A few months has passed since part one of the review of the QNAP TVS-682T. So after further tinkering and testing, how did it perform and will I be buying one?
It seems part one of the QNAP TVS-682T shared storage unit with two Thunderbolt connections went down well. Please take a read of part one of the QNAP review first if you haven’t done so already- it describes the loading of drives and setup.
The unit has been storing and sharing relatively quietly in the studio for a few months. This brings us to our first point, connectivity.
The QNAP has great connectivity, two 10GigE, four GigE, four USB 3 ports (One USB 3 on the front too) plus 3 HDMI and two audio jacks. And yes, you can use all of them at the same time.
Sharing via Thunderbolt is easy, but as the longest copper cables are 2 metres that means that you will be sitting the box in-between two edit stations. You can of course use optical cables, but these add a lot to the cost.
One note here, the Thunderbolt cable sockets on the QNAP are ‘upside down’ so don’t go trying to force a cable in the wrong way round.
Whilst we are talking about the physical attributes, I was concerned about the slightly flimsy plastic disk sledges. Over the summer, I installed a 12 bay Synology NAS and I would say that the QNAP has the edge on build quality of the sledges. The Synology twist locks for the drive seemed very fragile compared to QNAPs blue lever and knowing me, I’d lose the plastic key if it wasn’t gaffer taped to the back.
I spent a coupe of hours with a QNAP engineer who very kindly 'dialled in' to the unit. He first checked out that the unit was configured correctly (which it was) before building an NFS share.
This is quite straightforward, but as always if you don't know how to do it, it can be confusing. This gave the option to connect to the QNAP with NFS, which should give faster speeds. It was labled (Final Cut Pro) for identification.
There has also been a stream of updates for Qfinder Pro, opening the app prompts you to upgrade. This requires a restart of the host Mac after downloading and running the installer. The firmware on the unit also upgraded over the trial period. By logging in to the back end, a panel pops up with the available update. After the firmware downloads, the whole unit has to power cycle. It is good to see QNAP publishing the updates and the regular prompts for installation means everything stays up to date.
The recent update bought T2E, which is a way to share an Internet connection through the box and down the Thunderbolt cable. Handy to connect to a 1G ethernet connection without having to add another adaptor to a MacBook Pro.
Back to speed and did things go faster?
I'm beginning to think that the good ol' Blackmagic Speed Test isn't really up to the job anymore. Yes it will give you a straight read of speed, but with SSD hybrid drives there's a lot more going on.
I ran the speed test on the NFS share - which incidentally mounts automatically when selected from the Qfinder Pro app.
First up was with the SSD off and as you can see the results were 492 MB/s write and 681 MB/s read. This was higher than the previous test using 1Gig SMB which gave approximately 400 MB/s read/write.
Then I retested with the SSD caching on.
Rather a disappointing result at 307 MB/s write and 562 MB/s read- maybe the cache is getting in the way of the test. The previous test with SMB (Over Thunderbolt) was 519 MB/s write and 728 MB/s read. Let me stress again that with the SSD cache enabled, the results do tend to fluctuate and I've tried to grab the best result I can.
This needed more investigation and the QNAP tools provide some handy metrics.
The NFS mount allowed me store the FCPX library on the QNAP along with the cache set externally in the same folder. So how would the SSD cache react to different FCPX operations?
In the screen grab below, you can see two traces of SSD read and write performance. In the first sector during the Blackmagic test, the unit was writing to the SSD, but reading very little.
This changed when I started to open and close the FCPX Library repeatedly. The write rate went up by far more than the read rate.
Then I hit the render button and you can see at about halfway through the render, the read and write to the cache are about the same - 80% read/write.
So on this basis, hopefully accessing and rendering a project should be quicker with the SSD cache on?
A complicated FCPX project with the SSD cache off took 10 minutes and five seconds to render.
The same project with the SSD cache on took 10 minutes and five seconds to render.
Disappointing to say the least, does this really mean that the SSD cache has no overall performance boost for FCPX? It doesn't look like it, however I would say that opening and closing a Library a few times does seem snappier with the SSD cache on.
My naive testing reinforces the point that to get the best out of this machine you need to have it configured correctly by a dealer.
Both Bob Zelin and Filip Vandoorme have recommended using the 10GigE share, turning the SMB packet signing off (Didn't even know that existed) and connecting clients via a cheap 10GigE switch. Unfortunately, without a couple of Thunderbolt 2 to 10GigE converters, that's beyond my testing capabilities for now. From my limited trials, it does seem that an SMB share seems to provide constant results and of course FCPX Libraries can now be stored on SMB. (Although you could have an NFS AND an SMB share at the same time!)
Would I buy a QNAP TVS-682T? The answer is yes and no
Reasons for Yes:
- This has to be one of the smallest and cheapest Thunderbolt shared storage units on the market.
- 10 GigE could be the best connection for new Macs! Thunderbolt to 10GigE converters can cost $600. There are two 10GigE connections as standard. (A Synology Dual port 10GigE PCI Express card is $250)
- SSDs should help the caching of small files such as FCPX Libraries
Reasons for No:
- Can be complicated to tune perfectly for maximum speed
- Non-thunderbolt versions are cheaper -10 GigE could be the best connection for new Macs!
- Non SSD versions are cheaper
- Limit of 4 HDD giving a maximum capacity of 30TB using 10TB drives in RAID 5
- Inconsistent bandwidth - anywhere from 100 to 700 MB/s over Thunderbolt
I’ve been impressed with the QNAP and for a couple of edit stations working together the transfer speeds should be fine. Four spinning drives doesn’t really give enough headroom for more than two users, so I’d be looking to purchase the larger TVS-1282T with 8 HDD and 4SSD bays. However, having said that, I’m sad that the unit is going back.
Also remember that you don't need to populate the entire unit with disks. It might be more cost effective to buy a larger model say the new TS-1685 and half fill it with disks. That would give you a capacity of 6x10TB- combining the disks into RAID 5 results in 50TB of space overall. Then buy another set of 6 when you need more space as these can be added to the data pool.
What I would really recommend is purchasing a unit through a dealer who can configure it for you. Unless you are completely comfortable with IT and like many menus and options to wade through, let the experts like Bob Zelin (US) or Filip Vandoorne (Europe) do their thing.
QNAP has an extensive list of models ranging in price, capacity and connectivity. With 10GigE on the horizon as Apple’s high speed data sharing connection of choice, QNAP models with built in 10GigE connectivity are a good choice. I expect to see many more of them on my freelance editing travels.
Peter Wiggins is a broadcast freelance editor based in the UK although his work takes him around the world. An early adopter of FCP setting up pioneering broadcasts workflows, his weapon of choice is now Final Cut Pro X.