[Arm-netbook] crowdfunding page is live -- and different CPUs for the future

[Manuel A. Fernandez Montecelo]

Hi,

2016-07-18 15:34 Luke Kenneth Casson Leighton:
>On Mon, Jul 18, 2016 at 11:26 AM, Manuel A. Fernandez Montecelo
><manuel.montezelo at gmail.com> wrote:
>> 2016-07-18 01:38 Luke Kenneth Casson Leighton:
>
>>> ... but if you *don't do* that licensing, and instead try to replicate
>>> them all, you are immediately placing the entire project at risk.
>>> bear in mind that TSMC won't talk to you if you make a failed chip
>>> (first time) because you're wasting their time.  and it costs $USD 2
>>> *MILLION* for the production masks (the lithographic masks like an OHP
>>> plastic sheet)
>>
>>
>> I don't really have any idea about the fabrication processes, but
>> according to this:
>>
>>  https://dev.sifive.com/documentation/freedom-u500-platform-guide/
>>
>>  "The resulting customized U500 SoC is optimized for manufacture in a
>>  TSMC 28nm metal-gate process, and delivered as packaged tested parts
>>  by SiFive."
>>
>> and contains most of the technologies that you mention, except video,
>> but maybe the custom accelerators can substitute traditional GPUs.
>
> yeah the "except video" means it can't be used (as a SoC).
>connecting a GPU via PCIe.... mmm... you're at what... between 20 to
>1000 watts there, depending on the GPU?
>
> and up to *FOUR* DDR3/4 lanes?  WOW.  128-bit-wide memory access.
>yowser.  that's going to be something like 12-20 watts just on memory
>access.

The 32 bit version is more power-restrained, perhaps:

  https://dev.sifive.com/documentation/freedom-e300-platform-guide/

(but I don't think that 32-bits of a new architecture it's very
interesting / future-proof)


I suppose that using custom co-processors/accelerators is an alternative
possibility for video/display, but probably not easy.  Some uses of the
SoC (e.g. micro-servers) probably don't care anyway, but I understand
that it's part of the EOMA68 standard.


Other than that, I hoped that by providing the links would lead to some
quick evaluation of the platform just announced, but if they're not
useful nevermind, sorry for the noise.


> > Yeah, I agree.  I was only saying that if one's going to go out of
>> her/his way and consider IC1T for a future option, RISC-V can be a more
>> interesting and future-proof alternative *than IC1T* (not better than
>> ARM or MIPS at the moment).
>
> ... we still have to have the OS support.  so we still need to wait
>for debian, arch and fedora to catch up.

I'm quite sure that they'll come sooner than IC1T, though ;-)

The next FreeBSD release will come with support for RISC-V.

>> Yeah, Loongson would be also good, although I am not sure if they will
>> keep it active or if they'll abandon it in favour of others.
>
> well it's the one that the chinese government is pushing for their
>independent supercomputer - intel lost out there thanks to the NSA,
>congratulations U.S. Government you just f*****d your own economy well
>done!

Indian research agencies / government are investing heavily in RISC-V,
it seems.

I wouldn't be surprised if it picks-up pace also in other places like
China.  And in general, RISC-V is not that different from MIPS /
Loongson, after all.


Cheers.
-- 
Manuel A. Fernandez Montecelo <manuel.montezelo at gmail.com>