HELP. My fps has dropped significantly

[Janzen]

I just recently finished doing my water cooling loop and everything was installed correctly and everything is running fine. I have 5800x and rtx 3080 and I normally get around 150+ in warzone but now I am getting 80 or below and gameplay just isn't as smooth as normal. I also play on low graphics settings. I tried playing other games and some games runs fine and others stutter and have low fps. My temps seem normal, idle around 60c for both CPU and GPU and it goes up to about 70-75c for both when I start gaming . I have both water cooled. I'm pretty sure everything is getting enough power from psu, I have 850 watt psu. I updated my bios and all drivers already. I really don't know what else to try 

[Roswell]

If your components are truly idling at 60c, especially liquid cooled, there's something really wrong going on. Even more puzzling is that your delta over idle is only a 15c difference. It should be much higher...

 

You're probably heavily thermal throttling and those relatively good looking temps you see while gaming are the result of aggressive downclocking.

 

Check CPU and GPU usage and make sure you're actually idling at 60 and if you are, you need to sort out whatever's wrong. Dead pump, bad thermal paste applications, etc.

[Hairless Monkey Boy]

Since you recently changed your cooling solution, that is the obvious first suspect.

 

60C idle is very hot, especially for water cooling. Your idle temps should be only a degree or two higher than your idle coolant temp, which in a proper water cooling loop should be the ambient temp or a few degrees higher.

 

Going from 60C idle to 70C full load seems highly unusual. So I suspect these numbers to be inaccurate. Get HWinfo64. Power down your PC for several hours to let everything cool down to ambient temps. Boot it back up and run HWinfo64. Open up charts for the CPU and GPU Usage, temperatures, and clock speeds. Startup a game and play for a half hour or so, close the game, let the PC cool off for about 5 minutes, and then screen shot the charts and post them here.

[Janzen]

I will give that a try, I figured my temps weren't normal but I'm new to water cooling so I wasn't sure. Will my pc be safe to use for the time being? I  I am using it for school, obviously no gaming but I need it for the rest of the day 

10 minutes ago, Hairless Monkey Boy said:

Since you recently changed your cooling solution, that is the obvious first suspect.

 

60C idle is very hot, especially for water cooling. Your idle temps should be only a degree or two higher than your idle coolant temp, which in a proper water cooling loop should be the ambient temp or a few degrees higher.

 

Going from 60C idle to 70C full load seems highly unusual. So I suspect these numbers to be inaccurate. Get HWinfo64. Power down your PC for several hours to let everything cool down to ambient temps. Boot it back up and run HWinfo64. Open up charts for the CPU and GPU Usage, temperatures, and clock speeds. Startup a game and play for a half hour or so, close the game, let the PC cool off for about 5 minutes, and then screen shot the charts and post them here.

 

[Hairless Monkey Boy]

Just now, Janzen said:

Will my pc be safe to use for the time being?

I have no idea. If it suddenly shuts down, you'll have your answer. 

[Janzen]

Just now, Hairless Monkey Boy said:

I have no idea. If it suddenly shuts down, you'll have your answer. 

Okay thanks hopefully it doesn't   

[HanZie82]

Most components are protected from overheating.

But i would not test it, unless you have to much money to buy it again.

 

But sharing some info on the setup could help.

Like what pump and radiator as well as how you connected the coldplates?
Are you sure you used thermal compound of good enough quality?
Are the plates correctly tightened?

etc.

 

[Janzen]

2 minutes ago, HanZie82 said:

Most components are protected from overheating.

But i would not test it, unless you have to much money to buy it again.

 

But sharing some info on the setup could help.

Like what pump and radiator as well as how you connected the coldplates?
Are you sure you used thermal compound of good enough quality?
Are the plates correctly tightened?

etc.

 

I used the Corsair H305i kit.

