New build - Asus ROG Crosshair X670e Extreme pcie lanes

[Basilem84]

Hi all,
I’m looking to build a new AMD rig with the X670e chipset. What is really unclear to me is how the pcie lanes are handled between the GPU and M2 SSDs. I would like to buy the Crosshair x670e Extreme version but when I look at the manual it looks like if I add two Gen5 SSDs, the GPU will run on at x8 instead of x16. Am I right or do I miss something ?*
Thanks in advance and cheers.
Basile *

[Murph_9000]

Some configs will require PCIEX16_1 to run as x8, but it looks like you could use M.2_1 and GEN-Z.2_1 for a pair of M.2 NVMe 5.0 drives and still have x16 for the GPU. Using either M.2_2 or anything in PCIEX16_2 (e.g. a PCIe card with M.2 slots) will limit PCIEX16_1 to x8.

It's a shame that AMD seem to have rather limited the lanes on AM5. I'd rather hoped the new socket might have delivered a few more lanes and 256MB RAM, given that HEDT is kinda dead (for now). I'm assuming it must be an AMD thing, as I'd be very surprised if ASUS didn't make the best available use of lanes on their flagship ROG board.

[William_Clement]

Murph_9000 wrote:
Some configs will require PCIEX16_1 to run as x8, but it looks like you could use M.2_1 and GEN-Z.2_1 for a pair of M.2 NVMe 5.0 drives and still have x16 for the GPU. Using either M.2_2 or anything in PCIEX16_2 (e.g. a PCIe card with M.2 slots) will limit PCIEX16_1 to x8.

Anyone knows if it's possible to use like this? 1 5.0 x16 gpu + 1 5.0 4x nvme if it's then possible to add like 2 or 3 more nvme x4.0 without reducing the gpu lanes?

[Braegnok]

William Clement wrote:
Anyone knows if it's possible to use like this? 1 5.0 x16 gpu + 1 5.0 4x nvme if it's then possible to add like 2 or 3 more nvme x4.0 without reducing the gpu lanes?

AM5 Ryzen 7000, X670E motherboard supports 24 PCIe 5.0 lanes. With 16 lanes dedicated to GPUs @ x16, or x8/x8,..

The remaining 8-lanes are dedicated to NVMe SSD, or USB4, Thunderbolt4 controllers @ x4/x4.

[Old_Rager]

Hey Murph,

I too have the same question as I want to be able to run my two x PCIe5 NVMe's in a raid 0 config whilst not experiencing any lane throttling on both the NVMe drives as well as the PCIe x16 lanes that my GPU will be using.  So, I just wanted to clarify exactly what you are saying here, as I'm a bit cautious due to the way you said, "...it looks like...".  I guess, I need to get this right if you don't mind - and sorry for all of the detail as I have ASD. 

So if I understand you correctly, if I put my 1 x PCIe 5 NVMe drive into M.2_1 that is directly on the MOBO, I will then need to put my 2nd x PCIe 5 NVMe drive into the supplied expansion card's GEN-Z.2_1 drive bay which will occupy the bottom PCIe x16 slot on my MOBO if I want to use all 24 PCIe 5 lanes without any throttling to occur?  Is there no way to place both NVMe's directly onto the MOBO without impacting the 8 PCIe lanes that both of these drives will consume as well as not impeding on the 16 PCIe lanes that my GPU will be using in the top PCIe x16 slot on the MOBO?

Any help you can give me here is truly appreciated!!! 😉

[Basilem84]

Thanks a lot Murph for your clear answer, looks like you’re right and I totally agree it’s a shame that AMD limited the lanes on the new CPUs. Maybe I’ll wait a bit to see if anybody tried this and report if it’s working well.*
*

[xeromist]

FWIW, you're more likely to notice the improvement from the Gen5 SSDs than the reduction of lanes to the GPU. Remember that up until recently GPUs were getting gen4 x16 which is the same as gen5 x8. GPUs just can't utilize gen5 like NVMes can.

[Murph_9000]

xeromist wrote:
FWIW, you're more likely to notice the improvement from the Gen5 SSDs than the reduction of lanes to the GPU. Remember that up until recently GPUs were getting gen4 x16 which is the same as gen5 x8. GPUs just can't utilize gen5 like NVMes can.

You're correct about the 4.0 vs 5.0 performance, but it's not true for RTX 4000 series, as they are still PCIe 4.0, so would be running as 4.0 x8 (equivalent to 3.0 x16), same as RTX 3000 series. I'm not sure about the upcoming RX 7000 series from AMD. With RTX 3000 and RX 6000, I believe that 4.0 x8 was mostly ok for gaming, not sacrificing too much off the top (with the possible exception of 3090 and 6900, or quite specific workloads). I don't think there is much public analysis yet for the 407080, 4080, and 4090; for how much x8 will hurt them. Gaming performance will likely still be pretty good on x8, just unknown how much will be lost from the top end.

[xeromist]

Murph_9000 wrote:
You're correct about the 4.0 vs 5.0 performance, but it's not true for RTX 4000 series, as they are still PCIe 4.0, so would be running as 4.0 x8 (equivalent to 3.0 x16), same as RTX 3000 series. I'm not sure about the upcoming RX 7000 series from AMD. With RTX 3000 and RX 6000, I believe that 4.0 x8 was mostly ok for gaming, not sacrificing too much off the top (with the possible exception of 3090 and 6900, or quite specific workloads). I don't think there is much public analysis yet for the 407080, 4080, and 4090; for how much x8 will hurt them. Gaming performance will likely still be pretty good on x8, just unknown how much will be lost from the top end.

Ah, yes. Forgot about that. Good point. But, as you mention, testing has found a negligible difference between 4 at x8 and at x16. Several tech media outlets have done testing so if OP wants data it's out there.

If I had to guess it's not going to change much. Link speed only matters when you have to transfer data. The improvements we're seeing in GPUs don't necessarily require higher transfer speeds because a lot of the assets are already loaded into local memory. In fact, it almost seems like we're going the opposite direction. More A.I. upscaling means assets won't have to balloon to hit future resolution targets. The first earnest attempts at playable 8k are just going to be upscaled 4k.

[RobrPatty]

How do you get M-2 NVMe 4.0 card to show as Disk 0 on this board? I have 3 seperate sata drives & Asus NVidia 3080 GPU. GPU is installed in PCIEX16_1.