The Samsung PM981 SSD Review (512GB, 1TB): Next Generation Controller And 3D NAND
by Billy Tallis on November 30, 2017 9:02 AM ESTRandom Read Performance
Our first test of random read performance uses very short bursts of operations issued one at a time with no queuing. The drives are given enough idle time between bursts to yield an overall duty cycle of 20%, so thermal throttling is impossible. Each burst consists of a total of 32MB of 4kB random reads, from a 16GB span of the disk. The total data read is 1GB.
The burst random read performance of the Samsung PM981 is great by the standards of TLC SSDs, but is surpassed by several MLC-based drives, including the Phison E7-based Patriot Hellfire with planar MLC NAND.
Our sustained random read performance is similar to the random read test from our 2015 test suite: queue depths from 1 to 32 are tested, and the average performance and power efficiency across QD1, QD2 and QD4 are reported as the primary scores. Each queue depth is tested for one minute or 32GB of data transferred, whichever is shorter. After each queue depth is tested, the drive is given up to one minute to cool off so that the higher queue depths are unlikely to be affected by accumulated heat build-up. The individual read operations are again 4kB, and cover a 64GB span of the drive.
On the longer random read test that includes some higher queue depths, the PM981 comes a bit closer to the standard set by Samsung's MLC drives, and it outperforms all the non-Samsung drives.
 |
|||||||||
Both capacities of the PM981 show performance scaling with queue depth in the typical manner for a high-performance drive, though the 512GB model has passed an inflection point by QD32 and is approaching saturation.
Random Write Performance
Our test of random write burst performance is structured similarly to the random read burst test, but each burst is only 4MB and the total test length is 128MB. The 4kB random write operations are distributed over a 16GB span of the drive, and the operations are issued one at a time with no queuing.
There are a few MLC-based SSDs that offer substantially higher burst random write performance than the Samsung PM981, but it is on par with most high-end drives including the Samsung 960 PRO.
As with the sustained random read test, our sustained 4kB random write test runs for up to one minute or 32GB per queue depth, covering a 64GB span of the drive and giving the drive up to 1 minute of idle time between queue depths to allow for write caches to be flushed and for the drive to cool down.
On the longer random write test, the 1TB PM981 stands out with clearly higher performance than the Samsung 960 series could manage. The 512GB PM981 is slower but still definitely performing like a high-end drive.
 |
|||||||||
The random write performance of the 1TB PM981 scales very well with increasing queue depth. As compared to the Samsung 960 series, it also reaches its plateau around QD8, but is providing much higher throughput by that point. The 512GB model runs out of SLC cache during portions of this test so its performance is much lower and less steady.
Facebook
Twitter
RSS
53 Comments
View All Comments
mapesdhs - Thursday, November 30, 2017 - link
And Drazick, what do you mean by 2.5" drives? If you're referring to SATA, well then no, it's already at its limit of 550MB/sec, and producing something akin to SATA4 would be pointless when it's also hobbled by the old AHCI protocol.Also, "don't like" is an emotional response; what's your evidence and argument that they're a bad product somehow? Have you used them?
WithoutWeakness - Thursday, November 30, 2017 - link
By 2.5" drives I'm sure he means the same form factor as standard SATA 2.5" SSDs except using a newer, faster connection just like the U.2 connectors that Dan mentioned. We definitely hit the limit of what SATA 3 can deliver and it would be nice to have a new standard that can leverage PCIe NVMe SSDs in a form factor that allows us to use cables to put drives elsewhere in a case for better layouts and airflow. U.2 was supposed to be that connector but there are basically no drives that support the standard and very few boards with more than 1 U.2 port. There are a few adapters on the market that allow you to install an M.2 drive into a 2.5" enclosure with a U.2 connector on it but until motherboards have more than 1 U.2 port it won't be a real replacement for the ubiquity of SATA.msabercr - Friday, December 1, 2017 - link
Actually there are m.2 to U.2 connectors readily available from most MB vendors, and 7mm U.2 datacenter drives are starting to become a thing. See Intel SSD DC P4501. I wouldn't be surprised if AIC disappears after too long. Limiting the power draw would be the major hurdle in creating such drives but it's not impossible. The EDSFF is going to pave the way for many high density compact form factors for NVMe moving forward.sleeplessclassics - Thursday, November 30, 2017 - link
One more thing which I think will be different when these drives are launched as retail devices is the driver support for Phoenix controller. While, it is always difficult to pinpoint the exact bottlenecks on such bleeding edge technology, I think a driver that is better optimized for Phoenix controller will definitely produce better results (ceteris paribus)Also, there have been rumors of QLC-Nand. If that is true, that could be the differentiator between EVO and PRO series.
romrunning - Thursday, November 30, 2017 - link
Yes - QLC... more latency, lower endurance, slower writes - what's not to like? :-SSpunjji - Thursday, November 30, 2017 - link
Lower price..? Higher densities and increased production? That's what it's all about.If 3D QLC performs like 2D TLC then it'll do just fine for mass storage.
mapesdhs - Thursday, November 30, 2017 - link
Good point given the way in which most products seem to be abe to tolerate far more writes than for which they're officially rated, in which case it's likely most users will want something newer long before a QLC product's endurance has been reached. If one is doing something that will drain the endurance a lot faster, then one should be using something more suitable anyway.romrunning - Thursday, November 30, 2017 - link
Sure, but QLC is just like TLC - once you force it on enough people and you say it's "good enough", then the higher-performing but costlier flash (like SLC/MLC) slowly is removed from the product portfolio. I'm not in favor of these race-to-the-bottom "advances", just to reduce the price a bit for hte consumer but more for the mfg. You may get a slight bump in capacity, but for me, the performance/endurance trade-off with a slight reduction in price isn't worth it.Now, I suppose it doesn't matter anymore to me since I'll still be buying the 960 Pro until the Optane 900p reaches better pricing. But the slippery slope you encounter is that new product "advances" are usually better when comapred to to the "current" state of tech. If the current standard is QLC, then the new "improvement" might only be raising it to levels that SLC/MLC were at previously. So the possibility is that it may not be that much of an improvement.
bcronce - Thursday, November 30, 2017 - link
For read heavy mass storage drives, slower writes is fine. SSDs are getting fast enough that the IO or CPU is the bottleneck. Higher read latency for small queues will hurt performance, but not by a whole lot.The endurance is only an issue if you re-write your data a lot, like a paging file or a game drive that sees a lot of updates. A relatively static mass-media drive will probably be just fine.
sleeplessclassics - Thursday, November 30, 2017 - link
Latency can (till some extent) be handled with a bigger dram buffer. Also, controllers are the key here and not the NAND type. Today, even TLC can perform better than MLC/SLC just 2-3 generations ago due to better controllers.A couple of years ago and even last year, 500GB ssd was around $80. If the prices were sane, 64-layer 3D TLC would have been below $50 for sure.
And 96-layer QLC can give real competition to the HDDs.
As for lower endurance, that can be handled by slightly higher provisioning and slower writes....well they would be okay for 95% of the mainstream users.
Enthusiasts have optane and Z-NAND