Pretty much disappointing numbers. ARM architecture is crap 21 Окт 2015 г. 18:49 пользователь <i.sarf...@gmail.com> написал:
> Hi, > > > We are using Micron eMMC part MTFC4GMDEA-4M (4GB version) and we are > getting really low R/W numbers on custom vybrid board (Freescale Vybrid > MVF51NS151CMK50 > <http://www.freescale.com/products/arm-processors/vfxxx-controller/f-series/arm-cortex-a5-based-microprocessors-with-1.5mb-sram-lcd-security-2x-ethernet-l2-switch:VF5xx?&cof=0&am=0&tab=Buy_Parametric_Tab&fromSearch=false> > ). > We are using custom software (non-Linux) on this board. > > Micron eMMC: > 25MB data transfer in 256KB R/W blocks. Multi-block transfer enabled in > driver, using DMA. > Read *14*.0964 MBytes/sec : Write *5.5*438 MBytes/sec > > While on 1GB MMCplus card we getting R/W > Read ~ 23 MB/s , Write ~ 10 MB/s > > Can anyone point out any special fix we need to do for eMMC as MMCplus and > class 10 SD cards are performing double than micron eMMC. Although eMMC > datasheet claims 40 MB/s and 20MB/s for R/W but we are not even getting > half of that. > > Regards, > Sarfraz Ahmad > > > On Wednesday, June 25, 2014 at 11:50:56 AM UTC-4, David Paden wrote: >> >> Hi, >> >> Just tested on my BeagleBone Black A5C with 2GB micron eMMC. Tested with >> latest Debian Beta booting from SD card (Samsung EVO 32GB Class 10). >> Here's my numbers: >> >> SD Read: >> root@beaglebone:~# dd if=zero of=/dev/null bs=200M count=1 >> 1+0 records in >> 1+0 records out >> 209715200 bytes (210 MB) copied, 10.9215 s, 19.2 MB/s >> >> SD Write: >> root@beaglebone:~# dd if=/dev/zero of=zero bs=200M count=1 >> 1+0 records in >> 1+0 records out >> 209715200 bytes (210 MB) copied, 15.297 s, 13.7 MB/s >> >> 2GB eMMC Read: >> root@beaglebone:~# dd if=emmc/zero of=/dev/null bs=200M count=1 >> 1+0 records in >> 1+0 records out >> 209715200 bytes (210 MB) copied, 10.4306 s, 20.1 MB/s >> >> 2GB eMMC Write: >> dd if=/dev/zero of=emmc/zero bs=200M count=1 >> 1+0 records in >> 1+0 records out >> 209715200 bytes (210 MB) copied, 39.1959 s, 5.4 MB/s >> >> >> I have 4GB Micron part (MTFC4GMDEA-4M) on my custom AM335x board (similar >> to BeagleBone Black design) and get this on it's eMMC (booted to eMMC): >> >> 4GB eMMC Read: >> 1+0 records in >> 1+0 records out >> 209715200 bytes (210 MB) copied, 7.65176 s, 27.4 MB/s >> >> 4GB eMMC Write: >> ./dd if=/dev/zero of=zero bs=200M count=1 >> 1+0 records in >> 1+0 records out >> 209715200 bytes (210 MB) copied, 20.3072 s, 10.3 MB/s >> >> Thanks, >> Dave >> >> >> On Sunday, June 8, 2014 7:59:04 AM UTC-4, vescov...@gmail.com wrote: >>> >>> I was wondering if anyone has any number comparing the 2GB eMMCs >>> (original BBB) and the 4GB (Rev C) eMMC speeds? >>> >>> They claim it is faster but would like to see proof. >>> >>> When I compared the Micron 2GB eMMC vs. the SanDisk iNAND parts of the >>> same size the SanDisk parts smoked the Micron. >>> >>> I found boot and R/W times on the Micron parts are awful. So much so >>> even some off brand off-the-shelf standalone uSD cards show much faster >>> boot times (just the reading phase). >>> >>> >>> >>> Both the SanDisk and the Micron parts have the Double Data Rate (DDR) >>> feature which would speed thing up greatly but this ability on the AM335x >>> seems to be broken. >>> >>> TI claims DDR was not implemented and should never have been included in >>> the TRM ..which I see they now have amended. >>> >>> >>> >>> Any numbers? >>> >> -- > For more options, visit http://beagleboard.org/discuss > --- > You received this message because you are subscribed to the Google Groups > "BeagleBoard" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to beagleboard+unsubscr...@googlegroups.com. > For more options, visit https://groups.google.com/d/optout. > -- For more options, visit http://beagleboard.org/discuss --- You received this message because you are subscribed to the Google Groups "BeagleBoard" group. To unsubscribe from this group and stop receiving emails from it, send an email to beagleboard+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.