• XD5 pump
• XC7 CPU block
• XG7 GPU block
• Entho Evolv mATX case
• XR5 240 radiator 
• Cooler master V850 sfx gold PSU
• Gigabyte B550i Aorus pro AX mobo

It came with all the thermal compound applied. I think I'll probably double check everything when I get the chance since it seems like my temps should be lower 

[Hairless Monkey Boy]

9 minutes ago, Janzen said:

• XR5 240 radiator

• Entho Evolv mATX case

If you use only a single 240mm radiator for both a 3080 and any CPU in a case with notoriously poor airflow, you're gonna have a bad time.

[Janzen]

47 minutes ago, Hairless Monkey Boy said:

If you use only a single 240mm radiator for both a 3080 and any CPU in a case with notoriously poor airflow, you're gonna have a bad time.

So a second radiator would most likely fix my issues? I couldn’t fit another one but I could try to find a place for it

[Hairless Monkey Boy]

4 minutes ago, Janzen said:

So a second radiator would most likely fix my issues?

Not necessarily. Your primary issue may still be elsewhere. You'll need to provide the requested information before I'd be willing to make that claim.

 

That being said, 240mm would be my recommendation as a minimum size for the GPU alone, and that's assuming a case with decent airflow. Add the CPU into the mix and I'd recommend 360mm minimum.

[Janzen]

11 hours ago, Hairless Monkey Boy said:

Not necessarily. Your primary issue may still be elsewhere. You'll need to provide the requested information before I'd be willing to make that claim.

 

That being said, 240mm would be my recommendation as a minimum size for the GPU alone, and that's assuming a case with decent airflow. Add the CPU into the mix and I'd recommend 360mm minimum.

Here are the charts. I hope I got the right ones 

 

[Hairless Monkey Boy]

5 hours ago, Janzen said:

Here are the charts. I hope I got the right ones 

 

Awesome, thanks for providing those. Not exactly what I'm looking for, unfortunately.

 

I don't need to see "GPU Memory Usage" nor "GPU Memory Clock". And you are missing a GPU Utilization chart. A GPU memory temperature chart would be a welcome addition, as well.

 

The charts are not showing all the data I'd like to see. You might need to capture them a couple times. Here is what I want to see:

 

1. I want to see what the system looks like when it first boots up after being off for a while. So you would power up the system after it's been off for several hours, startup hwinfo right away and open up all the charts, once the charts are nearly full of data, capture a screenshot.
2. I want to see what it looks like when you first boot up a game. So after you take that first screen shot, launch a game and let it run for a little while. I want to see the transient between cold idle and running game, so let it run until is reaches "steady state", or until the moment you started the game in the chart starts getting close to the left edge of the chart.
3. And I want to see what it looks like after you close a game. This is what the charts you supplied show, though they aren't all the correct charts, so just do that again but with the correct charts open. Explicitly, close the game and allow the system to run idle until the charts look about the same as the ones you provided, then take a final screenshot. There should be a small amount of "game running" data and then a long "cooling down" tail.

Oh, and make sure you click the "auto fill" button on each chart before each screen shot.

[Janzen]

2 hours ago, Hairless Monkey Boy said:

Awesome, thanks for providing those. Not exactly what I'm looking for, unfortunately.

 

I don't need to see "GPU Memory Usage" nor "GPU Memory Clock". And you are missing a GPU Utilization chart. A GPU memory temperature chart would be a welcome addition, as well.

 

The charts are not showing all the data I'd like to see. You might need to capture them a couple times. Here is what I want to see:

 

1. I want to see what the system looks like when it first boots up after being off for a while. So you would power up the system after it's been off for several hours, startup hwinfo right away and open up all the charts, once the charts are nearly full of data, capture a screenshot.
2. I want to see what it looks like when you first boot up a game. So after you take that first screen shot, launch a game and let it run for a little while. I want to see the transient between cold idle and running game, so let it run until is reaches "steady state", or until the moment you started the game in the chart starts getting close to the left edge of the chart.
3. And I want to see what it looks like after you close a game. This is what the charts you supplied show, though they aren't all the correct charts, so just do that again but with the correct charts open. Explicitly, close the game and allow the system to run idle until the charts look about the same as the ones you provided, then take a final screenshot. There should be a small amount of "game running" data and then a long "cooling down" tail.

