Conclusion

The Intel SSD 670p is a great update to their consumer QLC product line. Intel hasn't eliminated the downsides of QLC NAND compared to TLC NAND, but they are continuing to reduce those downsides. With the 670p, peak performance has caught up to and in some cases clearly surpassed the performance of the QLC SSDs using the Phison E12 SSD controller. Worst-case performance has improved to the point that the larger models of 670p will no longer end up slower than hard drives during a torture test. Write endurance has increased again, and is now high enough that most consumers can stop worrying about whether QLC NAND will last long enough.

During ordinary consumer use and even some fairly heavy workloads, there won't be any of the performance problems that used to be a dead giveaway that a drive was using QLC NAND. The corner cases where performance plummets still exist, but they are getting harder to trigger with each generation. The most significant remaining performance downside to QLC is that random reads that cannot be served from the SLC cache, and in this instance they will be slower than a good DRAMless TLC drive. But even that weakness has to be put in context: the read latency outside of the SLC cache is still almost as fast as random reads from a TLC SATA SSD.

The big caveat to our conclusions is that we only tested the largest and fastest 670p model. The 512 GB model is doubtless faster than the 512 GB Intel 660p, but we still don't recommend any QLC drive smaller than 1TB. We're also not sure how compelling the performance improvements with the 1TB 670p will be: on paper it's slower than the 2TB model we tested, but not by much—especially for the benchmarks that are more likely to matter in real life.

The new 3D QLC NAND and new SSD controller introduced in the Intel 670p are both interesting from a technical standpoint. The combination works well, but both NAND and controller face an uncertain future - Intel is selling its NAND flash and SSD business to SK hynix. Intel is the last remaining NAND flash memory manufacturer using a floating gate memory cell design, which makes this 144L QLC unique within the industry. That also means SK hynix might easily decide to abandon this line of R&D and focus on the more popular charge trap flash the next time money gets tight.

The new Silicon Motion SM2265 controller provides greatly improved performance than the 660p/665p badly needed, but the SM2265 appears to be an awkward product that straddles two generations. Without even token PCIe 4.0 support, the SM2265's days are numbered. The advertising advantages of PCIe 4.0 make it very tempting to upgrade to the SM2267 controller when suitable drives are available. The SM2265 as used in the Intel 670p also fails to deliver anything like the impressive power efficiency advantages we saw with the SK hynix Gold P31 and its high-speed 4-channel controller.

NVMe SSD Price Comparison
March 1, 2021
  240-256 GB 480-512 GB 960 GB-1 TB 2 TB
Intel 670p
QLC
Launch
MSRP
  $89.99 (18¢/GB) $154.99 (15¢/GB) $329.99 (16¢/GB)
Updated
March 2
  $69.99 (14¢/GB) $129.99 (13¢/GB) $249.99 (12¢/GB)
Intel 660p
QLC
    $109.99 (11¢/GB) $224.99 (11¢/GB)
ADATA Falcon
TLC, DRAMless, 8ch
$37.99 (15¢/GB) $57.99 (11¢/GB) $99.99 (10¢/GB) $209.99 (10¢/GB)
Crucial P2
TLC, DRAMless
$49.99 (20¢/GB) $59.99 (12¢/GB) $104.99 (10¢/GB)  
Kingston A2000
TLC
$44.99 (18¢/GB) $60.54 (12¢/GB) $116.24 (12¢/GB)  
WD Blue SN550
TLC, DRAMless
$46.40 (19¢/GB) $59.99 (12¢/GB) $104.99 (10¢/GB) $224.99 (11¢/GB)
Inland Platinum
QLC, 8ch
    $96.99 (10¢/GB) $181.99 (9¢/GB)
Crucial P1
QLC
  $55.99 (11¢/GB) $104.99 (10¢/GB) $224.99 (11¢/GB)
Sabrent Rocket Q
QLC, 8ch
  $62.42 (12¢/GB) $109.98 (11¢/GB) $219.98 (11¢/GB)
         
ADATA S50 Lite
TLC, PCIe 4.0, 4ch
    $119.99 (12¢/GB) $229.99 (11¢/GB)
ADATA SX8200 Pro
TLC, 8ch
$44.95 (18¢/GB) $59.99 (12¢/GB) $119.99 (12¢/GB) $239.99 (12¢/GB)
HP EX950
TLC, 8ch
  $79.99 (16¢/GB) $124.99 (12¢/GB) $234.99 (12¢/GB)
SK hynix Gold P31
TLC, 4ch
  $74.99 (15¢/GB) $134.99 (13¢/GB)  
WD Black SN750
TLC, 8ch
$49.99 (20¢/GB) $62.99 (13¢/GB) $129.99 (13¢/GB) $293.73 (15¢/GB)

The retail consumer SSD market is a bit of a mess right now: the semiconductor shortages are being felt here as well, and much of what is in stock is starting to creep up in price. Even so, the recommended customer prices Intel is launching the 670p with are way out of touch.

