The Samsung 860 PRO (512GB And 4TB) SSD Review: Replacing A Legend
by Billy Tallis on January 23, 2018 10:00 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 512GB Samsung 860 PRO has the fastest burst random read speed among these SATA SSDs, about 5% faster than the 850 PRO. The 4TB model is the same speed as the 4TB 850 EVO.
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 involving some higher queue depths, the Samsung 860 PROs take a clear lead, and the 4TB model even outperforms the PM981 NVMe SSD.
The two Samsung 860 PROs offer the same power efficiency, which is a huge step up from the 850 PRO's efficiency and significantly better than any of the competition.
 |
|||||||||
Most of the Samsung drives hit the same limit of almost 400 MB/s around QD16. The two latest Crucial SSDs only just reach reach that level of performance at QD32, while the Intel 545s and Crucial MX300 never scale up that far. The PM981 NVMe drive scales far beyond any of the SATA drives, but at the cost of very poor power efficiency, especially at lower queue depths.
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.
The 4TB 860 PRO has the fastest burst random write speed, while the 512GB model scores slightly worse than the 512GB 850 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.
The sustained random write performance of the Samsung 860 PRO is a very slight improvement over their previous drives. Most of the competition is significantly slower on this test, but the Crucial BX300 is pretty close.
The Samsung 860 PROs are again the two most efficient SATA SSDs, and the 512GB model manages to match the efficiency of the much faster but more power hungry PM981.
 |
|||||||||
The 860 PRO reaches full random write speed at QD4 and is steady through the rest of the test. The Crucial MX500 and Intel 545s are much slower to get up to speed, and the SanDisk Ultra 3D plateaus at a mere 200MB/s after QD2.
Facebook
Twitter
RSS
64 Comments
View All Comments
Flunk - Tuesday, January 23, 2018 - link
I agree, the low price makes the mx500 a really good buy. It certainly qualifies as "fast enough" while delivering very low cost/GB.GreenMeters - Tuesday, January 23, 2018 - link
Looking like it. The 850 EVO was on sale over Black Friday for less than the MX500 (at 1 TB size anyway), but outside big discounts like that (and assuming there's no simultaneous discount of the MX500) then it looks like Samsung is about to be irrelevant when it comes to SATA. Disappointing in some ways (have four 850 EVOs in various systems now, two of them picked up at the aforementioned sale, and they've been great) but as long as PCI is becoming more affordable I guess it's not a big deal.Alistair - Tuesday, January 23, 2018 - link
You are comparing with the launch price of the MX500. I've bought 5 x MX500 1TB drives for $242 USD each. I'm pretty sure the Samsung 850 is more expensive.Samus - Wednesday, January 24, 2018 - link
If you are going to compare sale prices of Samsung drives to the competition, it becomes even more obvious Samsung is a bad buy when you see the sale prices of competitors. The BX300 256GB drives were on sale for $70 at one point. No Samsung 250GB drive has been under $90 in over a year, even on sale.bug77 - Wednesday, January 24, 2018 - link
SATA does not prevent performance improvements. 4k random reads are what give a drive its speed for home usage and we're not even at 100MB/s in this aspect. Plenty of room for improvement right there.In other news, if you have ~$250 to spend, you can either get a 512GB 850 Pro or a 1TB MX500. Imho, as good as Samsung is, there's no contest here.
zodiacfml - Sunday, February 11, 2018 - link
I agree. The only limit is in sequential. However, we have seen the performance/capabilities of the Intel Optane drives and that even that doesn't improve a desktop experience by any noticeable level.Round - Wednesday, January 24, 2018 - link
Sorry, but I disagree. What's so impressive about this, because I'm not seeing it? They improved the power spec, but for real world use, especially in a desktop, I'm just not seeing any benefit at those prices.I can't see buying any more Samsung drives (I have 6 850 Evos) or recommending them to anyone. The price/performance from Crucial is superior, and I doubt anyone is ever going to notice a performance difference between the MX500 and 850/860 Evos (the 860 Pro is priced ridiculously high and is not a wise purchase for any average user).
I find myself hoping Samsung gets punished in the market place....
StrangerGuy - Wednesday, January 24, 2018 - link
Yup, I fail to see whats so great about this SSD either for consumers either. MX500 beats it in 4K random IOPS while having a much higher GB/$. The extra endurance and warranty length is also completely irrelevant for 99.99% of consumers out there; I myself have a Crucial M550 1TB since 2014 and I still only have 11TB total writes on it.BurntMyBacon - Wednesday, January 24, 2018 - link
While I don't find this drive particularly impressive (not much room to impress on SATA anymore), it does have the distinction of likely being the last MLC drive available on SATA. While normal consumers can (in theory) use TLC drive with no negative effect, there are cases of people who have experienced a TLC SSD failure and aren't too eager to get another. I've personally been involved with 7 TLC SSD failures (3 different Crucial models, 2 Sandisks, and 2 Samsung). While the NAND was not likely responsible for any of these failures and this makes up a pretty low percentage of total TLC SSDs deployed within my purview, it does start to leave a less reliable image when compared to the zero MLC SSD failures (Crucial, Intel, Samsung, Sandisk, Corsair, etc.) I've seen in my client base. Granted, this is all anecdotal and I use global data (including HDD vs SSD failure rates) to color my recommendations. However, clients who've experienced the drive failures have universally decided that TLC was not an option for them. I haven't sworn off TLC drives personally, but entirely coincidentally, I have yet to purchase one since I burnt one out under heavy load (improper cooling on the controller I believe).chrone - Tuesday, January 23, 2018 - link
Could you guys perform synchronous write test in Linux as well?```
dd if=/dev/zero of=/tmp/test bs=4K count=100 oflag=direct,sync status=progress
dd if=/dev/zero of=/tmp/test bs=4M count=100 oflag=direct,sync status=progress
```
Sadly, the synchronous write for older Samsung SSD 850 Pro is similar to HDD. Synchronous write are used by OS and app for data consistency and reliability in Linux environment.