Oh, and make sure you click the "auto fill" button on each chart before each screen shot.

My mistake I thought you only wanted the last graphs but here are the new ones. 

 

This is the first start

 

Here are the ones for while I game

 

And these are after the game has been closed

[Hairless Monkey Boy]

Excellent job. Those are very informative. And good inclusion of the GPU power usage.

 

In short, you have a problem with the GPU's cooling. It's thermal throttling pretty hard. Steep drops in clocks, and only using a portion of it's power limit, when normally a 3080 would be using 100% of it's power limit unless there is a FPS cap enabled (vsync or otherwise). The GPU memory temps look fine(actually very good), so maybe you are using thermal pads that are too thick, or the GPU block is not installed properly, or it is not perfectly compatible with your specific card.

 

I can go into detail about how I came to those conclusions if you want.

[Janzen]

4 minutes ago, Hairless Monkey Boy said:

Excellent job. Those are very informative. And good inclusion of the GPU power usage.

 

In short, you have a problem with the GPU's cooling. It's thermal throttling pretty hard. Steep drops in clocks, and only using a portion of it's power limit, when normally a 3080 would be using 100% of it's power limit unless there is a FPS cap enabled (vsync or otherwise). The GPU memory temps look fine(actually very good), so maybe you are using thermal pads that are too thick, or the GPU block is not installed properly, or it is not perfectly compatible with your specific card.

 

I can go into detail about how I came to those conclusions if you want.

I can see how you got the conclusions. I don't have vsync or any FPS cap enabled and everything was already pre-applied to the waterblock do you think I should re-do the thermal pads? I can re-install the waterblock just to make sure but  I know it's compatible with my gpu, well that's what it says on corair's website at least. 

 

Or would another radiator fix the thermal throttling? 

[Hairless Monkey Boy]

11 minutes ago, Janzen said:

I can see how you got the conclusions. I don't have vsync or any FPS cap enabled and everything was already pre-applied to the waterblock do you think I should re-do the thermal pads? I can re-install the waterblock just to make sure but  I know it's compatible with my gpu, well that's what it says on corair's website at least. 

 

Or would another radiator fix the thermal throttling? 

If the block is compatible and it came with the appropriate thermal pads, then your primary issue seems to be a bad GPU cold plate mount. If it was mounted correctly, the GPU idle temp and the CPU idle temp would be within 1-2 degrees of each other (and ~0-5 degrees above the coolant temp).

 

Once you solve that issue, it may be prudent to add another radiator to the loop, but you'll have to correct the GPU cold plate mount before that determination can be made. Take apart the GPU. Take note of the thermal paste squish pattern (maybe take a picture and post it).

 

As a side note, I noticed that you are running your Windows in a high power mode, and you have the Nvidia Control Panel setting "Power Management Mode" setup with a "Global Setting" high performance mode. I'd recommend setting the NCP setting per-game rather than globally, that way the GPU will downclock when you are just on your desktop. As for the Windows Power plan, I tend to use balanced mode so the CPU downclocks, but that doesn't consume as much power or produce as much heat as an always active GPU, so you can use your own judgement on that one.

[Janzen]

56 minutes ago, Hairless Monkey Boy said:

If the block is compatible and it came with the appropriate thermal pads, then your primary issue seems to be a bad GPU cold plate mount. If it was mounted correctly, the GPU idle temp and the CPU idle temp would be within 1-2 degrees of each other (and ~0-5 degrees above the coolant temp).

 

Once you solve that issue, it may be prudent to add another radiator to the loop, but you'll have to correct the GPU cold plate mount before that determination can be made. Take apart the GPU. Take note of the thermal paste squish pattern (maybe take a picture and post it).

