• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Perfcap reason fluctuates between power util and vrel under load

Joined
Oct 7, 2007
Messages
226 (0.04/day)
Location
North of Iran
System Name Main system
Processor Intel Core-i7 4790K @4.5
Motherboard Asus Z97-Pro Gamer
Cooling Cooler Master Hyper 212 EVO
Memory 32 Gig (16 GigADATA XPG V2 DDR3 2400 + 16 Gig Geil Evo Veloce DDR3 1600 CL11)
Video Card(s) Gigabyte GTX1080 G1 Gaming
Storage Western Digital Blue Cavier 500 and 1 Terabyte + 1 Terabyte Green Cavier + Samsung SSD 860 EVO 500G
Display(s) ASUS ML238 LED Monitor
Case Green Viper X3
Audio Device(s) Onboard-Realtek
Power Supply Green 700watt (Its a rebranded psu from highpower/kolink I guess)
Software Windows 10 /Ubuntu 16.04
Hello everyone,
I bought a Gigabyte GTX1080 G1 gaming today . I was running one of my experiments just now which noticed
my gpu load suddenly goes to 0 from 100% and then comes back to 100% and again goes to 0. I noticed PerfCab reason fluctuates between 3 options, pwr, util and vrel.
Is something wrong with the card ?
here is the gpuz screenshot :
perfcab fluctuations.gif

I need to say that my former GTX980 (g1 gaming as well) did not have this problem. the load was constantly at 97 to 99% and I used to get Vrel/util most of the time and nothing else.


Any help is greatly appreciated
 
Joined
Oct 17, 2012
Messages
9,781 (2.35/day)
Location
Massachusetts
System Name Americas cure is the death of Social Justice & Political Correctness
Processor i7-11700K
Motherboard Asrock Z590 Extreme wifi 6E
Cooling Noctua NH-U12A
Memory 32GB Corsair RGB fancy boi 5000
Video Card(s) RTX 3090 Reference
Storage Samsung 970 Evo 1Tb + Samsung 970 Evo 500Gb
Display(s) Dell - 27" LED QHD G-SYNC x2
Case Fractal Design Meshify-C
Audio Device(s) on board
Power Supply Seasonic Focus+ Gold 1000 Watt
Mouse Logitech G502 spectrum
Keyboard AZIO MGK-1 RGB (Kaith Blue)
Software Win 10 Professional 64 bit
Benchmark Scores the MLGeesiest
My guess would be that it's the driver or the bios of the card throttling it's power and/or temperature. It's protecting the card so it doesn't overheat

For this reason I edited my 970 to have expanded power and temperature limits
 
Joined
Oct 7, 2007
Messages
226 (0.04/day)
Location
North of Iran
System Name Main system
Processor Intel Core-i7 4790K @4.5
Motherboard Asus Z97-Pro Gamer
Cooling Cooler Master Hyper 212 EVO
Memory 32 Gig (16 GigADATA XPG V2 DDR3 2400 + 16 Gig Geil Evo Veloce DDR3 1600 CL11)
Video Card(s) Gigabyte GTX1080 G1 Gaming
Storage Western Digital Blue Cavier 500 and 1 Terabyte + 1 Terabyte Green Cavier + Samsung SSD 860 EVO 500G
Display(s) ASUS ML238 LED Monitor
Case Green Viper X3
Audio Device(s) Onboard-Realtek
Power Supply Green 700watt (Its a rebranded psu from highpower/kolink I guess)
Software Windows 10 /Ubuntu 16.04
My guess would be that it's the driver or the bios of the card throttling it's power and/or temperature. It's protecting the card so it doesn't overheat

For this reason I edited my 970 to have expanded power and temperature limits
the temp stays at 65C , I dont think this is OK to immediately limit voltage in order to avoid getting hotter!
At this rate it seems I cant even use the normal gtx1080 power for my processing chores! let alone overclock it!
both memory controller, bus interface, everything gets clamped at zero! I have never seen this with my GTx980!!!
perfcab fluctuations2.gif
 
Last edited:

newtekie1

Semi-Retired Folder
Joined
Nov 22, 2005
Messages
28,472 (4.25/day)
Location
Indiana, USA
Processor Intel Core i7 10850K@5.2GHz
Motherboard AsRock Z470 Taichi
Cooling Corsair H115i Pro w/ Noctua NF-A14 Fans
Memory 32GB DDR4-3600
Video Card(s) RTX 2070 Super
Storage 500GB SX8200 Pro + 8TB with 1TB SSD Cache
Display(s) Acer Nitro VG280K 4K 28"
Case Fractal Design Define S
Audio Device(s) Onboard is good enough for me
Power Supply eVGA SuperNOVA 1000w G3
Software Windows 10 Pro x64
Util: Limited by GPU Utilization

This means the GPU is idle. This isn't throttling, it means the GPU is waiting for work to do.

