8:54
bob sherbin:
Morning (for those on U.S. time). It's an early fall day here in San Jose with plenty of sunshine. I'm Bob Sherbin, vp of corporate comms at NVIDIA, and I'm really jazzed to be live blogging.
Tuesday September 21, 2010 8:54 bob sherbin
8:55
bob sherbin:
The san jose convention center is filling up fast. we have about 10 minutes to go and the seats are more than four-fifths full.
Tuesday September 21, 2010 8:55 bob sherbin
8:56
bob sherbin:
I'm sitting in the press section where there are about 100 reporters from a bunch of business publications and lots and lots of specialized trades.
Tuesday September 21, 2010 8:56 bob sherbin
8:56
bob sherbin:
we've spent a good six months working toward today and it's really nice to be about to take off.
Tuesday September 21, 2010 8:56 bob sherbin
8:57
bob sherbin:
speaking of taking off, they're beginning to show some 3D demos to warm up the crowd. Tuesday September 21, 2010 8:57 bob sherbin
8:57
bob sherbin:
there are some fairly high end 3D glasses attendees have been issued. they don't fold but they offer terrifying fidelity.
Tuesday September 21, 2010 8:57 bob sherbin
8:58
bob sherbin:
It looks like we're in a space ship passing through a very high tech honey comb. Tuesday September 21, 2010 8:58 bob sherbin
8:58
bob sherbin:
there's some techno music going, no narration but the audience is pretty rapt. even the reporters have stopped typing. Tuesday September 21, 2010 8:58 bob sherbin
8:59
bob sherbin:
the big voice is coming on giving the usual warnings. Tuesday September 21, 2010 8:59 bob sherbin
8:59
bob sherbin:
everyone's been told to put their 3D glasses on for the intro video. they're showing 3D sonograms Tuesday September 21, 2010 8:59 bob sherbin
9:00
bob sherbin:
a list is releling out around the theme of Right Now. Right now GPUs are being used to prevfent cancer death from radiation, nOAA is generating weather modesl 35x faster than before. the planet's second fastest supercomputer uses GPUS Tuesday September 21, 2010 9:00 bob sherbin
9:01
bob sherbin:
Right now...researchers are using GPUs to find plaque in arteries, eradicating malarai-carrying mosquitos, to design, buildings, shampoo, pizza curst, antibiotics Tuesday September 21, 2010 9:01 bob sherbin
9:02
bob sherbin:
video's spooling down. and Jen-Hsun Huang is coming on stage Tuesday September 21, 2010 9:02 bob sherbin
9:03
bob sherbin:
jen-hsun, jhh hereafter, says gtc is about the amazing discoveries based on gpus and that there's an amazing conference ahead Tuesday September 21, 2010 9:03 bob sherbin
9:03
bob sherbin:
four things, jhh says: nvidia's focus, cuda's development, and then some amazing announcements, and then a peek into the future. plus therell be some great demos, he says Tuesday September 21, 2010 9:03 bob sherbin
9:03
bob sherbin:
NVIDIA's focused on three areas: visual computing, parallel computing and mobile computing. Tuesday September 21, 2010 9:03 bob sherbin
9:04
bob sherbin:
He talks about each of the company's main brands: quadro for workstations; tesla for supercomputing; and gforce and tegra for consumers Tuesday September 21, 2010 9:04 bob sherbin
9:04
bob sherbin:
Tegra, jhh says, is based on ARM because it's the instruction set of choice and the fastest growing chip on earth Tuesday September 21, 2010 9:04 bob sherbin
9:06
bob sherbin:
Mobile computing will be enormously disruptive. It's an area where great contributions will happen. Mobile computing isn't just about portability. it's the first computer equipped with all kinds of sensors -- cameras, other situation-aware devices. because of them, mobile computing will achieve and deliver experiences not possible before. Tuesday September 21, 2010 9:06 bob sherbin
9:06
bob sherbin:
JHH is beaming talking about tegra. He's wearing his trademark black shirt and jeans, and some really nice loafers Tuesday September 21, 2010 9:06 bob sherbin
9:07
bob sherbin:
Next slide up is a a view of an airforce fighter jet, flying right at the audience. He uses it to talk about the importance of tesselation. Fermi's really good at. 6x faster, in fact, than what it replaced Tuesday September 21, 2010 9:07 bob sherbin
9:08
bob sherbin:
Turns on the air force jet is from a computer game not yet released. It's called Hof, or something along those lines. it requires the representation of 4 billion triangles Tuesday September 21, 2010 9:08 bob sherbin
9:09
bob sherbin:
the jets are flying voer the middle east. you feel like a top gun pilot. the jets are shooting at each other in 3D.a parachute would be nice right now Tuesday September 21, 2010 9:09 bob sherbin
9:10
bob sherbin:
the really cool thing is how realistic the ground is, you can see individual trees. Tuesday September 21, 2010 9:10 bob sherbin
9:10
bob sherbin:
sorry, the name might have been hok 2. but i suspect i'm bobbling it a bit Tuesday September 21, 2010 9:10 bob sherbin
9:11
bob sherbin:
Next slide up is describing the GP GPU Revolution, which incorporates a CPU plus GPU Tuesday September 21, 2010 9:11 bob sherbin
9:13
bob sherbin:
