R420 and fast-14 technology from intrinsity

[rwolf]

I know that the license for fast-14 technology was announced in february, but what do you think the chances are that this technology is used in the R420? Is this the reason for the secrecy surrounding the R420? It could be that this technology will show up in six months in the refresh chip, I don't know.

ATI has a MIPS license and might be using MIPS based math units in the R3XX. Fast-14 technology is also MIPS compatible so it might be a nice fit for the R420.

Keep in mind that this is a .13 micron technology not a .9 micron technology which is why I think it is for this generation of graphics chips and not the next.

http://www.intrinsity.com/technology/technology_fast14_set.htm

 

[AlphaWolf]

I know that the license for fast-14 technology was announced in february, but what do you think the chances are that this technology is used in the R420? Is this the reason for the secrecy surrounding the R420? It could be that this technology will show up in six months in the refresh chip, I don't know.

ATI has a MIPS license and might be using MIPS based math units in the R3XX. Fast-14 technology is also MIPS compatible so it might be a nice fit for the R420.

Keep in mind that this is a .13 micron technology not a .9 micron technology which is why I think it is for this generation of graphics chips and not the next.

http://www.intrinsity.com/technology/technology_fast14_set.htm

I would think that they would need that IP before the design stage began, not a refresh. However I don't really know exactly what fast-14 involves.

 

[Sabastian]

I would think that they would need that IP before the design stage began, not a refresh. However I don't really know exactly what fast-14 involves.

Wouldn't a big customer like ATi be allowed to test the product before an actual purchase. Fast-14 has been around for a bit hasn't it? I just don't think ATi would purchase the IP if they were not sure about the effect.

 

[Geo]

I would think that they would need that IP before the design stage began, not a refresh. However I don't really know exactly what fast-14 involves.

Wouldn't a big customer like ATi be allowed to test the product before an actual purchase. Fast-14 has been around for a bit hasn't it? I just don't think ATi would purchase the IP if they were not sure about the effect.

This is *not* a hint, because I don't know anything more about the particulars on this subject than anyone else not under NDA. Having said that, I do know a little about business and IP. When the deal was publicly announced versus when the deal actually began do not, necessarily, have to coincide. Often one partner has a vested interest in *not* announcing a deal while the other has a vested interest in announcing it as soon as possible --often the timing of an announcement is a matter of negotiation.

 

[Sabastian]

This is *not* a hint, because I don't know anything more about the particulars on this subject than anyone else not under NDA. Having said that, I do know a little about business and IP. When the deal was publicly announced versus when the deal actually began do not, necessarily, have to coincide. Often one partner has a vested interest in *not* announcing a deal while the other has a vested interest in announcing it as soon as possible --often the timing of an announcement is a matter of negotiation.

Yeah, announcements like the one in question are often sensitive to both parties involved. I have wondered about this IP since the day it was announced on just which product it might be in effect on. I suspect heavily that it is licensed for the xbox2 gpu but can't rule out the possibility that ATi has intensions of implementing it in earlier products. At any rate since we don't know all we can do is speculate until it becomes more clear just where this IP is going wrt ATi product cycle. My argument becomes bias in favor of ATi in terms of the timing of the announced IP licensing because I simply don't believe that they would blindly purchase it without knowing the advantage they would gain from it. Of course I could be totally wrong in that assessment, thus I include my own variety of disclaimer for your bewildering pleasure

 

[TempestX]

Fast-14
http://www.intrinsity.com/technology/body/downloads/Fast14 Overview.pdf

 

[bloodbob]

I know that the license for fast-14 technology was announced in february, but what do you think the chances are that this technology is used in the R420? Is this the reason for the secrecy surrounding the R420? It could be that this technology will show up in six months in the refresh chip, I don't know.

ATI has a MIPS license and might be using MIPS based math units in the R3XX. Fast-14 technology is also MIPS compatible so it might be a nice fit for the R420.

