Well, I've checked out the official TI Kernel v3.2 and now I get speeds well beyond 20MB/s. Does anyone have a clue why the newer Kernels are so much worse?
On Sunday, November 10, 2013 9:47:32 PM UTC+1, thoma...@gmail.com wrote: > > Well, this page > http://processors.wiki.ti.com/index.php/AM335x-PSP_04.06.00.06_Features_and_Performance_Guide#USB_Drivercertainly > suggests that the BBW can do much better. I'll try to investigate > this further ... > > On Tuesday, November 5, 2013 3:53:42 PM UTC+1, Juanjo wrote: >> >> It seems 3.12 has DMA support enabled for USB (3.8 isn't) so it should be >> a bit better or at least not choke the CPU on interrupts. >> >> Anyway it seems USB isn't the brilliant part of the BBB :( >> >> Maybe the BBW with 3.2 runs faster on USB. >> >> On Saturday, November 2, 2013 12:22:45 PM UTC-3, thoma...@gmail.comwrote: >>> >>> I'm trying to use the BBB as a (very) lightweight NAS. Unfortunately, >>> I'm a bit disappointed with the transfer speed I can achieve over the USB >>> port. I've now tried different kernels such as Robert C Nelsons >>> 3.8.13-bone28 and 3.12.0-rc5-bone6, different USB storage devices (such as >>> a USB 2.0 flash drive and a USB 3.0 hard disk), and tried both directly >>> attaching the flash drive to the BBB as well as using a powered USB hub. >>> The HD of course only works on the hub. The BBB is running on a 5V external >>> power supply. >>> >>> Using different tools (hdparm and dd) the USB read speed always maxes >>> out at about 9MB/s which is far less than the same devices deliver on other >>> systems (e.g. about 25MB/s on my Raspberry Pi). >>> >>> Does anyone have an idea why this is the case or how this can be >>> improved? >>> >> -- 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/groups/opt_out.