StarTech Dual mSATA SSD to 2.5? SATA RAID Adapter Converter

Preamble: I don't work for StarTech, and I don't write reviews. I just had a few people express interest in this device so I'm posting about it.

As someone with a desire to put more than 1TB worth of SSD drivespace in their computer using only a single 2.5" drive bay, I purchased a StarTech Dual mSATA SSD to 2.5” SATA RAID Adapter Converter. It's a 2.5" board with two mSATA sockets and a JMicron JMS562 disk controller. I installed two 1TB Samsung Evo 840 mSATA drives into it, then installed it into the primary SATA III 2.5" drive bay (there are two 2.5" bays total) in my Dell Precision M6600 laptop computer.

First impressions: the board is very well built for being such a small, delicate thing. It's obvious a lot of thought was put into the design. It feels like it could withstand a drop from a decent height and still work fine - although I didn't try it. The board has a couple of LEDs that supposedly light up and indicate the state of the drives. However, when it's installed inside my laptop, I can't see them. Both of the Evo 840s snapped in with ease. I then set the jumpers to specify RAID0 (out of a possible RAID0, RAID1, JBOD, and BIG) and mounted it into the bay using the same screw holes a normal 2.5" drive would use.

Upon powering up my computer, I went into the BIOS setup, and it reported that I had a 2000GB drive installed (the Evo 840s are actually 1000GB as opposed to a "real" terabyte) and reported the drive manufacturer / model as a JMicron RAID0. Exiting the BIOS I booted from a USB 3.0 flash drive and proceeded to install Windows 7, which went extremely quickly.

After Windows 7 finished installing, I logged in and looked at disk manager: 1.81TB total drive space formatted as NTFS. Not bad considering it only took up the space of a single 2.5" drive bay. I ran a few CrystalDiskMark benchmarks on the array; they were not particularly impressive, considering it was supposed to be a RAID0. I would've thought they would've at least maxed out the bus and doubled the 4K random read/writes, but alas they scored no higher than my Crucial 960GB SSD and slightly less (on the 4K random r/w) than my Samsung PM841 512GB SSD. I'll post the exact numbers if there's interest but trust me when I say, they're nothing to write the internet about. (In retrospect, I'd like to acquire another PM841 and run the RAID0 tests again with drives that built for performance as opposed to the Evos which really aren't.)

I installed the Intel SSD Toolbox v3.2.1 utilities to determine if TRIM was enabled for the RAID0 setup. It was not. It was enabled for the PM841 so I know the utility worked properly.

Overall, I was happy to have nearly 2TB of SSD storage space in a single 2.5" drive bay. I could definitely see this device helping someone with a laptop that only has a single 2.5" bay, if really fast storage was needed. I was slightly disappointed regarding the performance. One would definitely see a "vast increase in system performance", as the marketing literature touts, if going from a Samsung Spinpoint M9T 2.5" mechanical hard drive to the SSD setup, but there was no increase as one may hope for going from a single SSD to RAID0. Again, that could only be because of the Evo 840s. Without actually testing different drives there's no way for me to know.

 

You can compare your performance with my results here for a single 1TB EVO mSATA. There's a little bit of headroom to improve performance within the limit of the SATA 3 interface (around 550MB/s). Or maybe the JMicron controller is a bottleneck.

John

 

The PM841 mSATA would be a step backwards. It's the first gen, OEM only, 840 TLC. EVO should dominate it.

 

I could've been more clear: the PM841 was slower in all benchmarks except for 4K/QD32, where it scored *4 times* higher than the RAID0. As John indicated, I think the bottleneck is the JMicron controller.

I bought an mSATA-to-2.5" pass through adapter. When I have time I'll rerun the benchmarks on just the mSATA drives against each other in the SATA III slot, without the StarTech adapter.


Sent from my Nexus 7 using Tapatalk

 

My experience with SATA controllers is Intel > Marvell > JMicron. I've never seen a JMicron controller actually give fast performance.

 

Thanks for taking the time to share, erblemoof. Can you post your resulting benchmarks with CrystalDiskMark or other? Sequential speeds should easily be able to saturate the bus, if it can't then I'd ditch that adapter right away. But even with RAID 0 typically SSD's vastly improve sequential read/write nearly double, but as the file sizes get smaller, the advantage is basically marginal. It is nice though to have an option for 2TB fast SSD storage, at least faster than a hard drive, for a reasonable price.

