Photographers and hobbyists alike continue to have a nearly insatiable need for “large capacity” memory cards with high speed and reliability and preferably low cost. The “traditional” people who claim to know better tend to stick with Sandisk and Lexar for a “no fuss” experience while some others are willing to explore and use other brands, sometimes all the way down to OEM unbranded cards.
I’m a little bit of both – but today’s contender is definitely not a no-name. It is the Toshiba Exceria Type 2 64Gb UHS-1 SDXC card. This was purchased for me by one of my friends in his time over in Taiwan, at a cost of around AU$1/Gb (which is roughly what a Kingston Ultimate card costs). It claims to have a read speed of 95Mb/s and write speed of 60Mb/s which is pretty good for the price.
Toshiba itself is a reputable name in storage, their Exceria Pro cards have recently come to acclaim as the fastest SDXC card on the planet. But don’t be fooled – there are many different types of Exceria with various capacities and performance profiles. I’ve opted for the Type 2 as it is more easily available, and has a good balance between read speeds and write speeds – it’s no secret that many cameras are CPU limited and unable to reach the maximum write speeds of many of the highest speed cards, especially for the lower end camera bodies! In some cases, cards faster than about 30-50Mb/s write see virtually no advantage in practice (barring any compatibility issues).
This, in itself, may be a controversial point but the evidence suggests it is the case. A primary source is Rob Galbraith’s database of card performance. When a Nikon D800 ($2700) camera can only manage 42MB/s to a 95MB/s Sandisk card, you know that your “measily” D3200 isn’t going to make the most of the faster cards.
The point is not to pay a premium where you won’t see the advantages. If in the future, you will keep your cards over to a new, better, faster camera – you might want some headroom but seeing that capacity and storage demand continues to increase, future-proofing in this regard is a bit of a fallacy. So in reality, I’d say any card that does about 30Mb/s write is sufficient for most consumer DSLR’s – but if you don’t need to pay a premium for a faster card, you might as well go faster!
That is the case with the Toshiba Exceria. Another encouraging sign is that the cards themselves are manufactured in Japan (yes, I’m partial to this) and that Toshiba actually has a partnership with Sandisk when it comes to flash memory. Bargain? It sure is!
It comes in a thin card box (pictured earlier) and is housed in a plastic “shell” inside a plastic tray inside the box.
There is no leaflet as such, but the box does feature a dotted line which encourages you to cut the box apart to read the speed disclaimer which is printed on the inside – the first time I’ve seen this sort of disclaimer!
Lets just say that there’s nothing too exciting about how an SD card looks like. In terms of verification features, there is a hologram on the box (pictured earlier), and a metallic label on the card itself. In fact, I haven’t heard many people purchasing the Exceria by name, so I think it’s probably “safe” from counterfeiting for now.
I have two samples of this card in my possession – the CID values vary due to different serial number and manufacture date, but the CSDs are identical. The device size has also been calculated for your convenience.
CID: 02544d534436344786f730566e00d9bb CSD: 400e00325b590001dbff7f800a40003f CID: 02544d534436344786fe31903600d1b7 CSD: 400e00325b590001dbff7f800a40003f 124,780,544 blocks -> 63,887,638,528 bytes
Interestingly, and rather depressingly a normal trend, the device size is below even 64Gb “weasel” gigabytes (i.e. 64,000,000,000 bytes) due to overprovisioning for flash management. It’s not the worst “difference” – my 128Gb Kingston turned out to be a 125Gb (that’s weasel gigabytes, mind you) card.
Of course, prior to commissioning any flash memory for serious usage, exercising it to check it performs as expected with your devices is an advisable move. It is standard practice for me to perform speed benchmarks, full surface random fills and verify passes using my preferred card reader followed by a run in my camera (Nikon D3200 RAW+JPG) in high-rate burst shooting.
Testing was performed with my Transcend RDF8 USB3.0 reader attached to a Renesas USB3.0 controller on Windows 7. Note that small block access speeds in hardware devices may be faster due to less latency overhead from transaction turnaround on the USB bus to the reader, and the reader to the card.
I would have to commend both cards for completing the torture test without any unexpected drop-outs, slow-downs or corruption (checksum failures). It seems that these cards are perfectly good for use which is great news for me!
Testing with CrystalDiskMark revealed an unexpected result, which is why the test was repeated on the first card twice to ensure it wasn’t an unexpected result. The filesystem was exFAT as provided on factory formatting.
Interestingly, Card 2 is a little slower than Card 1. Judging from manufacture date, Card 2 is a January 2013 card, whereas Card 1 is a August 2013 card. I think this kind of variance is “normal” as different flash chips and controllers will result in speed variations. For the most part, results were consistent across runs – but unusually, with a queue depth of 32, the card was twice as fast on read and half as fast on write for 4kB blocks.
I think this is an artifact of the benchmark program’s timing and is not “real”. The reason I say this is that SD cards don’t have an inherent command queueing capability and prior testing of cards have often revealed figures “within” error margin identical between the 4k and 4k QD32 results.
As expected, most large and speedy cards are pretty much optimized for large block work rather than small block accesses, but comparing the results relative to other cards is the key. To that end, these cards kick the pants off the Kingston Ultimate I tested earlier.
A second opinion can be had by using HDTune Pro to survey the read speed. In this case, I looked at both fresh-format and random-filled performance (as there have been several MB/s discrepancies with some cards with them reading faster over “clean” zeroed areas – I discovered this with some of my cards but none of those results have made it onto the blog due to a lack of time).
Card 1 Empty
Card 1 Random Filled
Card 2 Empty
Card 2 Random Filled
As you can see from these results, Card 2 being slightly slower is again borne out at least initially, but there is no meaningful difference between the “dirty” and “clean” states meaning the benchmarked read speeds are what you will experience in reality. After it was random-exercised, the speeds are much closer, within the margins for error. At 80Mb/s, downloading a full card surface will take 762 seconds or 12 minutes and 42 seconds (approximately). That’s pretty speedy!
Looking at the test results, the cards are likely to meet the stated performance levels. In fact, the writes were a little faster, and the reads were a little slower, but everything depends on the devices and readers you use the card with due to variation in clocking/timing and transfer mode support. I’m sure the Transcend RDF8 isn’t the “ideal” partner for any card, but if a $20 reader can extract this level of performance from a card, I’m pretty happy as it is. Whatever you do, try not to saddle it with a small thumb-size USB 2.0 reader as many of these top out at 7MB/s – a quality USB2.0 reader should get to at least 15MB/s.
It’s definitely a good addition to my collection of cards, and will likely take over from my 32Gb Samsung Class 10’s as my “primary” shooting card (as those really don’t go much faster than 12Mb/s write). The 64Gb capacity will go to reducing card clutter and card swaps, while not being so big as to suffer poorer performance (e.g. as with the 128Gb and above “affordable” cards on the market today).