Crucial MX100 vs. Samsung 850 EVO

[Techhog]

http://www.anandtech.com/show/8066/crucial-mx100-256gb-512gb-review/8
When idle it uses more, when being used it consumes less. Both consume at a third of the power of a standard HDD.

I think this page is pretty good at putting things in perspective:
http://techreport.com/review/25610/ocz-vector-150-solid-state-drive-reviewed/2

Mine in my laptop gets pretty hot. It's sitting at 44C right now. I think it's because it's not able to dissipate heat very well in here, though. It's barely touching any metal.

 

[hojnikb]

850 evo is indeed 3D NAND, but it is still the TLC kind (as opposed to the pro with MLC).. doesn't that mean the performance degradation of stale data problem which plagues the 840 Evo's can still affect the 850 evo.. because of the memory still being TLC ?

TLC is not the main culprint for performance degradation (=electron leakage).

small fab (ie tiny cells) combined with TLC makes this happen. Think of 3D TLC at 40nm as 20nm MLC.

At 40nm electron leakage is much much lower (as there are more electrons inside and thicker walls). Even if we want to differentiate 8 differente states (so TLC), its still gonna take a hell more leakage (so much much longer) for that to be apperent.

 

[Brado78]

The Kingston Hyper X is an ok drive and Kingston had an EXCELLENT RMA service

 

[hojnikb]

The Kingston Hyper X is an ok drive and Kingston had an EXCELLENT RMA service

But its sandforce and kingston did some shady moves in the past (v300).

 

[Hulk]

Nothing with sandforce is reasonable. Unless you want to live on the edge.

Yup, I'm waaay out there on the edge with my Intel 330 Sandforce based drive. Two years and not a single problem.

If you really want to be on the edge...

Looks to me like the drives to stay away from are Samsung 840 EVO and possible any Samsung TLC based drives

 

[Hulk]

As I understand it, it did with some but not others. Luck of the draw...

I don't believe the 850Evo is seeing this problem.

... yet.

 

[slumbarr]

But its sandforce and kingston did some shady moves in the past (v300).

Are you referring to this? I was surprised as Ive been using thier RAM and thought they had premium memory.

http://www.anandtech.com/show/7763/an-update-to-kingston-ssdnow-v300-a-switch-to-slower-micron-nand

 

[omega3]

http://www.anandtech.com/show/8066/crucial-mx100-256gb-512gb-review/8
When idle it uses more, when being used it consumes less. Both consume at a third of the power of a standard HDD.

I think this page is pretty good at putting things in perspective:
http://techreport.com/review/25610/ocz-vector-150-solid-state-drive-reviewed/2

Mine in my laptop gets pretty hot. It's sitting at 44C right now. I think it's because it's not able to dissipate heat very well in here, though. It's barely touching any metal.

Which numbers on the anand bench show idle temps.. like how do you conclude they're both 1/3th of standard HDDs?

What Techhog posted doesn't give me lots of confidence to put an MX100 in my HP laptop.. how does it get so hot.. i don't think even most standard 5400 rpm HDDs (which is in my laptop) get so hot.. can we conclude MX100 is not good for laptop.. seriously why so hot when IDLE ???

 

[Ketchup]

I just looked at the temp for the MX100 on my Lenovo laptop right now and I am at 25 C. I got it up to 27 when running Disk Cleanup. I don't know how one would heat one of these up to 44.

 

[Charlie98]

You know, it's funny... I put my M500 in my daughters laptop and it idled at 50C. I pulled it and replaced it with my old Agility3... And it idles at 35C. The M500 is on the HTPC now... And the temps are fine. Just odd.

 

[Charlie98]

... yet.

...so, the 850Evo is guilty before the fact?

 

[Hulk]

...so, the 850Evo is guilty before the fact?

I stand corrected. No not guilty but under close scrutiny.

I hope the 850 Evo doesn't show the read degradation issue, I really do. But until enough time goes by without any negative reports I'm going to steer clear of them.

I'm no expert by any means but I'm thinking the bug/issue *could* proceed in the three steps outlined below. And by issue let's be clear, after data is written to the drive and not rewritten for a few months the voltage in the cells in question drop and the controller has a hard time figuring out the state of the cell. This results in multiple "trys" or rereads, which slows performance considerably.

1. Hopefully the controller just wasn't programmed correctly for these TLC NAND cells. That is taking the degradation into account. In this case Samsung just needs to reprogram the controller. It's a tough thing to diagnose/fix because you've gotta wait a few months in between trials!

2. Since we're only talking about a handful of electrons separating one level of these TLC cells from the next it might be that the read program is too aggressive (too fast) and the read process must be changed quite a bit. This could restore long-term performance consistency at the cost of overall performance. Of course this is a bad solution because it's not what buyers paid for, a very fast drive.

3. There is a serious problem with the TLC NAND and the drive is FUBAR for any long term storage. Total recall.

Looks like they already tried option #1 and thus far it hasn't worked. They will most likely try it again before heading to option #2 which the shit starts to hit the fan. It's unlikely this issue will proceed to #3 because I don't see Samsung making that big of a mistake. But in that case they will have a full blown shit show on their hands.

 

[SSBrain]

Another possibility could be that Samsung "cheated" with their 19nm and 21 nm planar TLC NAND, which could be what allowed them to use it on their consumer SSDs much before other first-party manufacturers. What if these SSDs are actually working as intended and can only reliably show fast read speeds for a relatively short initial "grace period" time after the NAND is programmed, and read speeds are lowered on purpose for long term storage? And that the heads who designed these drives hoped that users wouldn't notice? Since first-gen TLC Samsung 840 are affected too and for some reason they're being completely ignored by the company, I'm starting to think this is not really a "bug", but the way these drives were intended to work.

