[ETA]MrSpadge

Member
  • Content Count

    799
  • Joined

  • Last visited

  • Days Won

    7

Everything posted by [ETA]MrSpadge

  1. What's this persistent write cache? An SSD cache or different firmware settings? In the latter case I couldn't see it improving performance and reliability at the same time (otherwise it would already be used everywhere). And I'm surprised these are running at 7.2k rpm! How are they doing it? The case sure looks to be full of platters. 5x1 TB or another 6 platter design, like Seagate announced? MrS
  2. [ETA]MrSpadge

    WD1003FZEX slow access times warning!

    Regarding the new Black 1 TB: it's sequential transfer rate is higher than any other Black, so it's pretty safe to say they're using a single 1 TB platter here. They're not doing this in the 4 TB model. I've long suspected that's because access times for these dense platters are difficult to handle. Seagates drives with 1 TB platters and 7.2k rpm perform well but don't have pretty access times either. From my point of view you can take the new 1 TB Black as indirect confirmation of this guess. They went from 500 GB to 1 TB platters, since anything below these points would be uneconomical, gained some STR but have trouble finding those tiny sectors quickly. Those are physical rather than firmware issues. I really doubt they could currently do it any better, because if they could they would have put this into the flagship model as well. I agree that it probably should have been called "Blue", unless that model performs even worse. Regarding general offerings: one could fault WD for not offering Blue HDDs larger than 1 TB any more. There's the Black, but you pay for its added complexity - which is mostly not worth it from my point of view. That would be foolish sicne these drives have plenty of good use cases. It's just that "main desktop drive" is not it. Here are a few: - the advantage in power draw and vibration matters if you're running many of them - you can reach higher platter densities at lower rpm, i.e. make bigger drives or achieve the same capacity cheaper - if your disk system is I/O limited by network or USB the higher speed of the 7.2k rpm drives won't gain you anything - for "cold data" the higher speed of the 7.2k rpm drives won't gain you anything *ehm* Only the 4 TB model transitioned back to 5.9k rpm, as I suspect because the vibration from the additional platters would make access times too bad at 4x1 TB 7.2k rpm. But the 1, 2 and 3 TB models stay at 7.2k rpm. Because the OP is obviously concerned about performance being affected by the high access time and sent the drive back because of this. Actually.. that the drive felt just as fast as his previous WD Blue doesn't necessarily mean the high access time is to blame - unless e.g. a 2 TB Black felt faster. MrS
  3. After 300 days of continous use my 3 TB Seagate has clocked in 2k load/unload cycles (163 start-stop), which is fine given the specs of ~600k cycles. But I'm running BOINC, which keeps the disk active with check-pointing and data transfers. What about the Toshiba 2x1 TB 7.2k rpm? I'd choose WD firmware for performance reasons any day, but I don't consider the performance differences between HDDs of the same class important any more. Actually before I got that Seagate I wanted a 2x1 TB WD Blue tuned for performance and was annoyed for a few years that I couldn't get this. But then I realized I could used the additional space of the 'Cuda.. and happy I am BTW: I think 1 TB platters at 7.2k rpm is still challenging given the current technology. It get's worse the more platters you add (that's why at 4 TB you either get reduced rpm or 800 GB platters). I suppose at lower capacity it may still be chaper to use 1 or 2 more heads and go for lower density platters, because the heads for 1 TB density would still be challenging. Whatever was used in the new mobile 3x667 GB Seagate could break this barrier, if it's cheap enough. It should be good for ~1.3 TB platters in 3.5" at low rpm and hence 1 TB at 7.2k rpm shouldn't be a problem with that any more. Not sure what they're doing to achieve this, though. MrS
  4. Idiots, for sure! It couldn't possibly be technical and financial limitations, right? Anyway, if you really want performance look into SSDs and or SSD caching. In both cases the HDD performance won't matter much any more, be it 667 GB or 1 TB platters. Any such solution is certianly a lot faster than a 2x1 TB 7.2k rpm HDD. Or if you want / need lot's of capacity go for the Seagate with 3x1 TB platters or the 4 TB models irregardles of speed. Platter density is not as important for PC performance as it used to be. Don't get obsessed with it. MrS
  5. I think so, especially because the previous 1 TB WD Black was officially still using 500 GB platters. MrS
  6. Indeed. That's not much of a secret, though: you can't introduce a new storage medium without thouroughly researching and documenting things like data retention and P/E cycles. It's something people cared about even in the first flash devices decades ago. And which is something which keeps quite a few alternative technologies off the shelf MrS
  7. [ETA]MrSpadge

    Expertise needed please!

