One SanDisk PRO-READER SD Express Dual Card SDDR-A451-GNPNN
to test the maximum sustained write speed of these cards. They’re specified at “Sequential Write Performance 650MB/s”. However, when I try to copy a single 16gb file with random data from an internal PCIe4 SSD to the sdcard (exfat, 128kb cluster size, empty) in Windows 11 Explorer, the data rate starts at around ~300MB/s, then quickly drops to a very constant 218MB/s:
I’ve tried this now with both cards on 4 windows11 machines, a linux machine, a linux machine with a direct PCI->RTS5261 chip (the same as used in the PRO-READER, just without the USB/PCI bridge).
I see that the card really gets quite warm while writing, but adding ridiculous amounts of airflow along the surfaces doesn’t change anything.
Reading doesn’t hit the advertised 880MB/s either, but it gets a little closer, and that tells me that the USB connection between CPU and cardreader isn’t the bottleneck:
I’ve contacted SanDisk support, and, to be frank, it’s either AI answers or they don’t have the attention span to tackle this ticket (Incident: 250605-000877). Thankfully someone called me back and said they had seen the video, then proceeded to ask questions answered in the video…
Trying to be constructive… has anybody seen SUSTAINED write speeds higher than 218MB/s while copying large files to the disk? Note, I’m not asking about synthetic disk benchmarks like CrystalDisk etc., I’m asking about the simple real-world case of storing files.
SD Express can reach 985 MB/s but the slot needs to be PCIe based to keep up. USB speeds are inadequate to max out the speeds. TB4 is barely able to keep up.
Faster speeds will reach 4GB/s with two PCIe 4.0 lanes. Support for this speed will likely require integrated slots on laptops.
@SBrown, I’m getting a net transfer rate of 660Mbytes/s while reading, so 5.28Gbit/s. That tells me I’m using a port with a signalling rate higher than 5GBit, of which USB 3.2 Gen2x1 with its 10GBit/s seems to be a likely candidate, as it’s the next faster one, matches my laptop port’s spec and the reader’s spec as well… and I’m using the cable sandisk shipped.
I haven’t yet seen a USB connection be faster in one direction than the other, so I assume that this connection can transfer at least 660MByte/s in both directions.
@Vegan , looking at the USB signalling rates at wikipedia’s USB page, I think you’d be wrong even if I complained about 985MB/s (which is the limit of PCIe Gen3 x1). But I’m complaining about 218MBytes/s, which is well below USB 3.2 Gen1x1 limits.
So I don’t think it’s the reader or connection - it’s the card. And in the meantime I have found Sandisk saying “210MB/s minimum sustained write speed” in the footnotes of the card’s product page. I’ve never seen it there before, so not sure if they just added it or I just didn’t catch that earlier.
Guess I’m just a bit disappointed that it degrades from only half the marketing-spec of “Up to 650MB/s write speeds” to a third of the marketing-spec so quickly while writing.
But seeing how the guaranteed-sustained speed is 100MB/s for the 128GB card and 218MB/s for 256GB card, I have hope that the 521GB card’s speed will scale the same way.
I’d still be interested to hear from anybody who can write to the card faster than me.
My Dell 3480 has USB4 and best I can get from a USB4 card reader varies depending on the card. Best so far is 230MB/s suggesting my reader is not obsolete yet. SD Express speeds may perk up over the next few years as TB5 and new slots on machines get modernized.