They're positioning the 670p in a higher price bracket than the QLC SSDs with the Phison E16 PCIe 4.0 controller. That also leaves a lot of room for reasonably-priced TLC SSDs to undercut the 670p - and not just the budget models. Basically anything significantly cheaper than a Samsung 970 EVO Plus is also cheaper than the 670p's recommended customer pricing.

Once the Intel 670p comes down to sane price ranges, it certainly has the potential to be a good product. It helps raise the bar for QLC SSDs and entry-level NVMe SSDs in general. The 670p probably can't quite come down in price far enough to match the 660p, but if it got close then it could be a very compelling bargain.

Update March 2, 2021: That didn't take long. Retail prices for all three capacities of the Intel SSD 670p have dropped considerably. It's still a bit expensive for a QLC SSD, but the 670p delivers better real-world performance than any previous QLC SSD. The updated prices are a lot more reasonable, especially for a just-launched product. The 670p is now priced to match most mainstream TLC SSDs with 8-channel PCIe gen3 controllers, which is the performance class the 670p competes in for most real-world workloads.

Mixed IO Performance and Idle Power Management
Comments Locked

72 Comments

View All Comments

  • Wereweeb - Tuesday, March 2, 2021 - link

    What are you smoking? Four bits per cell is indeed 33% more bits per cell than three bits per cell.
  • Bp_968 - Tuesday, March 2, 2021 - link

    He's smoking "math". Lol. 3 bits per cell is 8 voltage states, and 4 bits per cell is 16 voltage states, which is double. If your going to comment with authority on an advanced subject you should learn the basics, and binary math is one of the basics.
  • bug77 - Tuesday, March 2, 2021 - link

    Yeah, it doesn't work like that.
    3 bits is 3 bits. They hold 8 possible combinations, but they're independent of each other.
    4 bits is 33% more than 3.
  • Billy Tallis - Tuesday, March 2, 2021 - link

    3 bits per cell is 8 POSSIBLE voltage states, but any given cell can only exist in one of those states at a time. The possible voltage states are not the cell's data storage capacity. The number of bits per cell is the cell's data storage capacity.
  • FunBunny2 - Tuesday, March 2, 2021 - link

    "3 bits per cell is 8 POSSIBLE voltage states, but any given cell can only exist in one of those states at a time."

    which incites a lower brain stem question: does NAND and/or controllers implement storage with a coding along the lines of RLL or s/pdif (eliminate long 'strings' of either 1 or 0) in order to lower the actual voltages required? if only across logically concatenated cells, so 1,000,000 would store as 1,00X where X is interpreted as 4 0? I can't think of a way off the top of my head, but there must be some really smart engineer out there who has?
  • code65536 - Tuesday, March 2, 2021 - link

    Um, voltage states are not storage--it's instead a measure of the difficulty of storing that many bits. QLC is 4 bits per cell, and needs to be able to discern 16 voltage states to store those 4 bits. TLC needs to discern only 8 stages in order to store 3 bits. What that means is that QLC stores 33% more data at the expense of 100% more difficulty. Each bit added doubles the difficulty of working with that data. SLC->MLC was 100% more difficulty for 100% more storage. MLC->TLC was 100% more difficulty for 50% more storage. TLC->QLC was 100% more difficulty for 33% more storage. And QLC->PLC will be 100% difficulty for only 25% more storage.
  • Spunjji - Thursday, March 4, 2021 - link

    I think at this stage it's also worse than double the difficulty - the performance and endurance penalties are very, very high.
  • HarryVoyager - Monday, March 8, 2021 - link

    The 660p initially showed a significant price advantage; I was able to get a 2TB at $180, but it has since disappeared.

    That said, in day to day practical use, I haven't seen much difference between a 860 Pro, the 660p and an XPG Gamix X7.

    All of them have been considerable faster than my harddrives, and pretty much all of them can feed data faster than my CPUs or network can process it.

    I know at some point that will change, and we will see games and consumer software designed to take advantage of the sort of data rates that NVME SSDs can provide, but I'll likely get a dedicated NVME drive for that, when that day comes.
  • RSAUser - Tuesday, March 2, 2021 - link

    No one who looked at the actual daily write warranty said that.

    I've never had a TLC drive file as a host OS drive, only the two times I bought a QLC one after two years.

    So my Motto is TLC for host, QLC for mass storage.
  • yetanotherhuman - Tuesday, March 2, 2021 - link

    Nope, in my mind TLC is still a cheap toy, fit for a less important machine or maybe a games drive. MLC, 2-bit per cell, is still the right way to go, and QLC is so shit that it should be given away in cereal boxes.

Log in

Don't have an account? Sign up now