripped in HWinfo and flashes in GpuZ sensor Perfcap Reason Thrm, when the benchmark run finishes and you return to the menu, or during gaming session when you exit to menu, so basically when the load on the GPU drops.

first i thought that would be due to the VRAM temps that could not be monitored until recently (thanks HWINFO!!) but now i can see that when this happens the temps are well within limits and way below any throttling threshold.

during testing i found that this behavior started with the 457.x driver branch as the 456.xx did not exhibit this behavior.
i always did a clean install (using ddu) of the follwoing driver versions

461.09: initial driver after i received the card GPUZ sensor records thrm PerfCap Reason,HWinfo shows Performance Limit Thermal Flag tripped with YES
461.33: GPUZ sensor records thrm PerfCap Reason,HWinfo shows Performance Limit Thermal Flag tripped with YES
461.40: GPUZ sensor records thrm PerfCap Reason,HWinfo shows Performance Limit Thermal Flag tripped with YES, HWINFO got updated to include VRAM Temps readout and here i could see that the temps are well within limits (below 90 degrees Celsius)

after researching i found the specs for the mem modules
D8BGX list operating temps state 0-95, throttling should occur around 110 degrees according to Igors Lab research
D8DGX are rated for 1313 MHZ but RTX 3090 clocks them lower at 1219MHZ

I stumbled upon a EVGA forum thread from December, that basically mentions this behavior to be driver related, so i tested
EVGA GeForce RTX 3090 FTW3 XOC BIOS BETA

also on another forum (i posted there as well) other folks with rtx 3080/90 see this as well



456.98 hotfix: GPUZ sensor does not record PerfCap Reason Thrm, nor does HWINFO Flag the Performance Limit Thermal as Yes during the same testing sessions, also temps on the GPU where slightly lower with 2 degrees (68 instead of 70) however VRAM temps are the same as on the other drivers
456.71 whql: GPUZ sensor does not record PerfCap Reason Thrm, nor does HWINFO Flag the Performance Limit Thermal as Yes during the same testing sessions

i would have stayed with the 456.98 however RDR2 was freezing when entering the game world

out of curiosity I also tested the
457.51: same result as with the 461.x drivers

from the EVGA forum i take it that not only my RTX 3090 is affected but various 3080s as well .

it would be great if users would share their observations or can report if they experience something similar.

thanks!
1 - 10 of 10 Posts

· Facepalm
Joined
·
11,677 Posts
This is a known bug.
This started on the first 457.xx driver branch.
This is not exclusive to eVGA. Founder's Edition cards have this happen also. Yes it is indeed on a load change.
eVGA cards record a temp up to 6500C when this happens.
 
  • Rep+
Reactions: fozzybear

· Registered
Joined
·
30 Posts
Discussion Starter · #3 ·
This is a known bug.
This started on the first 457.xx driver branch.
This is not exclusive to eVGA. Founder's Edition cards have this happen also. Yes it is indeed on a load change.
eVGA cards record a temp up to 6500C when this happens.
many thanks for your reply!
has this been confirmed somewhere as a bug by any chance? did someone acknowledge that as AIB vendor or Nvidia ?
from what i have read in various forums i have seen Zotac, evga and MSI cards and now that you mentioned the FE cards, Nvidia as well
 

· Facepalm
Joined
·
11,677 Posts
many thanks for your reply!
has this been confirmed somewhere as a bug by any chance? did someone acknowledge that as AIB vendor or Nvidia ?
from what i have read in various forums i have seen Zotac, evga and MSI cards and now that you mentioned the FE cards, Nvidia as well
It's obvious it's a bug. 6500C is hotter than the surface of the sun.
 

· Registered
Joined
·
30 Posts
Discussion Starter · #6 ·
It's obvious it's a bug. 6500C is hotter than the surface of the sun.
yes of course, i am not arguing here. i just wanted to know if any vendor acknowledges this, that would give a larger incentive to nvidia to work on the driver and fix it there or in the vbios.

afaik evga is the only aib to have extensive temp monitoring.
msi and many other unfortuately do not.
so the only currently available software reading vram and gpu temps would be hwinfo, at least i am not aware of another, and no temps in "sun-surface" region are recorded.

it would be great if you can share where you got that valuable info so i can read up on it myself.
 

· High Clocker
Joined
·
3,763 Posts
Performance limit Thermal flag is set by,
Core going over set slider.
Mem going over 110C
VRM going over 120C
Driver triggers. (Nvidia made it so switching between full screen and windowed when in DX12 throws the flag, Starting Vulkan load throws the flag. And a few programs that were reported back in the nvidia forum for causing black screen crashes).
 

· Registered
Joined
·
437 Posts
Driver triggers. (Nvidia made it so switching between full screen and windowed when in DX12 throws the flag, Starting Vulkan load throws the flag. And a few programs that were reported back in the nvidia forum for causing black screen crashes).
This is exactly when it happens to me, thanks for the info! (3060Ti)
 

· Registered
Joined
·
30 Posts
Discussion Starter · #9 ·
Performance limit Thermal flag is set by,
Core going over set slider.
Mem going over 110C
VRM going over 120C
Driver triggers. (Nvidia made it so switching between full screen and windowed when in DX12 throws the flag, Starting Vulkan load throws the flag. And a few programs that were reported back in the nvidia forum for causing black screen crashes).
thanks for the information, i was aware of the temp limits for the GPU and VRAM however not the VRM and driver triggers

reading through


i think this is the thread Falkentyne is referring to, it seems that the 6500C reading comes from the power delivery?
that would be in line with the momentous Power% dip when i see the temp blip on MSI Afterburner monitoring graphs

anyway, should i be worried about the health of the card and the bug having a harmful effect, bother with a warranty claim or can i safely ignore this?

regarding Vulkan, rdr2 crashes on Vulkan with 456.98 driver, and MSI Kombustor also crashes if i dont rename the vulkan2 plugin dll since i installed a driver supporting rtx 30 cards, there seems to be going on something more

many thanks to all participants in this thread, it helps a lot for peace of mind, i am slowly getting there.

I have opened tickets with MSI regarding this but responses have been unhelpful at best, even a

"in order to save your time, suggest send it for repair service. thanks! "

without giving a reason what lead to this recommendation.

thanks folks!
 

· Banned
Joined
·
1,319 Posts
This does happen to me as well, but I have a 3060Ti lol :( (Gigabyte Gaming OC PRO)
I take this as a proof that MSI TORX FAN 4.0 this is a hoax, and the cooling system on the card this is not effective.
 
1 - 10 of 10 Posts
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top