Intermittent PCIe Recovery Count increase on MSI GPU – video signal loss in idle, stable under stress

[leo253315b002db]

MSI GeForce RTX5070 12Gb INSPIRE 3X OC

I would like to ask for clarification regarding an issue I am experiencing with my MSI graphics card.

Problem description:
After approximately 20 days of normal use, the GPU started showing intermittent instability related to the PCIe connection. The issue manifests as:

• Sudden loss of video signal (monitors go black)
• GPU fans ramping up to 100%
• System itself remains powered on
• PCIe Recovery Count in HWInfo increases during these events

I am attaching a screenshot showing the PCIe Recovery Counter.

Important observations:

• The Recovery Count increases during idle or light load (desktop usage, browser, YouTube, etc.)
• During stress tests (GPU fully loaded) the Recovery Count does NOT increase at all
• No PCIe CRC / Replay / Bad TLP / Lane errors are recorded
• The same system with a different GPU works perfectly stable and shows Recovery Count = 0
• Power supply: be quiet! 750W, same PSU used for both GPUs
• Different PCIe power cables were tested

Service center feedback:
The authorized service center reported that the issue did not reproduce under service conditions and explained that Recovery Count usually does not indicate a PCIe error.
However, the issue continues to reproduce consistently during real-world usage.

Why this seems unusual:
From my understanding:

• If the issue were related to PSU, cables, or motherboard, it would most likely appear under stress, not disappear
• Recovery Count increasing only during idle suggests instability during power-state or PCIe link-state transitions
• This behavior does not seem normal for a properly working GPU

Questions to MSI:

1. Is regular PCIe Recovery Count increase during idle considered normal behavior for MSI GPUs?
2. Can intermittent PCIe instability fail to reproduce during stress testing but appear during idle transitions?
3. Would this behavior indicate a possible hardware issue related to PCIe PHY or power-state handling?
4. Are there any known MSI GPU cases with similar symptoms?

Any insight or clarification would be greatly appreciated.

 

[Alan J T]

• The same system with a different GPU works perfectly stable and shows Recovery Count = 0
• Power supply: be quiet! 750W, same PSU used for both GPUs
• Different PCIe power cables were tested

What was the Test GPU
Try setting the PCIE link in BIOS to Gen 4.0. This will have little or no effect on the actual GPU performance
Also, are you able to test with a Different PSU

 

[leo253315b002db]

Thank you for the suggestions.

I have manually set the PCIe link to Gen 4.0, however this did not have any effect on the issue. The behavior remains the same — PCIe Recovery Count continues to increase during idle/light load and remains stable during stress tests.

I will also try to test the GPU in a completely different system. This is essentially the maximum level of testing I am able to perform on my side.

Could you please clarify the following:
Is it considered normal behavior for a properly functioning MSI GPU that the PCIe Recovery Count continuously increases during idle, but does not increase under load?

 

[Alan J T]

Could you please clarify the following:
Is it considered normal behavior for a properly functioning MSI GPU that the PCIe Recovery Count continuously increases during idle, but does not increase under load?

What is the Event ID Number

As far as I know, it should not be doing it

Or I may be completely wrong, But I have not seen the GPU power draw this low before. Nvidia may have done something with the Power Profiles

I will make my Contact aware and ask for testing

 

[leo253315b002db]

After setting NVIDIA Control Panel → Manage 3D Settings → Power Management Mode → Prefer Maximum Performance, the PCIe Recovery Count no longer increases. So far, I have not encountered any GPU crashes.

Is this considered normal behavior for a properly functioning GPU, that the issue only disappears when forcing Maximum Performance mode?

Additionally, I tested the GPU in a completely different system, and the Recovery Count also increased there, which indicates that the behavior is not system-specific.

 

[Alan J T]

Is this considered normal behavior for a properly functioning GPU, that the issue only disappears when forcing Maximum Performance mode?

I have put in a request with MSI to look into the issue, and let me know if it is expected behaviour or not

From waht I have been able to test it is due to the lower power setting by NVIDIA or Windows. It appears to trigger on the 5090 as soon as it drops below 10- watts

Opening and closing Steam and Discord windows also triggers the event

 

[Alan J T]

After setting NVIDIA Control Panel → Manage 3D Settings → Power Management Mode → Prefer Maximum Performance, the PCIe Recovery Count no longer increases. So far, I have not encountered any GPU crashes.

Is this considered normal behavior for a properly functioning GPU, that the issue only disappears when forcing Maximum Performance mode?

