Intel comming out with a 16 core Nehalem family.

On Sun, 8 Jun 2008 13:13:33 -0700, "Zootal"
<giganews@zootal.nospam.com> wrote:
>>
>>>From what I have read, Windows does not scale well above 4 cores. New
>>>languages will have to be perfected to take advantage of so many cores.
>>
>> Solaris/X86. Linux.
>>
>> -AH
>
>The linux 2.6 kernel has a lot of smp improvements - seperate read and write
>spin locks, seperate request queues for each cpu, etc., - but it really
>wasn't made with massive multi cores in mind, and can choke depending on
>what you do. Not sure about Solaris - the sparc processor itself is somewhat
>limited, but IIRC the OS scales up nicely to relatively large numbers of
>cores.

Over 100 processors in Sun's largest systems, with practically linear
performance improvement.

-AH

> Yes, but the app and cpu (and scheduler in your example) are still
> required to work under the rules and constraints of the OS. The point
> is, whatever functions an app embeds are by definition not part or no
> longer part of the OS. It's one or the other, and for good reason.
>

Of course there are rules and constraints - but that doesn't change the
concept of having apps running exclusively on cores managing the core
themselves instead of the "OS". Whether you call it part of the OS or not, a
cpu scheduler is a cpu scheduler, and if my app running on this core with
exclusive access to the core can use it's own cpu scheduler instead that of
the main "OS", that is what it will do.

.> What are we going to do with 16 cores?
.> From what I have read, Windows does not scale well above 4 cores.

Ray Tracing is perfectly cores-scalable. I’ve observed near perfect
speed increase on dual Penryn WS (8-cores) under windowsXP-64/32 and
under Fedora as well; doubling number of cores doubles the FPS
(~1.97x). Still, to achieve 100-FPS 1024x1024 of ray-tracing scene
exceeding the realism of current rasterised graphics will require
around 1000 Penryn-like cores. Probably we may see it in 10-years.

--sb





> From what I have read, Windows does not scale well above 4 cores. New
> languages will have to be perfected to take advantage of so many cores.
>

On Jun 5, 11:57 pm, "William" <nos...@pacifier.com> wrote:
> Lots of new stuff being announced out of Taipei Taiwan lately. See:http://www.tgdaily.com/content/view/37824/135/
>
> What are we going to do with 16 cores? Over clockable, 192-bit DDR3-1333
> and DDR2-1600 memory. Their are cooling solutions that will take the IC
> down to -100 degrees Celsius.
>
> From what I have read, Windows does not scale well above 4 cores. New
> languages will have to be perfected to take advantage of so many cores.
>
> Interesting times.
>
> William

<stefanbanev@yahoo.com> wrote in message
news:eb05a4ac-7bdc-470b-a44b-5d25a5c3c71e@e39g2000hsf.googlegroups.com...


> Ray Tracing is perfectly cores-scalable. I’ve observed near perfect
> speed increase on dual Penryn WS (8-cores) under windowsXP-64/32 and
> under Fedora as well; doubling number of cores doubles the FPS
> (~1.97x). Still, to achieve 100-FPS 1024x1024 of ray-tracing scene
> exceeding the realism of current rasterised graphics will require
> around 1000 Penryn-like cores. Probably we may see it in 10-years.
>
> --sb

100 FPS ray traced scene? It depends on how closely you want to approximate
the rendering equation, but I'm not sure 1000 cores would be able to handle
it. I'd be very curious to see some actual stats on this, is anyone doing
any current work with ray tracing? IIRC they demoed a realtime ray tracer at
SigGraph a few years ago, but it only gave a few fps.

If we used one of my 16x16x16 core cpu cubes, maybe we could push it up to
100 FPS :-)

On Jun 12, 1:07 pm, "Zootal" <gigan...@zootal.nospam.com> wrote:
> <stefanba...@yahoo.com> wrote in message
>
> news:eb05a4ac-7bdc-470b-a44b-5d25a5c3c71e@e39g2000hsf.googlegroups.com...
>
> > Ray Tracing is perfectly cores-scalable. I’ve observed near perfect
> > speed increase on dual Penryn WS (8-cores) under windowsXP-64/32 and
> > under Fedora as well; doubling number of cores doubles the FPS
> > (~1.97x). Still, to achieve 100-FPS 1024x1024 of ray-tracing scene
> > exceeding the realism of current rasterised graphics will require
> > around 1000 Penryn-like cores. Probably we may see it in 10-years.
>
> > --sb
>
> 100 FPS ray traced scene? It depends on how closely you want to approximate
> the rendering equation, but I'm not sure 1000 cores would be able to handle
> it. I'd be very curious to see some actual stats on this, is anyone doing
> any current work with ray tracing? IIRC they demoed a realtime ray tracer at
> SigGraph a few years ago, but it only gave a few fps.
>
> If we used one of my 16x16x16 core cpu cubes, maybe we could push it up to
> 100 FPS :-)

> is anyone doing
> any current work with ray tracing?