If it was throttling due to temperature it would say Thrm. If it was throttling due to the power limit it would say Pwr.
 
Joined
Oct 17, 2012
Messages
9,781 (2.35/day)
Location
Massachusetts
System Name Americas cure is the death of Social Justice & Political Correctness
Processor i7-11700K
Motherboard Asrock Z590 Extreme wifi 6E
Cooling Noctua NH-U12A
Memory 32GB Corsair RGB fancy boi 5000
Video Card(s) RTX 3090 Reference
Storage Samsung 970 Evo 1Tb + Samsung 970 Evo 500Gb
Display(s) Dell - 27" LED QHD G-SYNC x2
Case Fractal Design Meshify-C
Audio Device(s) on board
Power Supply Seasonic Focus+ Gold 1000 Watt
Mouse Logitech G502 spectrum
Keyboard AZIO MGK-1 RGB (Kaith Blue)
Software Win 10 Professional 64 bit
Benchmark Scores the MLGeesiest
Util: Limited by GPU Utilization

This means the GPU is idle. This isn't throttling, it means the GPU is waiting for work to do.

If it was throttling due to temperature it would say Thrm. If it was throttling due to the power limit it would say Pwr.

I noticed PerfCab reason fluctuates between 3 options, pwr, util and vrel.
 
Joined
Oct 7, 2007
Messages
226 (0.04/day)
Location
North of Iran
System Name Main system
Processor Intel Core-i7 4790K @4.5
Motherboard Asus Z97-Pro Gamer
Cooling Cooler Master Hyper 212 EVO
Memory 32 Gig (16 GigADATA XPG V2 DDR3 2400 + 16 Gig Geil Evo Veloce DDR3 1600 CL11)
Video Card(s) Gigabyte GTX1080 G1 Gaming
Storage Western Digital Blue Cavier 500 and 1 Terabyte + 1 Terabyte Green Cavier + Samsung SSD 860 EVO 500G
Display(s) ASUS ML238 LED Monitor
Case Green Viper X3
Audio Device(s) Onboard-Realtek
Power Supply Green 700watt (Its a rebranded psu from highpower/kolink I guess)
Software Windows 10 /Ubuntu 16.04
Util: Limited by GPU Utilization

This means the GPU is idle. This isn't throttling, it means the GPU is waiting for work to do.

If it was throttling due to temperature it would say Thrm. If it was throttling due to the power limit it would say Pwr.
if you look at the PerfCab Reason graph log section, you can see different colors , green, blue and gray. each of them as you know belong to a different category .
the data is cached in my memory so there is always something to be computed. thats why the gpu load is maxed out at 100% and then suddenly it becomes 0% and then again it becomes 100. this can be seen by literally looking at the the graphs drawn in red.
PWR.gif


