The Intel SSD 660p SSD Review: QLC NAND Arrives For Consumer SSDs
by Billy Tallis on August 7, 2018 11:00 AM ESTAnandTech Storage Bench - Light
Our Light storage test has relatively more sequential accesses and lower queue depths than The Destroyer or the Heavy test, and it's by far the shortest test overall. It's based largely on applications that aren't highly dependent on storage performance, so this is a test more of application launch times and file load times. This test can be seen as the sum of all the little delays in daily usage, but with the idle times trimmed to 25ms it takes less than half an hour to run. Details of the Light test can be found here. As with the ATSB Heavy test, this test is run with the drive both freshly erased and empty, and after filling the drive with sequential writes.
As with the Heavy test, the SLC cache of the Intel SSD 660p is extremely beneficial and brings the average data rate of the 660p up into high-end NVMe territory. When the drive is full and the SLC cache has been reduced to its minimum size, performance suffers and drops below the Crucial MX500 mainstream SATA drive but not all the way down to the level of the Toshiba RC100 DRAMless NVMe SSD.
The best-case latency scores from a freshly-erased 660p are acceptable for a high-end NVMe SSD and excellent for an entry-level drive. In the worst case of a full drive, the average latency is far higher but still low enough that the drive won't actually feel much slower. The 99th percentile latency climbs very high by SSD standards, but is still barely up into the average latency range of hard drives.
On the Light test, the average read latency of the 660p stays comfortably below that of SATA drives even for the worst-case full drive test run, and only the average write latency shows a serious problem from filling up the whole drive and not giving it enough time to empty the now-reduced SLC cache.
The full-drive 99th percentile write latency of the 660p on the Light test is almost as bad as the 600p or the Toshiba RC100. Otherwise, the 660p doesn't have any worrying QoS problems on this test and users won't notice serious pauses from the drive.
The energy usage of the 660p on the Light test is below most other NVMe drives when the test is run on an empty drive, and even with the extra background work and longer test duration of the full-drive test run the 660p is only a little less efficient than the average for this bunch of NVMe SSDs.
Facebook
Twitter
RSS
86 Comments
View All Comments
dromoxen - Friday, August 10, 2018 - link
You would hope these things would have even larger dram buffers than tlc. I will pass on these 1st gen and stick with with HD.Has intel stopped making ssd controllers?
To do some tests , write endurance, why not cool down the m.2 nand to LN2 temps, I'm sure debauer has some pots and equipment. I expect these will be even cheaper by jan 19
tomatotree - Tuesday, August 14, 2018 - link
Intel makes their own controllers for all their enterprise drives, and all 3DXP drives, but for consumer NAND drives they use 3rd party controllers with customized firmware.As for LN2 cooling, what would that show? That the drive might fail if you use it in a temperature range way out of spec?
351Cleveland - Monday, August 20, 2018 - link
I’m confused. Why would I buy this over, say, an MX500 (my default go-to)? This thing is a dog in every way. How can Anandtech recommend something they admit is flawed?icebox - Thursday, December 6, 2018 - link
I don't understand why everybody fusses about retention and endurance so much. Do you really buy ssd's to leave them on a shelf for months or years? Retention ? If it dies during warranty you exchange it. If it dies after it then it's probably slow and small in comparison with what's available than.You do have backups, right? Because no review or test or battery of tests won't guarantee that *your drive* won't die.
BTW that's the only way I saw ssd's die - it works perfectly and after a reboot it's gone, not detected by the system.
icebox - Thursday, December 6, 2018 - link
The day has come when choosing storage is 4 tiered.You have fast nvme, slow nvme, sata ssd's and traditional hdd's. At least I kicked hdd's off my desktop. I have a samsung nvme for boot and applications and sata ssd's for media and photos. Now I'm looking of replacing those with the 2tb 660p and moving those to the nas for bulk storage.
southleft - Tuesday, May 14, 2019 - link
It would be very helpful if the review would show just how full the drive can be before performance degrades significantly. Clearly, when the drive is "full" its performance sucks, but can we expect good performance when the drive is half-full, two-thirds full, three-quarters full? C'mo, Anandtech, tell us something USEFUL here!boozed - Monday, December 30, 2019 - link
There's something wrong with the 970 EVO's results on page 3. Full performance exceeds empty performance. This is not reflected in the 970 EVO review.