Congrats on 10k, @RoryMcCune

Another attempted impersonation of a high-rep user?

(@avid, @RoryAlsop)

@Terry, no- I think those are his real initials.

@RoryAlsop There are only 676 combinations of two initials, so there is bound to be some collisions.

If we see another Rory, now that would be a most improbable coincidence.

@TerryChia CPU consumption is quadratic with the clocking rate (all other things being equal). Therefore, if you want more computing power with less heat generation and longer battery life, it makes sense to use more cores -- provided that software is comfortable with the increased architecture parallelism.

@ThomasPornin No arguing that. My main quibble is that consumers just don't need that much processing power in a smartphone in today's context.

Hell, the Tegra-3 chip in my Nexus 7 can handle the most demanding tablet games out there just fine.

I'd rather they focus on reducing power consumption while maintaining the current level of processing power.

There again, multicores. Double the cores, halve the clock rate, and you get twice the battery life for constant CPU power.

@ThomasPornin Heh. In my country at least, a phone is much higher up on my need-list than running from lions.

Oh yeah, asian lions have become pretty rare. Also, Singapour would be too far south for tigers, I believe ?

@ThomasPornin I'm not getting the impression that manufacturers would be halving the clock rate. In any case, is the software optimization will still be the biggest issue.

Not counting the whole urbanization business (wild animals are more often seen, as it were, in the Wild).

@ThomasPornin Yeah, no wild tigers here. Almost nothing wild can be found here.

If software could be parallelized at will, we would already have 10000-core processors clocked at 4 MHz.

Won't there be engineering difficulties with 10000 core chips? ;)

ten thousands of 6809E CPU would be a totally awesome architecture

interconnect would have to be thought out carefully

A 6809E CPU is about 9000 transistors. 10000 of them would still be 90 millions of transistors, while modern x86 are now in the billions range.

More problematic would be the RAM. If you have many cores, you need to give each core a few kilobytes of private RAM, at least for storing the instructions for that core.

Each bit of RAM is 6 transistors (unless you find a way to cram DRAM in a CMOS chip, which is what IBM does with the Cell CPU; it seems to be doable but hard, and patented).

A 6809E with 8 kB of static RAM means half a million transistors or so.

mmh, more like 400000.

At that point, you'd better put a 32-bit 68000 or ARM core, it won't change things much

and 10000 cores with RAM would need 4 billions of transistors, which is in the range of the industrially doable.

Things are cheaper id the cores run in SIMD mode (many cores using the exact same instructions, synchronously). You need much less per-core RAM. At that point, you have a GPU.

And GPU do exist. I have a few of them at home.

Programming for SIMD is a bit hard, though.

@ThomasPornin What's this newfangled thing called the GPU?

@TerryChia It is an invention of the Devil (the Gimmick to Pervert Users). It makes them contemplate buying smartphones instead of sensible things like books.

@ThomasPornin and computers to play games instead of program :p

@LucasKauffman if you like it really smokey and peaty - give laphroige a try. my favorite...

if you find one smokier or peatier - or if @RoryAlsop says so, of course - I will go right out and buy it.

Laphroige is like drinking liquid bbq.

or drinking a cigar, if you like that.

and regarding the stress in particular - besides the whole dealing with crap and crap cow-irkers - I've realized that in professional services, there is always a big important urgent project.

there will always be another emergency.

my reaction - yeah, so?

I used to ask the salespeople - "is it urgent?" when they say yes, I respond with "Oh, we don't deal with urgent. When it gets moved up to burning, give me a call."