and also this log shows the card is already executing commands and its not idle for input data (data gets cached at the very beginning)
I0428 00:55:28.371031 22960 net.cpp:676] Ignoring source layer accuracy_top-5-train
I0428 00:55:28.693717 22960 blocking_queue.cpp:49] Waiting for data
I0428 00:57:00.091953 20704 data_layer.cpp:73] Restarting data prefetching from start.
I0428 00:57:00.169011 22960 solver.cpp:398] Test net output #0: accuracy_top-1 = 0.58596
I0428 00:57:00.169011 22960 solver.cpp:398] Test net output #1: accuracy_top-5 = 0.820521
I0428 00:57:00.169011 22960 solver.cpp:398] Test net output #2: loss = 1.74516 (* 1 = 1.74516 loss)
I0428 00:57:00.803012 22960 solver.cpp:219] Iteration 300000 (3244.9 iter/s, 92.4528s/20 iters), loss = 1.11276
I0428 00:57:00.803012 22960 solver.cpp:238] Train net output #0: accuracy_top-1-train = 0.71875
I0428 00:57:00.803012 22960 solver.cpp:238] Train net output #1: accuracy_top-5-train = 0.898438
I0428 00:57:00.803012 22960 solver.cpp:238] Train net output #2: loss = 1.11276 (* 1 = 1.11276 loss)
I0428 00:57:00.803012 22960 sgd_solver.cpp:105] Iteration 300000, lr = 1e-05
I0428 00:57:10.136579 22960 solver.cpp:219] Iteration 300020 (2.143 iter/s, 9.33272s/20 iters), loss = 1.43738
I0428 00:57:10.143584 22960 solver.cpp:238] Train net output #0: accuracy_top-1-train = 0.671875
I0428 00:57:10.143584 22960 solver.cpp:238] Train net output #1: accuracy_top-5-train = 0.898438
I0428 00:57:10.143584 22960 solver.cpp:238] Train net output #2: loss = 1.43738 (* 1 = 1.43738 loss)
I0428 00:57:10.143584 22960 sgd_solver.cpp:105] Iteration 300020, lr = 1e-05
I0428 00:57:19.493077 22960 solver.cpp:219] Iteration 300040 (2.13913 iter/s, 9.34958s/20 iters), loss = 1.15417
I0428 00:57:19.500082 22960 solver.cpp:238] Train net output #0: accuracy_top-1-train = 0.6875
I0428 00:57:19.500082 22960 solver.cpp:238] Train net output #1: accuracy_top-5-train = 0.90625
I0428 00:57:19.500082 22960 solver.cpp:238] Train net output #2: loss = 1.15417 (* 1 = 1.15417 loss)
I0428 00:57:19.500082 22960 sgd_solver.cpp:105] Iteration 300040, lr = 1e-05
I0428 00:57:29.842706 22960 solver.cpp:219] Iteration 300060 (1.93392 iter/s, 10.3417s/20 iters), loss = 1.36165
I0428 00:57:29.849216 22960 solver.cpp:238] Train net output #0: accuracy_top-1-train = 0.625
I0428 00:57:29.849216 22960 solver.cpp:238] Train net output #1: accuracy_top-5-train = 0.882813
I0428 00:57:29.849216 22960 solver.cpp:238] Train net output #2: loss = 1.36165 (* 1 = 1.36165 loss)
I0428 00:57:29.849216 22960 sgd_solver.cpp:105] Iteration 300060, lr = 1e-05
I0428 00:57:39.647658 22960 solver.cpp:219] Iteration 300080 (2.04126 iter/s, 9.79786s/20 iters), loss = 1.21606
I0428 00:57:39.653666 22960 solver.cpp:238] Train net output #0: accuracy_top-1-train = 0.710938
I0428 00:57:39.653666 22960 solver.cpp:238] Train net output #1: accuracy_top-5-train = 0.898438
I0428 00:57:39.653666 22960 solver.cpp:238] Train net output #2: loss = 1.21606 (* 1 = 1.21606 loss)
I0428 00:57:39.653666 22960 sgd_solver.cpp:105] Iteration 300080, lr = 1e-05
I0428 00:57:49.135305 22960 solver.cpp:219] Iteration 300100 (2.10946 iter/s, 9.4811s/20 iters), loss = 1.07624
I0428 00:57:49.141811 22960 solver.cpp:238] Train net output #0: accuracy_top-1-train = 0.695313
I0428 00:57:49.141811 22960 solver.cpp:238] Train net output #1: accuracy_top-5-train = 0.890625
I0428 00:57:49.141811 22960 solver.cpp:238] Train net output #2: loss = 1.07624 (* 1 = 1.07624 loss)
I0428 00:57:49.141811 22960 sgd_solver.cpp:105] Iteration 300100, lr = 1e-05
I0428 00:57:59.200597 22960 solver.cpp:219] Iteration 300120 (1.98837 iter/s, 10.0585s/20 iters), loss = 1.35399
I0428 00:57:59.200597 22960 solver.cpp:238] Train net output #0: accuracy_top-1-train = 0.71875
I0428 00:57:59.200597 22960 solver.cpp:238] Train net output #1: accuracy_top-5-train = 0.882813
I0428 00:57:59.200597 22960 solver.cpp:238] Train net output #2: loss = 1.35399 (* 1 = 1.35399 loss)
I0428 00:57:59.200597 22960 sgd_solver.cpp:105] Iteration 300120, lr = 1e-05
I0428 00:58:11.194643 22960 solver.cpp:219] Iteration 300140 (1.66765 iter/s, 11.993s/20 iters), loss = 1.33887
I0428 00:58:11.200647 22960 solver.cpp:238] Train net output #0: accuracy_top-1-train = 0.695313
I0428 00:58:11.200647 22960 solver.cpp:238] Train net output #1: accuracy_top-5-train = 0.875
I0428 00:58:11.200647 22960 solver.cpp:238] Train net output #2: loss = 1.33887 (* 1 = 1.33887 loss)
I0428 00:58:11.200647 22960 sgd_solver.cpp:105] Iteration 300140, lr = 1e-05
