RTX 4090 Running PCIE 4.0 x8 not x16

[Wizerbwski]

Hi all,

Another PCIE not running at x16 question.

Ok, I have a newish system and have recently installed a fresh Windows 11 Pro due to driver issues with other software.

My rig was running at x16 but now I cant get it passed x8 in GPUZ (latest version) running various games, Cyberpunk, Witcher 3 etc.

System specs:
MSI Suprim X RTX 4090 on a riser cable
MSI X670E Carbon WiFi (bios version 1.H0
Thermaltake Toughpower GF3 1200W PCIe Gen 5.0 ATX3.0 PSU
Corsair Dominator 64GB 6000Mhz DDR5
M.2 Slot 1 (CPU) = Crucial T700 Gen 5.0 2TB SSD
M.2 Slot 2 (CPU) = Seagate Firecuda 2TB Gen 4.0 SSD
M.2 Slot 3 (chipset) = Kingston KC3000 Gen 4.0 1TB SSD
AMD Ryzen 9 7800X3D

Before the Windows refresh GPUZ reported x16 speed when gaming period... not its wont go past x8

I have performed the below:
Removed GPU, cleaned and checked pins, reseated 3 times without riser cable.
Removed both gen 4.0 SSD's completely
Moved Gen 5.0 Windows drive to M.2 slot 3 (chipset)

I still cant get it to go to X16.

GPUZ screenshot attached.

Any help appreciated

Cheers
Wiz

 

[Wizerbwski]

Ok, I just updated the BIOS to V1.J3 from 1.H0 and its now running at x16 - LOLZ

Cheers thou!

 

[3x0]

At this point it's very likely a faulty card. Can you try the second PCIE x16 slot if you have not already? Do you have a spare GPU to try on that system?

 

[ir_cow]

Did you check to see if a M.2 slot shares the PICE bandwidth with Slot 1? That or you have something installed in PCIE slot 2. Just a hunch.

At this point it's very likely a faulty card. Can you try the second PCIE x16 slot if you have not already? Do you have a spare GPU to try on that system?

The 2nd slot is only 8x and shares the bandwidth with the 1st slot.

 

[JimminyCrumpet]

If you look at the tooltip, this is power management related. Turn off ASPM / PCIe power states in windows power settings and it'll stop happening.

 

[3x0]

If you look at the tooltip, this is power management related. Turn off ASPM / PCIe power states in windows power settings and it'll stop happening.

That's only for Gen speeds (1.1, 2.0 etc), not PCIe width

 

[Wizerbwski]

Its ok, it fixed itself after a BIOS update... which from my experience is rare!

 

[JimminyCrumpet]

That may have been the case at some point, but cards are allowed to initiate power-down of lanes when they enter L0s (it's always a card-initiated thing), and send an "upconfigure" message when they come back out of low power. Not sending data over 15/16 lanes saves a decent amount of power, more the faster PCIe gets. It has also introduced ugly latency and caused crashes in the past. I've never seen AMD cards do it. PCI-SIG's faq page has some answers about it (they can only activate lanes that were activated when the initial configuration was happening, etc).

 

[eidairaman1]

That may have been the case at some point, but cards are allowed to initiate power-down of lanes when they enter L0s (it's always a card-initiated thing), and send an "upconfigure" message when they come back out of low power. Not sending data over 15/16 lanes saves a decent amount of power, more the faster PCIe gets. It has also introduced ugly latency and caused crashes in the past. I've never seen AMD cards do it. PCI-SIG's faq page has some answers about it (they can only activate lanes that were activated when the initial configuration was happening, etc).

If a lga cpu heatsink is overtightened, it causes board warp which can create an air gap between the lga pins and the cpu pads exhibiting the same behavior. Also some boards drop the bus speed when certain pcie slots are used or nvme drives are installed...

In Afterburner there is an advanced setting that can cause this problem too where the bus doesnt ramp up to full speed.

Simple fix for the OP, case closed.