Knowledge VS edk2-rk3588

tkaiser has some additional information from his initial testing at https://github.com/ThomasKaiser/Knowledge/blob/master/articl... if people are curious to know more, though it should be noted these were early release samples and as such, software/firmware isn't exactly perfect at the moment

"Armbian recommends A1 rated SD-Cards only now (A2 rated cards need yet lacking driver support and therefore show lower overall and especially random IO performance)."

From: https://github.com/ThomasKaiser/Knowledge/blob/master/articles/Quick_Preview_of_ROCK_5B.md

Yes and no, in this articleguy tests A1 and A2 cards and on non compatible reader some A2 cards are slower than A1 cards.

One thing worth noting is that basically, even now (almost 10 months post announcement), almost no one has a 6000U laptop outs (a search on Amazon and Best Buy shows two 6800U laptop models total, one Asus and one Lenovo). Two niche vendors, XMG and Star Labs, have both publicly stated that they would have loved to have offered Ryzen 6000 laptops, but couldn't get any allotments. There are were also well documented chipset issues - even into the summer Lenovo and Asus talked about requiring firmware updates to enable their USB4 ports.

That being said, starting w/ Rembrandt, AMD now has full 40Gbps USB4 controllers built on-chip. I'm really looking forward to Ryzen 7040 because Phoenix looks great (Zen4 + RDNA3 on TSMC N4 - yes please) and hopefully USB4 support has matured enough on the AMD side that Framework is able to release something.

My understanding is that the I/O limitation is in the re-timers - currently the Framework uses 4X JHL8040R's (labeled as Burnside Bridge) directly connected to the iTBT: https://github.com/FrameworkComputer/Mainboard/blob/main/Ele... Apple used these for their first M1 MBAs as well: https://github.com/ThomasKaiser/Knowledge/blob/master/articl...

But both Apple and AMD are now using Kandou retimers:

* https://www.gizchina.com/2022/07/25/apple-completely-got-rid...

* https://kandou.com/matterhorn.html

* https://kandou.com/assets/downloads/product-briefs/KB8001-Pr...

Some early A2 cards were implemented poorly and actually had worse IOPS than A1 cards, as they used a worse onboard controller in the hopes of the system making up for the deficiency. See here. Reputable large A2 cards no longer have this issue but it's still a minefield.

In addition to this, the A2 standard requires hardware/driver support of Command Queuing and Write Caching, and without proper support A2 cards can and will perform worse than their A1 counterparts

FWIW, there is a port of EDK2 to the RK3588, which is what powers this board. Most of the peripherals work. So that's nice. But the thing is, it isn't UEFI so much as the device-tree/ACPI distinction you need to be mindful of. You can use either DTBs or ACPI to boot Linux on this device, though. https://github.com/edk2-porting/edk2-rk3588

I saw on the edk2 github (https://github.com/edk2-porting/edk2-rk3588) that ESXi Arm Fling version 1.12 was tested and working with the RK3588S. When I tried booting from the ISO, I got a purple screen that gave an error "Cannot enable LPI or ITS-based MSI." Does it work for the Orange Pi 5, and if so, what do I have to do to get it to install?

edk2-rk35xx/configs at master · edk2-porting/edk2-rk35xx · GitHub