Recent content by ATS

  1. A

    Microchip SMC 1000 For The Serial Attached Memory Future

    A lot of these proposals are software centric and completely oblivious to hardware. So, you are going to disaggregate? That means you are going to need a fairly high radix switch between memory and compute. Which means the latency through that switch is going to be measured in 10s to 100s of...
  2. A

    Microchip SMC 1000 For The Serial Attached Memory Future

    Lets see, higher latency than FB-DIMM, same or higher power... Don't really see this ever really being used. Nor do I see the pie in the sky dis-aggregation ever happening for performance, power, and cost reasons. HPE et al want it because it is a high value add for them that they can...
  3. A

    Toshiba RM5 Value SAS Adopted by Dell EMC

    It is primarily because customers and vendors have standardized on specific controllers and controller eco systems along with associated firmware. Know of multiple cases where large orgs have basically used bespoke firmware because V1.1x of the firmware worked but has an esoteric bug (security...
  4. A

    Chinese backdoors on Supermicro

    pretty easy to do a cut/snip to an I2C interface to insert a chip.
  5. A

    Can we expect Memory prices to go down anytime soon?

    I wouldn't expect DRAM prices to come down until either a density transition occurs or new capacity comes online. The market is largely capacity constrained atm.
  6. A

    AMD EPYC v Intel Xeon Scalable Taking Stock of Myths July 2017

    Most shocking that they don't have pcie hotswap working, for Intel that's a PRQ gate.
  7. A

    Did some write benchmarks of a few SSD's

    Power Loss Protection. For things like Sync Writes that is a killer feature. Consumer drives that don't have PLP, can't legitimately Sync until the data is actually on the flash, and enterprise drive with PLP can signal Sync the second it hits the write cache.
  8. A

    Scalable Xeons: AMD sure scared Intel into pulling out the big guns

    People keep posting that asspull number. No one that can talk has any idea what the yields are and those that know can't talk. Also the idea that large dies have poor yields is intrinsically false, fyi.
  9. A

    Intel Core X Series Launched: 18 Cores and Maddening Differentiation

    There are only 2 large scale suppliers of TIM pastes, Dow Corning and Shin Etsu. Arctic Cooling does not make TIMs. They are not a large scale industrial manufacturer. Their cooling solutions with almost 100% certainty start out in vats at either Dow Corning or Shin Etsu, as do pretty much...
  10. A

    Intel Core X Series Launched: 18 Cores and Maddening Differentiation

    There are only 2 large scale suppliers of TIM pastes, Dow Corning and Shin Etsu. Arctic Cooling does not make TIMs. They are not a large scale industrial manufacturer. Their cooling solutions with almost 100% certainty start out in vats at either Dow Corning or Shin Etsu, as do pretty much...
  11. A

    Intel Core X Series Launched: 18 Cores and Maddening Differentiation

    1) DC TIMs are the best you can get on the market for what matters: reliability. Their TIMs have virtually zero pump out, have free flowing fill capability, and are basically unaffected by either thermal cycling or time degradation. 2) TIM can be used on all size dies. Solder still has...
  12. A

    Intel Core X Series Launched: 18 Cores and Maddening Differentiation

    Yeah that's complete BS. It isn't low-quality TIM. In fact, it is one of the best most stable TIMs you can get on the market. That TIM can be thermally cycled close to infinite times with basically no pump out and has essentially no time based degradation. Also anyone that thinks they...
  13. A

    PCIe NVMe HBA FYI

    Can't tell exactly, but the default config for that card is to treat the PCIex16 as 4x4. It likely has to do with both PCIe lane allocation and bifurcation support in the bios.
  14. A

    Very excited, first Intel DC S3700 failing

    unlikely to see those outside of the vendors atm. In order to do reasonable releases you need a pretty high volume and a reasonable failure history. It is unlikely that any given customer at this point has had a large enough population for a long enough time to collect reasonable data to...
  15. A

    Very excited, first Intel DC S3700 failing

    Sometimes it is good to see failure, just so you know that everything around a failure actually works like say knowing there is a failure.