ATI and Fast14 - When?
- Thread starter Unknown Soldier
- Start date
I cannot suggest any detail on technical term since I am mech eng 
. Anyway, sometime annoucement on press-release likes this might lead some effect to their customer too, i.e., AGE press with NV that started debate here by Geo. But I believe ATi is using that technology in some way.
. Anyway, sometime annoucement on press-release likes this might lead some effect to their customer too, i.e., AGE press with NV that started debate here by Geo. But I believe ATi is using that technology in some way.JasonCross
Newcomer
I bring this up with ATI every six months or so when discussing whatever their latest chip is, and I always get a very vague response that sounds something like "that was only for the Xbox 360 chip." (only never that clear) And it doesn't seem to help me get a clear answer if I press them on the issue - they just sort of clam up.
Long story short: if it appears anywhere, it's in the Xbox 360's GPU, which is supposedly VERY power effecient for its size, complexity, and clock rate. I have no idea if it will appear in a desktop graphics chip. It seems to be something that is fundamental to the design on the gate/path level - in other words, not something where you can take an existing desktop chip and "apply" Fast14 to the notebook version. You need to use Fast14 libraries and tools from the start. Perhaps it'll be a part of the R600 family?
Long story short: if it appears anywhere, it's in the Xbox 360's GPU, which is supposedly VERY power effecient for its size, complexity, and clock rate. I have no idea if it will appear in a desktop graphics chip. It seems to be something that is fundamental to the design on the gate/path level - in other words, not something where you can take an existing desktop chip and "apply" Fast14 to the notebook version. You need to use Fast14 libraries and tools from the start. Perhaps it'll be a part of the R600 family?
I read last year that the 130 process fast14 chips were tapped out its still being experimented with. The fab is tring to work out all the kinks and has moved to tapping out 80n chips with fast14 process. No product names were mentioned but it was clear that r 520 never used this process.
Did't ati do more than licines the process I swear ATI bought the company . Somebody please comment on this .
Did't ati do more than licines the process I swear ATI bought the company . Somebody please comment on this .
This fast14 stuff reminds me of those isotopically pure wafers that were supposedly to give chips much lower power draw and higher frequency ceilings that AMD was said to put into production. Those were also endlessly discussed and speculated over on web forums and never appeared in reality.
I believe both of these belong squarely to pie-in-the-sky tech, if it really worked as advertised, people would be using this stuff already because of the obvious advantages. Hence, something must be wrong, and my guess is it's with the tech, not the rest of the entire world.
I believe both of these belong squarely to pie-in-the-sky tech, if it really worked as advertised, people would be using this stuff already because of the obvious advantages. Hence, something must be wrong, and my guess is it's with the tech, not the rest of the entire world.
arjan de lumens
Veteran
IIRC, those isotopically pure wafers failed in the marketplace mainly because the added cost of making such a wafer greatly outweighed the small savings you could get from making slightly smaller heatsinks/heat-spreaders. As anyone who has ever made an atom bomb can tell you, isotope separation is an extremely expensive process.
Yeah, it basically requires that you build a centrifuge that is capable of moving the heavier atoms within the solid to the outside. Not simple in the least.arjan de lumens said:
This is primarily why nuclear weapons have moved away from U235 and towards Plutonium: Plutonium is U238 (the more stable, and thus more common isotope), and bombarding it with neutrons (to make U239). The U239 quickly decays to Plutonium, which is about as easy to break apart as U235, all for a much, much smaller price (not to mention there's just a whole hell of a lot mor U238 around than U235, since most of the U235 has decayed).
The Fast14 tech is already being used by Ati, as far as I know.
And as some might have guessed there's no such thing as miracle improvement when it comes to semiconductor these days. Fast14 is not an exception.
They didn't issue any press releases for the first chip that use this technology for some reason. Some have some conspiracy theory about that, BTW.
And as some might have guessed there's no such thing as miracle improvement when it comes to semiconductor these days. Fast14 is not an exception.
They didn't issue any press releases for the first chip that use this technology for some reason. Some have some conspiracy theory about that, BTW.
Guden Oden said:This fast14 stuff reminds me of those isotopically pure wafers that were supposedly to give chips much lower power draw and higher frequency ceilings that AMD was said to put into production. Those were also endlessly discussed and speculated over on web forums and never appeared in reality.
I believe both of these belong squarely to pie-in-the-sky tech, if it really worked as advertised, people would be using this stuff already because of the obvious advantages. Hence, something must be wrong, and my guess is it's with the tech, not the rest of the entire world.
Dynamic logic was around before CMOS I believe and was mostly abandoned because it is very timing oriented and was harder to design for then CMOS. The advantage of it is that it switches much faster and uses less power. I think that intrinsity offers libraries that you can drop into your CMOS architecture to allow logic components like the ALU to run a much faster clock speeds then the surrounding CMOS architecture.
arjan de lumens
Veteran
What Intrinsity offers appears to be fairly well documented on their homepage. Their main additions over past dynamic-logic technology appears to be a new logic family "NDL" and a reasonably good automated dynamic-logic place & route tool - before this, you more or less had to draw each transistor by hand when doing dynamic logic. The continued lack of support for HDLs still limits the technology to a not-too-wide range of applications, where final performance is more important than development time.
Of course, if you have REALLY large resources available for high-clock-speed chip design, you will likely prefer something like this instead.
Of course, if you have REALLY large resources available for high-clock-speed chip design, you will likely prefer something like this instead.
Unknown Soldier
Veteran
I think we all waiting for that news.
Interesting though.. that news has disappeared off the site.
http://www.intrinsity.com/press/press_release_ati.htm
404
US
Interesting though.. that news has disappeared off the site.
http://www.intrinsity.com/press/press_release_ati.htm
404
US
DudeMiester
Regular
The less I hear "Fast14" the happier I become.
There are some rumours and misconceptions that never go away I guess.
There are some rumours and misconceptions that never go away I guess.
arjan de lumens
Veteran
Not really. Intrinsity has just changed the layout of its website: http://www.intrinsity.com/index.php?option=com_content&task=view&id=26&Itemid=41 appears to work for now.Unknown Soldier said:
Err, no.Unknown Soldier said:
They re-designed their whole site, which is why the old link is not working anymore.
There you are: http://www.intrinsity.com/index.php?option=com_content&task=view&id=26&Itemid=41
Similar threads
