Hi all,

 

I am looking to speed up my main Windows 10 VM. Currently, I have it in a vdisk with sitting on the cache pool (2 x Samsung PM883 in a BTRFS mirror pool). I have a dual socket setup and unfortunately the HBA and my GPU sits in PCIe slots assigned to different CPUs. The assigned CPU threads all belong to the GPU slot. So I am trying to speed up my VM as much as possible. There is probably some overhead associated with

a) HBA being on a PCIe attached to the other CPU

b) Dual file systems i.e. NTFS vdisk and BTRFS pool

c) HBA sharing the bandwidth between the array drives and cache drives

d) SSDs being SATA

 

I know above individually shouldn't have a huge impact but taken together, is probably significant. I can resolve all the above by adding a dedicated NVMe drive using a PCIe to M.2 adapter (my motherboard does not have a straight M.2 slot).

 

All my PCIe slots on my motherboard are PCIe 3. I can see the the PCIe 4 NVMe drives typically perform better. I wanted to check, if using the faster PCIe 4 NVMe will still benefit me even if I am attaching them to a PCIe 3 slot? Or should I just go with a higher end PCIe 3 NVMe drive?

Edited November 29, 20205 yr by newunraiduser5
Too trigger happy - posted an incomplete question. Edit to add the full question

...never tried, but according to this: https://en.wikipedia.org/wiki/PCI_Express#PCI_Express_4.0 PCIe-4.0 claims backward compatibility.

Hence it should be technically possible to go with a PCIe-3.1 host card in x8 to achieve/transport the same bandwidth for a M2.PCIe-4.0 x4 module (see table in here: https://en.wikipedia.org/wiki/PCI_Express#History_and_revisions).

But I don't see any Adapter cards available with host side PCIe 3.x and module M2 side in PCIe-4.0

The pcie4 nvme would just run at pcie 3.0 speeds, which means you aren't going to benefit from the gen4 device. It may run a smidge faster than a standard gen3 but I doubt it. I wouldn't pay for gen4 speeds when you're not going to benefit from it