    Another vote for the 840 Evo. It's got a very well balanced controller and makes clever use of that TLC NAND. In the real world you don't have to worry about its write cycles. And you're not going to fill it with 4k random data repeatedly, since you said you're going to place OS and programs there. The price premium for that Plextor is insane IMO. I didn't look up the Kingston V300, but the 840 Evo likely performs better enough to warrent the price premium over the V300. MrS
  8. I see it like this: warrenties are like insurances. Statistically you'll pay a bit more than you get back. That's what the provider makes a living from. And you also buy the safety of being secure should "that event" hit you. Due to these reasons I only like insurances for cases where being hit by "that event" really hurts or threatens existance, like a liability insurance. Consequently I'm not buying more expensive HDDs for longer warrenties. By the way.. I don't think there is such a thing like "HDDs not being able to run 24/7". I think this is primarily just there to scare us and make us buy more expensive HDDs and secondly to ease manufacturer specifications. Like for Seagate desktop HDDs: they are not saying the drives won't take 24/7, but rather that their specs assume 8h of usage per day. That's a very important difference, which most people overlook and is not something Seagate or WD will want to point out. That is not say that there are no differences in HDD quality and longevity. What I'm saying is that for the HDD it really doesn't matter if it ran 8h a day for 3 days or 24h on 1 day. Thermal equilibrium is reached after ~15 min, instantaneously compared to the timescales we're talking about. And the spindle? Well, it's turning at 5 to 7k rounds per minute. What the hell should change after 24h that didn't already happen after 8h? In the two cases I'm comparing the 24h HDD would actually have experienced less wear due to recieving only 1/3 the start-stop cycles per runtime. But of course after a certain time, e.g. 1 year, the 24/7 HDD has accumulated 3 times the overall runtime, which counterbalances the reduced number of start-stop cycles. MrS
  9. That review seems very strange: in 4k random read/write 7.2k rpm HDDs score in the range of 1 - 2 MB/s, translating into about 75 - 150 IOps. Either they got the unit of their scores screwed up or the test mode - neither of which is very reassuring. And they may say that "the Seagate Barracuda 7200.14 3TB is faster in every benchmark".. except in the PcMark 7 score they posted themselves (your link). Seriously? Looking at the numbers myself I'd first dismiss anything relying only on STR (sustained transfer rate), because once you're trasnferring at > 100 MB/s in a desktop environment your jobs is likely finished soon anyway, no matter if it's 150, 180 or 200 MB/s. More STR is nice, but doesn't really change the subjective performance unless the difference is massive. What counts is when random transfers are thrown into the mix and the HDD comes to a crawl - that's when you're sitting there, thinking "oh boy, why didn't I buy an SSD?!" That's where the WD Black shines compared to other HDDs (the very good access time, which always comes at the price of higher access noise), but falls behind SSDs far more than in STR limited scenarios. And that's what hardware.info didn't really test, hence the excellent performance of the Toshiba and Seagate with their STR advantage due to their 1 TB platters. Their PCMark scores are probably the most useful of the bunch, and here I'd say those 3 drives perform about equal: the difference is less than 5%! PcMark "still" relies rather heavily on STR - but in this case it's a god thing, as it tries to resemble a typical desktop workload and you shouldn't hit your HDD with much more random requests anyway. If you need to, you'Ve hopefully already got some SSD solution. And that's why I think the WD Black is poor value, despite its good performance. MrS
  10. Unless your storage system performance is directly related to your income (file servers etc.) then I'd say this price increase is absolutely not worth it. Sure, the WD is faster and at 10% higher price might get the nod - but 60% is far too much for little benefit. For the same price you could get 2 2TB Toshiba or Seagate in Raid 0. And the last time I looked the Seagate SSHD was cheaper than the WD Black. Performance-wise they trade blows (Seagate wins handily from cache, WD wins a bit uncached). Drawback: only available up to 2 TB with 7.2k rpm, whereas the 4 TB model is 5.9k rpm. Still, 2 x 2 TB Seagate SSHD are cheaper than a single 4 TB WD Black in Germany! MrS
  11. MRFS, in principle I agree with the argument "spend more on hardware to be more productive" for a net win. That's why I'm equipping all of our new or refurbished boxes with SSDs. However, the benefit obviously depends on how intense the workloads are. In research at university we're usually more limited by thinking rather than the storage subsystem speed (if it's already a decent single SSD). Of course there are also computationally intense tasks - that's what our simulation machines are for. Anyway, if your workloads are often more I/O intense, then by all means go for the Raid 0 I'm still not convinced it's the best solution for general office productivity. The reason is diminishing returns regarding I/O performance. Dropping transfer times from whatever into the single-digit second range is of obvious benefit. But when going from 2s to 1s I question people's ability to actually use that time for anything productive. It surely depends on how much you're limited by "thinking about what to do next" versus "hammering out standard procedures". And then there's the OS file cache and yourself simply leaving your usually needed programs open, which both mitigate the benefit from even faster storage further. Faster storage also makes CPU bottlenecks more pronounced for further diminishing returns. Disclaimer: I have never worked with a storage system faster than a single high end SATA3 SSD. MrS
  12. [ETA]MrSpadge

