576 cores, or 192 per board

haiticare2011 · 2014-06-07T11:06:08+00:00

GTC conference -see tekra k1 board - see introduction to parallel programming - free, online The Kepler architecture, building block of the US fastest supercomputer, was originally a GPU (graphics), but was adapted to general scientific problems via a process GPGPU - general programming on graphics processing units. Nvidia makes the IC with 192 cores on board - 3 of them can be easily strung together for 576 cores. That is currently $576 retail, coincidentally. (or not) The key to making this work is the c compiler which automagically does the parallelism. This then has been used to build a Python interface as well. These language tools were developed as part of the supercomputing effort. These will run neural networks very well, as the nets require dot products, which the cores feast on. Graphics pixel processing, machine learning, and data exploration are being pursued energetically. It turns out the key to making this technology work is to train programmers and scientists how to use it. The course ay udacity, given by experts, fills that need. This is free, and a Stanford Prof. plus Nvidia developer give the course. I believe there are 60,000 people taking it at any one time. SO that's it. My interest is to see what I/O capabilities there are. Nvidia claims high speed camera interfaces.

H

haiticare2011 12 years ago

There was one process called "dye transfer." It is a very interesting process, chemically. I seem to remember it involved getting the image in a gelatin layer, and then diffusing it onto a paper substrate. Some hobbyists were doing it, as they thought it was higher quality. I believe kodacolor or some variant was based on that. Older films have a 'dyed-in" quality. (1950 era)

BUT, you can't beat the modern video cameras, IMO. These cameras, used to shoot TV and video since 2000, are quite amazing. They can shoot in low light, and the pics look like 20 megapixel iso 3000 shots with hardly any noise. I bet they are quite cheap nowadays, maybe only $70k, but I don't know. Huge dynamic range, even in low light.

Vote

R

Robert Baer 12 years ago

Seems that with the correct processing (pre-fogging, etc) that one-photon sensitivity is possible, making it a viable "wave/particle" sensor in a 2-hole experiment.

Vote

H

haiticare2011 12 years ago

That's interesting. A photomultiplier can see one photon, I'm sure. I don't know what the limit is for a PD.

What is the one photon experiment?

This is fascinating stuff, quantum mechanics. The single photon will generate interference patterns, even though it goes through only one hole?

There is some paradx there, don't know if I've said it right. And the collapse: If you try to tell which hole the photon goes through, the interference pattern disappears?

Vote

R

rickman 12 years ago

Mama don't take my Kodachrome and leave your boy so far from home Mama don't take my Kodachrome away

Rick

Vote

J

josephkk 12 years ago

Supressing the asinine google groupers blank lines.

because

sensitized

process,

gelatin layer,

it,

variant was

to shoot

and

bet

dynamic range, even in low light.

No way in hell. Doesn't come anywhere close to the high lines per inch of Kodachrome. Try 300 megapixels or more for 135 film. Can be blown up to

11 by 14 fine if the shot is good enough.

?-)

Vote

J

josephkk 12 years ago

exposed er

the size

empirical.

guesstimate. I

guesstimate that

a 16

other

OK so i underestimated the pixel numbers a bit (or substantially). But at least 3 bytes per pixel. The resolution and color accuracy were phenomenal, unmatched AFAIK.

Well, in the first place those movies were *NOT* shot with Kodachrome. Or anything slightly like it. It never was used for movies AFAIK.

?-)

Vote

M

Martin Brown 12 years ago

Particularly because you have over estimated the old films MTF capability by about a linear order of magnitude. 100lp/mm was doable on Kodachrome 25 although the red layer did tend to stand out a bit.

Kodak used the number 24Mpixel per frame for their 35mm professional film scanning service and it was already overkill. There were a handful of monochrome films that could top 100lp/mm and T-MAX 100 was one.

formatting link

Kodakchrome 25 being colour would not have beaten this and in practice with the lenses of the day would almost certainly have been worse.

According to Kodak's own 1996 datasheet for KK25 it didn't perform all that well despite its having a dedicated following. Ektachrome was worse. Faster Fuji Sensia and Velvia would leave it and any other Kodak colour film standing so you never went back...

formatting link

Regards, Martin Brown

Vote

R

Robert Baer 12 years ago

Some fussy fuzzy stuff like that, only 2 holes AFAIK.

Vote

U

upsidedown 12 years ago

Do you have any pointers to what Nvidia is doing ?

Searching for "object segmentation" and "boundary detection" brings up mostly academic papers but not much references to actual real-time hardware.

Vote

576 cores, or 192 per board

Join the Discussion

Didn't find your answer?