NEC has implemented AcceleonG10 in display IC

finsider: you were faster. :)

anyways, as an "side topic" here's something which are actually "sitting" on my bookshelf right now:
picture1.jpg

picture2.jpg


I kept this under wraps for several days just linking these pictures only to MURC. What you are looking at is two different revisions of Bitboys AXE. Rev A has manufacturing fault by Infineon, but Rev B is fully working DX8 GPU with 12MB eDRAM. There's less than 10 AXEs outside the company and I am the lucky one having two of them.

benchmarks?? nope, I have only chips, not the cards.
 
I wonder about one thing though... "TFT-LCD controller driver integrating graphics engine on-chip". This is a seperate IC which functions as a LCD Controller... now when you are doing a highly integrated mobile phone why would you use an extra chip as an LCD controller when this is usually integrated with the ARM or other CPU ? Extra chips on the PCB result in more power usage and extra costs AFAIK.

ATI btw did the same thing with their imageon LCD controllers and they found their way into some PDAs where extra ICs are not that big a deal.

All IMHO of course and my own personal 2c...

K-
 
Kristof said:
I wonder about one thing though... "TFT-LCD controller driver integrating graphics engine on-chip". This is a seperate IC which functions as a LCD Controller... now when you are doing a highly integrated mobile phone why would you use an extra chip as an LCD controller when this is usually integrated with the ARM or other CPU ? Extra chips on the PCB result in more power usage and extra costs AFAIK.

ATI btw did the same thing with their imageon LCD controllers and they found their way into some PDAs where extra ICs are not that big a deal.

All IMHO of course and my own personal 2c...

K-

well, extra chips is better than no chips available at all... ;)
(sorry I cound't resist. ;) it actually WAS targeted under the belt. ah, I have waited this SO LONG! :D )
 
Nappe1 said:
well, extra chips is better than no chips available at all... ;)
(sorry I cound't resist. ;) it actually WAS targeted under the belt. ah, I have waited this SO LONG! :D )

You know I can't comment on that... 8)
 
Kristof said:
Nappe1 said:
well, extra chips is better than no chips available at all... ;)
(sorry I cound't resist. ;) it actually WAS targeted under the belt. ah, I have waited this SO LONG! :D )

You know I can't comment on that... 8)

that's why I did it. :devilish:
 
DaveBaumann said:
I hope thise images aren't actual size! ;)


EDIT: LOL! :) you kind of screwed (by editing your post) my quote and the answer... :) in any case, I don't know if there is chips based on MBX. at least those aren't released yet, as it seems on case of this one having G10 core.
 
Kristof said:
I wonder about one thing though... "TFT-LCD controller driver integrating graphics engine on-chip". This is a seperate IC which functions as a LCD Controller... now when you are doing a highly integrated mobile phone why would you use an extra chip as an LCD controller when this is usually integrated with the ARM or other CPU ? Extra chips on the PCB result in more power usage and extra costs AFAIK.

This is not a separate chip. Basically every TFT module has at least a dummy gate driver that handles the row data latching, row counters and so on. The problem with these modules is in fact power consumption as the LCD controller (usually integrated with arm as you mentioned) has to send constant stream of data to the display, even if the picture is static. This is a lot of data over the long wires to the display module itself. Now, this new chip replaces the dummy chip on the TFT module with integrated frame buffers so that only changes to the picture are sent from the main system. Screen refresh is done completely inside the module. Including graphics engine inside this chip is efficient, especially in terms of power consumption, as the data traffic from the cpu to the TFT module is dramatically reduced. To render for example a complex antialiased polygon to the TFT screen only requires handful of bytes to be transferred from the cpu to the TFT module.

The very nice thing about this solution is that to integrate this solution in to existing mobile device only requires putting this kind of TFT module in to the device and software update... No changes to the main system board like solutions from some other solution providers require.

- MikaT (Main architect of the Acceleon G10)
 
Congratulations for shipping products!
Mr.Trug, you are legendary coder of 2nd Reality(1993) just ten years ago.

Can this core render any 3d primitive?
 
Nappe1 said:
finsider: you were faster. :)


I kept this under wraps for several days just linking these pictures only to MURC. What you are looking at is two different revisions of Bitboys AXE. Rev A has manufacturing fault by Infineon, but Rev B is fully working DX8 GPU with 12MB eDRAM. There's less than 10 AXEs outside the company and I am the lucky one having two of them.

benchmarks?? nope, I have only chips, not the cards.

so there are cards running? what clock speed? how are the chances to see a preview of such a card?
 
12 megs of eDRAM? Whats the process and die size? Might give us a hint at the chips in the PSP..
 
tEd said:
so there are cards running? what clock speed? how are the chances to see a preview of such a card?

I have no information if there's working cards. Axe was specsed to run at 175MHz. (means total bandwidth of ~24GB/s in single chip and 48GB/s in dual chip) And Trug knows better if it is possible see some benchmarks some day. ;)

zurich said:
12 megs of eDRAM? Whats the process and die size? Might give us a hint at the chips in the PSP..

afaik, there isn't eDRAM on Acceleon G10 and AXE is/was desktop GPU.
 
moichi said:
Congratulations for shipping products!
Mr.Trug, you are legendary coder of 2nd Reality(1993) just ten years ago.

Can this core render any 3d primitive?

Thank you,

This core is not really targeted for full 3D stuff, but rather for SVG (Scalable Vector Graphics) rendering which is ideal for UI stuff. But it can also render triangles, so flat shaded 3D it can easily do.

Check www.acceleon.com for G10 details (the core used in this NEC chip)
 
Thanks, Mr.Trug.

In japan many cell phones already implements 3d primitive software renderer.
(Many cell phone uses hi corporation's software renderer mascot capsule engine.
http://www2.mascotcapsule.com/
Mitsubishi's cell phone use its own Z3D hardware renderer(see hwws03 paper)).
http://www.graphicshardware.org/presentations/Kameyama.pdf

For example, Namco Ridge Racer run in japanese cell phone(SHARP SH53J).
http://www.sharp.co.jp/products/jsh53/
http://www.namco.co.jp/mobile/jsky/ridgeracer.html

I expect to see LCD controller which implements G30 to accelerate OpenGL ES and JSR-184 with higher clock.
Japanese country is crazy about cell phone's 3d function.
So I think many business chances exist in Japan. :)
 
Back
Top