Quick Review: Samsung 850 EVO 1Tb (MZ-75E1T0B) Solid State Drive

Solid-state drives are something that I can rarely get enough of, as high performance storage has a big impact on the time spent waiting on a computer to get data-heavy tasks done. As someone who has enjoyed this, I had four assorted 256Gb or smaller SSDs in my computer which were used for working documents and temporary scratch space. Sadly, my board had two eSATA ports which I had converted to internal usage start to fail, so a migration of drives from those ports to internal ports were necessary. The easiest way to free the necessary ports? Consolidate three SSDs into one, and have more SSD space as an upgrade at the same time.

Ordinarily, I wouldn’t go splurging out on a new 1Tb-class SSD without good reason. This time, the reason was again cost. Amazon EU was offering the drive at AU$360 shipped, for a record-smashing 36c/Gb. I hadn’t heard of drives this cheap, let-alone well-regarded Samsung 3D V-NAND based drives, even if it was TLC-based. I couldn’t resist the offer and jumped right in. It took a painfully long journey via UPS iParcel, but now it’s here!

Unboxing

20151214-1330-0519 20151214-1330-0520

Being a more value-oriented model, it is clear that things are not quite as lavish as with the 850 PRO series. The box itself doesn’t have the special texture or black wash that the more expensive model does, and the 850 EVO offers 5-years warranty compared to the 10-years offered on the PRO.

20151214-1330-0521 20151214-1331-0523 20151214-1331-0522

The edges of the box show nothing spectacular, aside from being subtly crushed in transit. However, another key difference is seen, namely the 850 EVO was Made in China, as opposed to the 850 PRO being Made in Korea. This suggests the manufacturing might not have been done “at their home” and instead may be outsourced or produced by a secondary plant.

20151214-1333-0524

Despite this, you still get a CD-ROM with the software on it, not that you’d use it, and the paper guides which aren’t particularly useful. It seems they did omit the promotional stickers which come with the PRO, not that they’re by any means necessary. You don’t get a spacer for 9.5mm use, nor do you get a 2.5″ to 3.5″ cradle, cables or screws, which would have been much more useful.

20151214-1333-0525

The drive itself follows the same build as the PRO and indeed most later model SSDs. It has a metallic shell with a matte black lustre finish and a chromed beveled edge. The square on the top is silver, indicating EVO, whereas orange/red indicates PRO.

20151214-1333-0526

Power requirements are slightly less than the 850 PRO, being marked as 1.4A rather than 1.5A. The same proprietary screw is used to close the case, and therefore, no teardown will be performed.

20151214-1334-0527

And of course, it follows the regular screw hole patterning for a 2.5″ drive of 7mm height.

Testing and Benchmarking

Before I entrust my data to any device, I always test it thoroughly, and benchmarking just so happens to be part of the commissioning process. This drive was tested on the Asrock  Z97M-Extreme4 with overclocked Intel Pentium G3258 platform that the 850PRO was tested on earlier.

Samsung Magician

850-evo-magicianverified

The drive was verified as genuine and had the latest firmware installed. Another point of difference is that the 850 EVO appears to have 1000GB whereas the 850 PRO has 1000GiB of accessible storage. This is probably due to overprovisioning and the use of an SLC-like cache area for write-consolidation to improve endurance and write cache performance improvement. RAPID mode was not used throughout testing.

CrystalDiskInfo

850-evo-smart-before 850-evo-smart-post

As with most Samsung SSDs, the SMART attributes are well known, and provide good coverage of information related to the drive’s state, with the exception of total host reads perhaps. After testing, it can be seen that the drive was almost completely written twice.

HDTune Pro

14-December-2015_16-00 14-December-2015_14-17

The 850 EVO put in very solid figures for sequential access, scoring 519.8Mb/s average read and 499.1Mb/s average write. This is so close to the values returned by the 850 PRO that it seems the SATA III controller may have a greater impact on the observed performance.

14-December-2015_16-05

I checked the random access performance, although for some reason, it seems that the read performance results were not saved, leaving me with only this diagram of the write performance. Again, the performance is indistinguishable from the 850 PRO and in some cases, bests the 850 PRO slightly.

14-December-2015_16-07

In file tests, the drive really got into its stride for accesses of around 64kB and larger, and saw better IOPS figures than the 850 PRO.

14-December-2015_16-03 14-December-2015_16-04

Additional tests were done, although their results are probably more appropriate for hard drives than SSDs.

CrystalDiskMark

850-evo-cdm

The results from CrystalDiskMark indicate the drive has exceptionally good performance, trading blows with the 850 PRO. It’s not obvious that one is better than the other from a pure performance standpoint, although the use of TLC in the EVO is likely to reduce its endurance.

ATTO

850-evo-atto

ATTO seems to show that the drive is very strong in small-block performance with 8kB and larger seeing most of the drive’s performance with full performance by about 64kB accesses. This is the same as we see with the PRO and reflects the similarities in its controller and firmware optimization.

AS SSD Benchmark

850-evo-asssd

The timing issues that caused unrealistic benchmark values on the PRO do not appear with the EVO, hence the result is a much more realistic number. Because of the issues with AS SSD, I’m not inclined to trust the multi-threaded results as much, but they do reflect CrystalDiskMark results to some degree.

850-evo-asssd-comp

The compression benchmark shows no impacts of compressible data, as expected with modern SSDs.

850-evo-asssd-copy

The copy benchmark shows mixed results, again, trading blows with the PRO. The PRO has a significant lead on the first test case (ISO) but loses out by a smaller margin on the last two test cases (Program, Game). This may be due to the internal state of the SSD, but the differences are still minor. All timings were within 1s.

