Date: Sat, 14 Oct 2006 03:46:22 -0700 From: soralx@cydem.org To: freebsd-hardware@freebsd.org Subject: Re: Quiet computer Message-ID: <200610140346.22554.soralx@cydem.org> In-Reply-To: <78ED28FACE63744386D68D8A9D1CF5D4209C94@MAIL.corp.lumeta.com> References: <78ED28FACE63744386D68D8A9D1CF5D4209C94@MAIL.corp.lumeta.com>
next in thread | previous in thread | raw e-mail | index | archive | help
> single core. Yes, I realize that the woodcrest is faster than above > mentioned cpu, but a 100x speed difference? That doesn't seem realistic > to me (although if those are valid results, I'd be pretty happy with > that)... Even if VIA's true RNG is actually being used, then I see how this could be realistic. The pseudo-RNG in you machine constitutes a mathematical algorithm with an initial random seed (yarrow, if I'm not mistaken). You have a fast CPU, and it crunches numbers rapidly; thus, you get that impressive speed. For a true RNG, every bit provided is original, random (not a result of some transformation). So, if the hardware portion has limited bandwith (randomness 'regenerates' slowly -- bits don't change fast enough), or there's delay along the path, then RNG device speed will be limited -- and therefore encryption speed as well (if I'm not mistaken). But if the randomness source is slow, it's not a big problem. For example, a chip can be used that's meant to transform, or 'dilute', the truly random stream somewhat to increase speed. Hmm... OTOH, noone yet complained about quantum processes being too slow, so I suppose /dev/urandom is not what we want to test. When we see some performance data for the VIA CPU, then we'll determine whether it's TRNG spitting out the random bits or is it the yarrow that's increasing entropy :) BTW, David, test (if you have enough time for curiosity) /dev/urandom (or wherever the generator is) with different values of 'bs', from very small to large (1 byte, 512, 4k, 16k, 64k) to see how much it depends on block size. > bash-2.05b$ time dd if=/dev/urandom bs=1024 count=10240 of=/dev/null > 10240+0 records in > 10240+0 records out > 10485760 bytes transferred in 0.154649 secs (67803598 bytes/sec) >Bucky [SorAlx] ridin' VN1500-B2
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200610140346.22554.soralx>