Imagination IP for PC

[ngkmh]

Its been a while since that last Imagination thread so I'll kick one off....

Saw this posted in an investor's site (IMG had AGM yesterday):

Some very good demo's yesterday including the 8 pipe series 5 PC card which was there to demonstrate series 5 technology. Kristof claimed this tech was 50% faster than anything the competition as currently capable of....

Any comments Kristof?

 

[trumphsiao]

Its been a while since that last Imagination thread so I'll kick one off....

Saw this posted in an investor's site (IMG had AGM yesterday):

Some very good demo's yesterday including the 8 pipe series 5 PC card which was there to demonstrate series 5 technology. Kristof claimed this tech was 50% faster than anything the competition as currently capable of....

Any comments Kristof?

would you mind giving me a link ? thx

 

[Geo]

50% faster than the competition for the next 3 months? At which time it will be roughly as fast? And by the time it could be on the street it will be 50% slower?

I hate to pee in anyone's oatmeal, but is there any particular reason I shouldn't think of it that way?


Edit: I should mention that the roadmappery I've seen referenced elsewhere seems to suggest 2H 2007 for a desktop part hitting the streets, and this informed my (grossly unfair? Love to have someone with the cred to do so come in and say that) post above.

 

[chavvdarrr]

or 50% faster than 8-pipe cards from other vendors?

 

[trumphsiao]

50% faster than the competition for the next 3 months? At which time it will be roughly as fast? And by the time it could be on the street it will be 50% slower?

I hate to pee in anyone's oatmeal, but is there any particular reason I shouldn't think of it that way?

what I'm interested in is the Spec of "Transistor counts" and "Die Size" ?

G80 500M
R600 600M
High-End PVR5 will be ???

 

[Geo]

or 50% faster than 8-pipe cards from other vendors?

Oh, damn, I hadn't even considered that one!

Tho possibly that's not terrible, assuming it scales within acceptable current process limits.

 

[Megadrive1988]

50% faster in what?

clockspeed
shader performance
framerate
overall

?

 

[Rolf N]

Oh, damn, I hadn't even considered that one!

Tho possibly that's not terrible, assuming it scales within acceptable current process limits.

I disagree. That would be terrible. 50% more oomph against competing 8-pipers could put the validity of the whole approach into question.

Thankfully (?) we're missing the clockspeed from the equation. If a very slow-running early sample did that, things are looking good.
At similiar clocks and bandwidths it isn't unreasonable to expect at least 150% more bang against an IMR, IMO.

 

[Tim Murray]

Keep in mind that when the little birdeis were saying "Series 5 is about to come out if it's ever going to come out at all," all of those little birdies were saying "it's eight pipes." That was a few years ago.

Something to ponder.

 

[Ailuros]

Keep in mind that when the little birdeis were saying "Series 5 is about to come out if it's ever going to come out at all," all of those little birdies were saying "it's eight pipes." That was a few years ago.

Something to ponder.

Depends what you mean with "pipes" exactly. There are also crack pipes (sorry for the pun ).

 

[Lazy8s]

While the eight pipeline configuration at this point in their roadmap could be early work on an integrated PC and/or laptop solution, it might be a development prototype running on the PC representing either SGX530 or Thalia-L, their embedded cores for high-performance cellphones and for car navigation systems respectively.

 

[Ailuros]

What is a "pipeline" nowadays and how exactly do you folks define it? Even moreso in a unified shader core.

SGX530 was claimed to have an effective fill-rate of 1.2GPixels/s. I don't see it having more than 2 TMUs and that's already quite high for a core that's slated for high end PDA/mobile devices.

If there truly was demonstrated anything for the PC, I have severe doubts that it had anything to do with either SGX530 or Thalia-L. If it was something with say 8 TMUs for instance then it should be quite a bit apart from anything SGX or Thalia-whatever.

 

[Ailuros]

I disagree. That would be terrible. 50% more oomph against competing 8-pipers could put the validity of the whole approach into question.

Thankfully (?) we're missing the clockspeed from the equation. If a very slow-running early sample did that, things are looking good.
At similiar clocks and bandwidths it isn't unreasonable to expect at least 150% more bang against an IMR, IMO.

NV43 would be such an "8-piper" (again depending what anyone means with "pipe" exactly). If that should be a comparative measure then even 150% is more than just conservative IMHO. Since marketing usually works with "up to" figures there could be quite a few case scenarios where higher ballparks could be yielded. Since NV43 cannot combine for instance MSAA+ float HDR, imagine what a combination of even 2xSSAA + float HDR would result to.

 

[Geo]

http://www.theinquirer.net/default.aspx?article=33318

Well, Inq doesn't have any problems putting two and two together, and getting 22.

 

[Ailuros]

http://www.theinquirer.net/default.aspx?article=33318

Well, Inq doesn't have any problems putting two and two together, and getting 22.

Fuad said in a recent newsblurb that after licensing MBX, Intel might also license SGX. I simply mailed him and linked him to the licensing announcement from the 17th of May 2005 for Eurasia/SGX with Intel.

He assumes again too much obviously, since afaik G965 hasn't anything to do with IMG.

 

[Geo]

Nobody in this business believes in coincidence much.

Even less so after Monday.

People see both Intel and ImgTec seriously stirring after a long desktop graphics somnulence, and minds will go there, appropriately or not.

 

[Ailuros]

If then I don't see anything prior to GMA "4000" or whatever they'll call it.

 

[roninja]

good to see a thread started quoting me!

Yup nice in-house PC demo. IMG have yet to tie up a partner for Athena its desktop graphics chip however they do have a project for MUSE intended for UMPC/Laptop configurations.

 

[_xxx_]

what I'm interested in is the Spec of "Transistor counts" and "Die Size" ?

G80 500M
R600 600M
High-End PVR5 will be ???

Ok, I'll bite here. Less than 300 Mio.

 

[Ailuros]

good to see a thread started quoting me!

Yup nice in-house PC demo. IMG have yet to tie up a partner for Athena its desktop graphics chip however they do have a project for MUSE intended for UMPC/Laptop configurations.

Hmmm any ducks in that demo?

If IMG hasn't yet a partner for Athena, it looks as grey for a PC desktop sollution as it always did. Unless of course Intel truly is interested in standalone graphics, which I frankly will only believe if I see it officially announced.