4 ms·

> When copying files to the SSD, the temperature sensor reported a maximum of 60°C. Additional read operations in the form of an ZFS scrub hit 63°C, which was t

> When copying files to the SSD, the temperature sensor reported a maximum of 60°C. Additional read operations in the form of an ZFS scrub hit 63°C, which was the maximum I observed during my quick testing. Dunno at what temperature the drive starts throttling itself, but I'd think a good idea would be to drop a dab of heatsink goop (and possibly some metal coins/washers held in place if it's a big gap) between the important bits of the SSD and the casing. It'll conduct heat out much better without the air interface. I can't tell if the SSD maker put a sticker on there, or a heat spreader.

RedShift1 4y ago
A heatsink pad would be less messier and you can get those in different heights, so no coins needed.
justsomehnguy 4y ago
> and possibly some metal coins/washers held in place if it's a big gap Please, no! Thermal pads exists exactly for this. They wouldn't be worse than a coin/washer covered in a "heatsink goop".
wtallis 4y ago
I don't think I've ever seen a NVMe drive with a warning temperature threshold below 70°C. Given the bandwidth limitations of ExpressCard, it would be difficult to get most consumer NVMe drives to reach even 3W power consumption. But even if you did get a drive hot enough to start throttling, it probably wouldn't slow down noticeably unless you were able to get it all the way up to the critical temperature (usually around 85°C) — a NVMe drive that's throttling mildly is still generally faster than PCIe 2.0 x1.
- jeffbee 4y ago
  I have no idea what this vendor has set the critical temperature to, but the Phison PS5013-E13 datasheet says it operates up to junction temperatures of 125°C. In general I would say the heat issue with SSDs is overblown.
  - consp 4y ago
    Isn't the NAND more susceptible to heat than the controller? I'm no expert but this is what I picked up.
    - mafuy 4y ago
      No, it's the other way around afaik. The memory likes to be hot, at least during operation.
      - jeffbee 4y ago
        That is correct. Flash must be heated to be programmed. If you cool it, it just draws more power to make up the difference. Power-off retention time is directly proportional to programming temperature.
      - monocasa 4y ago
        No, temperatures higher than about 70C cause excessive wear on NAND cells during writes.
        jeffbee 4y ago
        Totally backwards. Higher operating temperatures lead to longer retention, lower read error rates, and even faster reads and writes. Every aspect of NAND performance is better at higher temperatures. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8538043/#sec6-micromachines-12-01152title https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8538043/#sec6-m... Note that the 70° rating of commercial flash chips is the ambient temperature. The junction temperatures will be much higher.
        monocasa 4y ago
        That's only to a point. That point is not much above ambient 70C. Hence why your citation doesn't bother testing above ambient 70C. For another example of what I'm talking about, here's a a Micron patent for NAND on die temperature alerting, explicitly providing both high temp and low temp alerts. https://patents.google.com/patent/US20190129648A1 https://patents.google.com/patent/US20190129648A1

RedShift1 4y ago

A heatsink pad would be less messier and you can get those in different heights, so no coins needed.

justsomehnguy 4y ago

> and possibly some metal coins/washers held in place if it's a big gap Please, no! Thermal pads exists exactly for this. They wouldn't be worse than a coin/washer covered in a "heatsink goop".

wtallis 4y ago

I don't think I've ever seen a NVMe drive with a warning temperature threshold below 70°C. Given the bandwidth limitations of ExpressCard, it would be difficult to get most consumer NVMe drives to reach even 3W power consumption. But even if you did get a drive hot enough to start throttling, it probably wouldn't slow down noticeably unless you were able to get it all the way up to the critical temperature (usually around 85°C) — a NVMe drive that's throttling mildly is still generally faster than PCIe 2.0 x1.

jeffbee 4y ago
I have no idea what this vendor has set the critical temperature to, but the Phison PS5013-E13 datasheet says it operates up to junction temperatures of 125°C. In general I would say the heat issue with SSDs is overblown.
- consp 4y ago
  Isn't the NAND more susceptible to heat than the controller? I'm no expert but this is what I picked up.
  - mafuy 4y ago
    No, it's the other way around afaik. The memory likes to be hot, at least during operation.
    - jeffbee 4y ago
      That is correct. Flash must be heated to be programmed. If you cool it, it just draws more power to make up the difference. Power-off retention time is directly proportional to programming temperature.
    - monocasa 4y ago
      No, temperatures higher than about 70C cause excessive wear on NAND cells during writes.
      - jeffbee 4y ago
        Totally backwards. Higher operating temperatures lead to longer retention, lower read error rates, and even faster reads and writes. Every aspect of NAND performance is better at higher temperatures. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8538043/#sec6-micromachines-12-01152title https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8538043/#sec6-m... Note that the 70° rating of commercial flash chips is the ambient temperature. The junction temperatures will be much higher.
        monocasa 4y ago
        That's only to a point. That point is not much above ambient 70C. Hence why your citation doesn't bother testing above ambient 70C. For another example of what I'm talking about, here's a a Micron patent for NAND on die temperature alerting, explicitly providing both high temp and low temp alerts. https://patents.google.com/patent/US20190129648A1 https://patents.google.com/patent/US20190129648A1