I0428 00:58:21.040182 22960 solver.cpp:219] Iteration 300160 (2.03268 iter/s, 9.83924s/20 iters), loss = 1.32039
I0428 00:58:21.047205 22960 solver.cpp:238] Train net output #0: accuracy_top-1-train = 0.648438
I0428 00:58:21.047205 22960 solver.cpp:238] Train net output #1: accuracy_top-5-train = 0.882813
I0428 00:58:21.047205 22960 solver.cpp:238] Train net output #2: loss = 1.32039 (* 1 = 1.32039 loss)
I0428 00:58:21.047205 22960 sgd_solver.cpp:105] Iteration 300160, lr = 1e-05
I0428 00:58:30.830806 22960 solver.cpp:219] Iteration 300180 (2.04419 iter/s, 9.78385s/20 iters), loss = 1.58379
I0428 00:58:30.837810 22960 solver.cpp:238] Train net output #0: accuracy_top-1-train = 0.617188
I0428 00:58:30.837810 22960 solver.cpp:238] Train net output #1: accuracy_top-5-train = 0.828125
I0428 00:58:30.837810 22960 solver.cpp:238] Train net output #2: loss = 1.58379 (* 1 = 1.58379 loss)
I0428 00:58:30.837810 22960 sgd_solver.cpp:105] Iteration 300180, lr = 1e-05
I0428 00:58:43.533479 22960 solver.cpp:219] Iteration 300200 (1.57546 iter/s, 12.6947s/20 iters), loss = 1.29555
I0428 00:58:43.533479 22960 solver.cpp:238] Train net output #0: accuracy_top-1-train = 0.640625
I0428 00:58:43.533479 22960 solver.cpp:238] Train net output #1: accuracy_top-5-train = 0.882813
I0428 00:58:43.533479 22960 solver.cpp:238] Train net output #2: loss = 1.29555 (* 1 = 1.29555 loss)
I0428 00:58:43.533479 22960 sgd_solver.cpp:105] Iteration 300200, lr = 1e-05
I0428 00:58:57.092994 22960 solver.cpp:219] Iteration 300220 (1.47509 iter/s, 13.5585s/20 iters), loss = 1.3301
I0428 00:58:57.092994 22960 solver.cpp:238] Train net output #0: accuracy_top-1-train = 0.664063
I0428 00:58:57.092994 22960 solver.cpp:238] Train net output #1: accuracy_top-5-train = 0.875
I0428 00:58:57.092994 22960 solver.cpp:238] Train net output #2: loss = 1.3301 (* 1 = 1.3301 loss)
I0428 00:58:57.092994 22960 sgd_solver.cpp:105] Iteration 300220, lr = 1e-05
I0428 00:59:27.482066 22960 solver.cpp:219] Iteration 300240 (0.658147 iter/s, 30.3884s/20 iters), loss = 1.21879
I0428 00:59:27.482573 22960 solver.cpp:238] Train net output #0: accuracy_top-1-train = 0.679688
I0428 00:59:27.482573 22960 solver.cpp:238] Train net output #1: accuracy_top-5-train = 0.898438
I0428 00:59:27.482573 22960 solver.cpp:238] Train net output #2: loss = 1.21879 (* 1 = 1.21879 loss)
I0428 00:59:27.482573 22960 sgd_solver.cpp:105] Iteration 300240, lr = 1e-05
I0428 00:59:54.203022 22960 solver.cpp:219] Iteration 300260 (0.748511 iter/s, 26.7197s/20 iters), loss = 1.53185
I0428 00:59:54.203022 22960 solver.cpp:238] Train net output #0: accuracy_top-1-train = 0.578125
I0428 00:59:54.203022 22960 solver.cpp:238] Train net output #1: accuracy_top-5-train = 0.835938
I0428 00:59:54.203022 22960 solver.cpp:238] Train net output #2: loss = 1.53185 (* 1 = 1.53185 loss)
I0428 00:59:54.203022 22960 sgd_solver.cpp:105] Iteration 300260, lr = 1e-05
I0428 01:00:31.217540 22960 solver.cpp:219] Iteration 300280 (0.54035 iter/s, 37.0131s/20 iters), loss = 1.22318
I0428 01:00:31.217540 22960 solver.cpp:238] Train net output #0: accuracy_top-1-train = 0.632813
I0428 01:00:31.217540 22960 solver.cpp:238] Train net output #1: accuracy_top-5-train = 0.882813
I0428 01:00:31.217540 22960 solver.cpp:238] Train net output #2: loss = 1.22318 (* 1 = 1.22318 loss)
I0428 01:00:31.217540 22960 sgd_solver.cpp:105] Iteration 300280, lr = 1e-05
I0428 01:01:23.414973 22960 solver.cpp:219] Iteration 300300 (0.383174 iter/s, 52.1956s/20 iters), loss = 1.28781
I0428 01:01:23.414973 22960 solver.cpp:238] Train net output #0: accuracy_top-1-train = 0.664063
I0428 01:01:23.414973 22960 solver.cpp:238] Train net output #1: accuracy_top-5-train = 0.890625
I0428 01:01:23.414973 22960 solver.cpp:238] Train net output #2: loss = 1.28781 (* 1 = 1.28781 loss)
I0428 01:01:23.414973 22960 sgd_solver.cpp:105] Iteration 300300, lr = 1e-05
I0428 01:02:11.234027 22960 solver.cpp:219] Iteration 300320 (0.418257 iter/s, 47.8175s/20 iters), loss = 1.27358
I0428 01:02:11.234027 22960 solver.cpp:238] Train net output #0: accuracy_top-1-train = 0.6875
I0428 01:02:11.234027 22960 solver.cpp:238] Train net output #1: accuracy_top-5-train = 0.867188
I0428 01:02:11.234027 22960 solver.cpp:238] Train net output #2: loss = 1.27358 (* 1 = 1.27358 loss)
I0428 01:02:11.234027 22960 sgd_solver.cpp:105] Iteration 300320, lr = 1e-05
I0428 01:02:53.814581 22960 solver.cpp:219] Iteration 300340 (0.469711 iter/s, 42.5793s/20 iters), loss = 1.19984
I0428 01:02:53.814581 22960 solver.cpp:238] Train net output #0: accuracy_top-1-train = 0.695313
 

