View Full Version : The cool new technology thread

Sir Penguin
08-08-2003, 02:35:43
1) The fogscreen (http://www.fogscreen.com):
Some Finns have created a prototype for projecting an image onto a gaseous screen. It can be projected from either the back or the front, and the presenter can walk through the screen and stuff. The website has some pictures, and promises improvements in the near future. It's not exactly new technology, but the inventors have recently formed a company to market it, so it should get more popular starting now.

2) "Human batteries" (http://www.smh.com.au/articles/2003/08/03/1059849278131.html):
Some Japanese guys are developing a way to get energy from glucose in blood. They say one could get up to 100W from a person, provided that that person doesn't actually do anything (like live, I guess). The Ars Technica article (http://arstechnica.com/archive/news/1060008809.html) I got this from suggests powering pacemakers and artificial hearts with this technology, as well as built-in insulin regulators for diabetics.


No longer Trippin
08-08-2003, 04:17:23
100 watts, that's just enough juice to get the prescott core to turn over.

08-08-2003, 04:26:37
I saw the inquirer blurb about Intel reaching the century mark. I imagine that someday PCs will replace water heaters in most houses. :)

No longer Trippin
08-08-2003, 06:03:59
I honestly don't know how they are going to keep it cool. My 2500 AAUIB barton (first release) pushes nearly 100 watts due to having to pump 1.9 volts into it to get it to 11 x 200 - used to go higher, but nb got picky - should have watercooled it. On that puppy I have an SLK800U, and 80 to 120 adaptor (industrial glue to attach it) and Delta 190 CFM screamer wire tricked down to about 100 CFM I'm guessing and rather quiet - though I have the fan tied to the truss across the chenming case for extra support). That's the only way I can keep the temps sensible without it sounding like a windtunnel in there. The Prometia has been sold, so that is out the question. Runs fast enough and the current setup works, so no complaints now.

I wonder what Intel is going to do for a heatsink. Even moving to the .09 micron process won't reduce the temps by the 25% or so it needs to run cooler as the die will be even smaller, thus less contact so the heat still will be a huge factor. Are they TRYING to melt the ZIF socket? Hell, most overclockers go to watercooling if they are even getting close to 100 watts... can't see dell shipping comps with watercoolers. You want the average user to fill the setup, bleed it properly, and fill it when it gets low and bleed it again? They also aren't going to ship them prefilled - either way is a nightmare waiting to happen.

Good luck on aircooling it - and then the faster ones - those will be drawing even more.

I'm just looking for the prescott to come out so I can grab a 3.2C for cheap after several months. Though the 2.4C's have been known to run at stock vcore to 3.2 - just as the new 2500 bartons will hit the 3200 ranking at stock vcore now as well. The Prescott is sadly just a slightly revised northwood core. For them to drop the temps due to the required wattage would mean they'd have to scrap the entire core and start fresh, and they aren't going to do that - they've already dropped the idea of adding the 64 bit interface to it, guess as that would just add more heat. Don't see how they are gonna market this things as it's gonna be loud as hell, or the cooler will be all copper and extremely massive along with having a huge fan attached as well. They've stolen the heater title from AMD with the 3.2C - now this will just cap it off if current rumors about it needed around 100 watts to even turn over and run are true.

08-08-2003, 12:51:20
The 100W figure was on the 90nm process.

The problem is gate leakage gets expontentially worse, a minor problem on 130nm is a major one on 90nm.

They're trying a new stepping to bring it down.

And the Prescott is more than a "slightly revised" Northwood core -- it's fully 64-bit internally (though not enabled), doubles the numbers of ALUs, splits up the FPU and SSE units, adds 13 more instructions, has a better branch predictor, 1MB of L2 cache, etc.

08-08-2003, 13:01:18
How many humans would we need to power a machine empire? Just curious.

Qaj the Fuzzy Love Worm
08-08-2003, 15:12:31
Wow, so now our futures could end up either as The Matrix, or Seaquest DSV. Or both! Neat :)

No longer Trippin
08-08-2003, 22:54:37
It's based of the northwood core... didn't know it needed 100 watts at the .9 nano process, i thought that was for the .13 micron - though admittedly I haven't been keeping track of it that much. I just knew it was based on the northwood core and anything to reduce the wattage would require scrapping the core most likely. So 32bit - sounds like the zeons will be the first the to get the 64 bit interface - though heard rumors that they'll be at a 667 fsb (the C core did have a drop in gain from fsb around that point going by the anandtech review of the 3.2C IIRC, also heard those zeons with a 64 bit interface are due in 2005 sometime, again from anandtech.

.9 nano at 100 watts - man that is going to be a cooling nightmare for stock. If it was at .13 - it would be doable, maybe cheaper than I have thought previously due to the extra die area - not to mention they might have to leave it without a heatspreader as those are known to raise temps a few degrees - or perfect it a bit more. Still may have run a little noisy compared to current stock sinks, but at .9 micron - it's going to be louder than most normal users want to hear unless they design some small breakthrough in air cooling. .13 will require more juice, though more contact area.

Your right about the elctromigration - the smaller the process, the harder it gets to isolate. And the more current it's drawing just adds to that problem. SNDS comes to mind again.

Adding the 1mb cache isn't that much of a change. That's what turned the B core pentiums into being competive with AMD's XP's at the time as an 1800XP would generally beat a 2ghz A core in quite a few benches - after going to the B core though, the XP was about par, a bit slower, but a lot cheaper... though once you got to around the 2400 ranking or so, the Intel chips of 2.4ghz at the C core especially, AMD lost a good deal in the rankings... I mean the 2500 barton was the only one that ran close to it's nearest couterpart, the 2.4C in benches, and the C would kill it in some, though the barton still held it's own. Just the 3200 is a joke. A 2.8C stomps it due to the fact that it's only 220 mhz faster than the 2500... that doesn't equate at all to the approximate 30% greater usage of clock cycles AMD has at all. If the A64 and Opteron don't take off and get them a few quarters of being running the black (though they aren't gonna be out of dept, that's for sure), AMD looks like it's going to be eaten up by IBM - they won't keep AMD, but take the best of the engineers at the least. At the Dresden plant maybe, but by then it may be a liability as thats a decent ways down the road. AMD needs its next crop of chips to succeed.. else investors will eventually stop feeding them money.