One further note. With the Microcenter card the benchmark shows seek times of 0.0ms, which is probably an error. The SanDisk one shows seek times of approximately 1.7ms. > On Feb 9, 2015, at 7:23 AM, Justin <shadow@darksideresearch.com> wrote: > > Ok, today I swapped the crappy Microcenter brand 8GB “extreme speed” model for a 16GB SanDisk Ultra rated for 48MB/s. With a high quality SD card reader, such as the ones built into some of the current MacBook Pro and Air models, these cards will write at their rated speed for sustained writes. I can say that random write performance fall off on them as a function of size (bigger equals worse random write). > > Peak sustained write and read speeds are so close as to be indistinguishable, roughly 1800KB/s read and about 1400 KB/s write. Where you start to see differences is with random writes, and with number of operations per second. The SanDisk will do nearly 3x the number of random 512 byte writes per second (a bit over 300 vs about 100), and its peak random write speed is 5x higher as a result of performance increasing up to the 20KB write size. Both perform the same up to about 5KB writes, then the Microcenter card falls down to near 0. The SanDisk peaks at 20KB write size, with the SanDisk doing over 1500KB/s. There is a hard fall off above 20KB/s, all the way from 1536KB/s down to near zero. At that point they both are writing at maybe 20KB/s, which slowly moves up to maybe 60-80KB/s by the time you get to write sizes of 100KB. > > Based on this performance, there seem to be some hard limits within the SCS2SD board. Based on the benchmarks posted as the “current” performance limits for the SCSI2SD it looks like even with a high quality card I’m not getting the larger block size write performance that is described. I will try to make adjustments so that I can manually enter the manufacturer data in with the new GUI (which is required to talk to the newest firmware but lacked the Mac compatibility switch that spoofs an Apple SCSI drive from the older one) to see if that frees up random write performance for larger write sizes. > >> On Jan 9, 2015, at 7:53 PM, silverdr@wfmh.org.pl wrote: >> >> >> >> On 2015-01-09 22:21, Justin wrote: >>> I wanted to follow up on this and let everyone know how this thing worked out. >> >> Justin, thank you for the follow-up. Interesting read. Please do check with a truly good card. If you mind the cost, maybe get a smaller but still a fast one from a proper brand manufacturer and rated for speeds higher than the SCSI limits so that we can clear the doubts. >> >> -- >> SD! >> >> Message was sent through the cbm-hackers mailing list > > > Message was sent through the cbm-hackers mailing list Message was sent through the cbm-hackers mailing listReceived on 2015-02-09 13:01:20
Archive generated by hypermail 2.2.0.