newtekie1

Semi-Retired Folder
Joined
Nov 22, 2005
Messages
28,472 (4.25/day)
Location
Indiana, USA
Processor Intel Core i7 10850K@5.2GHz
Motherboard AsRock Z470 Taichi
Cooling Corsair H115i Pro w/ Noctua NF-A14 Fans
Memory 32GB DDR4-3600
Video Card(s) RTX 2070 Super
Storage 500GB SX8200 Pro + 8TB with 1TB SSD Cache
Display(s) Acer Nitro VG280K 4K 28"
Case Fractal Design Define S
Audio Device(s) Onboard is good enough for me
Power Supply eVGA SuperNOVA 1000w G3
Software Windows 10 Pro x64
if you look at the PerfCab Reason graph log section, you can see different colors , green, blue and gray. each of them as you know belong to a different category .
the data is cached in my memory so there is always something to be computed. thats why the gpu load is maxed out at 100% and then suddenly it becomes 0% and then again it becomes 100. this can be seen by literally looking at the the graphs drawn in red.

Then something else is making the GPU wait to do work.

Grey = Util = GPU Idle Waiting for Work to Do
Blue = VREL = Normal When Full Load
Green = Pwr = Full Load Is Causing the Card to Hit the Power Limit

Even when the card hits the power limit, it won't just stop doing work. The load will stay at 100%, or close to it, the clocks and voltages will just lower to keep the card under the power limit. Example:



In that you'll see the part in the middle where the GPU starts turning green. At that point, I changed the power limit of the card from 100% to 50%. You'll notice the power consumption droped to about 50% and the clock speed and VDDC graphs dropped significantly. I started to get the green Pwr under PerfCap Reason, but the GPU load never went down. Then I raised the power limit back up to normal and everything went back to normal.

