GPU RTX 4080 (laptop) running at x16 1.1 instead of x16 4.0 during stress test

[peterj]

I have a very weird issue with my laptop which has a capability of running the GPU on X16 gen 4.0.

GPU after a while only runs in x16 gen 1.1 , and not just in idle but under load as well. When I disconnect the ac adapter from the laptop and plug it back, GPU-Z shows it is running in x16 4.0 but then never goes back to idle, it is constantly running in gen 4.0, for about an hour or so or even more, then it suddenly switch back to x16 1.1 again, until I do the same process again with the ac adapter plug or simply just restart the laptop. This does not affect my CPU performance at all. And just to confirm, the GPU runs x16 1.1 not just on light stress on GPU-Z but during Furmark / Timespy tests as well. I checked on GPU demanding games and running on gen 1.1 causes way less FPS than running on gen 4.

What I've tried so far without any success:
- There is no advanced BIOS to this laptop, which allows me to change the pcie settings
- Latest BIOS and older versions as well.
- Checked the temperatures, no thermal throttling happens
- Checked the voltage, it is stable at max 0.975V no fluctuations
- Checked with latest nvidia driver as well as old ones
- Fresh windows install
- Latest intel chipset driver

Other important thing I would mention is that this happened on my old laptop motherboard and happens again with the new warranty replacement one. My laptop is Lenovo Legion Legion Pro 7 16IRX8H.
So if it's not the motherboard, is it possible that my AC adapter cannot give enough power to the GPU?

 

[agent_x007]

GPU-z "stress test" isn't stressful enough to get your GPU going full tilt (same goes for max. 3D clocks card can reach under load actually).
Try Unigine Superpostion in Window mode (or Heaven 4.0).

 

[DarkDreams]

Check what your windows power plan -> advanced power settings -> pcie link speed management says and try different settings there maybe.

 

[peterj]

GPU-z "stress test" isn't stressful enough to get your GPU going full tilt (same goes for max. 3D clocks card can reach under load actually).
Try Unigine Superpostion in Window mode (or Heaven 4.0).

As I wrote in my post, this happens during other furmark / timespy tests etc. as well.

Check what your windows power plan -> advanced power settings -> pcie link speed management says and try different settings there maybe.

Tried that one as well, no luck.

 

[agent_x007]

Are there PCI-e related options in you BIOS ?

 

[peterj]

Are there PCI-e related options in you BIOS ?

As I wrote in my original post, no. No advanced BIOS available for this laptop model.

 

[agent_x007]

Well, the only option remaining is to either try to force PCI-e operation through hexedit hack (I think there was a method to do this, but I can't find it now ), or try multiple DDUs and driver reinstalls (along with multiple earlier versions) - to check if that changes anything.
If all this fails, you are stuck

At that point I can only suggest checking 3DMark's PCI-e feature test (or similar tool) to confirm slow PCIe operation.

 

[DarkDreams]

Tried that one as well, no luck.

Regarding the power adapter: official page says it comes with a 330W one, so that should be sufficient. I would also highly doubt that the pcie link speed would make a big enough difference in power draw to cause this very specific issue. However I would still always recommend using the correct or at least powerful enough adapter otherwise you might kill the adapter if it is maxed out all the time.
To me this really sounds like a bios issue and the fact that there are not even advanced options would indicate to me that the bios wasn't a high priority for Lenovo. Best option might be to contact customer support and see what they say. It does negatively affect performance after all.

edit: also just found for advanced bios this reddit thread might be worth a look

 

[Naki]

1) Do (m)any games run noticeably slower than they should on similar specced laptops?
(i.e. same RAM amount, same or similar CPU, same GPU)

2) Test with some good, common benchmarks - 3DMark (includes various tests), Unigine Valley, Unigine Superposition, Monster Hunter Online benchmark.
If your benches results are commonly near those of same/similar laptop configs, you might be just fine.

If they are fairly or very lower than that, you might be having issues that need fixing.

