On Tue, May 9, 2017 at 5:13 PM,  <ronwirring@safe-mail.net> wrote:

> How can you make an arm gpu for 150000usd?

you don't.  ARM is a registered trademark, and copying what they're
doing is asking for trouble.

instead you take one of the "open gpus" or parts of them and use that.
there's several i've been tracking: MIAOW, Nyuzi, the ORSOC graphics
accelerator - there's surprisingly quite a lot out there.

> How would you use it?

 it would be on the same silicon, on the same memory bus as the RISC-V
64-bit core.  anything else is too power-hungry.

> Would you put it additionally on a bord and not use the gpu located on the processor socket?

 no.

> If you reverse engineer the latest mali gpu, what you hold against it is, that when next version of an arm socket is for sale it will have a new mali gpu and require a new reverse engineering?

 correct.  so whatever you get it's guaranteed to be "old".  this is
the sad fact of reverse-engineering: all that effort, with *no
guarantee of success*.... just to get something that's years
out-of-date.


> Lets say 50000 people would buy the filrefly rk3399. Same people would pay 3eu each for reverse engineering the gpu. Then we would have a source code arm computer?

no.  ARM has acted so unethically in slandering luc verhaegen and
blackmailing companies that pay him that i am not interested in
supporting their business any more than is absolutely necessary.

> It takes a lot of coordination to provide gnulinux distributions I assume.

 yes.  a _lot_.  however amazingly people are actually doing just
that... even on risc-v where there isn't yet even any actual hardware.

> It is unfortunate people are not able to coordinate such that we can get source code hardware.

 it's still quite a young area of focus.

l.

_______________________________________________
arm-netbook mailing list arm-netbook@lists.phcomp.co.uk
http://lists.phcomp.co.uk/mailman/listinfo/arm-netbook
Send large attachments to arm-netbook@files.phcomp.co.uk