At the same time, hitting the power limit is also not something to be alarmed by. Different loads will cause the card to consume more or less power, even if both are cuasing 100% load.

What you are experiencing is not a GPU problem. Something is not putting load on the card correctly.
 
Joined
Oct 7, 2007
Messages
226 (0.04/day)
Location
North of Iran
System Name Main system
Processor Intel Core-i7 4790K @4.5
Motherboard Asus Z97-Pro Gamer
Cooling Cooler Master Hyper 212 EVO
Memory 32 Gig (16 GigADATA XPG V2 DDR3 2400 + 16 Gig Geil Evo Veloce DDR3 1600 CL11)
Video Card(s) Gigabyte GTX1080 G1 Gaming
Storage Western Digital Blue Cavier 500 and 1 Terabyte + 1 Terabyte Green Cavier + Samsung SSD 860 EVO 500G
Display(s) ASUS ML238 LED Monitor
Case Green Viper X3
Audio Device(s) Onboard-Realtek
Power Supply Green 700watt (Its a rebranded psu from highpower/kolink I guess)
Software Windows 10 /Ubuntu 16.04
Then something else is making the GPU wait to do work.

Grey = Util = GPU Idle Waiting for Work to Do
Blue = VREL = Normal When Full Load
Green = Pwr = Full Load Is Causing the Card to Hit the Power Limit

Even when the card hits the power limit, it won't just stop doing work. The load will stay at 100%, or close to it, the clocks and voltages will just lower to keep the card under the power limit. Example:



In that you'll see the part in the middle where the GPU starts turning green. At that point, I changed the power limit of the card from 100% to 50%. You'll notice the power consumption droped to about 50% and the clock speed and VDDC graphs dropped significantly. I started to get the green Pwr under PerfCap Reason, but the GPU load never went down. Then I raised the power limit back up to normal and everything went back to normal.

At the same time, hitting the power limit is also not something to be alarmed by. Different loads will cause the card to consume more or less power, even if both are cuasing 100% load.

What you are experiencing is not a GPU problem. Something is not putting load on the card correctly.
Thank you very much, that was a great explanation .
by the way should I be alarmed by the TDP%? I saw the card hit 140% TDP!
Thank you very much :)
 
Joined
Oct 2, 2004
Messages
13,791 (1.94/day)
Actually, VRel doesn't mean "normal under load", but is expected to happen and is normal (it's related to the boosting thing). VRel means Voltage reliability. It means GPU is adjusting voltages for optimum performance/stability. I wanted to say you can see the voltage changes under VDDC which allign with these events, but then I remembered this is Pascal which has much more granular boosting and voltage control, meaning you can't since graph is a up and down "mess". It's easier on my Maxwell which only makes basic steps up and down and it's easier to see.
 
Joined
Mar 10, 2014
Messages
1,793 (0.49/day)
What is with that memory clock? It sound too slow(2*1128.4MHz*4=9.0Gbps, does gpu-z read that gddr5x memory clock wrong or have you under clocked memory. Vanilla gtx1080 should have 1250MHz mems)
 

newtekie1

Semi-Retired Folder
Joined
Nov 22, 2005
Messages
28,472 (4.25/day)
Location
Indiana, USA
Processor Intel Core i7 10850K@5.2GHz
Motherboard AsRock Z470 Taichi
Cooling Corsair H115i Pro w/ Noctua NF-A14 Fans
Memory 32GB DDR4-3600
Video Card(s) RTX 2070 Super
Storage 500GB SX8200 Pro + 8TB with 1TB SSD Cache
Display(s) Acer Nitro VG280K 4K 28"
Case Fractal Design Define S
Audio Device(s) Onboard is good enough for me
Power Supply eVGA SuperNOVA 1000w G3
Software Windows 10 Pro x64
by the way should I be alarmed by the TDP%? I saw the card hit 140% TDP!

No need to be worried at all, it is ok for the TDP to spike like that.

Actually, VRel doesn't mean "normal under load", but is expected to happen and is normal (it's related to the boosting thing). VRel means Voltage reliability. It means GPU is adjusting voltages for optimum performance/stability. I wanted to say you can see the voltage changes under VDDC which allign with these events, but then I remembered this is Pascal which has much more granular boosting and voltage control, meaning you can't since graph is a up and down "mess". It's easier on my Maxwell which only makes basic steps up and down and it's easier to see.

