GPU stuck at PCIe x8 instead of x16

[rylo123]

I just wanted to share what I ran into recently in case it helps anyone else pulling their hair out over PCIe link speeds.

• Motherboard: ASUS PRIME B760-PLUS D4
• CPU: Intel i7-14700F
• GPU: (installed in the top PCIEX16 slot, labelled G5 in BIOS) 4090
• SSD: M.2 NVMe in the top M.2_1 slot (above the GPU)

I noticed in GPU-Z and HWiNFO that my GPU was only running at PCIe 4.0 x8, not x16. This was the case even when the GPU was under load.

So I:

• Updated BIOS and drivers.
• Checked the BIOS — PCIEX16 (G5) was on Auto so I changed it to Gen 4.
• Still stuck at x8.
• M.2_1 was showing x4.
• Reseated the GPU multiple times
• Moved the SSD to a different M.2 slot
• Tried running with no SSD at all
• Checked the slot and GPU for dust or damage

How can I fix this.

This is the link to the manual of the motherboard: https://www.asus.com/uk/motherboard...helpdesk_manual?model2Name=PRIME-B760-PLUS-D4

My Bios:

https://imgur.com/a/thTCMvp

 

[maxmax123]

Yeah I have the same issue with. Im 99% sure my problem is bad CPU contact pins. I would try that if it bothers you.

 

[iameatingjam]

I had this problem on my last pc. It drove me up the wall because I could not for the life of me figure out the problem. Luckily for you (and me), you have pcie 4, not pcie 3 and a 4090 having enough vram that the bus wont have to do much memory swapping, so you don't lose much performance.

And yeah I was never able to figure out that problem. I knew it was with the mobo ( or cpu I guess) because it happened with any gpu I plugged in.

So yeah you could just say f it and not worry about it. Or you could, try reseating the cpu or turning off link state power management ( it helped me solve a similar problem with a m.2 drive so maybe it will help you, kinda shot in the dark but its easy to do from the edit power plan part of windows, under pcie). Its happened to that drive 3 times, one time just turning it off was enough, one other time I just had to reseat it, one time I had to do both, fyi. So maybe turn it off. Then turn your computer off and reseat card one more time, but before that you might also want to try:

Cleaning all the pins on the gpu with iso and blowing compressed air into the pcie socket incase there is something obstructing the connection.

 

[LabRat 891]

I had this problem on my last pc. It drove me up the wall because I could not for the life of me figure out the problem. Luckily for you (and me), you have pcie 4, not pcie 3 and a 4090 having enough vram that the bus wont have to do much memory swapping, so you don't lose much performance.

And yeah I was never able to figure out that problem. I knew it was with the mobo ( or cpu I guess) because it happened with any gpu I plugged in.

So yeah you could just say f it and not worry about it. Or you could, try reseating the cpu or turning off link state power management ( it helped me solve a similar problem for me with a m.2 drive so maybe it will help you, kinda shot in the dark but its easy to do from the edit power plan part of windows, under pcie). Its happened to that drive 3 times, one time just turning it off was enough, one other time I just had to reseat it, one time I had to do both, fyi. So maybe turn it off. Then turn your computer off and reseat card one more time, but before that you might also want to try:

Cleaning all the pins on the gpu with iso and blowing compressed air into the pcie socket incase there is something obstructing the connection.

My personal favorite:
Memory channels and I/O dropping out due to-

A. Thermal Paste on a pin/pad.
I have 2 semi-functioning AM4 boards, that 'kindasorta died' due to TIM getting into the socket.
Even aerosolized drycleanchem (see pic below) didn't fix it, but the issues 100% started after TIM was introduced to the PGA socket.


B. Overtightening/mis-loading of the cooler.
I've had issues back on 1155 w/ overtightening, myself. More recently, my roommate's AM5 rig was not stable w/ his 7900XTX until he got his CPU Secure Frame kit (see pic below) installed.

I just wanted to share what I ran into recently in case it helps anyone else pulling their hair out over PCIe link speeds.

