RTX 2080 low performance
Good day. From desperation in trying to solve the problem below, I have already broken my head. In short: the memory of the video card heats up to 109 degrees and sometimes starts not only strongly, but throttling. I check all the indicators, looking for solutions to the problem - to no avail. NVIDIA support says that this is a normal temperature for Micron chips, but this, you understand, is such a thing.But to the main problem: the most unpleasant thing, even without paying attention to temperatures, is the 3DMark benchmark results, which are twice as low as the average for 2080, ~ 11k against the average ~ 23k https://www.3dmark.com/3dm/72858986.
The results under load at 99% in the same WoT in the SD client on the Ultra graphics give out only 40-50 fps, Elden Ring gives out 40-45 even at low ones, but Doom Eternal managed to maintain 40-40 in quality with full ultras and RTX + DLSS. 50fps. Some kind of game.
In total, the only thing I can start from and show here at the moment is:
- GPU-Z and HWInfo indicators for memory temperature under load in the amount of 108-109 degrees
- an almost constant emphasis on Power in terms of PerfCap -
TDP even under load in around 50-60%
-jumping core voltage I attach a
link to the HWInfo and GPU-Z logs: https://drive.google.com/drive/folders/1pXLwe7eOOcbf5L9Gisrq7uoHCZ_jQxop
System configuration:
CPU: Intel Core i5-12400F
Video Card: NVIDIA GeForce GTX 2080 Gigabyte Windforce 8G
RAM: Crucial Ballistix DDR4 8GBx2pcs 3000MHz PC24000
SSD: WDC WDS480G2G0B-00EPW0 (480GB M.2 SSD)
Motherboard: ASRock B660M Pro RS
Power Supply: Cooler Master MWE 750W
I would be very grateful for your help, thank you.
Frendzy
Come on, I5 12th generation does not reveal the map, well, you're driving!
Is there any backstory? Where did the video card come from, was it opened or not? Was it normal before or suddenly became bad? Have you tried scanning the system for miners?
If there are no miners, you need to disassemble, look at the state of the thermal pads in memory. Or take it under warranty to a service for verification.
Loken
took the card from the tenacious paws of the huckster, but no jambs were noticed behind it, while I played Arex on the ultras for three hours in a row and there were no drawdowns or departures. An autopsy showed the absence of any external changes in the textolite or chips, everything is as it should be. There can be no miners, because I just inserted it into a computer assembled from scratch. There is a slight suspicion of pads, but to the touch, and in general, they look fresh. For some reason, I still have suspicions towards either the PSU or some software problems such as settings in the afterburner or in the windows itself. Although the PSU is already at 750 and this, in theory, should be more than enough. For some reason, I also think in the direction of the voltage settings either in the tweak or in the BIOS of the board, but in this I am a complete zero. Already stupidly ready to pay anyone who at least finds a problem and, if possible, eliminates it.
nineteensixtynine
By the way, the GPU Hot spot is not the temperature of the memory, but the temperature of the hottest spot in the GPU.
I don't see the temperatures and voltages of the video memory on your screen at all.
Although 110 degrees hotspot at standard voltage and 4400 turns of turntables is already a reason to think about whether everything is fine with the cooling system.
Loken
in the attached link to google drive there is a screenshot from GPU-Z, where it is clearly shown that the temperature of the memory matches the temperature on the hotspot. and what, sobsna, could be wrong with the cooling system? I checked the tubes, recently changed the paste. there were gaskets to check, but it is doubtful enivey
nineteensixtynine
Can you record logs from the Fire Strike Extreme dx11 test with Grafic test 1 and Grafic test 2 enabled and everything else disabled?
In particular, they are interested in the parameter logs GPU frequency, video memory frequency, GPU temperature, hotspot temperature, video memory temperature and video card fan speed.
Here is an example of my logs from this test. I recorded via HWINFO with a 100ms polling rate.
Spoiler
nineteensixtynine
I can't understand much from the logs you posted. Your GPU frequency jumps there, then in the middle all the temperatures rise sharply, and then they fall just as sharply. At the same time, your turntables, even at 4400 rpm, cannot keep temperatures up to throttling values.
Spoiler By 
the way, are you sure you didn't mix up the sensors? It hurts that the readings of the hot spot and the temperature of the memory are identical. Maybe it's data from the same sensor?
