• 2 Posts
  • 16 Comments
Joined 2 months ago
cake
Cake day: August 7th, 2025

help-circle

  • Moore’s law was about the technology – but I am talking about the application of the technology. It was unusual for most businesses to base their purchasing / refreshing decisions around the idea that the technology would be good for 2-3 Moore’s law cycles. This was especially true back in the days of Mainframes and later “Mini” computers (shrunk down versions of Mainframes – not Mini PC’s) where companies like DEC and IBM went to great lengths to ensure that upgrading to a newer system didn’t impact other operations in a business.

    Most of this carried on with Vax and Unix Systems (like Sun workstations, SGI, etc.) in the same lifecycle.

    When PC’s started coming into the business world, the thought was that they would fit that same lifecycle – and many of them did. This set the mark for early PC’s when IBM brought them to the consumer market. The IBM PC was, after all, the consumer version of a business computer.

    Apple, Commodore, TI, Atari, et al. were a bit different – coming at things more from the entertainment, education, and hobby side of things.

    I see what Steve is doing here is attempting to push things back towards the business lifecycle, and with good reason: it’s better for the planet if fewer machines are abandoned due to the arbitrary whims of some marketer’s concept of profitability.




  • Meant to comment on this earlier… I’m implementing an LVM cache – which is filesystem / device level caching. Having a failure with something at this level could mean corrupting a 42TB storage device. This would be a far cry from losing an application level set of cache files…

    That’s why I am being a lot more cautious about this drive. A failure here could be non-recoverable.


  • Yeah, I’m quite aware of a lot of the junk on Amazon – and I normally would stick to a well known brand like Samsung, WD, or Crucial… But there were no listings for m.2 SSD’s in the 32-64G range. At first I ordered a “Kingdata” drive (an obvious play on Kingston), but later I saw a listing for a drive from Transcend – which I recalled from my IT days, and a quick check of their website confirmed they were the company I was thinking of.

    So, this is why I am fairly certain that this is some kind of labeling / packaging mistake. Transcend is reasonably well-known, and afaik aren’t scammers.

    And, to top it off, I ran some additional tests on the drive… And for what it is, it is performing exactly how I would have expected: 420MB/s read/write, with 0.1msec access times – with extreme consistency. (Given that this is installed on a PCIE adapter that only has 1 lane available.)


  • Okay - wild… The results of f3probe:

    Good news: The device `/dev/sda’ is the real thing

    Device geometry: Usable size: 931.51 GB (1953525168 blocks) Announced size: 931.51 GB (1953525168 blocks) Module: 1.00 TB (2^40 Bytes) Approximate cache size: 0.00 Byte (0 blocks), need-reset=no Physical block size: 512.00 Byte (2^9 Bytes)

    Probe time: 16.12s

    Oops - misstated something before. This is an MLC NAND drive, the cache is supposed to be DDR4 DRAM. I suspect, however, this is a mis-labeled drive…


  • I agree - I wouldn’t trust it either…and, surprisingly, this one came from Amazon, and not some fly-by-night AliExpress store. (I rarely purchase something there without seeing reviews first…

    But the other thing about this is that I checked out the website for the product. They are a company that specializes in enterprise and embedded products. I was pretty certain I had heard of them before in the enterprise world, which is why I purchased the drive.

    The reason I bought this drive was because it specifies having a NAND cache on it (MLC, but beggars can’t be choosers with drives like this), whereas the others I looked at didn’t have (or at list didn’t have specs which listed having) any form of NAND caching.

    @nao@sh.itjust.works - thanks f or the pointer to f3 – I’ll grab it and check the drive before I return it.










  • This is definitely something that has been coming for some time. Yes, I am a little surprised it has taken this long for a major channel know for hardware reviews for gamers to take this step… I would have thought that the popularity of the Steam Deck, and all of the handhelds that are now running Linux would have been a motivator.

    But it seems that Steve is really seeing that there is more of a progression of people not wanting to go to Windows 11, and the issues surrounding Microsoft’s insistence on adding creepy features that no one asked for (like Recall) as the push they needed.

    And I agree, Bazzite is probably one of the best choices that they could make. The immutability of the system will allow them to have consistent images that won’t change on them randomly. That is a definite requirement when dealing with this type of benchmarking.


  • I don’t know if you were joking or not… But in case you weren’t: the Intel guys typically have information about upcoming / unreleased products before the details are out in the open. Yes, the drivers can be maintained by the community when the information is available… But, day one driver support won’t be there (since they are generally developed in-house by Intel, and then pushed up to the kernel for release), and community development would (likely) take significantly longer…

    And on the Enterprise customer side, there might be some hesitation about adopting newer Intel products that don’t have drivers officially supported by the company…at least in environments where Linux would be the (logical) choice… That might lead Enterprises to look at Windows instead of Linux.