PC severely underperforming

[Copi]

Hello, I just finished fixing my computer's heating and then later boot issues. My computer boots up normally, but my CPU, and mainly my GPU are performing a lot worse than before. For example, I get 1/4 of the frames I normally get in Destiny 2, and according to userbenchmark my CPU is in the 27th percentile and my GPU in the 0th. I've made sure all the PSU connections are secure and in the right place, and that something else isn't throttling it like XMP not being on. My full PC specs are as follows: 
CPU: Intel Core i7-12700KF

GPU: Gigabyte Eagle RTX 3070 TI (LHR)

RAM: 2x8 gb Trident Z RGB (3600mhz xmp)

Storage: A 512g intel optane nvme and a 512g teamgroup nvme)
PSU: EVGA 750w GA
BIOS: American Megatrends International, LLC. F6, 1/27/2022

OS version: Microsoft Windows 10 Home 10.0.19044 Build 19044.

Any ideas on what to do would be greatly appreciated.

[KangUsapEmenq]

1 minute ago, Copi said:

fixing my computer's heating

We need more details as of what did you do onto your computer (cleaning? repasting? or else?) as to more detailed answers either.

As for cleaning and repasting, I assume the thermal paste either being too much, not spreading nicely, or it's not a good one.

[RONOTHAN##]

Run an actual benchmark (something like 3DMark Fire Strike or Time Spy, they've got both a GPU and CPU component to them and are free in the 3DMark demo on Steam, and they're actually consistent and good unlike UserBenchmark) and have HWiNFO64 running in the background logging during the run. That should give all the info we need in order to figure out what's going on. 

[Copi]

6 minutes ago, TukangUsapEmenq said:

We need more details as of what did you do onto your computer (cleaning? repasting? or else?) as to more detailed answers either.

As for cleaning and repasting, I assume the thermal paste either being too much, not spreading nicely, or it's not a good one.

I moved the AIO and cleaned out the rad, also ignore the solution stuff, I misclicked.

[LIGISTX]

11 minutes ago, Copi said:

I moved the AIO and cleaned out the rad, also ignore the solution stuff, I misclicked.

What are your CPU temps and what speed is it boosting to? Maybe you unplugged the pump or fans by mistake? Or caused the AIO to get slightly dismounted by bumping it. 

[Copi]

1 hour ago, LIGISTX said:

What are your CPU temps and what speed is it boosting to? Maybe you unplugged the pump or fans by mistake? Or caused the AIO to get slightly dismounted by bumping it. 

CPU temps are in the 50c ish range, and the AIO seems to be working.

[NaClKnight]

23 minutes ago, Copi said:

CPU temps are in the 50 ish range

During the benchmark? Or while idling? That's high for an idle temp and low for a benchmark temp. Let's get HWINFO64 running and confirm that your Aio is reporting a pump "fan" speed in addition to the speed reported by the radiator fans

 

From there let's run firestrike or time spy and get some numbers to compare to, as mentioned above

[Copi]

10 hours ago, NaClKnight said:

During the benchmark? Or while idling? That's high for an idle temp and low for a benchmark temp. Let's get HWINFO64 running and confirm that your Aio is reporting a pump "fan" speed in addition to the speed reported by the radiator fans

 

From there let's run firestrike or time spy and get some numbers to compare to, as mentioned above

I can't seem to find a pump fan speed in HWINFO, but here's the results of my timespy benchmark:

 

[Copi]

12 hours ago, RONOTHAN## said:

Run an actual benchmark (something like 3DMark Fire Strike or Time Spy, they've got both a GPU and CPU component to them and are free in the 3DMark demo on Steam, and they're actually consistent and good unlike UserBenchmark) and have HWiNFO64 running in the background logging during the run. That should give all the info we need in order to figure out what's going on. 

I did that just now, the results are in a reply to another comment below. 

[NaClKnight]

27 minutes ago, Copi said:

but here's the results of my timespy benchmark:

Ok, those GPU frequencies are absurdly low, both for GPU core and GPU memory (VRAM). Per the listed specs that thing should be clocking between 1500 and 1750 MHz for both the GPU and the VRAM

 

Is it possible that there's a setting limiting the power supplied to GPU, either in the Windows settings or the Nvidia settings? Or a potential Nvidia driver error? "GPU running at extremely low frequencies" is a problem unto itself

37 minutes ago, Copi said:

I can't seem to find a pump fan speed in HWINFO

That's strange. It will be listed in one of the motherboard header sections, likely at the bottom of a section. This is mine, for instance, but we're interested in confirming that the pump is running and the fans are too. 

