<div class="gmail_quote">On Thu, Dec 17, 2009 at 12:52 PM, Brian Wood <span dir="ltr"><<a href="mailto:beww@beww.org">beww@beww.org</a>></span> wrote:<br><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
<div class="im">On Thursday 17 December 2009 10:47:12 am Mike Perkins wrote:<br>
<br>
><br>
> Also if you're doing a lot of commflagging or transcoding, I would guess<br>
> that extra RAM could make life easier as well.<br>
><br>
</div>I wouldn't argue against the fact that Myth "can use" more RAM, but it doesn't<br>
"need" it. Certainly I wouldn't go with a 64-bit system solely so Myth could<br>
have more than 4GB of RAM, unless you needed it for something else in<br>
addition to Myth.<br>
<br>
I also wonder if more RAM on the graphics card might be more useful than main<br>
system RAM, with VDPAU.<br>
<div><div></div><div class="h5">____________________________<br></div></div></blockquote></div><br>Caveat: I'm assuming we're talking about x86 vs x86-64<br><br>People often forget that the x86-64 provides twice as many general purpose registers as 32-bit x86. They have twice as many SSE registers as well. Depending on workload, performance can be either worse (due to larger pointers) or better (due primarily to the increase in registers). <br>
<br>If I recall, transcoding is one of the poster children for running 64-bit. I don't have numbers on hand but I definitely noticed an improvement. I think 15% improvement is the general rule of thumb although I've heard of 30% improvements when transcoding as well. As with all such things, your mileage will probably vary :)<br>
<br>--Jack<br>