I don’t understand what you are asking - can you rephrase it?
sorry for confused I am looking for the same card but 4th generation for my pc:
AMD Ryzen 9 5900X
Asus TUF GAMING B550-PLUS
AORUS NVMe SSD- 2TB
The only 4th gen cards that I know of are the:
ASUS Hyper M.2 v4
MSI XPander-Aero v4
and you can’t either of them except for hugely marked up ones on eBay right now.
And in one moment, I suddenly understand the whole point of staggering starts… so the damn SSD → SATA copy doesn’t ALL happen at the same time across all the plots. I’ve been staring at this for 20mins
(sdm and sdn are SATA drives)
Yeah staggered starts is mandatory, otherwise I/O is the bottleneck. I want the CPU to be my bottleneck if possible. Means I could not have scaled out SSD or memory any more.
Agreed - I’m seeing if I can get my hands on a few Intel 4510s to see if they make a big diff.
QUICK UPDATE: Intel hasn’t released an RST (proprietary RAID tech they integrate on some mobos) driver for the Linux Kernel, so on Ubuntu 20.04.2 you can’t see any arrays you create in the BIOS.
Well I’m glad you started this thread! I was going to look into the same thing haha. I started on a single nvme, but not the right one, and it was SLOW. So setup a raid (lost everything in the process) and am trying that. It helped some with sustained speed, but I think switching to dual 980 pros will offer even better speeds than my firecudas now and then have them run parallel.
I have done some research and it looks like nvme’s in raid 0 suffer in 4K read and write speeds pretty heavily. I’m not sure what size files chia writes, but that can have a negative effect as well if they are larger (which I’d imagine they are).
If your card doesn’t need bifurcation that means it uses a PCI-e switch, that adds latency and is (or used to be now??) more expensive.
Also your card is PCI-e 3.0 and the 980 PROs are PCI-e 4.0. 3.0 Should be enough for the steady state performance of the 980 PROs, you just can’t burst as high.
That’s about a hardware raid controller that doesn’t pass through trim commands. Linux software raid does passthrough trim commands and has done for years.
Staggering isn’t just needed for the SSD-> SATA copy. Phase 1 of the plotting is multithreaded, the other phases as not. So if you start 7 plots at the same time with “-r 4”, they’ll use 28 cores at most during phase 1 but only 7 during the other phases. So you’re not using your CPU optimally.
Never use your mobo’s raid, even if it does work. It’s just software raid with the disadvantage of hardware raid that you’re limited to certain hardware (sometimes even revision).