Replicating 850EVO 1TB test

[Rossik]

Hi all

Was recently thrilled to see the Samsung 850EVO 1TB mSATA option for my 2012 MacBook Pro which didn't cost the earth (via an adaptor) so went ahead and bought. Sadly only after did I come across the performance consistency problem raised by AT in their review http://www.anandtech.com/show/9023/the-samsung-ssd-850-evo-msata-m2-review/2

I did read in another forum of someone replicating the test but not encountering the issue. However, in running the Blackmagic Disk speed test I have seen some unusual behaviour. When I run the 5GB stress test the write speed performs consistently for quite a few cycles at around 500MB/s but if left for longer it then starts to slow down quite dramatically with each new cycle, quickly ending up in the sub 100s. I did a check with an external RAID0 setup (4x4TB HD) to make sure it wasn't a software issue but it consistently wrote at around 500MB/s and never experienced any drop-off when left cycling the test.

Would the above test point seem to be pointing to the same issue as identified by AT in their test? Also could someone tell me the specific software they use so I could replicate it (apologies if I missed that in the article, am new here)? Or a Mac equivalent in case their test software only works on Windows? It is specifically the 4KB random write (QD32) performance test over time which reveals the issue.

Many thanks for any help

Ross

 

[Elixer]

The issue with the 1TB mSATA is actually worse than I expected because it's literally stopping for seconds in a frequent manner. The pauses can even be over 50 seconds, so this isn't just some normal garbage collection that's happening in the background. I find this to be very alarming because it may have dramatic impact to user experience and it's simply something that no modern SSD should do anymore. I did let Samsung know about my findings before publishing this review, but I wasn't really able to get any comment from them regarding this issue and whether Samsung has noticed something similar in its internal tests. Adding over-provisioning seems to help as the pauses become much more infrequent, but for now I would still advise against buying the 1TB mSATA version until there's a fix for the IO consistency.

As there has been no update, I guess there is no fix yet.

Though, I do wonder, is it throttling because of heat?
Not sure QD32 is all that common for consumers, that means there are 32 requests in line with 4k of data.
The best way to test that is just hammer the drive with tons of write requests, and see how it does.

 

[Rossik]

You might be on to something with the heat comment. Had another go this morning with the test after the laptop having been asleep and no significant processing. It is doing much better at maintaining a higher write rate. Tends to consistently hit high 300s low 400s after initially going up to 500MB/s. Certainly the use of an adaptor is not ideal and may contribute to the heat issue.

 

[Elixer]

Since @Hellhammer isn't around anymore (the one that did that review), the new SSD guy, @Billy Tallis might be able to get more information from Samsung about this. (Which is funny, since @Hellhammer works for Samsung now).

 

[Rossik]

Would be nice to hear latest thoughts from either of them.

It's doing that slow down again but am in the middle of processing a hefty import from Aperture to Photos, although writing on my external. Will keep playing around with the heat idea. Also running Mac Sierra Beta so sometimes with little stalls I'm thinking is the mSATA issue or just beta release teething.

 

[Hellhammer]

Since @Hellhammer isn't around anymore (the one that did that review), the new SSD guy, @Billy Tallis might be able to get more information from Samsung about this. (Which is funny, since @Hellhammer works for Samsung now).

I'm still around, but I don't post that often since my view is biased in many topics now

Anyway, the problem was confirmed to be the mSATA adapter I used to test (or the voltage regulator in the adapter to be exact). The voltage regulator couldn't provide enough power, hence the drive would "throttle" every now and then. In other words, the drive itself is completely fine. I also gave Billy a couple of new adapters that Samsung uses for internal testing to avoid the issue in the future.

Since you are using an adapter, the adapter is most likely the cause of the issue. The most common ones use linear voltage regulators, which have poor efficiency and thus need a higher current to power the drive, and in case of the 1TB drive the current seems to exceed the SATA spec. You need to find an adapter with a switching voltage regulator, but as far as I can see nobody mentions the type in the specs, so you would have to contact the manufacturers/sellers to confirm prior to buying one.

