• Content Count

  • Joined

  • Last visited

  • Days Won


Posts posted by johnw42


    PCper's review uses HDTune to find a sequential read speed range of 64.9 to 150.5 MB/s. I would have preferred to see the actual HD Tune trace, since sometimes there is a bit of fluctuation with the measured speed, and the MIN and MAX may be a transient blip rather than the main trendline. But at least they did measure the minimum sequential speed, which gives PCper a leg up on SR's review.

  2. For NAS use this kind of test isn't relevant. If it's just something neat you want to see though, we'll see about getting those numbers for you.

    Of course it is relevant. The sequential read/write speed of the disk can drop by about a factor of 2 as it fills up and you get to the inner portion. That affects read, writes, and rebuilds. The question is exactly how much it drops for these HDDs as compared to other HDDs.

    And of course I want to see it, that is why I posted about it.

  3. I'd love to see an interview from someone at Toshiba about their 3.5" HDD plans for the rest of this year.

    Is there any chance SR could interview someone at Toshiba? I'd like to know what consumer or prosumer 3.5" HDDs Toshiba plans to sell in the near future, and when they plan to release them. Most of HGST's large 5400rpm HDDs seem to be disappearing from the market, and I wonder whether they will reappear under a Toshiba label, and if so, when?

    • Like 1

  4. Interesting, but companies producing LSI/SF drives with current firmware support (Mushkin, Patriot, newer Kingston, lines), Intel, and Crucial will be the ones getting my money while OCZ treats data storage products like gamers overclocking their rigs to point of failure.

    I certainly agree about avoiding OCZ, but I prefer the more stable non-Sandforce SSDs from companies that make a strong effort for quality and reliability. SSDs such as the Plextor M3P, the Samsung 830, and the Crucial m4 (although the m4 is slower than the Plextor and Samsung models, it is also cheaper at the moment).

  5. Good improvement overall. Maybe it's now starting to show some promise of the "special" (new?) Marvell controller. I can't understand why they released the drive with the older/poor firmware in the first place, though.. doesn't make any sense.

    I don't think it is much of an improvement. It still trails the Plextor M3P, and the sequential read and write speeds of the V4 still don't look very good (except for the claimed sequential write speed on the 128GB model which looks good, if true, which is a big if with OCZ).

    As for why OCZ released it before it was ready, do you really have to ask? This is OCZ we are talking about. Quality is someone else's job. OCZ is all about hype and misleading the customer.

  6. The big question will be, once production meets demand again, will prices fall back down to what they were before? Even if they do, that in my book means that they are still significantly higher than what prices would be had the flood never happened and prices had continued to drop as normal.

    Probably not. I think the previous prices were unsustainably low. In my experience, Samsung was making the best HDDs. And Samsung is a HUGE company. And yet they found that the HDD business was not profitable enough for them to continue, so they sold their HDD business. This indicates to me that the HDD prices were too low to sustain the required R&D and production equipment investments while maintaining a profit margin equivalent to Samsung's other business units.

  7. Actually, johnw42, those who understand MORE than freshman or "basic economics" know that when there is a supply shock of decreased supply, prices CAN & OFTEN, but NOT always adjust higher in order to match demand to the decreased supply.

    In an unrestricted market, ALWAYS. If the supply of a product suddenly goes down, then in an unrestricted market, the only way for demand to go down to match supply is for the price to rise.

    In a restricted market, when one group tries to force another group to sell at certain prices, the price may not go down but you will have shortages, rationing, bribes and corruption and underground markets.

  8. Is this a drive that´s particularly suitable for video- and raw photography editing working in Macbook Pro environment, or should I be looking elsewhere?

    In my limited experience, an important quality for video editing is high sequential write speed for incompressible data.

    I think the Plextor M3P has the highest incompressible data sequential write speed of any consumer SSD, so in that sense it should be a good choice for video editing.

    I don't have any experience with SSDs in a Mac environment, so I cannot comment on that aspect of your question.

  9. Great job on the steady-state portion of the review!

    But there were two major problems, the first is by far the worst:

    1) Terrible choice of comparison SSDs. All 4 competitors were nearly equivalent Sandforce SSDs! What were you thinking?

    The competitors should have been: Intel 520, one other Sandforce SSD (Vertex3, HyperX, or KC100), Samsung 830, Crucial m4, and Intel 320.

    2) Your "real world" tests are flawed, and not at all representative of real world usage. For example, look at the "HTPC disk capture" test. Video files are a perfect example of incompressible data. And yet your Sandforce SSDs report rates of 450MB/s, which is close to the easily-compressible sequential write rates you measured with IOMeter of 460MB/s, and nowhere near the incompressible sequential write rates of less than 300MB/s you measured with IOMeter. Clearly your "real world" test is writing UNrealistic highly compressible files instead of real video files.

  10. In SR reviews, a standard feature is the IOMeter synthetic tests: database, file server, web server, and workstation.

    This is useful data, but for many users, the most important part is the low QD portion, basically QD=1 and 2. Most users will very seldom have their SSDs reach QD > 4.

    But it is difficult to compare SSDs at QD=1 and QD=2 in these graphs because they tend to overlap each other since the scale of the graph is much larger to encompass the QD=32 results.

    It would be nice if you could present the data in such a way that the QD=1 and QD=2 results are easier to compare among SSDs. Possibly it could be done by changing the y-axis to a log scale, but I'm afraid that still might not be enough to see the low QD results clearly. Perhaps the best thing would be to add an additional graph for each of the four tests that shows only QD=1 and 2, with the vertical scale enlarged to show detail at the low QDs.

  11. We understand that all drives get lower sequential R/W rates. That isn't the issue. A back-to-back test of the same drive with 256 vs. 512 may be very telling. ;)

    This comment makes no sense. The 128GB M3P has 256MB of cache, how could they possibly test it "back-to-back" with 512MB of cache?

    Besides, the sequential write speed of the 128GB M3P would be exactly the same as it is, even if the RAM cache were increased in size.

    And why do you keep writing "sequential R/W"? The sequential read speeds are the same to within 1%.

  12. They are only using 256 MB. cache on the 128 GB. drives instead of 512 GB. like on the larger drives - which appears to acct. for a high percentage of the drop in sequencial R/W performance.

    No, you are mistaken. First, there is no "drop in sequential" read performance that correlates with cache size. And the lower sequential write speed for the 128GB model is almost certainly due to lower flash parallelism in a lower capacity model, NOT due to lower DRAM cache on the SSD. As I already said, all SSDs have lower sequential write performance in the lower capacity models. That is because the bottleneck for sequential writes is the flash itself -- with the higher capacity SSDs you can get more channels and interleaving writing in parallel, so higher write speeds.

  13. What cache issue are you talking about?

    All SSDs have slower sequential writes at lower capacities. The higher capacities typically have more parallelism, which allows faster sequential writes.

    As for sequential read, the spec on the M3P is 535 MB/s, 540, 535, for 128GB, 256GB, and 512GB models. That looks fine to me.

    Some people are mislead by Sandforce SSDs which often spec unrealistic sequential write speeds, especially on the lower capacity models, but those speeds can only be obtained with highly compressible data (like a stream of zeros). On real data, the Sandforce sequential writes are typically about the same as a Crucial m4.

    If the 128GB M3P really can write at 350 MB/s, it will be faster than any 120GB Sandforce drive when fed real data.