 

As a side note, I noticed that you are running your Windows in a high power mode, and you have the Nvidia Control Panel setting "Power Management Mode" setup with a "Global Setting" high performance mode. I'd recommend setting the NCP setting per-game rather than globally, that way the GPU will downclock when you are just on your desktop. As for the Windows Power plan, I tend to use balanced mode so the CPU downclocks, but that doesn't consume as much power or produce as much heat as an always active GPU, so you can use your own judgement on that one.

Okay that make sense. I will take apart the block when I get the chance and I will update on here

 

 

[Janzen]

On 5/24/2022 at 8:04 AM, Hairless Monkey Boy said:

If the block is compatible and it came with the appropriate thermal pads, then your primary issue seems to be a bad GPU cold plate mount. If it was mounted correctly, the GPU idle temp and the CPU idle temp would be within 1-2 degrees of each other (and ~0-5 degrees above the coolant temp).

 

Once you solve that issue, it may be prudent to add another radiator to the loop, but you'll have to correct the GPU cold plate mount before that determination can be made. Take apart the GPU. Take note of the thermal paste squish pattern (maybe take a picture and post it).

 

As a side note, I noticed that you are running your Windows in a high power mode, and you have the Nvidia Control Panel setting "Power Management Mode" setup with a "Global Setting" high performance mode. I'd recommend setting the NCP setting per-game rather than globally, that way the GPU will downclock when you are just on your desktop. As for the Windows Power plan, I tend to use balanced mode so the CPU downclocks, but that doesn't consume as much power or produce as much heat as an always active GPU, so you can use your own judgement on that one.

So I was planning on taking apart the block today but when I booted up my GPU and CPU idle temps are how you said they would be if the cold plate was mounted correctly. GPU is idling 38-40 and CPU idling at 40-42. It's still throttling but not as bad as before. I also changed the power modes like you said. Should I still take apart the block? or do you think it might be something else? 

 

Let me know if you want me to post the graphs on here again 

[Hairless Monkey Boy]

1 hour ago, Janzen said:

So I was planning on taking apart the block today but when I booted up my GPU and CPU idle temps are how you said they would be if the cold plate was mounted correctly. GPU is idling 38-40 and CPU idling at 40-42. It's still throttling but not as bad as before. I also changed the power modes like you said. Should I still take apart the block? or do you think it might be something else? 

 

Let me know if you want me to post the graphs on here again 

The thing is, the hot spot temp was also higher compared with the GPU temp than I would expect. Typically, and in my experience, the gap should be around 10C, but yours was closer to 30C+. This indicates uneven cooling of the GPU die, which would indicate a poor mount.

 

You can go ahead and post more graphs if you wish. I'll certainly look at them and let you know what I think.

[Janzen]

13 minutes ago, Hairless Monkey Boy said:

The thing is, the hot spot temp was also higher compared with the GPU temp than I would expect. Typically, and in my experience, the gap should be around 10C, but yours was closer to 30C+. This indicates uneven cooling of the GPU die, which would indicate a poor mount.

 

You can go ahead and post more graphs if you wish. I'll certainly look at them and let you know what I think.

The hot spot temp changed to. Here are the graphs, this is from a little bit gaming and then after closing the game. I didn't include first boot cause my computer has been for a bit. 

 

I can still open the block if you think I should I just thought it was weird that my GPU temps changed that much

[Hairless Monkey Boy]

6 hours ago, Janzen said:

I just thought it was weird that my GPU temps changed that much

Idle temps have changed thanks to the change in performance mode. Nothing else has changed.

[Janzen]

14 hours ago, Hairless Monkey Boy said:

Idle temps have changed thanks to the change in performance mode. Nothing else has changed.

I opened up the block, here are the pictures 

[Hairless Monkey Boy]

Textbook bad mount. The GPU die isn't even completely covered in thermal compound (which it absolutely must be).

[Janzen]

11 minutes ago, Hairless Monkey Boy said:

Textbook bad mount. The GPU die isn't even completely covered in thermal compound (which it absolutely must be).

The thermal paste was pre-applied I should’ve just re-applied it myself and I also put the thermal pads on the board itself just to be safe