What's new
Carbonite

South Africa's Top Online Tech Classifieds!
Register a free account today to become a member! (No Under 18's)
Home of C.U.D.

980ti Fluctuating FPS, Memory clock, GPU voltage & Core clock on load

mmbrajohan

Epic Member
Rating - 0%
0   0   0
Joined
Sep 2, 2010
Messages
95
Reaction score
9
Points
3,365
Age
41
Hi guys.

To start, my specs:

AMD Ryzen 2700x
ASUS Prime A320M-K
1 x 16GB RAM
Gigabyte 980ti Gaming G1
512MB NVME
3TB 7200 HDD
Antec VP700P

The history:
At my old job this card acting funny and we preplaced it with a 1080ti, which had no issues on the exact same system. It was a i7 6th gen, with 64GB RAM, 800W PSU, Etc..... Decent machine. And no problems with the PSU.

The now:
I decided to get the card and test it on my machine. It does exactly the same thing. I run a GTX980 on my current system with absolutely no issues.

So in other words, the card has been tested on 2 decent setups, which won't cause any bottlenecking from the CPU or being underpowered by the perfectly working PSU's.

And before you think it is the thermal paste, it is not. This does not show any thermal capping or any thermal issues as you will see in the graphs.

Logically I have tried reinstalling the latest drivers, did a driver clean, flashed the GPU with the latest official BOIS for that card, tried custom BIOS from here, cleaned it, reseated it many times, the 12V rails work work fine, etc etc... Please note that I have some experience with this, so any basic thing has been looked at...

Here the issue:
When not under any load, with the Geforce settings set at optimal power, it idles at low clocks just fine... Please look at Picture 1

1 - No Load - No OC - GeForce Optimal.JPG



When not under any load, with the Geforce settings set at maximum performance, it idles at high clocks just fine... Please look at Picture 2

2 - No Load - No OC - GeForce Max Performancel.JPG



NOW


When I apply load to the GPU on any of the above settings, the memory clock, GPU voltage & Core clock, reaches a maximum and then immediately drops down to about half of that, and then steadily climbs back up to max, where it just drops down again. This seems to be a very uniform repetitive cycle... Please Look at Picture 3

3 - Furmark Load - No OC - GeForce Max Performancel.JPG



This happens to any load put on the GPU, from any app. BFv, Lumion, Furmark, etc. And this isn't linked to any 100% CPU usage bottlenecking, or any RAM issues.

As mentioned, this happens when loaded with official and custom BIOS.

I have also tried lowering and increasing the memory clock, core clock, core voltage and power limit values, but same results...

It feels like this can be one of the following two scenarios:
Software related - The software, bios or drivers or something else, as a failsafe reverts back to stable/safe settings when a certain voltage or clock speed is reached.
Hardware related - A cap or shunt resistor isn't working properly, causing a drop in voltage on the GPU.

Has anybody ever managed to fix this, or have a simple solution I haven't thought of yet? And please don't tell me it is dying, cause duh, or I need to buy a new graphics card, otherwise I would have already... (it's a money thing)

I have also done a PerfCap test:
No Load

4 - PerfCap - No Load.JPG




Under Load
5 - PerfCap - Furmark Load.JPG
 
I say power delivery. VRM chip dying or solder issue.
 
I call Memory overheating. Get some thermal pads and cool the mem chips.
It can't be temp related, seeing this problems happens when at its coldest and hottest. The same result for cold and hot...
 
Tried manually setting voltage and clock speed in afterburner? Basically slight undervolt and then lock in the clock speed, save profile and let it run
 
I say power delivery. VRM chip dying or solder issue.
The power (voltage) at idle speeds when maximum power, stays stable without any fluctuation though. From 0.8V to 1.2V, stays stable when not under load... It's ONLY under load where it starts with its nonsense... I can push the usage up high without any load and the temp also increases, so it is working, but the moment a load is applied, it starts.
 
Tried manually setting voltage and clock speed in afterburner? Basically slight undervolt and then lock in the clock speed, save profile and let it run
Did that... Exactly the same results.
 
The power (voltage) at idle speeds when maximum power, stays stable without any fluctuation though. From 0.8V to 1.2V, stays stable when not under load... It's ONLY under load where it starts with its nonsense... I can push the usage up high without any load and the temp also increases, so it is working, but the moment a load is applied, it starts.

Then there's only one thing left to do...
 
Did that... Exactly the same results.
Does the manual voltage you set also fluctuate with performance dips? Did you lower the clock speed as well? Tried locking the stock voltage but lowering the clock speed by maybe 100mhz?

Can you do a ctrl+f on afterburner and screenshot the graph
 
I had similar issue and it was caused by shunt resistor's solder that was eaten by liquid metal. Should be easy to clean off and re-solder

The shunt resistor came off easily with a tweezer because the solder was completely brittle

 
Hehehe. I did that with a gtx460, and when I picked the card up, all the caps fell out!!!

I also read that if you use your oven for that, you can never use the oven again, because of the poisonous fumes form the card settling in the oven...
 
I had similar issue and it was caused by shunt resistor's solder that was eaten by liquid metal. Should be easy to clean off and re-solder

The shunt resistor came off easily with a tweezer because the solder was completely brittle

Best answer yet!!!! That's what I think as well. Something on the MB isn't putting through enough voltage... I will open the card tonight and check.
 
Hehehe. I did that with a gtx460, and when I picked the card up, all the caps fell out!!!

I also read that if you use your oven for that, you can never use the oven again, because of the poisonous fumes form the card settling in the oven...

