No announcement yet.

PHOENIX FM-25S2S-120GBP1 ruinning slow!!!help

  • Filter
  • Time
  • Show
Clear All
new posts

  • #31
    Originally posted by sam350 View Post
    i havent used HDD erase yet thought id post what iv found.... found that using crystal disk mark 3.0 results depends hugley with what options you use... under file-test data- you have three options defult data, fill 1, fill 0.....if u use fill 0 or 1 i get the following results which are really really good (which setting should be used??)

    Sequential Read : 278.738 MB/s
    sequential Write : 271.960 MB/s
    Random Read 512KB : 272.267 MB/s
    Random Write 512KB : 269.893 MB/s
    Random Read 4KB (QD=1) : 38.807 MB/s [ 9474.4 IOPS]
    Random Write 4KB (QD=1) : 87.781 MB/s [ 21430.9 IOPS]
    Random Read 4KB (QD=32) : 199.849 MB/s [ 48791.3 IOPS]
    Random Write 4KB (QD=32) : 61.122 MB/s [ 14922.3 IOPS]

    but if i use defult setting i get much lower performance

    Sequential Read : 224.055 MB/s
    Sequential Write : 73.071 MB/s
    Random Read 512KB : 213.570 MB/s
    Random Write 512KB : 73.530 MB/s
    Random Read 4KB (QD=1) : 30.715 MB/s [ 7498.8 IOPS]
    Random Write 4KB (QD=1) : 68.916 MB/s [ 16825.3 IOPS]
    Random Read 4KB (QD=32) : 155.751 MB/s [ 38025.1 IOPS]
    Random Write 4KB (QD=32) : 75.507 MB/s [ 18434.4 IOPS]

    using atto benchmark my results are pretty much on par with your from website which is good....

    using as ssd benchmark my results are low.. write speeds suffering..

    AS SSD Benchmark 1.5.3784.37609
    Name: FM-25S2S-100GBP1
    Firmware: 2.1
    Controller: iaStor
    Offset: 103424 K - OK
    Read: 209.46 MB/s
    Write: 75.57 MB/s
    Read: 21.04 MB/s
    Write: 72.63 MB/s
    Read: 119.18 MB/s
    Write: 75.77 MB/s
    Access Times:
    Read: 0.194 ms
    Write: 0.213 ms
    Read: 161
    Write: 156
    Total: 402

    is this just a problem with the benchmark programs giving inconsistant results or should i still use hdd erase.... and c how we go?? thanxu in advance for your help
    You need to understand something very important about the sandforce controller, which will explain the huge differences in the CDM benchmarks. The SF controller *COMPRESSES* data. This is how it gets such high speeds - but only in certain cases. The results largely depend on whether or not the data can be compressed, and to what extent it can be. Thus, when you use totally random data, you will get much lower speeds. However, when you use all 00's, you will see the best case scenario since the controller will be able to compress the data to the greatest extent possible.

    Do some testing on your own, and you'll see what I'm talking about very clearly. Try doing a sequential read/write on a large file that is already heavily compressed. Then compare that with a similarly sized file that isn't and is highly compressible.


    • #32
      Well said, svg.

      I would just like to point out one thing. Normally when you see CrystalDiskMark benchmarks, they are using the Default (random data). This is why I asked Ranxerox if he could post his results in Default mode too, to show that your results are not so much different from his.



      • #33
        These are my data after passing HDDERASE.EXE and fresh install with Windows 7.

        My best regards
        Last edited by Ranxerox; 07-30-2010, 01:04 AM.


        • #34
          Just got my 60GBP2 yesterday. Was able to do a fresh Win7 install and run a few benches.
          I'm concerned that some of my speeds are lower than what they "should" be. Sorry, I don't have the print screens (maybe later).

          Using CDM and AS SSD.
          My sequential reads are ok, just under 200MB/s.
          What concerns me most: random 4K reads are about 13MB/s only!. Best I've seen is 17.
          Sequential writes peak at under 100 MB/s.
          Also, my access times are slower than any results I've seen: .197 and .388.

          I realize all these numbers are still pretty good, and I'm pleased with the drive, but I'm just wondering what could account for the difference and if there is something I could do about it?

          Another question: Do you guys run the benchmarks from another drive - or from the same drive as you're testing? I realize that the mfr's/reviewers run their benches on perfectly clean drives, so I don't expect to hit those numbers w/ win7 installed.

          Core i3/Gigabyte GA-H55M-UD2H. I installed the Intel RST. AHCI in the bios. Optimized w/ the usual tweaks. Disk is aligned (1024K).

          Going to try overclocking my bus and see if that helps.


          • #35
            Try to update your BIOS.
            Maybe there is chance to improve performance.


            • #36
              newsflash be


              Thanks for the reply.

              Yes, I will try a newer bios version.

              I disabled ALL of the CPU heat/power stuff in the bios (EIST, C3/C6/C7,Thermal Monitor, C1E, ProcHot) and my 4K random read speeds show significant improvement (22MB/s app). Still not as good as most benchmarks I've seen, but much better than before. I tried overclocking too, but it made NO difference.


              • #37
                OK..a few things to report.

                First, I updated my bios and it had no effect whatsoever in terms of speed and the "smart command failed" message.

                But I read a terrific sticky at the OCZ forums which confirmed what I had already realized in my very short amount of time spent using this drive. Namely, that the benchmarking process itself is actually CAUSING the write speed degradation!


                You see, these sandforce controllers are a very different animal compared to mechanical drives and other (Intel/Indilinx) SSDs. I knew that already, but it takes a little bit of getting used to and a greater understanding of precisely how the SF controller works.

                So I used hdparm from my linux platform to "security erase" the drive. (Much easier than using hdderase). After that, I restored my windows and linux images and ran ONE cdm and ONE atto benchmark just to see the difference. I changed CDM to use 100MB instead of 1000MB and only ran the test 3 times instead of 5. I also narrowed the range of testing in ATTO. The tests confirmed that my original performance has been restored - and I will NOT be running any more tests - especially not the ones with non-compressible data.

                My CDM results:
                211 / 110
                203 / 100
                30 / 85
                183 / 98

                I'm pleased to see the original numbers have come back (my sequential writes were under 70 MB/s before cleaning the drive). I'm not so happy that some of my times are quite a bit lower than other ones I see posted, but they are certainly good enough. Turning off the CPU power mgmt. options in the bios changed my random 4K reads from under 20 to 30 and lowered my r/w access times. I have setup and will use my mechanical drive for any and all large sequential writes. O/S and applications ONLY for the SSD.

                Bottom line: don't bombard your Sandforce SDD with non-compressible writes because that will slow down your drive bigtime until it has a chance to recover (which will take quite some time).
                Last edited by svg; 07-29-2010, 08:48 PM.


                • #38
                  Thanks! svg,
                  The tweak about cpu power setting in BIOS is really a useful information.
                  Do you mind if i create a thread about tweaks for Phoenix/Phoenix Pro and add this information?


                  • #39
                    stomach cheroot

                    No problem. Glad to help.