What is the effect of a faster ssd on model loading?

Would there be any noticeable improvement with a 6000MB/s ssd vs a 3000MB/s one? Or would the processor or something else cause a bottleneck?

I can’t speak to the “model” part but I think it depends on what you mean…
Are you talking about an SATA II 3Gb/s device vs a SATA III 6Gb/s device?
If so, since there isn’t an SATA controller on the NX DevKit the type of bridge comes into play… Are you using a USB <> SATA bridge or a Mini-PCIe <> SATA bridge?
There’s also the matter of the SATA device itself. A cheap SATA III SSD may advertize 6Gb/s but the underlying storage silicon may not be able to be read that fast.

BTW, SATA speeds are specified in bits per second (Gb/s) not bytes per second (GB/s)

You’re right I forgot about the megabit versus megabyte. It’s a pcie 4x m.2 nvme drive. It’s supposed to have sequential read speeds of 6200 megabits per second. But I ended up going for a slightly slower drive that’s 3500Mb/s but has a much higher write endurance, 2TB drive rated at 1500 terabytes written. That way I can make sure to have some thing stable for a very long time and with a high number of rewrites

Ah gotcha. Again, I can’t speak from the model perspective, but I just ran a test on a Samsung_SSD_950_PRO_256GB NVMe stick which is rated at 2,200MB/s Seq. Read.

# time dd if=/dev/nvme0n1 of=/dev/null bs=1M count=5000 iflag=sync
5000+0 records in
5000+0 records out
5242880000 bytes (5.2 GB, 4.9 GiB) copied, 4.88312 s, 1.1 GB/s

real	0m4.890s
user	0m0.021s
sys	0m4.841s

That translates to about 8800Mb/s but it did peg a core for the 4.8 seconds it took.