I didn't mean to say the VRel means normal under load, just like Util doesn't mean the card is waiting for work to do, it means Utilization. What I meant was that seeing VRel is normal under load.
 
Joined
Sep 17, 2014
Messages
20,697 (5.96/day)
Location
The Washing Machine
Processor i7 8700k 4.6Ghz @ 1.24V
Motherboard AsRock Fatal1ty K6 Z370
Cooling beQuiet! Dark Rock Pro 3
Memory 16GB Corsair Vengeance LPX 3200/C16
Video Card(s) ASRock RX7900XT Phantom Gaming
Storage Samsung 850 EVO 1TB + Samsung 830 256GB + Crucial BX100 250GB + Toshiba 1TB HDD
Display(s) Gigabyte G34QWC (3440x1440)
Case Fractal Design Define R5
Audio Device(s) Harman Kardon AVR137 + 2.1
Power Supply EVGA Supernova G2 750W
Mouse XTRFY M42
Keyboard Lenovo Thinkpad Trackpoint II
Software W10 x64
Beyond your own experimenting, start up a game and do some actual gaming with it. Look for artifacting, and stuttering.

If you have neither of those and you also don't see these utilization drops while ingame, your card is OK.

All I can deduct from your GPUz screens is that your own experiment is causing this behaviour, or something else your PC is doing while loading the GPU. Temps, voltage, clocks are very stable, so that does not point to a problem.
 
Joined
Oct 7, 2007
Messages
226 (0.04/day)
Location
North of Iran
System Name Main system
Processor Intel Core-i7 4790K @4.5
Motherboard Asus Z97-Pro Gamer
Cooling Cooler Master Hyper 212 EVO
Memory 32 Gig (16 GigADATA XPG V2 DDR3 2400 + 16 Gig Geil Evo Veloce DDR3 1600 CL11)
Video Card(s) Gigabyte GTX1080 G1 Gaming
Storage Western Digital Blue Cavier 500 and 1 Terabyte + 1 Terabyte Green Cavier + Samsung SSD 860 EVO 500G
Display(s) ASUS ML238 LED Monitor
Case Green Viper X3
Audio Device(s) Onboard-Realtek
Power Supply Green 700watt (Its a rebranded psu from highpower/kolink I guess)
Software Windows 10 /Ubuntu 16.04
How can I increase the power limit? currently it just hits the power limit pretty much very often, if I use the gaming or OC profile this happens even more,
However, there is an update for bios (F2) I dont know if this is going to solve this issue or it will create even more issues !
I dont even know if other G1 gamings experience what I experience to know whether this is an issue in first place! (facing with power limit just nullifies not any basic overclocking capabilities, but also the boost as well )
 
Joined
Jan 8, 2017
Messages
8,810 (3.35/day)
System Name Good enough
Processor AMD Ryzen R9 7900 - Alphacool Eisblock XPX Aurora Edge
Motherboard ASRock B650 Pro RS
Cooling 2x 360mm NexXxoS ST30 X-Flow, 1x 360mm NexXxoS ST30, 1x 240mm NexXxoS ST30
Memory 32GB - FURY Beast RGB 5600 Mhz
Video Card(s) Sapphire RX 7900 XT - Alphacool Eisblock Aurora
Storage 1x Kingston KC3000 1TB 1x Kingston A2000 1TB, 1x Samsung 850 EVO 250GB , 1x Samsung 860 EVO 500GB
Display(s) LG UltraGear 32GN650-B + 4K Samsung TV
Case Phanteks NV7
Power Supply GPS-750C
MSI afterburner has a power limit slider , you can also force a custom curve for voltages/frequency although 2 Ghz is pretty much the limit for most pascal cards. Your card behaves exactly like other 1080s. I don't really understand what your issue is , are you getting stutters performance issues ?
 
