I just bought a Canon EOS 5D Mk3. Because its continuous drive speed is nearly twice that of the Mk2, I wanted to test the benefits of using different speed memory cards (i.e. 30Mb/s over 45Mb/s over 95Mb/s, etc).
Cameras, in continuous shooting mode, carry on recording files at the displayed rate (i.e. 6 frames per second, or 'fps') until the memory (or 'buffer') is full. Once full, the shot rate falls off dramatically. To something like one fps (depending on camera model). As the camera records, the red 'busy' light comes one. How quickly your memory card performs can be gauged by the time it takes for that red light to go out. Until the red light goes out shooting is either seriously slowed or, for slower (30MB/s) cards, stopped completely.
The time taken for the red light to go out (to clear the buffer and start rapid shooting once again) is where you see the difference between a 30MB/s card and a fast 95MB/s card (pictured above). Expect to get significantly different times between shooting/saving RAW and JPEG + RAW files.
What interested me was that the Extreme Pro CF 90MB/s card recorded significantly faster than the (supposedly faster) Extreme Pro SDXC 95MB/s card.
CF Extreme Pro = 45Mb/s [JPEG] and 60.6 Mb/s [RAW]
SDXC = 17.9Mb/s [JPEG] and 19.3 Mb/s [RAW]
So (I thought) what's going on? Why the significant record difference between identical cards? I did a little digging and discovered that it's not so much the card that's slow, it's the compatibility with the latest SD card speed rating - UHS-1. That's where the bottleneck occurs. If the card is UHS-1, the device that reads it must also be UHS-1 compliant otherwise it defaults to a slower speed. While Canon supports the latest UDMA speed standard for CF cards, it appears NOT to support the latest UHS-1 standard for SDXC cards.
Why support UDMA and NOT UHS-1? Back to my original question. What's the point of having two card slots if they perform at significantly different rates?
Moving through the forums and I found several individuals voicing the opinion that this might be a form of protectionism from Canon. The Canon 1DX has twin high speed CF slots. If the 5D Mk3 had twin slots that performed at the same speed as the top-of-the-range Canon DSLR, it might eat into that pro market. Speculation.
Here are some results from my simple tests:
SanDisk Extreme III 4Gb Card = 12 RAW, 27 secs to clear buffer
SanDisk Ultra 30Mb/s 8Gb = 12 RAW, 16 secs to clear buffer
Lexar Professional UDMA 300x 8Gb = 12 RAW, 12 secs to clear the buffer
SanDisk Ex Pro UDMA 7 32Gb 90Mb/s = 12 RAW, 26 secs to clear(?) buffer Lexar Professional UDMA 7 1000x 16Gb = 24 RAW, 4 secs to clear buffer
I did a few more tests and concluded that my new SanDisk cards were certainly not what the label stated (bought off eBay for $100 each. Sounded cheap at the time...).
Another problem facing users of fast cards is in reading the data from card to computer. Regular old fashioned USB 2.0 readers work, but not optimally. If you have a Mac, you'll appreciate that it doesn't support USB 3.0 - so I looked around for a decent Firewire CF card reader - these are hard to find.
I eventually tracked down one - made in Germany (Delock) and distributed here by Borge Imaging. Not cheap, at $69, but fast, at up to 800Megabits/s. One gripe: once installed, I still had to download a new (Mac) driver to get it working at the right (fast) speed. For PC users, USB 3.0 readers are rated as being even faster.
To get an idea of how your memory card should be performing, check out Rob Galbraith's Website [www.robgalbraith.com).
On it you'll find a comprehensive database of CF and SD card tests.
It's a great resource.