It is hot research area now. Intel/NVIDIA “discussion“ is quite
reviling as well. I’ve played with ray-tracing library from
IPP_5.3.095; now IPP 6.0 Beta is available at: http://www.intel.com/software/products/ipp/beta

1000-Penryns core estimation (~100 x E5450) was based on my
experiments with this library. Taking into account that Nehalem would
be ~30% more efficient per core and real-time ray-tracing speed
optimization tricks may be speeded up by factor 5…10 relatively Intel
IPP_5.3 implementation, I would not surprise to see ray-tracing takes
over the real-time 3D market in 10 years.

--sb

Adding more cores is about the only direction Intel can go in. They went
down a rabbit hole with the 'higher clock frequencies' strategy producing
CPUs that ran very hot.

Luckily, they had a skunk works in Israel that came up with the 'lower the
frequency but add more cores' strategy.

Somehow, I get the feeling they will go too far with the multi-core strategy
as well.

--g

On Jun 12, 1:07 pm, "Zootal" <gigan...@zootal.nospam.com> wrote:
> <stefanba...@yahoo.com> wrote in message
>
> news:eb05a4ac-7bdc-470b-a44b-5d25a5c3c71e@e39g2000hsf.googlegroups.com...
>
> > Ray Tracing is perfectly cores-scalable. I’ve observed near perfect
> > speed increase on dual Penryn WS (8-cores) under windowsXP-64/32 and
> > under Fedora as well; doubling number of cores doubles the FPS
> > (~1.97x). Still, to achieve 100-FPS 1024x1024 of ray-tracing scene
> > exceeding the realism of current rasterised graphics will require
> > around 1000 Penryn-like cores. Probably we may see it in 10-years.
>
> > --sb
>
> 100 FPS ray traced scene? It depends on how closely you want to approximate
> the rendering equation, but I'm not sure 1000 cores would be able to handle
> it. I'd be very curious to see some actual stats on this, is anyone doing
> any current work with ray tracing? IIRC they demoed a realtime ray tracer at
> SigGraph a few years ago, but it only gave a few fps.
>
> If we used one of my 16x16x16 core cpu cubes, maybe we could push it up to
> 100 FPS :-)

It is hot research area now. Intel/NVIDIA “discussion“ is quite
reviling as well. I’ve played with ray-tracing library from
IPP_5.3.095; now IPP 6.0 Beta is available at: http://www.intel.com/software/products/ipp/beta

1000-Penryns # of cores estimation (~100 x E5450) was based on my
experiments with this library. Taking into account that Nehalem would
be ~30% more efficient per core & per GHz and real-time ray-tracing
speed
optimization tricks may speed it up by factor 5…10 relatively to
Intel
IPP_5.3 implementation, I would not be surprised to see ray-tracing
takes
over the real-time 3D market in 10 years.

--sb

"geoff" <nospam@nospam.com> wrote in message
news:1Bw4k.32833$qk1.22954@fe02.news.easynews.com. ..
> Adding more cores is about the only direction Intel can go in. They went
> down a rabbit hole with the 'higher clock frequencies' strategy producing
> CPUs that ran very hot.
>
> Luckily, they had a skunk works in Israel that came up with the 'lower the
> frequency but add more cores' strategy.
>
> Somehow, I get the feeling they will go too far with the multi-core
> strategy as well.
>
> --g
>
>

That is a fascinating observation. If you can not go faster, or wider in
cpu architecture, than what is left? (Optical processors?) Something unknown
and not discovered?

Even with new exotic chemicals, substrates, packaging and cooling the speed
of light is in the way for much more advancement in speed.

Multi-core parallel processing is all their is. This is both a hardware and
software challenge for the future. Who knows where this will lead us.

William

It is tough to say, Forbes had an article in the late 90s about Intel which
showed the cost of developing/manufacturing a new chip as opposed to the
revenue brought in.

The two lines crossed in like 2002 and the conclusion was Intel would need a
completely new technology for chips.

That never happened, Intel kept on manufaacturing and there is no new
technology in the marketplace. Seems in 1995 or so, Intel talked about
switching over to bio chips after 2000.

Someone needs to visit the aliens at Roswell to get the next breakthrough.

--g