7 ms·

Peak power on these recent-ish high performance recent drive reviews shows 9w, but most drives peak consumption is 5W. https://www.tomshardware.com/reviews/sams

Peak power on these recent-ish high performance recent drive reviews shows 9w, but most drives peak consumption is 5W. https://www.tomshardware.com/reviews/samsung-990-pro-ssd-review/2 https://www.tomshardware.com/reviews/samsung-990-pro-ssd-rev... Absolutely some drives won't qualify, especially PCIe 5. But many will be fine. Also, PCIe 3 x4 is 4GBps or 32Gbps which isn't that far from saturating usb4 anyways, so it's not like we even can get drives running full tilt over USB at present. Many of these controllers probably top out at PCIe 3 or even less, if they don't support 40Gbps (most cheap-ish ones won't). Hopefully we start seeing some usb-pd capable systems & drives. The want to plug in 2 or 4 drive 3.5" enterprise drive arrays to my laptop & have it just work. Even just a single 3.5" of power would be so helpful. And it'd mean you could alternatively be charging your phone well too, which is probably the more common everyday ask.

megous 4y ago
5W is still a lot if you add controller power requirements, and 5V->3.3V DC-DC conversion losses. Also I have one of those Power-Z Type-C <-> Type-C power meters, and even with less power hungry western digital nvme, it didn't fit into 4.5W. And it consumed the most power not even during initial write test, but during a minute after kernel reported sync() success so after everything was supposedly written to the device and no power hungry activity was needed. Then it jumped to 7-8W. I guess these dram-less nvme's do some reshuffling from faster flash to slower flash locations during idle time. Which is probably terrible for portable use, where you may want to sync() and then unplug the device once the OS tells you everything is synced up. Creating regular power loss situations during these reshufflings just feels like putting too much trust into nvme firmware, IMO. :D

megous 4y ago

5W is still a lot if you add controller power requirements, and 5V->3.3V DC-DC conversion losses. Also I have one of those Power-Z Type-C <-> Type-C power meters, and even with less power hungry western digital nvme, it didn't fit into 4.5W. And it consumed the most power not even during initial write test, but during a minute after kernel reported sync() success so after everything was supposedly written to the device and no power hungry activity was needed. Then it jumped to 7-8W. I guess these dram-less nvme's do some reshuffling from faster flash to slower flash locations during idle time. Which is probably terrible for portable use, where you may want to sync() and then unplug the device once the OS tells you everything is synced up. Creating regular power loss situations during these reshufflings just feels like putting too much trust into nvme firmware, IMO. :D