That's why you should use sand to support the underneath. Watch the vid in the yooootoooob link I just posted
 
Tried manually setting voltage and clock speed in afterburner? Basically slight undervolt and then lock in the clock speed, save profile and let it run
I can't manually set the voltage and clock speed in afterburner, just assign + and - values. If you look at the pics in OP, you will see. I tried the latest version as well. Couldn't do it. But if I set the clock speeds to very low and undervolt it, it still does the same thing... Just less intense.
I can't add images (or I don't know how) without uploading the images to the interwebs first.
 
I had similar issue and it was caused by shunt resistor's solder that was eaten by liquid metal. Should be easy to clean off and re-solder

The shunt resistor came off easily with a tweezer because the solder was completely brittle

Urgh. Is there no way to get to it without removing the bloody heatsink? I don't have any thermal paste...
 
So I opened it up and the soldering looks perfect. My father in law, who is an electronic engineer, also said the the soldering looks good. The resistors also looks good and are soldered properly to the board. The caps also doesn't show any sign of popping, but I don't want to test the solid ones now... The thermal looked a bit iffy, so I will replace it, but I didn't get any temp spikes, so it couldn't be that.

Is there a way I can test the vram or vrm chip?

Should I maybe try a different custom BIOS?
 
I just noticed something and I don't know much about it... The memory clock shoots up to over 3500. I see on other posts, they overclock it about 1750. Is this normal?
 
So I opened it up and the soldering looks perfect. My father in law, who is an electronic engineer, also said the the soldering looks good. The resistors also looks good and are soldered properly to the board. The caps also doesn't show any sign of popping, but I don't want to test the solid ones now... The thermal looked a bit iffy, so I will replace it, but I didn't get any temp spikes, so it couldn't be that.

Is there a way I can test the vram or vrm chip?

Should I maybe try a different custom BIOS?

It was worth a try, and always a good idea to replace the thermal paste on an old card. Make sure to spread it over the whole die before assembling the cooler again. To check that you have good contact on the cooler you can compare the core temp with the hotspot temp in GPU-Z. They should not be very far apart (<15'C)

I notice now also in your GPU-Z screenshot that your PCIe slot and PCIe pin 1 and 2 voltage is drooping badly as the load goes up. It's normal if the core voltage fluctuates as the card throttles or whatever but the input voltages aren't supposed to fluctuate.

Perhaps your father in law can provide possible reasons why the input voltage would droop like that. I would guess a short somewhere but I'm far from an electrical engineer. You should probably check with another card also if it's input voltage stays constant under load on this same PSU
 
I just noticed something and I don't know much about it... The memory clock shoots up to over 3500. I see on other posts, they overclock it about 1750. Is this normal?
Yeah this sounds odd. Can you set a clock or clock offset on memory in afterburner? Although if memory serves you can't set a negative offset... Perhaps Google for a way to set a fixed clock for GPU memory?

I notice now also in your GPU-Z screenshot that your PCIe slot and PCIe pin 1 and 2 voltage is drooping badly as the load goes up. It's normal if the core voltage fluctuates as the card throttles or whatever but the input voltages aren't supposed to fluctuate.

Perhaps your father in law can provide possible reasons why the input voltage would droop like that. I would guess a short somewhere but I'm far from an electrical engineer. You should probably check with another card also if it's input voltage stays constant under load on this same PSU
AFAIK the voltage range tolerance for ATX is 5%, which would correspond to a drop to 11.4V. So the displayed voltages of 11.5V and 11.6V are within spec. Although I agree they look a tad lower than I am used to.
 
It was worth a try, and always a good idea to replace the thermal paste on an old card. Make sure to spread it over the whole die before assembling the cooler again. To check that you have good contact on the cooler you can compare the core temp with the hotspot temp in GPU-Z. They should not be very far apart (<15'C)

I notice now also in your GPU-Z screenshot that your PCIe slot and PCIe pin 1 and 2 voltage is drooping badly as the load goes up. It's normal if the core voltage fluctuates as the card throttles or whatever but the input voltages aren't supposed to fluctuate.

Perhaps your father in law can provide possible reasons why the input voltage would droop like that. I would guess a short somewhere but I'm far from an electrical engineer. You should probably check with another card also if it's input voltage stays constant under load on this same PSU
Mmmm.

I will have a look at it...
 
Yeah this sounds odd. Can you set a clock or clock offset on memory in afterburner? Although if memory serves you can't set a negative offset... Perhaps Google for a way to set a fixed clock for GPU memory?


AFAIK the voltage range tolerance for ATX is 5%, which would correspond to a drop to 11.4V. So the displayed voltages of 11.5V and 11.6V are within spec. Although I agree they look a tad lower than I am used to.
I already tried that. Pushed both the sliders to max, reduced the core and memory clock and increased it, with various increments of 5, with and without the custom bios, but it still does the same thing...

I don't have an app to set the values to a constant, just afterburner. Maybe I should google it, like you said...
 
Someone else in another forum suggested a shunt mod, where you basically short the resistor...
 
So is the voltage fluctuating with a set frequency? Almost like a capacitor?
 
I already tried that. Pushed both the sliders to max, reduced the core and memory clock and increased it, with various increments of 5, with and without the custom bios, but it still does the same thing...

I don't have an app to set the values to a constant, just afterburner. Maybe I should google it, like you said...
Either way if I were you, I would focus my googling on this question of the memory. If it is abnormal for the memory to do that, then that could help you narrow down your solutions.

One thing I don't think was covered in this thread - you notice this behaviour while monitoring the GPU under load. What does it look like when using the GPU to play games? Does it cause issues with frame rates or stuttering? Because the gamer part of me would say, if it doesn't detract from your gaming experience, then just minimise Afterburner and try forget it's happening.
 

Users who are viewing this thread

Latest posts

Back
Top Bottom