Another option, of course, would be to return the mSATA version (if you still can) and go for the standard 2.5" one since it won't have this problem.

 

[Rossik]

Many thanks for the update and clarification @Hellhammer. Maybe @Billy Tallis could add a little update to the article as it does say they would get back if there was a response from Samsung. That's good to know the drive itself is sound. I might have a bit more of a challenge with the solution as I am using it in a Retina MacBook Pro so no room for the 2.5" drive. Also means makers of adaptors specifically for Apple's proprietary connector will be niche. Hopefully I can find one who has a switching voltage regulator. Will report back here if I do.

I came to be aware of the issue when I handed my Mac in for a display change at an Apple store and in their standard hardware test it said the drive was failing. I explained about the drive mod I had done and he said he had heard about problems with that mod drive (although we now know it is the adaptor) although he also said he had heard of the motherboard ending up being damaged. Do you think this voltage issue could cause wider problems to the motherboard? I can appreciate I am doing a hack here and the risk that comes with that.

 

[Hellhammer]

Ahh, I hadn't heard of that mod so I assumed it was the last (i.e. 2012) classic MBP.

There is a possibility that the mod increases the chance of failure since it pulls more power than the system is designed to provide. This is especially true for Macs as Apple only validates specific drives with certain power draw, whereas PCs are often designed and validated for a larger variety of drives. That doesn't mean that it's certain to happen, but you might want to rethink about it, unless you can find an adapter with a better voltage regulator.

 

[Rossik]

Thanks @Hellhammer for that further warning. It may be I have to go back to the 500GB drive and find another solution to my storage wows. Or pay the Apple premium and have them install a 1TB drive (can't actually remember if this model came with that option or perhaps it was a 750). OWC does a 1TB version but Sandforce drives do underperform with compression and encrypted drives. Thanks again and I'll post here if I have luck with another adaptor. I've written to all the sellers/manufacturers on Amazon.

 

[Hellhammer]

If you can't find a suitable adapter, you might want to consider Transcend's JetDrive lineup. The drives are SM2246EN based, which isn't 850 EVO good but miles ahead of SandForce and the track record is solid. Pricing is also reasonable compared to what Apple would charge you.

 

[Rossik]

Thanks @Hellhammer. Had temporarily forgotten about Jet. Actually they were the ones I was going to get as an alternative to the OWC but then I read a feedback comment from someone suggesting the adaptor route for half the price and forgot all about them. Think I will pursue the Jet route if I come to a dead end with the adaptors.

 

[john3850]

voltage regulator info
https://www.dimensionengineering.com/info/switching-regulators

linear voltage regulation generates a lot of waste heat.

 

[Rossik]

Thanks John

 

[Rossik]

Just a little update @Hellhammer and others on responses from suppliers I contacted. Most are coming back with as expected comments about just being linear voltage regulators. However, I am getting a couple of responses about not having a regulator which I'm not sure how to interpret. My thoughts are that this is a bit 'lost-in-translation' and ultimately just linear like the others but what do you think? Here they are:

"Hello！friend,
I t has no voltage regular, only pcb wire convert."

and:

"They replied,this adapter can fit for Samsung EVO850 1TB mSATA and 2012 Retina MacBook Pro (A1398 A1425). Motherboard slots voltage is 3.3v,and this adapter don't have voltage regulator,so not linear voltage regulator and a switching voltage regulator."

I believe last bit should be read as "so not linear voltage regulator nor a switching voltage regulator". If it's true that there isn't any regulator is that actually a good thing or in fact worse or impossible?

 

[Rossik]

Had a electrical engineer friend look at my adaptor. He confirmed there is no regulator on it so if motherboard slot is 3.3V as above reply states and Samsung drive is too (which it is) should be okay in terms of not frying the motherboard. Still I do experience that slow down occasionally. Perhaps a heat issue due to poor adaptor connecter. Am tempted to buy a couple of the other non-regulator ones to see how they perform.