 

[peterj]

Regarding the power adapter: official page says it comes with a 330W one, so that should be sufficient. I would also highly doubt that the pcie link speed would make a big enough difference in power draw to cause this very specific issue. However I would still always recommend using the correct or at least powerful enough adapter otherwise you might kill the adapter if it is maxed out all the time.
To me this really sounds like a bios issue and the fact that there are not even advanced options would indicate to me that the bios wasn't a high priority for Lenovo. Best option might be to contact customer support and see what they say. It does negatively affect performance after all.

edit: also just found for advanced bios this reddit thread might be worth a look

But if it's a BIOS issue, my problem should be very common to other users as well on this laptop but it seems it's not the case. Moreover, the GPU runs on gen 4.0 after I unplug and plug back my AC adapter. Interestingly, it doesn't go back to 1.1 on idle after this, but it goes back eventually after 1-2 hours later and never turns back to 4.0. Problem with SREP is that there is no config file for my laptop although it works with the one for pro 5. However, I wouldn't mess with that again, as I bricked my laptop last time.

 

[eidairaman1]

If you have msi afterburner go into its advanced settings and restore everything to default, go into Power properties in Windows and change pcie link state to max performance.

 

[peterj]

If you have msi afterburner go into its advanced settings and restore everything to default, go into Power properties in Windows and change pcie link state to max performance.

I've tried these already, no luck.

 

[eidairaman1]

Are you using the lenovo provided power adapter?

 

[peterj]

Are you using the lenovo provided power adapter?

Yes.

 

[Naki]

Did you contact Lenovo tech support about it?

You did not respond, so have to ask again - are games running too slowly/badly, or do they run fine?

 

[TumbleGeorge]

Why did you decide that a stress test has load must necessarily demand increased and constant data throughput on the PCIe bus, which would also cause an increase in the reported protocol version?

 

[peterj]

Why did you decide that a stress test has load must necessarily demand increased and constant data throughput on the PCIe bus, which would also cause an increase in the reported protocol version?

Not just stress test, but on GPU demanding games when GPU only runs on gen 1.1, the FPS is approx 30% lower than 4.0. This is not right and my GPU should normally run on 4.0 when fully or nearly fully utilised. My laptop motherboard PCIE is capable of providing 4.0. As I wrote earlier, unplugging the power adapter connector from the laptop and plug it back resolves the issue temporarily for 1-2 hours and after that the GPU only runs again in 1.1, regardless of under load or just idle.

 

[eidairaman1]

Not just stress test, but on GPU demanding games when GPU only runs on gen 1.1, the FPS is approx 30% lower than 4.0. This is not right and my GPU should normally run on 4.0 when fully or nearly fully utilised. My laptop motherboard PCIE is capable of providing 4.0. As I wrote earlier, unplugging the power adapter connector from the laptop and plug it back resolves the issue temporarily for 1-2 hours and after that the GPU only runs again in 1.1, regardless of under load or just idle.

Replace the power adapter

 

[oinkypig]

When pice 4.0 is active run 3dmark gpu bandwidth test and check it against when it runs pcie 1.1
You should have reinstalled windows by now or had another windows install from another ssd/partition ect to test it on(last resort)
if you have a laptop with a 4080 it should have mux switch settings in bios. change it to dgpu.

Laptop manufacturers ship out the laptops with the latest stable builds, windows, drivers, ect
when you start updating the operating system and it messes around with drivers and operating system files, the build typically is no longer stable UNLESS the laptop manufacturer updates their bios for a particular reason such as your having.

I can recommend a few things though, turn off CPU core isolation, turn off app and browser control (search bar it) - Smart App control and turn this off. Enabling this will shut off any programs such as a manufacturer specific management software that governs inhouse software designs of GPU and CPU monitoring. Smart app control scans processes, that specifically arent windows based, constantly that windows doesnt recognize and either marks it safe or shuts it down overal. Most of the time motherboard control centers, especially for laptops, get put in the not safe zone. that would explain why pcie- 4.0 reverts to pcie 1.0 after a few hours. Also the ongoing scanning causes a huge uneccessary strain when the potential program is completely safe to use.

