On Sun, Jun 21 2015, James wrote:

> Alan McKinnon <alan.mckinnon <at> gmail.com> writes:
>
>> Allan Gottlieb writes:
>>
>> > The hardest decision is size vs performance, but I know you can't help
>
> Maybe, maybe not.

Perhaps I was unclear.  I simply meant that clearly higher performance
is better and so is a smaller/lighter laptop.  The trade-off seems to be
a matter for personal preference.

>> > 1. Graphics.
>> >     I can afford a high-end graphics co-processor, but prefer the
>> >     software/administrative simplicity of intel graphics.  I do not
>> >     play high speed games or otherwise run graphics intensive
>> >     applications. Am I correct in believing that Linux (the kernel)
>> >     supports (the dell option)
>> >        Intel Core i7-5600U Processor, UMA graphics, Smart Card
>> >     directly with no extra gentoo package needed?
>
> Cuda on nvidia is well seasoned, but expensive. Gentoo distros such as
> Pentoo, use cuda for smokin fast passwd cracking. Many/most apps
> will benefit, in the near future, with the deployment of GCC-5.x as RDMA via
> gcc% will allow for using that smoking GPU (a simd processor) and the DDR5
> ram as if it was part of the CPU/ram resources.  If you read up on all the
> advances with GCC 5 you will  see most gpu (amd, Intel etc) will/should be
> supported. How long for stabilization, is unknown, at this time. But
> for very few dollars it's the biggest thing to hit hardware, since the FPU
> was integrated onto the same die, imho. YMMV.
>
> Check whatever GPU you select for the amount of its own (discrete) DDR5
> memory on the GPU (card).
>
>> So on the whole, my experience with higher-end Dell is that hardware is
>> pretty much well-supported across the boards with very few gotchas. The
>> only two exceptions would be wifi cards (cheap to fix) and maybe GPU
>> co-processor (if you are unlucky to get an unsupported cutting edge one
>> and need to wait a bit for Linux support to catch up).
>
> I'd check around on the precise details of the GPU before purchase.
> Some GPU use the general system ram, and that is a severe
> (buss-bandwidth) bottleneck that really dampens performance on many
> softwares. The looming gcc-5 is a game changer on using video
> resources, as general system resources...

I doubt that gcc 5 (or 6) will extract much parallelism that Cuda can
exploit for my primary use cases: compiling (largely gentoo) sources and
running emacs.

I have learned that the high end graphic coprocessors (or GPUs as they
are now called) carry a significant administrative overhead, at least
for gentoo.

Thanks for responding,
allan

Reply via email to