[Copi]

3 minutes ago, NaClKnight said:

Ok, those GPU frequencies are absurdly low, both for GPU core and GPU memory (VRAM). Per the listed specs that thing should be clocking between 1500 and 1750 MHz for both the GPU and the VRAM

 

Is it possible that there's a setting limiting the power supplied to GPU, either in the Windows settings or the Nvidia settings? Or a potential Nvidia driver error? "GPU running at extremely low frequencies" is a problem unto itself

That's strange. It will be listed in one of the motherboard header sections, likely at the bottom of a section. This is mine, for instance, but we're interested in confirming that the pump is running and the fans are too. 

these are the only ones with a fan icon that I can find

[NaClKnight]

4 minutes ago, Copi said:

these are the only ones with a fan icon that I can find

Could System 2 correspond to any case fan/splitter you have connected?

If not, then those fans are strangely connected. 

 

By default, the fans on your AIO radiator should be connected to the CPU Fans header

The AIO Pump should connect to any other header on your motherboard. Ideally CPU_OPT or PUMP but it's not a huge deal. The important part is that those two should each have their own header on the Motherboard, and that the AIO Pump should be set to run at 100% voltage or duty cycle at all times. 

 

Then your other case fans should be attached to the remaining headers however possible. 

 

The important part though is that your Pump and AIO fans are running to keep your CPU cool during testing to avoid throttling during usage and testing.

[Copi]

2 minutes ago, NaClKnight said:

Could System 2 correspond to any case fan/splitter you have connected?

If not, then those fans are strangely connected. 

 

By default, the fans on your AIO radiator should be connected to the CPU Fans header

The AIO Pump should connect to any other header on your motherboard. Ideally CPU_OPT or PUMP but it's not a huge deal. The important part is that those two should each have their own header on the Motherboard, and that the AIO Pump should be set to run at 100% voltage or duty cycle at all times. 

 

Then your other case fans should be attached to the remaining headers however possible. 

 

The important part though is that your Pump and AIO fans are running to keep your CPU cool during testing to avoid throttling during usage and testing.

System 2 is probably my 3 intake fans. The pump itself seems to be working as it is pretty warm to the touch and feels like something is flowing through. The fans connected to the rad are both spinning. I'll try giving the pump its own header by itself. I am using a hub for my fan and rgb headers, which may have something to do with the problem. How would I check that the AIO pump is running at 100% voltage?

[NaClKnight]

10 minutes ago, Copi said:

System 2 is probably my 3 intake fans. The pump itself seems to be working as it is pretty warm to the touch and feels like something is flowing through. The fans connected to the rad are both spinning. I'll try giving the pump its own header by itself. I am using a hub for my fan and rgb headers, which may have something to do with the problem. How would I check that the AIO pump is running at 100% voltage?

Ohhhhh, if you're using a Hub that changes things.

Does your Hub connect via software? You may be able to check or change the Pump AIO speed there.

[Copi]

3 minutes ago, NaClKnight said:

Ohhhhh, if you're using a Hub that changes things.

Does your Hub connect via software? You may be able to check or change the Pump AIO speed there.

no, it just came with the case.

[NaClKnight]

Ok, in that case, move the Pump to its own header on the motherboard. You can leave all the LEDs and other fans alone for now. Just the Pump header is fine. 

[Copi]

4 minutes ago, NaClKnight said:

Ok, in that case, move the Pump to its own header on the motherboard. You can leave all the LEDs and other fans alone for now. Just the Pump header is fine. 

All right, I'll do that. Thank you for the advice. Any suggestions about the low GPU clock speed?

[NaClKnight]

6 minutes ago, Copi said:

All right, I'll do that. Thank you for the advice. Any suggestions about the low GPU clock speed?

Try this thread: 

 

The idea is to check your GPU power/wattage limits (using MSI Afterburner or somethinng similar) and see if that may be causing the low GPU results

 

[Copi]

1 hour ago, NaClKnight said:

Try this thread: 

 

The idea is to check your GPU power/wattage limits (using MSI Afterburner or somethinng similar) and see if that may be causing the low GPU results

 

It may be. I don't know if I'm looking at the right numbers or not, but the power limit seems to be 0 or almost nothing.

[NaClKnight]

5 minutes ago, Copi said:

It may be. I don't know if I'm looking at the right numbers or not, but the power limit seems to be 0 or almost nothing.

