Alan Greene's List: Computer

I posted this about the Transcend 32 GB SDHC card that was getting trashed in reviews by Canon owners. It also applies to the Sony card (read down), and readers may find it helpful in evaluating SD cards in general.

The 1-star threads on the Transcend are like watching a committee of blind men describing an elephant. Unsurprisingly, no one has the complete picture, but a lot of Canon owners think these cards are bogus. SD card technology is very complex, the Class system is often misunderstood, and there is a long history of compromised cards appearing on eBay -- or even from reputable dealers who have been hoodwinked. Kingston cards are most often faked (or taken from the midnight production run); that's because Kingston has about a third of the flash card market and doesn't forge their own chips, but any manufacturer can be spoofed. It doesn't pay to buy cheap cards on eBay or less reputable sources.

An SD card contains a controller chip and flash memory chips, even the microSD versions. Realize that SD means Secure Digital, and that security comes from crypto managed by the controller (MMC cards didn't have that bloat, but Hollywood DRM requirements made sure they faded). The controller can be pretty powerful: the Samsung SD controller is a 32-bit ARM TDMI chip with 128 k of code space -- that's cell phone power. It handles I/O and fading (when the card slowly wears out its NAND sites after about 100,000 hits so writing is randomly distributed and kept track of), the factory self-test, and a host of other functions, but it can also be programmed to report a false storage capacity. Sometimes the firmware on the controller or its crypto or something else on certain cards leads to problems with certain hardware, as it has, for example, with the Samsung Galaxy III and maybe the Canon cameras. Usually the card is found to confirm to SDcard dot Org specifications and the hardware is to blame, but you never know.

To test the actual capacity of your card, use H2testw, which is free. It writes the full amount of data to the card and then reads it back (this nukes whatever was on the card). This can take more than an hour with a 32 GB card but it tells you if the card indeed holds 32 GB or if it has problems. H2testw also gives you read and write speed numbers, but it's unclear whether the numbers are for random read/write, or sequential read/write or a mix. Sequential write is what photographers and videographers most care about.

Sometimes a Class 6 card might appear to test faster than a Class 10 card, when using computer read/write tests or even in a camera that wasn't designed with higher capacity cards in mind. That's because such cards use smaller block sizes (there's no cache on an SD card). Explaining the significance of this is getting too far into the weeds for an Amazon review, and modern devices and their firmware should not have that limitation.

To test random and sequential read/write speed, use CrystalDiskMark, also free. Run the full suite; it won't take long. CrystalDiskMark requires that the card be formatted, and the most reliable way to do that is to use the free tool from sdcard dot org. A 32 GB card will may report 39.9 GB before formatting and 29.8 GB after; don't worry, you haven't been ripped off.

Unfortunately, there's no easy way to read the manufacturer's ID off the SD card, even in Linux, to determine who actually made the card, because you have to have the card connected directly to a motherboard; an SD-to-USB adapter won't work because it doesn't pass through disk data, and that's what's in nearly all computers. But what do you care, so long as the card's as big and fast as claimed?

Now, what about the Transcend 32 GB SDHC C10 card? Here are full test results from CrystalDiskMark:

Transcend 32 GB SDHC C10
-----------------------------------------------------------------------
CrystalDiskMark 3.0.2 (C) 2007-2013 hiyohiyo
Crystal Dew World : [...]
-----------------------------------------------------------------------
* MB/s = 1,000,000 byte/s [SATA/300 = 300,000,000 byte/s]

Sequential Read : 19.953 MB/s
Sequential Write : 13.786 MB/s
Random Read 512KB : 19.618 MB/s
Random Write 512KB : 13.827 MB/s
Random Read 4KB (QD=1) : 3.003 MB/s [ 733.2 IOPS]
Random Write 4KB (QD=1) : 1.414 MB/s [ 345.1 IOPS]
Random Read 4KB (QD=32) : 3.491 MB/s [ 852.3 IOPS]
Random Write 4KB (QD=32) : 1.842 MB/s [ 449.8 IOPS]

Test : 50 MB [F: 0.0% (0.0/29.3 GB)] (x5)
Date : 2013/05/12 18:19:47

So you can see that the sequential write speed, what matters to a photographer or videographer, is over 13 MB/s, more than 30% above the C10 spec. The other speeds aren't too shabby, either. There's no way that this card is too slow for a contemporary camera, still or video, that is functioning and designed properly.

Also on Amazon for a very attractive price is a Sony 32 MB SD card. Here are the full CrystalDiskMark test results for it:

Sony 32 GB SDHS USH-I

Sequential Read : 19.998 MB/s
Sequential Write : 14.115 MB/s
Random Read 512KB : 19.772 MB/s
Random Write 512KB : 13.173 MB/s
Random Read 4KB (QD=1) : 3.682 MB/s [ 899.0 IOPS]
Random Write 4KB (QD=1) : 2.302 MB/s [ 561.9 IOPS]
Random Read 4KB (QD=32) : 4.517 MB/s [ 1102.8 IOPS]
Random Write 4KB (QD=32) : 2.323 MB/s [ 567.0 IOPS]

So the Sony is slightly faster, 40% over spec, on its sequential write. H2testw also reported full capacity and no errors for this card, so they are both top notch (The Sony doesn't come with a little case and the case for the Transcend case is about twice as big as it needs to be and so won't fit in the little pockets in a modern camera bag. Big deals.)

Just for reference, here are the results for a contemporary USB thumb drive:

SanDisk Ultra 32 GB USB thumb drive

Sequential Read : 22.508 MB/s
Sequential Write : 7.985 MB/s
Random Read 512KB : 22.365 MB/s
Random Write 512KB : 1.767 MB/s
Random Read 4KB (QD=1) : 3.719 MB/s [ 908.1 IOPS]
Random Write 4KB (QD=1) : 0.520 MB/s [ 126.9 IOPS]
Random Read 4KB (QD=32) : 3.871 MB/s [ 945.0 IOPS]
Random Write 4KB (QD=32) : 0.482 MB/s [ 117.7 IOPS]

As you can see, this USB drive is faster than the SD cards for reading, such as for playing music, running software, or looking up data, which shows that it is nicely optimized for its intended functions.

As an aside, anyone who thinks they have lost data on an SD card can recover pictures using the free tool at z-a-recovery or recover everything(!) with the free PhotoRec tools.

So, if people want to whine about these SD cards in their particular cameras, they should run these simple, free tests and reach their own conclusions about whether the cards are "too slow." The cards test above spec objectively. And they both work flawlessly on my still (not Canon) and video cameras. I think they have gotten a bad rap on Amazon from people who don't have the full picture, so to speak.