Realistic differences between Crucial MX200 and Samsung 850 Pro?
- Thread starter ItsChrisG
- Start date
Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
Micron/Crucial makes Eco drives. Good burst performance, interesting long term performance. If you don't beat on the drives much and let them sit idle often enough to run trim or gc you should be fine. But any sort of demanding workload requires better drives. Throughput may be good from short tests... but latency kills. Throughput also suffers in longer runs.
Even in short runs the terrifyingly bad micron fw gc routines show a high max latency, in line as does samsungs bottom tier evo drive.
What workload are you planning on? Lightly used enterprise SSDs can be had on ebay for cheap and will offer a much better stable performance than consumer drives.
Even in short runs the terrifyingly bad micron fw gc routines show a high max latency, in line as does samsungs bottom tier evo drive.
What workload are you planning on? Lightly used enterprise SSDs can be had on ebay for cheap and will offer a much better stable performance than consumer drives.
Per Anandtech's review there's very little air gap between the different sizes, performance wise.
Random & Sequential Performance - Samsung SSD 850 Pro (128GB, 256GB & 1TB) Review: Enter the 3D Era
I wouldn't expect it anyway at the 512GB tier. It's not uncommon in some product ranges for the 128GB to be the performance runt of the litter due to package count and on-die bandwidth, but not the 512GB.
What's interesting to me about the MX200 is the adaptive SLC cache size; where the controller in effect decides on the fly to use some NAND as SLC instead of MLC. Evidently this is not so impressive the way Micron implemented it in this case, but the implications are tantalizing: What if manufacturers exposed this ability to the consumer? I'd give a lot to be able to take a decent consumer 512GB MLC SSD @ $.40/GB and use it as a 256GB SLC @ $.80/GB, with all the speed, stability and endurance that implies.
MLC Nand used AS SLC is not as robust, you are just programming one of two states (0/1) to a 4-state bit so you can be a little more wreckless about it (and thus speed up), but if a 4-state(2bit) MLC node drops bad, it will likely cause both bits to rot at the same time. You have to look at the picture of the states to truly understand what i'm saying. The size of SLC nand is what gives it is endurance. the ability to store 1 bit in the space of 2 or 3 bits (MLC/TLC) !
That makes sense, and yes, if they use the existing 00/01 states which have a small charge differential then there's not much of a value-add in extending the MX200 approach, at least not for endurance purposes, since cell deterioration would make these values indistinguishable after the same number of P/E cycles as the standard MLC.
But if the controller had a mode where for the entire disk it treated MLC states 00/11 --- or whatever bit pair is represented by the maximum charge differential -- as 0/1 and read/wrote accordingly then you'd have something. Effectively a cheap SLC on commodity MLC NAND, for only 2x the price/bit of MLC.
Mind you this is just wild speculation on my part based on popular descriptions of the technology... I don't pretend to know how it actually works and I'm probably missing some devil in the details out of sheer ignorance.
Back in the days we'd overprogram EPROM's with more voltage as we continued to recycle them. +5V not working? Go to +12V? Not working goto +21V! The same thing happens with nand currently but not to such extremes.
I'd say anything SLC or EMLC or SAS interface would be considered enterprise SSD!
I'd say anything SLC or EMLC or SAS interface would be considered enterprise SSD!
The actual cell arrays between SLC and MLC/TLC are almost always the same. The only difference in the parts if there are any are on the drive side and the receive/sense amp side. Generally, a company will make 1 part that will be SLC or MLC/TLC based on bin characteristics and market requirements.
And cell wise there is absolutely no difference between an SLC cell storing a 0 and an MLC storing 00 or an SLC cell storing 1 and an MLC storing 11. MLC is using voltage domain multi-level signalling generally with the following correspondance: SA0 = 00, SA.33 = 01, SA.66=10, SA1 = 11.
Thanks for confirming my vague impression / dumb guesswork. It's pretty clear to me the reason that manufacturers don't make cheap SLC (i.e. roughly 2x the price/bit of MLC) when they obviously could is that they have very little incentive to. Their focus is on the race for density and staying profitable amid downward pressure on prices in the consumer market. Lowering prices on the smaller, pickier SLC market doesn't really serve their interests.
Well, they wouldn't have enough nand to meet demand if they focused on SLC. They literally cannot make the stuff fast enough.
That and there really aren't any markets that truly need SLC. The market has largely accepted eMLC (really just MLC with different signal levels and changed retention vs endurance priorities, generally the exact same die as normal MLC) and drives in the 10-25 DWPD range for a 5 year span. If you can meet the performance requirements (relatively easy with small amounts of DRAM), reliability (PLP via caps, et al), and endurance with MLC, why would you ever want to use SLC?
It would take a rather highly specialized application to actually require SLC NAND these days. So specialized that I can't actually come up with an actual application for SLC NAND. Many of the MLC drives these days have both higher performance, better steady state, and as high if not higher write endurance as the older SLC drives.
Last edited:
With that said...
I've been wondering.
Intel has: End-to-End Data Protection and then they also have: Enhanced Power Loss Data Protection
The 730 has E-PLP while drives like the S3500 and S3700 have both, and drives like the Intel 1500 Pro only have End-To-End Data Protection.
Someone care to go into depth on the difference between these? I thought the whole point of "End to end data protection" from intel was that it was more advanced PLP but now I realize it's listed as separate things, and some drives have 1 not the other, etc...
I've been wondering.
Intel has: End-to-End Data Protection and then they also have: Enhanced Power Loss Data Protection
The 730 has E-PLP while drives like the S3500 and S3700 have both, and drives like the Intel 1500 Pro only have End-To-End Data Protection.
Someone care to go into depth on the difference between these? I thought the whole point of "End to end data protection" from intel was that it was more advanced PLP but now I realize it's listed as separate things, and some drives have 1 not the other, etc...
They really are two different things. One is about recovery and reliability across power events and the other is about making sure that the data written and/or the data requested is the correct data.With that said...
I've been wondering.
Intel has: End-to-End Data Protection and then they also have: Enhanced Power Loss Data Protection
The 730 has E-PLP while drives like the S3500 and S3700 have both, and drives like the Intel 1500 Pro only have End-To-End Data Protection.
Someone care to go into depth on the difference between these? I thought the whole point of "End to end data protection" from intel was that it was more advanced PLP but now I realize it's listed as separate things, and some drives have 1 not the other, etc...
E2E-DP basically means that the data is protected by an error coding at all times and that the data at rest is also generally tagged with relevant metadata such as LBA to provide additional protection that what is actually being read is what was intended to be read. Some interface protocols allow this information to go further up and down the stack. For example, the SCSI formats allow the appending of 2B CRC, 4B Reference, and 2B Tag to be appended to all data. The Reference normally contains some address information (which may or may not include LBA) while the Tag allows the upper layers to appended basically any information they want/need. The additional SCSI informatio is generally only useful/important for large scale storage systems connected to multiple requestors through intermediaries. As things like the S3X00 series and Pro are SATA only devices, the end to end obviously refers to only on drive end to end. For the actual on interface/programer visible SCSI solution, iirc, the term is "T10 DIF" which uses 520 byte sectors and is why pulled used drives from many enterprise storage systems are formatted to 520+ byte sectors.
So for the SATA devices, it just means that the drive maintains data protection at all times and also generally includes non-user visible LBA information along with data on the drive.