Post a screenshot. Power limits may be presented as Wattage or Percentage offsets from the stock value, or as raw numbers. 

[Copi]

8 minutes ago, NaClKnight said:

Post a screenshot. Power limits may be presented as Wattage or Percentage offsets from the stock value, or as raw numbers. 

This is what I can find in my afterburner, the other ones seem to be related to the CPU.

[NaClKnight]

2 hours ago, Copi said:

This is what I can find in my afterburner, the other ones seem to be related to the CPU.

I'm not an expert on NVidia GPUs but that screen looks like it's missing a lot of data. Aside from Afterburner it seems clear that the GPU is being power limited somehow, either by SW or by Windows. The question now is lifting that limit.

 

To confirm that, lets go back to HWINFO64 for a bit. The GPU sensors section should be towards the bottom, and one of the items (with a lightning bolt) should mention PPT or Power or something similar witj a measurement in W: 

 

Run another TimeSpy benchmark while HWINFO is open and take a look at the "Max" GPU values after Timespy completes. My AMD GPU has different labels and sensors, but your GPU will have equivalents. Time Spy should push your GPU as hard as it can go, so after it runs, your "Max" power values should all be in the high 200's, between 250W and 315W.

 

Take a screenshot of the GPU section of HWINFO after running Timespy and let's review if that hunch is correct

[Copi]

32 minutes ago, NaClKnight said:

I'm not an expert on NVidia GPUs but that screen looks like it's missing a lot of data. Aside from Afterburner it seems clear that the GPU is being power limited somehow, either by SW or by Windows. The question now is lifting that limit.

 

To confirm that, lets go back to HWINFO64 for a bit. The GPU sensors section should be towards the bottom, and one of the items (with a lightning bolt) should mention PPT or Power or something similar witj a measurement in W: 

 

Run another TimeSpy benchmark while HWINFO is open and take a look at the "Max" GPU values after Timespy completes. My AMD GPU has different labels and sensors, but your GPU will have equivalents. Time Spy should push your GPU as hard as it can go, so after it runs, your "Max" power values should all be in the high 200's, between 250W and 315W.

 

Take a screenshot of the GPU section of HWINFO after running Timespy and let's review if that hunch is correct

here you go

[NaClKnight]

36 minutes ago, Copi said:

here you go

Are you sure you took these numbers **after** running the TimeSpy benchmark? If so,

 

Well I have some unfortunate news...

 

Nothing about this screenshot makes sense.

Everything from the PCIe Link Speed at the bottom (Should be 8.0 GT/s) to the total GPU Power never rising about 25% to the GPU power maxing out at 35W and the absymal (under 1000 MHz) clock speeds. Everything is weird here. 

 

At this point there may be an incorrect BIOS setting limiting the GPU. Im also curious about the "Performance Limit - U" there under GPU Fan2, the one that says Yes. Can you let your mouse hover on the that category to get a full view of what it says?

 

Realistically though, your next step may be creating a new thread with all of the images and info you've provided to me, and asking for help in the GPU forum specifically about why your GPU specifically is running at such reduced settings 

 

We know why it's underperforming: the clock speeds and power usage are all greatly reduced. The question is what setting or software could be causing this.

[Copi]

20 minutes ago, NaClKnight said:

Are you sure you took these numbers **after** running the TimeSpy benchmark? If so,

 

Well I have some unfortunate news...

 

Nothing about this screenshot makes sense.

Everything from the PCIe Link Speed at the bottom (Should be 8.0 GT/s) to the total GPU Power never rising about 25% to the GPU power maxing out at 35W and the absymal (under 1000 MHz) clock speeds. Everything is weird here. 

 

At this point there may be an incorrect BIOS setting limiting the GPU. Im also curious about the "Performance Limit - U" there under GPU Fan2, the one that says Yes. Can you let your mouse hover on the that category to get a full view of what it says?

 

Realistically though, your next step may be creating a new thread with all of the images and info you've provided to me, and asking for help in the GPU forum specifically about why your GPU specifically is running at such reduced settings 

 

We know why it's underperforming: the clock speeds and power usage are all greatly reduced. The question is what setting or software could be causing this.

Here's a screenshot of that, the cut off lines are things like power, thermal, max operating, voltage. I'll go make a thread in the GPU forum now, thank you for all your help. These numbers came from maybe 20-30 seconds after the timespy benchmark.