Last edited:
Joined
Oct 7, 2007
Messages
226 (0.04/day)
Location
North of Iran
System Name Main system
Processor Intel Core-i7 4790K @4.5
Motherboard Asus Z97-Pro Gamer
Cooling Cooler Master Hyper 212 EVO
Memory 32 Gig (16 GigADATA XPG V2 DDR3 2400 + 16 Gig Geil Evo Veloce DDR3 1600 CL11)
Video Card(s) Gigabyte GTX1080 G1 Gaming
Storage Western Digital Blue Cavier 500 and 1 Terabyte + 1 Terabyte Green Cavier + Samsung SSD 860 EVO 500G
Display(s) ASUS ML238 LED Monitor
Case Green Viper X3
Audio Device(s) Onboard-Realtek
Power Supply Green 700watt (Its a rebranded psu from highpower/kolink I guess)
Software Windows 10 /Ubuntu 16.04
MSI afterburner has a power limit slider , you can also force a custom curve for voltages/frequency although 2 Ghz is pretty much the limit for most pascal cards. Your card behaves exactly like other 1080s. I don't really understand what your issue is , are you getting stutters performance issues ?
Actually I'd never seen the power limit on my former GTX980, and when I saw it happening, and the clock decreased this kind of worried me.
I thought (prior to you saying this is normal with all 1080s), when they create a G1 gaming card and offer automatic boosting or even profiles for mild overclocking, then this should all be OK , at the very least in the default mode (no oc!) (meaning clocks will always be at the boosted/overclocked rate and continue to stay that way unless the temps go up or something like that occurs. )
I am just about running some benchmarks, my experiments are finished and meanwhile I can run some games/tests .

Edit :
Thank you every one. I ran some games/benchmarks, and in them I never saw any power limit. it seems (as already been pointed out by you guys) this only happens with my experiments .
 
Last edited:

newtekie1

Semi-Retired Folder
Joined
Nov 22, 2005
Messages
28,472 (4.25/day)
Location
Indiana, USA
Processor Intel Core i7 10850K@5.2GHz
Motherboard AsRock Z470 Taichi
Cooling Corsair H115i Pro w/ Noctua NF-A14 Fans
Memory 32GB DDR4-3600
Video Card(s) RTX 2070 Super
Storage 500GB SX8200 Pro + 8TB with 1TB SSD Cache
Display(s) Acer Nitro VG280K 4K 28"
Case Fractal Design Define S
Audio Device(s) Onboard is good enough for me
Power Supply eVGA SuperNOVA 1000w G3
Software Windows 10 Pro x64
The older Maxwell cards didn't often hit the power limit with stock settings, and usually not for extended periods of time, because nVidia was more conservative with them. Plus, GPU Boost on Maxwell was not as agressive as Pascal. Pascal cards, in my experience hit the power limit a lot more often due to nVidia being more aggressive with the clock speeds and GPU Boost.
 
Joined
Oct 17, 2012
Messages
9,781 (2.35/day)
Location
Massachusetts
System Name Americas cure is the death of Social Justice & Political Correctness
Processor i7-11700K
Motherboard Asrock Z590 Extreme wifi 6E
Cooling Noctua NH-U12A
Memory 32GB Corsair RGB fancy boi 5000
Video Card(s) RTX 3090 Reference
Storage Samsung 970 Evo 1Tb + Samsung 970 Evo 500Gb
Display(s) Dell - 27" LED QHD G-SYNC x2
Case Fractal Design Meshify-C
Audio Device(s) on board
Power Supply Seasonic Focus+ Gold 1000 Watt
Mouse Logitech G502 spectrum
Keyboard AZIO MGK-1 RGB (Kaith Blue)
Software Win 10 Professional 64 bit
Benchmark Scores the MLGeesiest
This goes back to what I mentioned in the beginning of this thread. In the current version of afterburner you can increase the power/temperature limit to 90° or something. Or if you're familiar with modifying a bios you can increase the limitations there Which is what I chose to do. It really gave my reference 970 a massive power increase. Now I can overclock 25% stable. NP.

Full disclosure though ,I had help editing the bios since it wasn't something I was willing to take a chance on
 

newtekie1

Semi-Retired Folder
Joined
Nov 22, 2005
Messages
28,472 (4.25/day)
Location
Indiana, USA
Processor Intel Core i7 10850K@5.2GHz
Motherboard AsRock Z470 Taichi
Cooling Corsair H115i Pro w/ Noctua NF-A14 Fans
Memory 32GB DDR4-3600
Video Card(s) RTX 2070 Super
Storage 500GB SX8200 Pro + 8TB with 1TB SSD Cache
Display(s) Acer Nitro VG280K 4K 28"
Case Fractal Design Define S
Audio Device(s) Onboard is good enough for me
Power Supply eVGA SuperNOVA 1000w G3
Software Windows 10 Pro x64
Or if you're familiar with modifying a bios you can increase the limitations there Which is what I chose to do.

Good luck doing that on a Pacal card.
 
Top