Intel Chip Overhaul Coming

[pop pop]

From the Wall Street Journal (http://online.wsj.com/public/article/0,,SB112380688787511502-nooFSw53SaO_RyjT6yfrOcG4UMw_20050911,00.html?mod=b logs)

For those who don't read the Journal, yes, they do have a good technology section.

Here's a couple of exciting quotes:

"A big emphasis is going to be performance per watt," said Bill Calder, an Intel spokesman. "That is a very big deal."

Can you say Cool :) And Mobile. :)

And this:

"I assume they will be able to go from two to four to eight cores on a chip in a fairly smooth and nondisruptive manner,"

The time to market is estimated at 18 months.

About a year ago I said to some fellow engineers at work that it's my belief that we are on the crest of a wave in computing power that is simply staggering. To put it into perspective, consider this: most of us already have more computing power just on our desktop than all of NASA did when the US first put a man on the moon. :eek: :eek:

Now, if we could only think of something more useful than games to use all this power for. :confused:

[jlreich]

Now, if we could only think of something more useful than games to use all this power for. :confused:
What could possibly be more useful than games? :p

It is staggering to think not much more than a year ago, an Athlon XP 3800 was a top end machine. Since then we have went from 32bit to 64bit, and now to dual core 64bit. Terabyte HDD's, GT7800, HDD's the size of a quarter. :eek: Now if the software end of things could just catch up a little.....

The advancement of technology over the last year or so has simply blown me away. I look forward to what the next year or so might bring. :cool:

[saphalline]

I look forward to what the next year or so might bring.That's easy - check the roadmaps! Intel will begin to lean more heavily on the 1066 FSB by the end of next year, and will open up a few more avenues in variation. They're still planning on making a 32-bit dual-package version of the Celeron for the mid-range, based on the belief that two 32-bit cores will benefit consumers long before 64-bit becomes necessary. I agree with them. :D Intel will also continue the P4 line and expand the PD line. The single-core Celeron will be kept (for now) as the cheap baseline CPU. On the mobile end, next-gen Centrino stuff is looking really cool, but there's also a lot of different rumours out there. Only time will tell. But mobile versions of the PD will hit DTR's before too long (perhaps even by this holiday season).

AMD is going to attack the dual-core arena more aggressively, and they've already started with the introduction of the X2 3800+. The A64 line will, of course, continue. And their Semprons have finally earned back 64-bit support! Performance seems to be lack-luster right now, but with the cut-backs in the Sempron and the single-channel RAM thing, this is to be expected. Besides, Semprons are dirt-cheap, so who cares? On the mobile front, the Turion64 is having problems penetrating the market. I'm not surprised, and neither is AMD. They knew this would happen. After all, it took them how long to gain enough of a following to compete with Intel? It's predictable that the mobile market isn't warming up to them quite yet. But they continue to push on, and the Turion64 isn't at a total disadvantage to the PentiumM, but its target is a small one (between PM and P4-M).

In the coming years, expect 64-bit across the board, and I'm guessing that quad-core CPU's will be released even before the last single-core CPU is discontinued. But that's just a guess...

[jlreich]

Cool, quad-core. :cool:

What about more channels for system ram as you spoke about on v-ram? Any chance of that in the near future? Sounds like the technology is there, just a matter of is it practical for system ram. Perhaps a bios setting or even jumpers to switch between dual and quad as needed depending on how many sticks you have.

Wouldn't it be nice to have quad channel ram. :D

[saphalline]

Oooh, I don't think so. System RAM and vid card RAM are two entirely different beasts. Implementing virtual channels and other optimizations used on vid cards would complicate the memory controller beyond current capabilities. One of the reasons that system RAM is so much slower than vid card RAM is because of its flexibility. You can use many different speeds of RAM in many different sizes with all sorts of combinations of modules. If you start using optimizations that vid cards use, you run into serious limitations in terms of how much RAM you can have, what speed they run at, their latencies, their chip density, number of banks, etc. Not only that, but have you noticed how often vid cards change their RAM type? :p We're on GDDR3 now! DDR2 and DDR are already gone! :eek: If you want faster system RAM, you gotta be prepared to replace it very often and shop for a looong time to find the exact sticks you would need. Besides that, but you need a new memory controller every time something is added. AMD had enough problems tweaking their K8 onboard memory controller for full RAM loads, and there's still no word on when they'll have DDR2 support. Intel meanwhile has been releasing chipsets faster and faster as new technologies surface, forcing them to design a new memory controller almost every 9 months! It isn't easy to maintain this level of cut-throat advancements. And quad-channel itself would only add to the misery, as hundreds of people out there wonder why they need to buy 4 sticks of RAM every time they upgrade. :rolleyes: I have too many problems already trying to explain dual-channel RAM to the newbies!! ;) Quad-channel might make their heads explode! :eek: :D

More RAM bandwidth is definitely needed soon. We're gonna start feeling the pinch as dual-core picks up momentum, and quad-core will only hurt us more. But I don't think virtual channels are the right move. I don't think quad-channel is a good idea, either, but if DDR2 can't pick up the pace here, we might end up with quad-channel...