    Seagate Desktop SSHD Review Discussion

    I definitely agree with that. I always wondered why it took them so long, and why the first attempts were so half-heartedly. Of course they wouldn't sell too well at 500 + 4 GB and a significant price premium... MrS
  13. I'm taking a completely different approach to SSDs: unless you've got a very write-heavy workload don't mind the write cycles. Don't overspend on the SSD (current sweet spot for me: 120 GB for 80-90€) and place large files elsewhere. If it's a modern drive which is not being completely filled the write cycles won't be a problem for many years. At that point the future solutions will probably be that much better that you won't care much about the old drive any more. This is certainly much more economical than buying 2 or more drives for Raid 0 just to get higher write endurance. Consider what new drive 90, 180 or 270€ will get you in a few years, if you really went through those write cycles before any other reason made the drive unusable! MrS
  14. They may be manufactured in China, but it's surely still the same design as under the Hitachi flag. And I wouldn't call 1 1 TB per platter 7.2k rpm driver "old generation". There is nothing better, despite some models being newer. Sure, the new WD Black is the fastest HDD on the block again - but for a huge price premium for a performance difference which pales in comparison with SSDs. Go for the 'Cuda 3 TB if you don't trust those Toshibas. Won't an HDD break eventually, too? Are you running some super write intensive server workload? If not: never mind the write endurance. And you don't need the absolute fastest SSD either, as e.g. the 840 Evo is significantly cheaper than the Pro and does really well for desktop workloads. Personally I'm very happy running a 3 TB 'Cuda (7.2k rpm) with an Agility 3 60 GB as cache drive using Intel SRT. Both drives have been good value nad most of the time the combined performance far exceeds a WD Black. Nowadays I'd get the 120 GB 840 Evo, use 60 GB for the OS partition and 60 GB as cache. MrS
  15. [ETA]MrSpadge

    Is alignment really a thing on SSDs?

    Thanks for providing these hard numbers, MRFS! That should be exactly what jtsn was looking for. MrS
  16. The drives are not even listed in a standard german price comparison tool. It's got a "good tradition" in the HDD space to announce and only ship months afterwards - but this doesn't make it any more acceptable. It's OK to paper-launch, if they tell is so and provide an expected shipping date. Thanks for looking into it, Brian. MrS
  17. [ETA]MrSpadge

    Seagate Desktop SSHD Review Discussion

    That's good to know, thanks for sharing! MrS
  18. [ETA]MrSpadge

    Is alignment really a thing on SSDs?

    Just create an intentionally misaligned partition, benchmark it and compare with an aligned partition. The difference in random write should be obvious, whereas it may not be as present in sequential data. I don't know about those capability bits, but I've read that the physical blocks in NAND are 4k (or maybe even 8k) - just as FastMHz said. MrS
  19. [ETA]MrSpadge

    Seagate Desktop SSHD Review Discussion

    I still prefer my Seagate 3 TB and 60 GB Agility 3 as cache.. but nice to hear this drive works well for you Some aspect people are often forgetting when dismissing the relatively small cache sizes is that for any game or program you practically never need the entire data at once, so saying "but program x doesn't fit into this cache" is pretty meaningless here. MrS
  20. [ETA]MrSpadge

    WD Black2 SSD/HDD Review Discussion

    @Chris: agreed. Pre flood we could get 500 GB drives (1 platter, 2 heads - so could be 1 TB now) for 17€ from an OEM. Sure, that included a massive discount compared to retail prices, but still: I'm pretty sure the OEM and Seagate still made some profit in this deal. It'll take quite some more time for SSDs to catch up with this. MrS
  21. [ETA]MrSpadge

