• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

complete system on chip?

J

jhu

Lifer
now that the .13 micron process is ramping up, how come we don't see anyone working on a system on a chip? i'm talking about everything but the connectors on said chip. perhaps a highly clocked 486, memory controller, 64mb memory, video, etc., all on one chip.

while we're at it, how about multi-processing using the pentium core? it only has about 3million transistors. you can probably fit at least 4 of these things in the same space that a pentium 4 takes up
 
ok, so there are systems on a chip devices already (ah, the wonders of google). but what about putting multiple pentiums on a die and having multiprocessor capability?
 
The main problem of SoC is testing....you're increasing integration, but you've got less primary inputs/outputs, which leads to a higher testing/verification time.
 
The highest integration SoC there is currently is SiS 550. This is their earlier 540 all-in-one chipset, with a Rise M6 CPU pulled in. You need RAM, a BIOS ROM chip, and if you still need the legacy ports, a super-IO chip. Plus a handful of physical interface chiplets for audio, LAN, 2nd display head and other minor stuff.

regards, Peter
 
The problem is that companies that are good at designing processors are not always the same companies who are good at designing graphics chips.
I think eventually it will be more and more integrated.
 
What everyone else said plus the fact that it would be slow as hell.

Carlo



<< now that the .13 micron process is ramping up, how come we don't see anyone working on a system on a chip? i'm talking about everything but the connectors on said chip. perhaps a highly clocked 486, memory controller, 64mb memory, video, etc., all on one chip.

while we're at it, how about multi-processing using the pentium core? it only has about 3million transistors. you can probably fit at least 4 of these things in the same space that a pentium 4 takes up >>

 
Also, as seen with the Hammer, once you start putting 5 - 10 components on the chip, if one fails then it is worthless.
 
Also, as seen with the Hammer, once you start putting 5 - 10 components on the chip, if one fails then it is worthless.
 
I believe the multi-processor has been tried as well. If not, it was on some drawing boards. Integration has been counter to the PC worlds
idea of incremental upgrade. You also need to have control over a large range of technologies such as video, chipset, processor. These sorts
of endeavours usually fail miseably. Look at mediaGX and all the wanted to do was do some multimedia integration. Even the great Intel
had a miserable SoC showing.

I still think this will slowly become more and more "normal". Look at the number of discrete chips on a motherboard today compared to 5 years ago.
Chipsets typically consist of one to 2 major chips now. AMD is beginning to bring more things "onboard". Like the memory controller in Hammer.
It will happen mainly because economics will drive it, not for any more noble reason. The less chips you have the cheaper things can be. I think
the major drive will be towards a single chip on each major component. Meaning processor, single chip motherboard, single chip graphics card.
Which we are very close to today.

The biggest issue is making small steps towards integration. As has been said, you don't want to lose performance or degrade reliability.
 
Well ... Intel's approach has been cancelled because it apparently was technically broken (part of the initial RAMBUS desaster), and/or delayed so far that it became pointless for its intended market. Cyrix MediaGX suffered from the latter problem mainly, but it has survived and found its market niche where power consumption, cost and size matters - after National Semiconductor adopted it, put it on a more modern fab process to bring its power consumption down and speed up (it's at 333 MHz now!). I've been involved in development of a board using it, and must say it's surprisingly decent.

Integrated-CPU chip solutions don't make sense in the normal PC market - this one moves too fast. Single-chip mainboard chip"sets" have been around since 1997, with an increasing amount of stuff pulled into the chip. Single-chip graphics are popular in the slow-moving notebook and industrial market, but useless on the desktop ... chips with enough integrated RAM to be bought by the enthusiast would be way too hard to make, too expensive, and always too late.

Hammer pulling in the memory controller is about performance.

regards, Peter
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top