Keep in mind that this is a .13 micron technology not a .9 micron technology which is why I think it is for this generation of graphics chips and not the next.

http://www.intrinsity.com/technology/technology_fast14_set.htm

They were too far in the developement for the r420 to be switching over design tools in my opion. So its a no for the R420. Now yes it is aimed at .13 but they originally used it on .18 tech though some of the technology might not scale well ( I'm talking about the singally delay evening stuff they got for the interconnectors ). Also about the talk of the physcial block copyign technology they got a while before the Fast-14 would say its unlikely they used the Fast-14 stuff yet. Gotta say I'm quite interested on hearing what products they plan to use it on I'm guess they might be using it on some of their .11 cores they are developing.

 

[martrox]

It's very common in the IT world that tools are used for a long preiod of time before they are purchased - but only in a test invironment. The investment is so large that the buyer has to know that the product works.... Of course, ATI could have used the R420 as the "test" invironment.....Lets suppose for a few moments.....

ATI is working on R400.....
ATI learns of Fast14 and checks it out......
ATI tries Fast14 on a new "test" project - R420..... and it works way beyond expectations
ATI decides to purchase Fast14, pushes R400 project back in order to use R420...... Hmmmmm.............

As Judy Tanuda used to say...... "Hey...IT could happen......."

 

[Sabastian]

It's very common in the IT world that tools are used for a long preiod of time before they are purchased - but only in a test invironment. The investment is so large that the buyer has to know that the product works.... Of course, ATI could have used the R420 as the "test" invironment.....Lets suppose for a few moments.....

ATI is working on R400.....
ATI learns of Fast14 and checks it out......
ATI tries Fast14 on a new "test" project - R420..... and it works way beyond expectations
ATI decides to purchase Fast14, pushes R400 project back in order to use R420...... Hmmmmm.............

As Judy Tanuda used to say...... "Hey...IT could happen......."

Hence my protest/confuddelment. I am thinking that it is licensed for xbox next... but who knows.

 

[karlotta]

Or ATI has been renting the tech for sometime now and found it to be in its best interest to just go ahead and buy it. Opps they havent bought them yet...?

 

[JoshMST]

Just my opinion, but promising custom cell performance on a bulk process is one hell of a promise! Not only that, but it would seem that the software would have to be modified significantly to apply to any single process from any manufacturer (eg. IBM's 130nm process is significantly different than TSMC's 130 nm process, even though each one shares quite a few tools). If anything, ATI would probably test this software out on a lower end product (just as they have done with process jumps, like using the RV350 on TSMC's 130 nm Low-K). So, if we see a new low end variant that is clocked at 800 MHz, then we can probably guess that ATI in fact did use the Intrinsity 14 software, and they have very good results with it.

 

[jolle]

It's very common in the IT world that tools are used for a long preiod of time before they are purchased - but only in a test invironment. The investment is so large that the buyer has to know that the product works.... Of course, ATI could have used the R420 as the "test" invironment.....Lets suppose for a few moments.....

ATI is working on R400.....
ATI learns of Fast14 and checks it out......
ATI tries Fast14 on a new "test" project - R420..... and it works way beyond expectations
ATI decides to purchase Fast14, pushes R400 project back in order to use R420...... Hmmmmm.............

As Judy Tanuda used to say...... "Hey...IT could happen......."

I thought it was more like:
ATI works on R400, realizes it will take to long to realize it and renamed it R500
ATI picks up R3xx core and revamps it into R420..

Dunno tho, but a totally new core isnt necessary since we are still stuck
with DX9...

 

[rwolf]

Just my opinion, but promising custom cell performance on a bulk process is one hell of a promise! Not only that, but it would seem that the software would have to be modified significantly to apply to any single process from any manufacturer (eg. IBM's 130nm process is significantly different than TSMC's 130 nm process, even though each one shares quite a few tools). If anything, ATI would probably test this software out on a lower end product (just as they have done with process jumps, like using the RV350 on TSMC's 130 nm Low-K). So, if we see a new low end variant that is clocked at 800 MHz, then we can probably guess that ATI in fact did use the Intrinsity 14 software, and they have very good results with it.

The Intrinsity test chip was fabricated by TSMC using the .130nm process.

 

[Megadrive1988]

this fast-14 technology is probably much more likely to be
used with R500(R400++) Xbox 2 VPU, R600 (old R500?) as
well as every lowend VPU and refresh inbetween.

then again, I know absolutely nothing on it, so who knows, maybe it was used with R420.