• Motherboard: ASUS PRIME B760-PLUS D4
• CPU: Intel i7-14700F
• GPU: (installed in the top PCIEX16 slot, labelled G5 in BIOS) 4090
• SSD: M.2 NVMe in the top M.2_1 slot (above the GPU)

I noticed in GPU-Z and HWiNFO that my GPU was only running at PCIe 4.0 x8, not x16. This was the case even when the GPU was under load.

So I:

• Updated BIOS and drivers.
• Checked the BIOS — PCIEX16 (G5) was on Auto so I changed it to Gen 4.
• Still stuck at x8.
• M.2_1 was showing x4.
• Reseated the GPU multiple times
• Moved the SSD to a different M.2 slot
• Tried running with no SSD at all
• Checked the slot and GPU for dust or damage

How can I fix this.

This is the link to the manual of the motherboard: https://www.asus.com/uk/motherboard...helpdesk_manual?model2Name=PRIME-B760-PLUS-D4

My Bios:

https://imgur.com/a/thTCMvp

Do you have any other GPU at all you could use to test? If it's consistently handshaking only 8 lanes, then I'd suspect the mobo/CPU.

IMO...
Get a contact frame for your CPU, carefully reseat the CPU.
Inspect the pins and pads closely with a flashlight for bent pins, burnt pads, thermal paste, hair, skin flakes, etc.
[Ex. I've had a literal speck of dust get into an LGA775 socket and NoPOST the build until blown-clear]
Lastly, make sure the cooler gets tightened down evenly, in a crossing-pattern.

If that still doesn't fix it, all the parts you have will still carry-over to a (prospective) replacement board.

 

[AleXXX666]

My personal favorite:
Memory channels and I/O dropping out due to-

A. Thermal Paste on a pin/pad.
I have 2 semi-functioning AM4 boards, that 'kindasorta died' due to TIM getting into the socket.
Even aerosolized drycleanchem (see pic below) didn't fix it, but the issues 100% started after TIM was introduced to the PGA socket.
View attachment 410041

B. Overtightening/mis-loading of the cooler.
I've had issues back on 1155 w/ overtightening, myself. More recently, my roommate's AM5 rig was not stable w/ his 7900XTX until he got his CPU Secure Frame kit (see pic below) installed.
View attachment 410040



Do you have any other GPU at all you could use to test? If it's consistently handshaking only 8 lanes, then I'd suspect the mobo/CPU.

IMO...
Get a contact frame for your CPU, carefully reseat the CPU.
Inspect the pins and pads closely with a flashlight for bent pins, burnt pads, thermal paste, hair, skin flakes, etc.
[Ex. I've had a literal speck of dust get into an LGA775 socket and NoPOST the build until blown-clear]
Lastly, make sure the cooler gets tightened down evenly, in a crossing-pattern.
View attachment 410043
If that still doesn't fix it, all the parts you have will still carry-over to a (prospective) replacement board.

never had ANY issues of such.
- lots of "paste leak" to socket - zero issues
- never used stupid frame with LGA1700 or AM5 - zero issues

 

[LabRat 891]

never had ANY issues of such.
- lots of "paste leak" to socket - zero issues
- never used stupid frame with LGA1700 or AM5 - zero issues

Can't blame ya.

Prior to AM4, I'd NEVER seen TIM 'kill' a socket. I've seen 2 on AM4.
The 1st instance, I found TIM upon inspection on a 2nd hand 'freebie' part from a friend. No POST until I sprayed the socket out, even then, dead RAM channel (R5 3600 and R5 5600 both tried).
The 2nd, I clearly saw a little ball of paste impaled on the pin of the R5 1600AE upon removal for reseating. That board wouldn't even try to POST until the CPU and socket were thoroughly cleaned w/ an ultrafine toothbrush and high-strength isopropanol.
I'd bet it has something to do with the both insulative *and* capacitive nature of many pastes, and the fine pitch to PGA1331

Esp. for AM5 (and its thicc IHS), I thought the frame was purely a money grab, or maybe an aid for specific XOCing, etc.
In my experience-since, it helps prevent the consequences of Ugug, He-maning the cooler down crooked