16 Oct
2011
16 Oct
'11
9:17 a.m.
On 16 October 2011 07:05, Dave McGuire <mcguire at neurotica.com> wrote:
On 10/15/2011 09:54 PM, Liam Proven wrote:
Alongside Jochen's post with its excellent points, I'd also say that,
yes, I reckon economies of scale will mean server CPUs too will end up
with onboard GPUs that will often go unused. Just as in many cases an
awful lot of servers don't use their onboard FPUs, either.
--
Liam Proven ? Info & profile: http://www.google.com/profiles/lproven
Email: lproven at cix.co.uk ? GMail/GoogleTalk/Orkut: lproven at gmail.com
Tel: +44 20-8685-0498 ? Cell: +44 7939-087884 ? Fax: + 44 870-9151419
AIM/Yahoo/Skype: liamproven ? MSN: lproven at hotmail.com ? ICQ: 73187508
OTOH, NeWS sounded fascinating and I've love to see a modern, FOSS
reimplementation of it. I think that perhaps it might not be a very
good fit for the modern display pipeline though - of a very fast
multi-gigaHertz multicore CPU connected closely to a cluster of
high-performance very-parallel MIMD GPU cores.
Right now, usually, they're on a PCIe card, but AMD is showing the way
- its Llano and Bulldozer "Fusion" APUs integrate the GPU onto the
same die as the CPU cores. This is the way all CPUs *will* go; it's
just that the current Intel offerings have crappy GPUs.
?Be careful with those "all CPUs" assertions; server processors are a pretty
big part of the market, and a GPU would be a very expensive useless piece of
silicon real estate there.