Announcement

Collapse
No announcement yet.

Samsung 980 PRO PCIe 4.0 NVMe SSD Linux Performance

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #21
    Originally posted by Michael View Post

    All a matter of the drives I have available for testing with generally stuck buying them for testing.
    hmmm....OK.
    I was sure you bought previous generations for testing (960 Pro and 970 Pro), but it seems it's not the case.
    My first idea was that you thought they re too old to be relevant in testing, or you were too lazy to search for them in your equipment ;-)

    Comment


    • #22
      Originally posted by evil_core View Post

      hmmm....OK.
      I was sure you bought previous generations for testing (960 Pro and 970 Pro), but it seems it's not the case.
      My first idea was that you thought they re too old to be relevant in testing, or you were too lazy to search for them in your equipment ;-)
      I did have a 960 PRO that failed one me but afaik don't have any other one around. I do have a 970 PRO or two but are racked up in other systems in use, i.e. without getting review samples often my SSD supplies are very limited / being utilized elsewhere.
      Michael Larabel
      http://www.michaellarabel.com/

      Comment


      • #23
        "Sequential writes" achieving almost twice the speed of "sequential reads" in this test tells me the writes went only into some SLC or DRAM cache, but not into the relatively slow TLC Flash.
        For SSDs, I consider all tests not including a full write to the entire storage space incomplete.

        Comment


        • #24
          Originally posted by Melcar View Post
          I wouldn't be surprised if the market at large just moves to QLC for their mainstream products. I was hopping QLC would be a budget alternative to TLC drives, but something tells me that won't be the case.
          There are Penta-Level Cells (PLC) in development.

          Comment


          • #25
            Originally posted by Michael View Post
            .
            Crucial P2 is labelled as a Corsair product in the graphs..?

            Comment


            • #26
              Originally posted by Jumbotron View Post
              Although I didn't see a price rundown I would imagine the Sabrent is the best bang for the buck by far as well as taking the geometric mean crown. Well done Sabrent!
              sabrent and all others current "pcie4" ssd except samsung are made on same chinese factory with same chinese controller and same flash memory, just with different labels. and that chinese controller is pcie3 ssd controller with bolted on pcie4 bus support, it can't saturate pcie4, so all those "pcie4" ssds have same 4/5gb/s specs. they can have different levels of cheating firmware though.
              i'm waiting for real pcie4 ssd to put between optane and hdds and these benchmark results aren't encouraging

              Comment


              • #27
                Originally posted by bachchain View Post
                How does it manage to be worse at SQL than its own last gen?
                last gen was mlc, this one is tlc. why do you think it's cheaper than last gen?(if we are talking about pro)

                Comment


                • #28
                  Originally posted by intelfx View Post
                  Bullshit. Modern SSDs don't work that way. LBAs are not related to internal geometry or allocation patterns whatsoever.
                  modern tlc ssds put some free space into slc mode cache and they work at advertized speeds only in that slc cache. if you exceed it, 980 has maximum linear write speed 2gb/s instead of 5

                  Comment


                  • #29
                    Originally posted by pal666 View Post
                    modern tlc ssds put some free space into slc mode cache and they work at advertized speeds only in that slc cache. if you exceed it, 980 has maximum linear write speed 2gb/s instead of 5
                    Precisely, but it does not matter one bit which "lba range" you use. As long as you write less data at once than the capacity of SLC cache, it will end up in the SLC cache.

                    Comment


                    • #30
                      Just wandering why you didn't put results for the firecuda in for direct comparison?

                      Comment

                      Working...
                      X