My only occassion pcie 4.0 ran at pcie 1.0 is when the CPU socket had a bent pin. It was an open box item. I returned it and got BNIB. Fixed the issue. It went through tons of troubleshooting before returning.

 

[peterj]

When pice 4.0 is active run 3dmark gpu bandwidth test and check it against when it runs pcie 1.1
You should have reinstalled windows by now or had another windows install from another ssd/partition ect to test it on(last resort)
if you have a laptop with a 4080 it should have mux switch settings in bios. change it to dgpu.

Laptop manufacturers ship out the laptops with the latest stable builds, windows, drivers, ect
when you start updating the operating system and it messes around with drivers and operating system files, the build typically is no longer stable UNLESS the laptop manufacturer updates their bios for a particular reason such as your having.

I can recommend a few things though, turn off CPU core isolation, turn off app and browser control (search bar it) - Smart App control and turn this off. Enabling this will shut off any programs such as a manufacturer specific management software that governs inhouse software designs of GPU and CPU monitoring. Smart app control scans processes, that specifically arent windows based, constantly that windows doesnt recognize and either marks it safe or shuts it down overal. Most of the time motherboard control centers, especially for laptops, get put in the not safe zone. that would explain why pcie- 4.0 reverts to pcie 1.0 after a few hours. Also the ongoing scanning causes a huge uneccessary strain when the potential program is completely safe to use.

My only occassion pcie 4.0 ran at pcie 1.0 is when the CPU socket had a bent pin. It was an open box item. I returned it and got BNIB. Fixed the issue. It went through tons of troubleshooting before returning.

I don't have the PCI express feature test for testing the GPU bandwidth, however, I have run a Timespy test and the GPU ran only at x16 1.1 with low GPU score (17335). I tested this on fresh installed windows, same results. Mux switch settings are available in my laptop but it has been set to discrete GPU only so my laptop doesn't use the integrated GPU.

I will now try with older bios to see what happens, as suggested here before it may be the BIOS is messing with my settings. However, I have manually installed the latest chipset and nvidia drivers, but this also happens with the older ones which are on Lenovo's driver list. I will give an update if this happens with my earlier version of BIOS.

Core isolation is disabled for me by default.

With regards to CPU socket bent pins, this issue also happened with my previous motherboard before Lenovo replaced that one during warranty repair. It also happens with the new replacement motherboard as well, exact same issue. The odds here I guess are very small that both motherboards had problems with CPU socket bent pins.

Can an unstable voltage from the wall socket cause this kind of anomaly? The reason I am asking that I have noticed that there are small approximately 0.5s outages at home where I notice this when light turns off for a quick 0.5s randomly.

 

[eidairaman1]

I don't have the PCI express feature test for testing the GPU bandwidth, however, I have run a Timespy test and the GPU ran only at x16 1.1 with low GPU score (17335). I tested this on fresh installed windows, same results. Mux switch settings are available in my laptop but it has been set to discrete GPU only so my laptop doesn't use the integrated GPU.

I will now try with older bios to see what happens, as suggested here before it may be the BIOS is messing with my settings. However, I have manually installed the latest chipset and nvidia drivers, but this also happens with the older ones which are on Lenovo's driver list. I will give an update if this happens with my earlier version of BIOS.

Core isolation is disabled for me by default.

With regards to CPU socket bent pins, this issue also happened with my previous motherboard before Lenovo replaced that one during warranty repair. It also happens with the new replacement motherboard as well, exact same issue. The odds here I guess are very small that both motherboards had problems with CPU socket bent pins.

Can an unstable voltage from the wall socket cause this kind of anomaly? The reason I am asking that I have noticed that there are small approximately 0.5s outages at home where I notice this when light turns off for a quick 0.5s randomly.

Yes

 

[Naki]

Try using the laptop with a (good!) UPS if your electric power is not quite stable.