Ohhh this is great news, I was trying to figure out if this was the case, thank you for clearing that up.
I've been using a 3.0 hub from Anker until now. So I can pick up a Thunderbolt 4 hub/dock and just use a C <-> A into the Highpoint 1244A, and I should have better results?
Yes. Thunderbolt 4 hubs/docks will usually have some USB 10 Gbps ports (check specs to see how many). They are generally more expensive than USB-C hub/docks though.
For another test, I connected the CalDigit Element Hub to the USB-C port of a Radeon Pro W7500 in an eGPU connected to my Mac mini 2018. I got 914 MB/s ((that's going through one hub, the Element Hub, and Thunderbolt - since the W5700 is connected via Thunderbolt). The W7500 does not have Thunderbolt, just like the USB-C port of the HP Thunderbolt Dock G2. The W5700 uses an AMD XHCI controller, not an Intel XHCI controller that is part of the Thunderbolt controller, just like the ASMedia 3142 in the Rocket1244A is not Intel.
I'm trying to understand how connecting a device to a Thunderbolt hub that is then connected to the Highpoint USB card, is faster than a device connected directly to the card with just C<->A cable.
What did I say that gave you that incorrect conclusion? My tests show that the longer the hub chain (and Thunderbolt chain), the slower the bandwidth. But it's not a lot slower, so a hub is good for adding more ports.
Your tests are using different NVMe and benchmarks so you can't compare them to mine.
Interesting, I thought it would cap a bit since the Highpoint 1244A is a x8 card and it's in a x4 slot.
There's no cap if you're only reading or writing to one USB at a time, since USB 10 Gbps (≈1000 MB/s) is less than PCIe 2.0 x4 (≈1600 MB/s).
Just making sure you saw that bit. Am I missing something or are you getting those slightly higher speeds because of using a titan ridge card?
I suppose different XHCI controllers may have differing performance. For example the USB4 ports of the M1 Macs do not perform as well for USB devices as Intel's Thunderbolt ports. The ASM1142 can only do 8 Gbps. The M1 iMacs with separate USB ports have an ASM3142 that performs better than both of those. The AMD XHCI might not be as good as the ASM3142 but it's better than the ASM1142 and maybe the M1.
Between the ASM3142, ASM3242, Alpine Ridge, Titan Ridge, Maple Ridge, we can't know for sure which is best without using the same device and benchmark for all tests. The difference is probably not noticeable in every day use.
Bonus question, what else do you have in that Netstor? I'm curious about how slot 2 bandwidth gets divvied up and what we can get away with on our systems using something like that.
GC-ALPINE RIDGE, GC-TITAN RIDGE, ThunderboltEX 4, Titan X (Maxwell) currently not being used. I use the Netstor for easily accessible slots. Also, I can get full speed from PCIe 3.0 devices, up to x8, since the PCIe 2.0 x16 host slot has similar bandwidth. Bandwidth is not divided. It is shared. A device that isn't being used doesn't take any bandwidth. PCIe packets have limited length, so no one device can take all bandwidth. It's similar to Ethernet and USB.
Other benefits : the slots are wide, and there's PCIe power (300W) for each slot, so testing or changing GPUs is super easy.
These are pretty great speeds! I thought I had a good card? It's the Amfeltec Squid x4. I know this last bit is getting a bit off-topic for this thread but if you don't mind - what should I look into to make sure I'm tapping into the full potential of this card / nvmes? My Sabrent 4TB is def not pulling those speeds!
I think you should be happy with your speeds since you probably wouldn't notice the difference if you had faster NVMe drives. If your using Black Magic Speed Test, then try ATTO Disk Benchmark or AmorphousDiskMark which will show higher numbers.
My card has a gen 4 PCIe switch so it can allow gen 4 NVMe to perform at full speed. But even with your card that has gen 3 PCIe switch, a gen 4 NVMe will probably perform better than your current gen 3 NVMe.
