3 ms·

Shouldn’t we need to use volume instead of area ?

tromp 5y ago
I think cooling requirements will keep use of the 3rd dimension severely limited in the foreseeable future. Also, the required number of lithographic steps might make it economically infeasible.
- londons_explore 5y ago
  NAND manages 100+ stacked transistors. Things like processor caches have similarly low average switching rates, so it doesn't seem out of the realm of possibility to see use of the third dimension for logic.
  - selectodude 5y ago
    NAND isn’t ever all being used at the same time and even then it can get very hot. SSD thermal throttling is a genuine issue.
    - xxs 5y ago
      SSD thermal throttle the controller =not= the NAND, itself... which actually doesn't even like to be cold.
      - wtallis 5y ago
        Depends on the drive. Sometimes it really is the NAND getting too hot.
      - merb 5y ago
        if you use the latest drives m.2 pci4, etc. high end the whole thing gets really hot. high end boards thus have heat sinks. keep in mind nand needs to be warm to reach peak perf, but not too warm. the whole heat sinks can reduces the heat by as much as 7°C degree.
- sterlind 5y ago
  I've heard of proposals to do cooling via microfluidic channels, but the lithographic steps problem seems inevitable. at least unless you can somehow pattern multiple layers at once, which would destroy the feature size.
- dehrmann 5y ago
  > I think cooling requirements will keep use of the 3rd dimension severely limited in the foreseeable future. It's fun to go look back on old CPU coolers. They started around Pentium-era CPUs, then kept getting bigger. Around 100W TDP, they stopped. I think that's the largest practical air-cooled cooler.
  - verall 5y ago
    A consumer tower (air) cooler can clear 150W on the stock fan and if you stick some 2k-3k rpm push-pull it will probably clear 300W. Fancier vapor chambers and thicker, higher RPM fans can clear up to thousands in server environments.
pjc50 5y ago
They're still not stacked, as far as I'm aware - still needs to be in a "well" on the substrate.

I think cooling requirements will keep use of the 3rd dimension severely limited in the foreseeable future. Also, the required number of lithographic steps might make it economically infeasible.

londons_explore 5y ago
NAND manages 100+ stacked transistors. Things like processor caches have similarly low average switching rates, so it doesn't seem out of the realm of possibility to see use of the third dimension for logic.
- selectodude 5y ago
  NAND isn’t ever all being used at the same time and even then it can get very hot. SSD thermal throttling is a genuine issue.
  - xxs 5y ago
    SSD thermal throttle the controller =not= the NAND, itself... which actually doesn't even like to be cold.
    - wtallis 5y ago
      Depends on the drive. Sometimes it really is the NAND getting too hot.
    - merb 5y ago
      if you use the latest drives m.2 pci4, etc. high end the whole thing gets really hot. high end boards thus have heat sinks. keep in mind nand needs to be warm to reach peak perf, but not too warm. the whole heat sinks can reduces the heat by as much as 7°C degree.
sterlind 5y ago
I've heard of proposals to do cooling via microfluidic channels, but the lithographic steps problem seems inevitable. at least unless you can somehow pattern multiple layers at once, which would destroy the feature size.
dehrmann 5y ago
> I think cooling requirements will keep use of the 3rd dimension severely limited in the foreseeable future. It's fun to go look back on old CPU coolers. They started around Pentium-era CPUs, then kept getting bigger. Around 100W TDP, they stopped. I think that's the largest practical air-cooled cooler.
- verall 5y ago
  A consumer tower (air) cooler can clear 150W on the stock fan and if you stick some 2k-3k rpm push-pull it will probably clear 300W. Fancier vapor chambers and thicker, higher RPM fans can clear up to thousands in server environments.

pjc50 5y ago

They're still not stacked, as far as I'm aware - still needs to be in a "well" on the substrate.