If this is the case, then the only way they could "solve" it would be by making these SSDs' wear leveling algorithms work more aggressively, shuffling cells around more often, which they clearly aren't currently doing. I wonder if that would really increase wear that much.

 

[Brado78]

hojnikb- The only SSD manufactures that I know of who, don't use Asynchronous NAND are Intel and Samsung. I know aside from Kingston, others like Crucial. PNY, OCZ and a few others use the Asynch nand in some of their models. The only thing Kingston did was the first to use it, and not tell anyone ....

 

[SSBrain]

I'm no expert by any means but I'm thinking the bug/issue *could* proceed in the three steps outlined below. And by issue let's be clear, after data is written to the drive and not rewritten for a few months the voltage in the cells in question drop and the controller has a hard time figuring out the state of the cell. This results in multiple "trys" or rereads, which slows performance considerably.

By the way, on my non-EVO Samsung 840 250GB, which has had 93 P/E cycles performed on average so far, it takes about two weeks of time (on a PC powered on 24/7) for sequential read speed on static data to start degrading.

 

[hojnikb]

hojnikb- The only SSD manufactures that I know of who, don't use Asynchronous NAND are Intel and Samsung. I know aside from Kingston, others like Crucial. PNY, OCZ and a few others use the Asynch nand in some of their models. The only thing Kingston did was the first to use it, and not tell anyone ....

Which is a big no-no in this industry, especially if this means 2x performance reduction.

Async flash on its own is fine though (and current gen is pretty fast, just look at partriot ignite).

 

[Hulk]

By the way, on my non-EVO Samsung 840 250GB, which has had 93 P/E cycles performed on average so far, it takes about two weeks of time (on a PC powered on 24/7) for sequential read speed on static data to start degrading.

Whoa! 2 weeks? That's disconcerting.

 

[SSBrain]

Yep, although keep in mind I really mean about that to start degrading in two weeks of time (=> drops in sequential read speed from HDTune occurring always in the same LBA ranges where static data lie), not reaching significant levels of degradation.

 

[omega3]

http://www.anandtech.com/show/8066/crucial-mx100-256gb-512gb-review/8
When idle it uses more, when being used it consumes less. Both consume at a third of the power of a standard HDD.

I think this page is pretty good at putting things in perspective:
http://techreport.com/review/25610/ocz-vector-150-solid-state-drive-reviewed/2

Can you explain those idle comparison numbers a bit between Crucial, Samsung and standard HDD.. is that the "slumber power" bench?

My concern is that alot of people say their Crucial MX100 becomes quite hot in a laptop, so is that something to worry about in case of a laptop?

 

[Ketchup]

Can you explain those idle comparison numbers a bit between Crucial, Samsung and standard HDD.. is that the "slumber power" bench?

Although I didn't write the article, I assume the "idle" is referring to "sitting on a desktop, not actively running an application. (meaning that background disk activity is still a possibility)"[/quote]

My concern is that alot of people say their Crucial MX100 becomes quite hot in a laptop, so is that something to worry about in case of a laptop?

I discussed this in post #82. I personally don't have a problem with this. Other laptop models may not do so well.

Here is one theory: a traditional hard drive is not air tight. So, there is air circulation, no matter what the ventilation is like in that bay. If you have a laptop that mounts the hard drive a little too close to a CPU or dedicated GPU, that lack of air circulation (since there are no moving parts in an SSD) could cause it to get fairly warm.

Keep in mind the max operating temp for most SSD's is in the 70s (C), so I haven't seen a single post here that describes anything dangerous.

 

[omega3]

Although I didn't write the article, I assume the "idle" is referring to "sitting on a desktop, not actively running an application. (meaning that background disk activity is still a possibility)"

But I don't see the link you gave, ie. this one mention anything about idle consumption unless you meant the "slumber power bench" (what is that anyway)?

 

[coercitiv]

Although I didn't write the article, I assume the "idle" is referring to "sitting on a desktop, not actively running an application. (meaning that background disk activity is still a possibility)"

AFAIK idle means the disk is ready, no read/write operation are being performed, and no AHCI Link Power Management is available (HIPM, DIPM or both).

For modern SSDs which support power management, the so called Slumber Power figures are more important than idle power since the SSD will go into a power saving state when not needed (a simple analogy would be a modern CPU entering power saving states at idle).

 

[Brado78]

hojnikb- I agree 100% with you on the whole bait and switch. I wouldn't be surprised if Crucial's BX 100 uses Asynchronous. The B stands for Budget and this is an entry level SSD just as kingston's V300. Asynchronous is cheaper to make and is more readily available...It's all about the dollar,they seem to forget about performance and customer loyalty.

 

[ClockHound]

hojnikb- I agree 100% with you on the whole bait and switch. I wouldn't be surprised if Crucial's BX 100 uses Asynchronous. The B stands for Budget and this is an entry level SSD just as kingston's V300. Asynchronous is cheaper to make and is more readily available...It's all about the dollar,they seem to forget about performance and customer loyalty.

Wondered about that too. It appears not, from thessdreview review:

"By using the Micron FBGA Decoder, we can identify the NAND flash as having the product number MT29F512G08CKCCBH7-10:C. It is 128G-bit 16nm synchronous MLC NAND flash memory with each module being 64GB in capacity." Review Link

Maybe they know someone at Micron and can get a good price on the good stuff...Or they saved some money on the controller swap.

 

[omega3]

If you have both an SDD and standard HDD in your pc, do you set sleep mode on or off?

Is it really required for SSD to avoid sleep mode.. on the other hand the HDD will use alot more energy if it never goes to sleep i guess