Re: 2017-06-20-netbsd-raspi-earmv6hf.img

[Mark Millard <markmi%dsl-only.net@localhost>]

On 2017-Jun-24, at 4:04 PM, Mark Millard <markmi%dsl-only.net@localhost> wrote:

> On 2017-Jun-24, at 3:47 PM, John Klos <john at @ziaspace.com> wrote:
>> 
>>> Not the best of examples:
>>> 
>>> 25 * 1024*1024 = 26214400
>>> 
>>> 26222722
>>> vs.
>>> 26214400
>>> 
>>> could well be in the noise for the time accuracy
>>> or how exact the clock rate is vs. nominal.
>>> 
>>> (I'm not saying that you are wrong, just that the
>>> example is not clearly faster than a nominal 25 MB/s.)
>> 
>> True. But considering there's overhead, and if the clock really is set to 50 MHz (sc->sc.sc_clkbase = 50000), then this should be showing a number greater than the maximum possible at that clock rate. I don't have any faster cards than a Samsung EVO +.
> 
> Well for a 32GiByte EVO+ on a macOS machine via dd:
> 
> 8388608000 bytes transferred in 100.740420 secs (83269536 bytes/sec)
> 
> 83269536
> vs.
> 26214400
> 
> (A clear difference.)
> 
> I used bs=4m like you did. I did not wait for
> the full /dev/rdisk3 to complete and so it
> is a partial transfer result.

In:

https://www.sdcard.org/press/past_evens/pdf/SD_Standards_and_Technology_GWTaipei_Oct2014.pdf

the macOS speed would be an example of:

UHS-I SDR104 208MHz 1.8V 104MB/s (spec version: 3.01)

It takes the extra pins on UHS-II to get beyond that
category (as of that document anyway).


===
Mark Millard
markmi at dsl-only.net