The Samsung SSD 850 EVO mSATA/M.2 Review
by Kristian Vättö on March 31, 2015 10:00 AM ESTRandom Read Performance
One of the major changes in our 2015 test suite is the synthetic Iometer tests we run. In the past we used to test just one or two queue depths, but real world workloads always contain a mix of different queue depths as shown by our Storage Bench traces. To get the full scope in performance, I'm now testing various queue depths starting from one and going all the way to up to 32. I'm not testing every single queue depth, but merely how the throughput scales with the queue depth. I'm using exponential scaling, meaning that the tested queue depths increase in powers of two (i.e. 1, 2, 4, 8...).
Read tests are conducted on a full drive because that is the only way to ensure that the results are valid (testing with an empty drive can substantially inflate the results and in reality the data you are reading is always valid rather than full of zeros). Each queue depth is tested for three minutes and there is no idle time between the tests.
I'm also reporting two metrics now. For the bar graph, I've taken the average of QD1, QD2 and QD4 data rates, which are the most relevant queue depths for client workloads. This allows for easy and quick comparison between drives. In addition to the bar graph, I'm including a line graph, which shows the performance scaling across all queue depths. To keep the line graphs readable, each drive has its own graph, which can be selected from the drop-down menu.
I'm also plotting power for SATA drives and will be doing the same for PCIe drives as soon as I have the system set up properly. Our datalogging multimeter logs power consumption every second, so I report the average for every queue depth to see how the power scales with the queue depth and performance.
Random read performance has always been Samsung's strength and particularly the 500GB and smaller capacities do well thanks to the faster MGX controller.
Power consumption is also good, although the 1TB model sips quite a bit of power.
 |
|||||||||
The performance scales nicely with the queue depth too.
Random Write Performance
Write performance is tested in the same way as read performance, except that the drive is in a secure erased state and the LBA span is limited to 16GB. We already test performance consistency separately, so a secure erased drive and limited LBA span ensures that the results here represent peak performance rather than sustained performance.
Random write performance is equally strong, which is mostly thanks to TurboWrite.
Power consumption is decent as well, and while the larger capacities are more power hungry the difference to competing drives isn't substantial.
 |
|||||||||
Since the 120GB SKU has less parallelism due to having less NAND, its performance doesn't scale at all with the queue depth (QD1 is already saturating the available NAND bandwidth), but the other models scale pretty nicely. You do see a slight drop in performance after the TurboWrite buffer has been filled, but in client workloads it's unlikely that you will be filling the buffer at once like our tests do.
Facebook
Twitter
RSS
58 Comments
View All Comments
Flunk - Tuesday, March 31, 2015 - link
This is what makes M.2 such an annoying standard. They tried to accommodate everything and ended up with compromises that don't make sense and will probably be written out of the standard in a future version.setzer - Tuesday, March 31, 2015 - link
Also don't forget about the single and double sided thing as noted in the article there are some laptops that only accept single-sided.Also there is nothing to prevent a manafucturer to put a B+M keyed M.2 socket but only connect the USB traces. See toshiba's Z30's laptops for a pratical example.
The joys of M.2 are great :P
ilkhan - Wednesday, April 1, 2015 - link
Answer: Ports should be wired and keyed for sata and pci-e.devices can be whatever they need.
The keys are there to prevent a pci-e device in a sata host.
rtho782 - Tuesday, March 31, 2015 - link
I still don't see a reason to replace my ageing 256GB Samsung 830s in RAID 0.I really want a decent PCIe NVMe M.2 or SATAe SSD of about 500GB, preferably Samsung and 3D nand. But nothing :(
MrCommunistGen - Tuesday, March 31, 2015 - link
Looks like the 500GB model is the performance sweet spot.I'm not that surprised with the different performance profile on the 1TB model since it is using the older MEX controller. Could the 1TB's stuttering under steady state load be due to thermal throttling of the controller?
I was not expecting the smaller capacity drives, particularly the 120GB model to have such (relatively) low performance. Still, compared to drives of yesteryear, performance is still quite good. My HTPC has an old 96GB Kingston V+100 but still feels pretty snappy. I'm sure that even the 120GB 850 Evo would run circles around that drive - and as such have plenty of performance for an average user.
sonicmerlin - Friday, April 3, 2015 - link
Ha I have that exact same Kingston drive in my desktop. I can only install like 1 or 2 games at once, but it's totally worth it. I doubt any SSD upgrades would make my computer feel even faster than it already is.Mrduder11 - Tuesday, March 31, 2015 - link
I can't remeber where I read it but should we be concerned about these drives getting too hot where it affects performance?Mecharon1 - Tuesday, March 31, 2015 - link
Is this drive bootable? More specifically, can I install my OS on the 120GB M.2 version and use something else for bulk storage?foxtrot1_1 - Tuesday, March 31, 2015 - link
That depends on the motherboard, but Windows 8.1 and Windows 10 should allow you to boot from M.2 no problem. Your BIOS is the issue.This is a golden age for PC hardware (at least, it will be this fall) but the proliferation of specifications and standards is really stupid. Get your act together, OEMs.
Kristian Vättö - Wednesday, April 1, 2015 - link
SATA is always bootable regardless of the form factor and OS, and the 850 EVO is a SATA drive (M.2 supports both SATA and PCIe). The bootability issue only applies to PCIe M.2 drives.