Anvil’s Storage Utilities

850-evo-anvil

Anvil seems to prefer the PRO with a score of 5301.27, compared to the EVO with 5199.15. The minor difference in score was more to do with the read performance, although the EVO did best the PRO on write performance on this benchmark.

H2testw

850-evo-h2tw

Just as with the 850 PRO, the drive achieved roughly 400Mb/s both ways with no corruption witnessed. From this, it seems ready for deployment.

Conclusion

Samsung currently dominates the SSD market, and it’s clear why that is the case. The 850 EVO offers performance parity with the 850 PRO at nearly half the price, and in fact, is closer to a performance SSD in its performance than it is to a “value” line SSD. At the price I paid, you would end up paying more to get a worse SSD – namely one with 300MB/s style speeds and no 3D V-NAND.

Furthermore, its use of 3D V-NAND offers a way around the low-endurance and data-loss slowdown issues around planar TLC operation, some of which made the 840 EVO much less desirable, without increasing the cost unreasonably. Free of those constraints, the 850 EVO is likely to offer sufficient endurance for very demanding users, and makes it a great choice for mainstream users. Seeing how well the 850 EVO performs, it’s hard to justify the premium for the 850 PRO.

About lui_gough

I'm a bit of a nut for electronics, computing, photography, radio, satellite and other technical hobbies. Click for more about me!
This entry was posted in Computing, Flash Memory and tagged , , , , , , . Bookmark the permalink.

4 Responses to Quick Review: Samsung 850 EVO 1Tb (MZ-75E1T0B) Solid State Drive

  1. Mark says:

    A word of warning about putting SSDs into an Apple Mac Mini (with OS X El Capitan):

    I recently put a Samsung 850 Pro into a mid-2010 Mac Mini (this also applies to the 2012 model). Wow! It made it a new machine. Boot time to a stable desktop went down by a factor of 10. I enabled the TRIM command using Apple’s trimforce utility. No problems there.

    After a month, I checked the detailed SMART attributes (using Samsung Magician in a Windows 7 bootcamp partition, you can also use an OS X utility like Smart Reporter). I noticed that in 400 hours of power-on usage the drive showed 20,000+ power cycles!!!

    Turns out when the Mac Mini goes to sleep, it effectively turns off power to the drive. But 20,000+ sleep cycles? No way. My system wakes up from sleep maybe 4 or 5 times a day. A little poking around in the console logs, I determined that my cheap (but very nice feeling) optical mouse was waking up the system… not a full wake-up, but enough to wake the drive. It looks like noise in the optical sensor was being interpreted as tiny mouse movements.

    The solution was to change the mouse. Also disabled the “sleep disk drives whenever possible” in system preferences.

    The bogus power cycles should not be a problem, except that Samsung SMART counts 100,000 power cycles as end-of-life. Not sure what happens when it gets to 100,000 cycles… and I don’t intend to find out.

    • lui_gough says:

      The SMART values are generally informative only when it comes to power cycles. It doesn’t really cause any wear on the drive in general, as long as the power cycles were done properly.

      In the case you are seeing the POR Recovery Count attribute start counting upwards, then it might be cause for concern, as this indicates improper shutdowns of the drive which has required a check of the internal flash-translation table and could have potentially (but not necessarily) resulted in data loss.

      It should be okay, but thanks for the heads up as I did just put the 850 PRO I had reviewed earlier into an iMac for a friend – we will keep an eye on that.

      – Gough

  2. Mark says:

    No, Samsung uses a separate SMART attribute (235) for unexpected power fails (like power cord yanks).

    SMART attribute 12 is their normal power cycle counter. With 20,000 power cycles in the raw data field, the attribute itself was at 80 (i.e. 80% of expected lifetime). I don’t know what happens when it gets to 0 and hope to never find out. There is a very good chance that the system and/or drive will report a failing drive. That might not lock up the system, but it would mask legitimate drive problems. OS X does not report detailed SMART status unless you use a third-part utility.

    My final config to mitigate the issue was to change the mouse and set the energy saving preferences to not “sleep the hard drive whenever possible”. I did not investigate the effects of the drive sleep energy saving preference in detail… it just seemed like a good thing to do. The mouse was the main source of power cycles to the drive.

    Disabling “wake on network access” reduced the power cycle countdown by a couple of times a day, but the system had to negotiate a new DHCP address when I woke it up in the morning.

    • lui_gough says:

      Then I don’t think you need to worry, as the Power Cycle Count attribute 0x0C (12 decimal) has a threshold to failure of 0, meaning it is not a pre-failure attribute as the value can never be less than the threshold of 0 (SMART attribute values, as opposed to the raw value, start positive and decline with age/declining health). It is a purely statistical/informational value, although I do agree it is incrementing alarmingly from what you have said. Eventually, it will get to zero, but the threshold is also zero. Some software may incorrectly claim you have a caution indication, but SMART warnings should not be issued unless the smart attribute value is strictly less than the threshold, termed a threshold exceed condition. See: http://www.easis.com/smart-value-interpretation.html and http://ma.juii.net/blog/interpret-smart-attributes

      The non-zero threshold values are all at 10 for attributes 0x05 Reallocated Sector Count, 0xb3 Used Reserve Block Count, 0xb5 Program Fail Count, 0xb6 Erase Fail Count and 0xb7 Runtime Bad Block and these should be the only attributes which can actually cause a pre-failure backup and replace warning.

      – Gough

Error: Comment is Missing!