Additionally, I tested the GPU in a completely different system, and the Recovery Count also increased there, which indicates that the behavior is not system-specific.

can you do a screen shot the GPU+Z Bios information page, please and post it
This one as shown

 

[leo253315b002db]

No problem, maybe something else?

 

[Alan J T]

No problem, maybe something else?

Unknown at the moment, but MSI is looking into it now that they know about it
This way, when they test the GPU, they can place the same V-Bios onto it to match your system as much as possible

 

[citay]

To chime in, someone I know has the same issue. New ASUS GeForce 5070 on an MSI MPG Z390 GAMING EDGE AC. He has to use this card on such a rather old PCIe 3.0 system now, because RAM prices etc. are a bit too high for an upgrade right now. I helped him install it, on first boot we get a black screen and reboot before we could even do anything. We chalked it off as a one-time thing because of the drivers or something. But later he told me he had it happen another two or three times. Monitor goes blank and loses signal, and shortly thereafter, the PC reboots. One time during gaming, but when the map changed and it was loading.

I briefly looked at the HWinfo Sensors and saw some PCIe errors being maxed out (65K), but according to this thread, only "Fatal Error Count" and perhaps "Lane Errors" should be critical, while some others can be normal, so I wasn't sure how relevant these errors were. I will try to take a screenshot soon, for sure it looked worse than on the screenshots here (BTW, please set HWinfo to English).

After setting NVIDIA Control Panel → Manage 3D Settings → Power Management Mode → Prefer Maximum Performance, the PCIe Recovery Count no longer increases. So far, I have not encountered any GPU crashes.

The problem with that workaround is the high idle power draw, 30W instead of 6W.

Additionally, I tested the GPU in a completely different system, and the Recovery Count also increased there, which indicates that the behavior is not system-specific.

Recovery count on its own is probably not indicative of a problem. According to the HWinfo author:

The "Recovery Count" counts the number of changes from L0 to Recovery. It triggers for example during a change in speed, width, or other possible reasons that usually don't mean a PCIe error occured.

L0 is a lower-power state of the GPU. So when you set High Performance power plan, it will not enter those low-power states anymore, it will keep clocking higher and use more voltage. The advantage is, any problems with entering and coming out of low-power states are prevented, the disadvantage is, it's wasting power for nothing, causing more heat and more energy cost.

So it's possible that the 5070 cards can trigger some kind of issue with the power-saving in idle, perhaps specifically on MSI boards. I doubt there will be another BIOS update for the Z390 boards. So perhaps it has to be solved in another way. Right now I don't see a good way that also preserves the lowest power draw in idle.

 

[Alan J T]

So it's possible that the 5070 cards can trigger some kind of issue with the power-saving in idle, perhaps specifically on MSI boards

I'm running a 5090, and Performance is not having any problems, still passing a Stress test
It's just a strange reset problem. I have not had any problems with the GPU, such as a black screen or other types of errors

 

[Alan J T]

Could you please clarify the following:
Is it considered normal behavior for a properly functioning MSI GPU that the PCIe Recovery Count continuously increases during idle, but does not increase under load?

Have heard back, and it is normal behaviour at low power mode


THis is also verified by the Author of HwInfo64

 

[leo253315b002db]

Thank you for the information. I previously had an issue where the graphics card would suddenly stop working and the fans would ramp up to 100%. The PCIe Recovery Count was the only suspicious indicator I could observe at that time.

However, the service center stated that the card is functioning properly, and after the service the issue has not repeated so far. I am not sure whether everything will remain stable, but for now I will not focus on this indicator as much.

 

[citay]

Ok, I solved it for the case I had here. First we tried around with increasing CPU SA Voltage and PCH voltage, since that was mentioned in another thread of the HWinfo forums as a possible solution, but there was no change, the Receiver Errors were stuck at 65K permanently (maxed out counter), and the Recovery Count also rising in record time.

But in the past, I had another Intel system that gave us trouble after a BIOS update, leading to high DPC latency in LatencyMon and audio dropouts during DAW work, and I knew that I could solve these things by enabling "Native PCIe Enable", which defaulted to [Disabled] after the update. So I tried it here, and enabling it immediately solved the issue:

So enabling this setting is what might help on an Intel system. I'm not sure if AMD-based boards have this option too.

Also, seeing how that system just now has started to look like the initial screenshot in this thread, it could've been an unrelated issue that the system additionally had. I will keep in contact with the owner if there are any further crashes/reboots of if this fixed those as well for him.

 

[Alan J T]

So enabling this setting is what might help on an Intel system. I'm not sure if AMD-based boards have this option too.

Not a setting I recall seeing, but I will have a look at the PCIe settings that are available for AMD Click BIOS X