By: Ricardo B (ricardo.b.delete@this.xxxxx.xx), May 15, 2013 9:31 am
Room: Moderated Discussions
RichardC (tich.delete@this.pobox.com) on May 15, 2013 8:53 am wrote:
>
>
> I think that's "lots of people" meaning 10K or maybe 100K people. But the installed
> base of desktops/laptops is hundreds of millions (the vast majority of whom are doing
> either no video transcoding, or video transcoding using the pretty-good very-fast
> hardware transcode engine).
The fact that millions don't do encoding is an argument for the encoding hardware not to be present, it's not an argument against SMT.
The fact that those who do encoding have encoding hardware does not mean they actually use it.
> >I've
> > already explained, that there is an increasing number of game engines that can use 8T.
>
> Gaming is a subset of the total market; out of that subset, gaming rigs which aren't
> limited by a low/midrange GPU are a smaller subset; and out of that smaller subset
> you're picking those which happen to be running the few gaming engines which show
> a useful speedup on 4C/8T over 4C/4T.
>
> > And ironically, when you're asked for common needs of straight out single thread
> > performance, all you can come up are examples which in the common case won't
> > stress even a single core, like web browsing and word processing.
>
> Web browsing seems to stress a single core pretty darn often. Pages don't render
> instantly.
This has gotten totally ridiculous.
You're just minimizing users that don't fit your argument and exagerating those who do.
Page rendering is limited 1st by network by far, 2nd by I/O. CPU is a far 3rd order effect.
The amount of people who want a CPU faster than a low-range Ivy Bridge so their webpages render faster are a tiny amount compared to those who do transcoding, play games, etc, etc.
>
>
> I think that's "lots of people" meaning 10K or maybe 100K people. But the installed
> base of desktops/laptops is hundreds of millions (the vast majority of whom are doing
> either no video transcoding, or video transcoding using the pretty-good very-fast
> hardware transcode engine).
The fact that millions don't do encoding is an argument for the encoding hardware not to be present, it's not an argument against SMT.
The fact that those who do encoding have encoding hardware does not mean they actually use it.
> >I've
> > already explained, that there is an increasing number of game engines that can use 8T.
>
> Gaming is a subset of the total market; out of that subset, gaming rigs which aren't
> limited by a low/midrange GPU are a smaller subset; and out of that smaller subset
> you're picking those which happen to be running the few gaming engines which show
> a useful speedup on 4C/8T over 4C/4T.
>
> > And ironically, when you're asked for common needs of straight out single thread
> > performance, all you can come up are examples which in the common case won't
> > stress even a single core, like web browsing and word processing.
>
> Web browsing seems to stress a single core pretty darn often. Pages don't render
> instantly.
This has gotten totally ridiculous.
You're just minimizing users that don't fit your argument and exagerating those who do.
Page rendering is limited 1st by network by far, 2nd by I/O. CPU is a far 3rd order effect.
The amount of people who want a CPU faster than a low-range Ivy Bridge so their webpages render faster are a tiny amount compared to those who do transcoding, play games, etc, etc.