    Advice on SSD 840

    Generally no really valueable data should be trusted to any single drive. They can always fail due to various reasons, with SSDs not pretecting from things like theft. So if a backup "only" mirrors data from somewhere else, I wouldn't see much of an advantage for a SSD based solution. In fact, each bit in an SSD consists only of a few 100 captured electrons. Over time they escape, loosing the data. Consumer drives usually hold their data for more than 1 year - still, it's not exactly an advantage for backups. Does your laptop currently have an HDD or SSD? If you nevertheless want to use the 840 for backup it's a bad idea to use Samsung Magician. It's mean for single-time transfers of the complete system (last time I used it didn't erase the old one, though). To backup simple videos and pictures you could easily copy them over in the explorer. Although this gets messy and slow if you do it repeatedly. I suggest using a tool like the freeware "DirSync" instead. Here you can set up synchronization jobs, which you can then run via planned tasks or manually each day/week/ipon changes or when ever. The tool compares source and destination and only transfers changed files, which makes the process very quick for most data (seldomly changing). You can also set it to "mirror" mode, where files deleted in the source are also deleted in the backup destination - keeps things clean. MrS
  22. I especially like "Shield" - sounds like this is trying to make really smart use of the flash cells. From 3k to 18k write cycles for the same technology node. Very impressive! MrS
  23. Yes, if you didn't change it the pagefile will be on C:. I was referring to your statment "And yes I have multiple hard drives in use for scratch disk, etc..)" .. but just realized that this surely means the Adobe scratch disk and not the OS pagefile. MrS
  24. No, I don't think so (without knowing your exact "typical" usage pattern). Regular HDD use is actually far more sequential and with on average much larger blocks than 4k. If there are purely random 4k accesses (like on servers, databases etc.) then HDDs start to crawl - which is why a few of them hurt badly, and which is why (good) programmers have ever since tried to avoid them as much as possible. If you look at typical transfer rates of HDDs under typical desktop loads you'll see "a few" to "a few 10's" of MB/s - which is far higher than random 4k rates. Regading the typical QD: I don't have hard numbers on this, but so far I've read it seldomly exceeds 1 or 2 for typical desktop loads. And if it does, HDD performance comes to a crawl anyway, so even with 40% faster than "sloooow" we'd be still at "slooow". You could observe QD yourself in Win ressource monitor. Regarding which HDD to keep: I'm sure the new Black is better.. although I wouldn't want to spend much, or anything at all if I already had the previous model, for the upgrade. This obviously depends on how much you usually wait for your HDDs and how much reducing those waiting times is worth to you. BTW: if your SSD is at least Sandforce 2200 class performance I'd put the page file on it, because that's the access pattern where it will provide the most benefit over HDDs. MrS
  25. There at least 4 aspects to be considered here. I'd say the first is STR (sustained transfer rate). That's the transfer rate achievable with sequential accesses, where the heads don't have to move to any random positions. It's the maximum a HDD can transfer, except for data from the cache. Here you're seeing 167 MB/s vs. 147 MB/s, i.e. a 14% improvement. In SR's review they got 187 MB/s vs. 153 MB/s, i.e. a 22% improvment. This difference could easily be due to the way the different tools measure, or maybe caused by variation among drives (although such a difference would be unusually large). The 2nd aspect is random access. Your benchmarks shows both drives doing 81 ops/s at 4k random read. Notice how the old Black gets 103 IOps in SR's review - you're clearly not testing the same thing! The most obvious explanation would be the differences in queue depth, where SR tests at a minimum of 2 threads with 2 queued items each, i.e. an effective queue depth of 4. At a queue depth of 1, as HD Tune is said to use according to someone before me, there's only so much you can do to improve access times without increasing RPMs or short stroking. And 3rd: WD themselves claim their performance improvements largely come from firmware tuning. This will show up in real world applications (like the file and web server tests), but not in cases where you're purely mechanically limited. Which are exactly the 2 corner cases you tested: completely sequential access (STR) and completely random at QD 1. That's a bit like judging a car's performance on a race track from just a dyno run to extract peak HP and torque. I'm not sure if there are any good free HDD benchmarks (that's why simple reviews always stick to these simple low-level tools).. but the one you used is certainly not it. And finally "up to xx% performance increase" doesn't mean this will manifest itself equally in all benchmarks. And I'll admit SR's review and conclusion do sound a bit enthusiastic - but that's because such performance gains are almost unheard of, for HDDs within the same technological generation! It's still an HDD, though, and if you read the article in any other way.. I don't know why. SSD performance is miles above HDDs, no matter if the HDD got 48% faster or not. And this gap is only going to grow over time. MrS