Discussion 1080 ti mini stuttering

Diamorif

Cable-Tie Ninja
Original poster
Jun 9, 2020
211
105
I just built a r9 3900/1080ti b550 pc in a dan a4 and am getting a bunch of stuttering heaven and furmark in 1080p and getting way lower score than a card of this caliber should be getting. It doesn't go over 34 degrees during the benchmark and is generally severely under performing. I intalled a kraken g12 to the card but the ram has been covered with copper sinks and the contact between the die and the cold plate are fine. Wondering if anyone might have an idea wtf might be going on
 

darksable

Chassis Packer
Jun 3, 2017
18
22
Just to be clear and check the, "flip the switch on the back of the power supply," troubleshooting...

You installed the Nvidia drivers, right?
 

Phuncz

Lord of the Boards
SFFn Staff
May 9, 2015
5,827
4,902
Just an FYI, not meant as an answer or critique: Furmark shows mainly power limitations. For performance benchmarking Unigine Superposition or 3DMark will be a better idea. What Furmark can show, if power isn't being limited, is performance without the need for RAM to VRAM bandwidth.

Concerning your issue: since both Heaven and Furmark are showing stuttering, it isn't PCIe bandwidth limiting you. Possibly it is locked at a low clock. Are you using OC tools ? check those.
 

tinyitx

Shrink Ray Wielder
Jan 25, 2018
2,279
2,338
Concerning your issue: since both Heaven and Furmark are showing stuttering, it isn't PCIe bandwidth limiting you. Possibly it is locked at a low clock. Are you using OC tools ? check those.
This could be.
One big abnormality is that 34C temp during benchmarking. This is unusually low. This temp might mean the card is not under stress at all.
Maybe one thing to do is do a clean reinstall of the driver.
 

Diamorif

Cable-Tie Ninja
Original poster
Jun 9, 2020
211
105
all, thanks for the responses., I appreciate the input. I've done all of the hardware checks and driver updates and only reference the Furmark and Unigine benchmarks as that's where the problem became apparent. I've done a complete unplug/replug, installed and reinstalled the drivers and disabled afterburner (OC software being used) to no avail. The only other thing I can think of is one of the unsleeved 8 pin power cables I'm using is kinda janky, not sure if that could be the problem. Other than that, I was thinking it might be something with the aio, and please correct me if I'm wrong, but I surmise that that would result in a huge temperature spike, not keep me pegged at a near idle temp
 

Phuncz

Lord of the Boards
SFFn Staff
May 9, 2015
5,827
4,902
What are your clocks like before running the stress test and during ?
Power related issues are also theoretically possible though the card should warn you at boot that something is wrong. I've never seen it with an iffy 8-pin cable though it might explain if it also remains at low clocks: due to power constraints.
 

Diamorif

Cable-Tie Ninja
Original poster
Jun 9, 2020
211
105
at idle, im getting somewhere between 300 and 700 mhz with spikes up to 1500 or so when i open a couple of chrome tabs. As soon as i try either benchmark, it locks at around 1000 and the image is stutter city, even at 1080p
 

rfarmer

Spatial Philosopher
Jul 7, 2017
2,588
2,702
I would open up GPU-Z. On the Sensor tab toward the bottom will be PerfCap Reason which will hopefully show you what is limiting your card.
 
  • Like
Reactions: Phuncz

Diamorif

Cable-Tie Ninja
Original poster
Jun 9, 2020
211
105
ok so its showing idle and power as the perfcap issues during the furmark gpu stress test. only thing that doesn't check out is the card ran fine on the same power cables with teh shroud on. I've got a low noise adapted attached to the 92mm fan on the AIO, could that be interfering?

edit: sorry, the kraken, aio is just plugged in
 

fish.ch

Caliper Novice
Jan 26, 2020
32
31
Did card run fine with G12 before? Or it started after the install? If after the install -- is there any fan attached to GPU PCB?
 

Diamorif

Cable-Tie Ninja
Original poster
Jun 9, 2020
211
105
ran fine without, started after the install. Two 92mm (one on aio and one on the bracket) one has the LNA.
 

fish.ch

Caliper Novice
Jan 26, 2020
32
31
Two 92mm (one on aio and one on the bracket) one has the LNA.
I mean attached as plugged in to headers on PCB?
My guess is that GPU sees no fan and set very low power limit (low enough to not cook running passively with heatsink it came with).
Second guess is there's no heatsinks on MOSFETs and they throttle?
 
  • Like
Reactions: Phuncz and rfarmer

Diamorif

Cable-Tie Ninja
Original poster
Jun 9, 2020
211
105
copper heatsinks have been applied to the mosfets with thermal tape, and yes the fans are attached to the onboard 4 pin with a mini 4 to dual 4 pin splitter
 

Diamorif

Cable-Tie Ninja
Original poster
Jun 9, 2020
211
105
welp, I'm a nitwit. it turns out, for your reference, that all of the die needs to be adequately covered for optimum performance. one of the die corners was semi-exposed and that was causing the stutter. thank you all for your input and hopefully I will recover cognitive function at some point in the near future
 

rfarmer

Spatial Philosopher
Jul 7, 2017
2,588
2,702
I am just glad you got it sorted, can be really difficult to trouble shoot this kind of a problem.
 

tinyitx

Shrink Ray Wielder
Jan 25, 2018
2,279
2,338
welp, I'm a nitwit. it turns out, for your reference, that all of the die needs to be adequately covered for optimum performance. one of the die corners was semi-exposed and that was causing the stutter. thank you all for your input and hopefully I will recover cognitive function at some point in the near future
This is rare and weird. You mean, the AIO waterblock does not cover the entirety of the delidded CPU die?
Could you show a pic, if possible?
And, how are you going to fix this?
 

Diamorif

Cable-Tie Ninja
Original poster
Jun 9, 2020
211
105
its the GPU die and the paste I reapplied the last time I troubleshot it didn't entirely cover the die area. just did an even spread with a thermal paste spatula and voila, the clock was good.
 
  • Like
Reactions: Phuncz and tinyitx

tinyitx

Shrink Ray Wielder
Jan 25, 2018
2,279
2,338
its the GPU die and the paste I reapplied the last time I troubleshot it didn't entirely cover the die area. just did an even spread with a thermal paste spatula and voila, the clock was good.
My bad...I typed too quickly without thinking...lol

In the early years, I used to adapt the pea method when applying thermal paste.
Now, I too use a TIM spatula to spread the paste manually myself over the entire applicable area.