 

This is great little idea but only if its built with some good RAID controller.4k definitely do not scale in RAID like seq. speed .In fact 4k read even drop a little(maybe 300iops) ,while 4k write stay same.On Intel RAID controller you can even triple 4k write with enabling "write-back cache" ,but like I said only with Intel RAID controller.
So EVO are not problem ,problem is JMicron RAID controller.See here what two EVO msata can do on Intel RAID controller.
RAID Report: Native mSATA RAID 0 w/ Samsung 840 EVO 250GB SSD's (TweakTown Exclusive)

Good RAID controller should cost at least 100$ without PCB ,connectors and everything else.Problem is that Startech probably didn't have any better soulution.AS Media have great AHCI controller ,but they don't have RAID controller,Marvell is great in building SSD controllers ,but they cant repeat that with RAID controllers ,so only option is Intel ,but Intel RAID controller is built in chipset so it doesn't work for this.
But this adapter is GREAT idea and I hope they make more expenciv version if they manage to find good controller.

 

Here's an AS SSD benchmark for the Evo 840 1TB drives in RAID0:

 

A *decent* RAID controller should cost at least $500 on a desktop, about $2k for a *good* one. Onboard Intel RAID is decent though but anything under $500 will usually rely heavily on the CPU.

 

Not only is this RAID controller not "decent," but I think it significantly *slows* the performance of the drives. Here's a benchmark of a Crucial M500 960GB drive in the other SATA III drive bay in the same computer as the StarTech RAID0 setup:


When I compare the two drives head-to-head using a popular SSD comparison tool ( AnandTech | Bench - SSD) the Evo should be much faster. However, in the RAID0 setup, they are significantly slower (see benchmark above).

So much so that I think this warrants an email to StarTech. It's ridiculous to not have the same performance as a single drive, let alone worse performance when using the adapter.

 

I am not talking about retail price of RAID PCI-E cards,I am talking about how much would Startech pay for that controller and then in the end this product would cost 220$ and not 60$
You cant putt LSI controller on this thing ,did you see cooler of LSI controller.
Don't forget you are talking about RAID cards that have RAM cache on it,where do you plan to solder that on this product.
I am taking about small simple controller that will utilize sata3 connection and that shouldn't be problem,its not like you need to scale 10 drives.
Problem is that JMicron don't know how to make one.I believe ASMedia can do it for 100$.
Maybe even Marvell 9128,9218 or 9230 can do thing,it didn't show it self in the past,but only because it was on pci-e card that use only one pci-e line so it was bottleneck in start.
note:9218 was connected with two pci-e line,but still it wasn't connect directly to chipset like Intel ports are.
Who knows this is definitely interesting product ,they only need to try harder.

 

FYI: I benchmarked the 1TB drives in BIG mode, which concatenates drive data as opposed to striping it. Performance was exactly the same as in RAID0 mode (see a few posts previous for benchmark graphic).

Conclusion: this adapter is only good for getting 2TB worth of SSD drives into a 2.5" space. While the setup will be faster than a rotating 2TB drive (Samsung M9T as the time of this writing) the adapter causes you to lose a SIGNIFICANT amount of SSD performance.

 

erblemoof, Thanks for the sharing.
If you're still tinkering with it, could you please try JBOD mode, and see if it's compatible with your laptop chipset. You don't mind to tell the chipset used in your laptop, do you?

Thanks a bunch

 

A quick Internet search on 'dell precision m6600 chipset' says it has an Intel QM67 chipset.

Setting the adapter to JBOD causes the adapter to report 1TB total - even with 2TB installed - and performance is exactly the same as in RAID0 or BIG.

 

That was quick!...
At least I know that when I buy this thing, I might not be able to use JBOD mode.
Thanks,

 

Update: I mentioned a while back that the performance (specifically, the lack thereof) of the Startech adapter warranted an email to the manufacturer. After some back-and-forth, and a considerable amount of waiting, the final email response is thus:

As such, if you're interested in this adapter, please go back through these posts and carefully read my performance results. Doing so should set your expectations of the adapter. As someone mentioned, it's really a great idea, but unfortunately, poorly implemented. Maybe someday, some company somewhere will produce a better performing adapter. But by then, drive sizes might be bigger anyway, making it redundant.

In case anyone is curious as to how I finally addressed my performance / storage issue: I ditched the adapter entirely and run the drives by themselves. I increased capacity - significantly - by installing Windows 8.1, then adding the Data Deduplication feature from Server 2012 (do an Internet search for it) and de-duped the drives. These were my results:

Capacity...UsedSpace FreeSpace SavedSpace TotalData
---------- --------- --------- ---------- ---------
x296.99 GB 124.75 GB 172.24 GB 304.69 GB 429.44 GB
1000.20 GB x91.46 GB 908.74 GB 129.27 GB 227.73 GB
x512.10 GB 160.31 GB 351.79 GB 241.59 GB 401.90 GB

Obviously, I'm only de-duping one of the 1TB drives. as you can see, the dedupe feature is extremely efficient. My D:\ drive, for example, is storing 429.44 GB on a 296.99 GB drive, and still has 172.24 GB free! Plus, I get the full performance of the drives. In conclusion, I think the best way to get the best storage and performance is to use the bare drives, and use Win 8.1 with the dedupe feature. Kind of a strange conclusion considering how I started, but in the end I think this was the best way to go, for me.