nineteen-sixtynine
In order to immediately immediately reduce all problems to zero, you do this: you go to the DNS, take the same card, or its analogue in terms of performance, for example 3060, insert it into your computer and look at the result. I've done this before. Because according to the law you can return the goods back within two weeks FOR ANY reason. A complex technical product cannot. For example, as circulars, chippers, punchers. And you will return the card 1000% back. Pray that the problem was in the system, not in the card. But I have a feeling that when you put another card, your troubles will go away. I'm telling my funny story of Myshivura: 2013. I wanted to take the second 560ti card for sli mode, but first test it separately. When I began to check it, I noticed its extremely high temperature, the inadequate hum of Carlsons, and most importantly, that its performance is ten times less, than another 560ti in games. Inserted back the old 560ti from Gigabuta. The performance is as it should be for this card. And he took it back. Now get up out of your chair and think. Your computer is built from scratch, it means it’s new, Windows is installed on a clean hard drive, the card is most likely in the first word Pisiay 16x, blockade at 750V cotton, more than enough of it, percent imb, it will work fine with my card, new drivers, power pins are all connected to your video card, the cable from the monitor is connected to the internal card, that is, to the 2080, and not to the embed. Make a conclusion. Your problem can be solved in minutes. No one needs to pay money for help. You decide. The main thing is to take another card and check. And then you will return it back to the dns, or you will leave it with you so that you don’t sit on the embed. If the problem is really in your 2080, then most likely it is software. Because you didn’t have a blueberry, as I understand it. You don’t have artifacts, hard freezes, frequent crashes from games and other graphical delusions due to the chip’s failure. Engage in its repair and hand over to specialists, nuuuuuu is such a thing. This is such a product that Uncle Vanya's intervention into the inside of the card will not give anything special, and the problem may soon appear again with it. On my memory pitch, it never happened that a problem card was repaired in a service center from a software or technical problem and it served for many more years. Therefore, only replacing it with a working card. The most important thing in this or that pitch problem is to reduce certain experiments to zero, so as not to waste time in vain and not waste your brain and nerves. you also don’t have frequent crashes from games and other graphics due to the chip’s failure. Engage in its repair and hand over to specialists, nuuuuuu is such a thing. This is such a product that Uncle Vanya's intervention into the inside of the card will not give anything special, and the problem may soon appear again with it. On my memory pitch, it never happened that a problem card was repaired in a service center from a software or technical problem and it served for many more years. Therefore, only replacing it with a working card. The most important thing in this or that pitch problem is to reduce certain experiments to zero, so as not to waste time in vain and not waste your brain and nerves. you also don’t have frequent crashes from games and other graphics due to the chip’s failure. Engage in its repair and hand over to specialists, nuuuuuu is such a thing. This is such a product that Uncle Vanya's intervention into the inside of the card will not give anything special, and the problem may soon appear again with it. On my memory pitch, it never happened that a problem card was repaired in a service center from a software or technical problem and it served for many more years. Therefore, only replacing it with a working card. The most important thing in this or that pitch problem is to reduce certain experiments to zero, so as not to waste time in vain and not waste your brain and nerves. that Uncle Vanya's intervention inside the card will not give much and the problem may soon appear again with it. On my memory pitch, it never happened that a problem card was repaired in a service center from a software or technical problem and it served for many more years. Therefore, only replacing it with a working card. The most important thing in this or that pitch problem is to reduce certain experiments to zero, so as not to waste time in vain and not waste your brain and nerves. that the intervention of Uncle Vanya inside the card will not give much and the problem may soon appear again with it. On my memory pitch, it never happened that a problem card was repaired in a service center from a software or technical problem and it served for many more years. Therefore, only replacing it with a working card. The most important thing in this or that pitch problem is to reduce certain experiments to zero, so as not to waste time in vain and not waste your brain and nerves.
Loken nope
, I have a lot of advanced edition in 3dmark. only basic, and there you can not choose. I can also drive through the usual one or stupidly another benchmark
MOSHIVURIA lapwing
, you understand what a problem is, against the backdrop of all this war, it is stupidly absent in local dnss, plus there are so many attendants, albeit temporarily, not. alas, this does not solve the problem. I, nevertheless, tend to think that the problem is in the core of the card and it is necessary to raise the temperature limits. as soon as the card heats up to 80, the pandemonium begins immediately with lags and an orb of screws. I think it would be necessary to have dropsy on the chip, but what kind of dropsy is when the percent of video is under the backplate
Loken
listen, about temperatures, I xs, what GPU-Z shows, then I tell here
nineteensixtynine wrote:
An autopsy revealed
More details here.
Did you open it yourself after buying it at home? Or was it opened by the huckster himself in front of you?
If she flew with him in Areche on ultras, then what could happen to her upon arrival home?
Even if she was killed, then in theory you should have had problems after some time (a day, a week, a month, six months, depending on the huckster), and not immediately after arrival.
But if you immediately opened it upon arrival, then there is a suspicion that the jamb is at this very moment.
nineteensixtynine
Yes, any test lasting one and a half to two minutes would be possible. Although it’s already clear that you have throttling both memory and the GPU itself, moreover, synchronously, if the sensors do not lie. Temperatures rise so sharply and strongly, as if there is no cooling at all.
I can assume that the person who was servicing the card installed the wrong thickness of the thermal pad or did not tighten the heatsink enough. Because of what there was a skew or loose contact of the heat-removal pad with the memory and the GPU.
Xs, try to sort out the cooling system, make sure that everything is tightened tightly and fits well to the cooled elements. Your turntables are threshed to the maximum, the voltages are normal while throttling both the memory and the GPU. I can not imagine any other problem, except with the contact of the cooling system.
nineteensixtynine wrote:
I still tend to think that the problem is in the core of the card and it is necessary to raise the temperature limits.
Well, you understand that this, in principle, should not initially exist and has a place to be in the problem of the map. And you want to temporarily somehow bypass this crap. But then she reappears. Yes, perhaps it's better than sitting without a computer. A great option is to put dropsy on the map. Leave the textolite only, but on the graph. install the pump. In short, as I did earlier with the second 1080. The problem is that dropsy for cards cost in this situation beyond a huge amount of money! I have a feeling that a smart-ass fox got caught and sold the card to you after mining! For a while, she more or less worked, and then all this garbage began, and you can’t prove it to that freak.
We did not open ClearSkyForUs
together with him, because we thought that we would break it. I assembled the computer two months later, inserted the card and away we go. after she lay down she slowed down slightly, but it quickly passed. I opened it after a couple of weeks when I changed the paste
Loken
well, I'll change the pads, since I'm also thinking about them. True, and here I'm afraid to do something wrong, because different people say differently which and where to put. on the nvidia forum, the celik wrote that they only need 0.5 and 1mm on both sides of the board, and on reddit the celik already says about 0.5mm, 1, 1.5 and 2.0. who to believe? no idea at all
MOSHIVURIA lapwing listen, here I don’t even know what to say, everything worked great with me, when he played Arex for three hours in a row, and then even xs. I think dropsy will somehow fix the situation, but then the ril will have to remove the screws with the radiator, and this is not comme il faut