Conclusion, Pricing and Final Thoughts
Conclusion:
Pros:
- Write speeds greatly improved(*) over prior generation Micron SSDs.
- Reduced write power consumption when in SLC mode (*).
- 16nm Micron flash should result in lower cost/GB.
- Very high endurance rating for a consumer / OEM SSD.
Cons:
- * Dynamic Write Cache performs inconsistently in actual use.
- * Worst case write speeds (while die shuffling) are painfully slow.
- Availability may be limited to OEMs.
Before wrapping up, I wanted to present a bit of a data dump for those wanting to directly compare the various write and read speeds seen through out this testing. Italicized figures are best guess approximations based on what we know of controller configurations that were not specifically tested:
SLC write | MLC write | TLC write | Mode change write | Duration | Read | Read (frag) | |
---|---|---|---|---|---|---|---|
M600 128 | 443 | 154 | 53 | Variable | 460 | 201 | |
M550 128 | 350 | 453 | 271 | ||||
MX100 128 | 150 | 450 | 201 | ||||
840 EVO 120 | 396 | 128 | 3GB | 444 | 444 | ||
M600 256 | 466 | 305 | 77 | Variable | 458 | 266 | |
M550 256 | 464 | 451 | 355 | ||||
MX100 256 | 300 | 450 | 266 | ||||
840 EVO 250 | 485 | 253 | 3GB | 459 | 459 | ||
M600 512 | 467* | 424* | 70* | Variable | 450 | 413 | |
M550 512 | 464 | 450 | 382 | ||||
MX100 512 | 462 | 447 | 382 | ||||
840 EVO 500 | 485 | 400 | 6GB | 461 | 461 | ||
M600 1TB | 467* | 424* | 56* | Variable | 449 | 413 | |
M550 1TB | 464 | 450 | 382 | ||||
840 EVO 750 | 485 | 400 | 9GB | 460 | 460 | ||
840 EVO 1TB | 484 | 404 | 12GB | 456 | 456 |
For those looking for specific write / read speed minimums for Windows-based file copy operations, the above table should answer any questions. The left three columns represent observed write speeds of the various models. 'Mode change write' is the worst-case write speed seen when the M600 is caught freeing up flash memory space by shifting data and dies from SLC to MLC. 'Read (frag)' is the speed observed when reading back a test file that was in-place fragmented (more detail at the bottom of page 4).
* Edit* Although the 512GB and 1TB 2.5" SATA models of the M600 are not supposed to support DWA, there were definite speed grades / inconsistency noted during testing (here), so I'll leave the chart as is, but star those results to indicate the potential disparity. Micron also states that the MSATA and M.2 versions of the M600 support DWA in capacities up to 512GB. We're not sure why all M600's don't come with DWA given Micron's push towards lower power consumption when the flash is operating in SLC mode.
Pricing:
No MSRP's here as this is an OEM targeted product. Approximate costs (not adjusted for high volume shipment):
- 128GB – $80 ($0.63/GB)
- 256GB – $140 ($0.55/GB)
- 512GB – $260 ($0.51/GB)
- 1TB – $450 ($0.44/GB)
Final Thoughts:
The Micron M600 SSDs were certainly challenging to properly evaluate. The dynamic flipping from SLC to MLC flash meant we had to come up with an entirely new routine based on evaluating performance consistency as test samples are filled to capacity. Once we worked out a realistic routine and applied it to a round of competing samples, we were a bit disappointed by the inconsistent results seen from Micron's first stab at this challenging and apparently tough to execute feature. The new testing method also revealed some additional shortcomings of the Marvell controller not seen in competing SSDs. All of that said, the M600 should bring a competitive cost/GB to OEMs who choose to include it in their system builds. As for my recommendation, whether or not it deserves a place on your 'avoid this laptop' list really depends on your tolerance for potentially inconsistent write speeds, coupled with how much you intend to fill your SSD. On paper, Dynamic Write Acceleration is a novel approach, but its current implementation doesn't appear to best the competing static SLC cache solution.
If there are more reviews
If there are more reviews like this where people are not able to get their heads around the Micron controller concept, they should simply release the successor to the MX100 line with their low cost standard controller (upgraded of course). This would become the go to SSD for millions. A consistent 256GB SSD for $80 sounds much better than the new dinky M600 for anything.
The M600 looks like a lemon to me at the moment.
There’s the rub. Testing in
There's the rub. Testing in this manner revealed that the MX100 has issues as well – just different ones. See the bottom of page 4 for details and explanation.
Makes one wonder if the
Makes one wonder if the marvel controller’s quirk is exclusive to the 88SS9189. I know sandisk uses previous revisions of the controller in their ssd’s.
Different companies, and different firmwares though. Probably not likely.
I’m a SSD neophyte, my
I’m a SSD neophyte, my primary usage: Photoshop, Lightroom, Audio recording, (minimal video)
I’m going to replace my 1TB Boot HD with a 1/2TB SSD (480,500,512). I’m leaning to the Crucial M550 over the M100 (only $20>), some say the M550 “is built for heavier use”. (?) I was looking at the Samsung but not after Twits “Padre SJ” and this review discuss slowdown issues.
Do the M550’s have the any slowdown issues? Or is this only the M600 due to the different/new controller?
Allyn M. talked about the M550 on July 25, 2014. (no “review”)
Q: Are the potential specs of the M600 series worth waiting for it to come out, or should I just pull the trigger on the M550 and stop waiting?
Thanks,
Dokk
ALLYN
A SANDISK ULTRA 2-Thru
ALLYN
A SANDISK ULTRA 2-Thru the same tests would be a great addition,
as the third variation of this tech……………..
The Sandisk Ultra II drive
The Sandisk Ultra II drive uses the Marvell 88SS9187 instead of the 88SS9189 controller and uses different firmware. So in my opinion it’s probably doubtful. Gonna take some months to also test whether or not sandisk figured a way around the leaky tlc problem.
My info tells me Sandisk is
My info tells me Sandisk is using-
9190-4ch for 120 and 240 GB drives,
and 9189 for larger drives……..
But it’s the tech i would like to see compared.
Sammy has a static cache,
Micron is using dynamic,
Sandisk is using on chip copy……………
Hmmm on closer inspection it
Hmmm on closer inspection it does seem that Sandisk likes to variate which Marvell controller is used on a drive or even capacity basis.
Example, the sandisk x300s drive uses the 9189 controller.