I have been testing out my new Komputer Bay 128GB 1050x card in my 5d3. Running benchmarks in ML, I am seeing some strange results. At first it made me think that the 72MB/sec max write speed I was seeing was a flaw in the CF card, but on making a few tweaks to settings on the camera I am seeing varying results on the benchmark test - so its making me think that maybe the card is capable of the advertised 100MB/sec write speeds, and maybe the overheads imposed by the firmware/operation are causing the loss of speed in benchmarks. This is what I ran and my findings:
Test 1: Global Draw - on, All other overlay function settings - off, liveview mode - on
So max write speed here is 77.7MB/sec
Test 2: Global Draw - OFF, All other overlay function settings - n/a, liveview mode - on
Test 3: Global Draw - OFF, All other overlay function settings - n/a, liveview mode - OFF (so the switch on the back was flipped to regular viewfinder mode as you would have generally for stills):
Bizzarely, the benchmark managed 92.3MB/sec here.
This shows that the display options have a huge effect on ML's write performance, and I'm thinking now that maybe the Komputer Bay card is capable of its advertised 100MB/sec write speed, but the software is just not capable of being able to write data that fast due to overheads.
Should I send the card back due to thinking it is not capable of attaining the advertised write speeds? The performance when shooting RAW does appear to be rather random/erratic.
Comments on a postcard please...
