Slashdot Mirror
Intel's Answer to AMD's Hammer - Yamhill
bdolan writes: "Today's San Jose Mercury News is reporting that Intel is going to put a 64 bit architecture extension in upcoming Pentiums if it turns out the Itanium doesn't take off. Hmm. Apparently they intend to only turn this on if AMD's 64 bit processor make major inroads against the Itanium architecture. Aren't we glad that competition is keeping everyone on their toes."
-- www.globaltics.net
Political discussion for a new world
Yeah MicroSoft is so fast and easy, accept when it gives you the blue screen of death, forceing you to do a hard reboot ½ the time and lose all the data that you have been working on! I really liked Netscape when it was 4.0 it was actually stable! Internet explorer just blew up over and over. Windows is such a falty program its pathetic, they are finaly getting it more stable on what there 7 edition??? but even that had a major bug! woohoo "time for me to stop ranting"
Stop overclocking your 3 year old generic motherboard, and maybe you won't blue screen or have problems. I blame the hardware manufacturers for writing bad drivers before I blame M$ for writing bad software. Its bloated, but it runs fast enough and stable enough for me. I haven't lost a project or work time in probably 6 years since 95 came out.
One of my companies maintains about 600 computers within 15 different organizations, all running different, badly written software. We get maybe 1 call a month about a BSOD, and even that's overstating it...
Intel have the technology to use 12" wafers, on which they etch the processors. Now, most of that packaging that you see, when you look at a CPU is just that. Packaging. Space to put all the pins, for the most part. The chip inside is unlikely to be much bigger than a square inch, if that.
(For the sake of argument, let's call it a square inch.)
This means a wafer will hold (at most) 452 processors, if you can get the shape right, and 144 processors if you're as good at geometry as Dan Quale is at spelling.
"But they can't make wafers of that quality!"
Bullshit! You put a bit of extra work into purifying the silicon, and then let it cool just a little bit slower. It's not that hard to grow crystals.
Even if you give them the benefit of the doubt, they still get a 95% success rate, which would give you 136 processors per wafer.
TWO OR THREE processors on a chip? They have the capability of putting well over a hundred of the damn things on a chip! What I want to know is why SMP architectures are so pathetically small! AMD can only manage two processors at a time!
Let's say you didn't want quite that many processors, but wanted a bit more pipelined cache, instead. Would three terabytes of cache be sufficient? That's what COULD be put on a processor, using nothing more than existing facilities, existing techniques and existing know-how. Or maybe you'd rather use a 4096-bit architecture, instead.
When people talk of Moore's Law "failing" at some point, they forget (or ignore) the margin between what is commercially sold and what is technically achievable, with no additional effort. The worst-case scenario I gave was a 136-node processor. That would be a third again as powerful as the entire rendering cluster used in the "Titanic" movie, without the networking bottlenecks, and squished into something the size of an old-fashioned vinyl turntable on a record player.
When I see AMD and Intel talking about "improving" their chips to support 64-bits, or supporting SMP just a little bit further, I have to laugh. Those poor, pathetic fools. SMP isn't particularly good for anything, anyway. SIMD is a horrible architecture for anything but trivial number-crunching.
But a 136-node SIMD/MIMD processor... Now, THAT would be a killer system. There would be nothing anyone could build to touch it, for a long time. A home computer would have the same power as a nuclear research facility does today. Windows might even become usable.
Will this super-proc ever be built? Nah. Not a chance. If someone -did- build it, the company selling it couldn't make money off selling upgrades for another two years, at least! More importantly, it would (temporarily) transfer so much computing power to individuals that current encryption schemes would seem very fragile.
So, what's the point to the speculation? To put the current technology into perspective. To show how we, as users of this technology, are being suckered along. Pet rocks were closer to the real thing than these lumps of half-melted beach sand. We could be doing better. We SHOULD be doing better. But enough people will buy into these quasi-64-bit regurgitated coral coasters, with their marketting ploys, that the chip manufacturers have no blasted need to GIVE us any better.
Am I done with this rant? Not quite. One more point, and then I am. Current sound cards and CD players use 16-bit, 2-channel, 144 KHz technology. The best ADC and DAC devices today are 24-bit, 8-channel, 920 KHz. Why the frigging hell are we being sold stuff that was obsolete, over a decade ago?!?! 20-bit ADC/DAC systems were already in wide circulation in 1991, for everything from synths to scientific instruments, and was already looking dated.
As consumers, we're being sold the Eiffel Tower, not just once, but every bloody time we upgrade. It's always bogus, there's always "better on the horizon", and we always fall for it. My voice isn't worth a damn, but if it was, I'd say "screw AMD -AND- Intel, give me a chip plant, and I'll show you what you COULD be using."
It's a small world and it smells funny; I'd buy another if it wasn't for the money; Take back what I paid (SoM)