JHH introduces CUDA, which he says involved two decisions. we recognized that we didn't want to replace the CPU but augment it with a many-core GPU. Second decision was to ride back of high volume GPU biz that was GForce. Tuesday September 21, 2010 9:13 bob sherbin
9:14
bob sherbin:
THere's another slide introducing a new concept for NVIDIA. it's called "Computing: The Third Pillar of Science." It underpins auto design, medical imaging, weather forecasting, seimsic imaging and drug design Tuesday September 21, 2010 9:14 bob sherbin
9:15
bob sherbin:
Researchers all over are embracing GPUs to accelerate their insight to discovery. CUDA makes that more accessible. And more affordable. Tuesday September 21, 2010 9:15 bob sherbin
9:16
bob sherbin:
I should have desscribed the stage. It's black, of course, with a bright green logo on the upper right, and a sign on the left that's green with white lettering stating, "GPU Technoloyg Conference." there are also some really cool green blow-up shots of GPU circuitry Tuesday September 21, 2010 9:16 bob sherbin
9:18
bob sherbin:
JHH is talking about eh important of a new approach based on parallelism, drawing on work from the great Berkeley engineering department. Tuesday September 21, 2010 9:18 bob sherbin
9:19
bob sherbin:
CUDA is that approach and it's fast adoption is evidence of that. He's going to show some demos of CUDA in action. Tuesday September 21, 2010 9:19 bob sherbin
9:19
bob sherbin:
JHH introduces Tony Tomasi, svpr of content and technology. tony has three things to show, he says Tuesday September 21, 2010 9:19 bob sherbin
9:20
bob sherbin:
first up is a tesselation demo in 3D. it shows a post-apocolyptic city. 1.3 billion polygons a frame are being rendered. there are over a thousand light sources. this is done through procedural tesselation. Tuesday September 21, 2010 9:20 bob sherbin
9:21
bob sherbin:
whoah, there are some terrifying bird statues on the top of each building. they're turning tesellation on and off. when on, there 500x the fidelity. the geometry, tony says, is an order of magnitude more complex than a current computer game. Tuesday September 21, 2010 9:21 bob sherbin
9:22
bob sherbin:
this isn't a city you'd want to live in. it's metallic, with swirling clouds. Tuesday September 21, 2010 9:22 bob sherbin
9:23
bob sherbin:
the next demo is volumetric simulation in ufll 3d. there are two crash dummies. one's blowing nvidia-green smoke at the other who's fighting it off with a fan and imparting velocity as he does. the smoke is swirling in a way an artist couldn't capture it. this takes 20 trillion ops per second, tony said. that's 1 billion theads per second. Tuesday September 21, 2010 9:23 bob sherbin
9:26
bob sherbin:
the third example of simulation in graphics is fluid dynamis. there's decaying old lighthouse on a stormy night, with a crashing sea and debris moving around. it's all being rendered in real time. the light from the lighthouse is casting light in different direction. The demo guy is moving around a boat and it's changing the course of the waves. It's fully volumetric. turns out this can be done on a single gpu Tuesday September 21, 2010 9:26 bob sherbin
9:26
bob sherbin:
this latest is 5000x faster than with cpus. very realistic. enough so that you'd want to put a wool blanket around your knees. Tuesday September 21, 2010 9:26 bob sherbin
9:27
bob sherbin:
that's it for tony, who heads off stage and back into the crowd. Tuesday September 21, 2010 9:27 bob sherbin
9:29
bob sherbin:
Next up, JHH says, is three demos of how CUDA's being used. Momentum's building, he said, CUDA SDK cummulative downloads has doubled in a year to 670K; the number of OEMs offering Tesla processors has increased in a year from one to nine. and the number of GTC papers submitted has grown fivefold to more than 330 Tuesday September 21, 2010 9:29 bob sherbin
9:30
bob sherbin:
it's pretty rare to sit in a presentation where there's no chitchat from the audience, no sarcastic comments. everyone's listening to jensen reeling off factoids, factlettes and factettes about cuda. Tuesday September 21, 2010 9:30 bob sherbin
9:32
bob sherbin:
it's clear, jhh says, that cuda's making deep inroads into most areas of science and engineering. Tuesday September 21, 2010 9:32 bob sherbin
9:32
bob sherbin:
a decent-size announcement, jhh hints. Tuesday September 21, 2010 9:32 bob sherbin
9:33
bob sherbin:
here it is: through pgi, we're announcing cuda x86. so you can deploy cuda apps on any computer or server in the world. this will open its use in a very big way. Tuesday September 21, 2010 9:33 bob sherbin
9:34
bob sherbin:
JHH poses a question: how do we put the power of cuda in the hands of every engineer designer and researcher in the world. What else can we do to spread hpc? Tuesday September 21, 2010 9:34 bob sherbin
9:35
bob sherbin:
the answer is: killer apps, he says Tuesday September 21, 2010 9:35 bob sherbin
9:37
bob sherbin:
matlab is used by millions, as a general purpose numerical computation package. it has stats tool kits, image-processing toolkits, data-acquisition toolkits, informatic toolkits. matlap will support cuda-accelerated GPUs with the parallel computing toolkit. Tuesday September 21, 2010 9:37 bob sherbin
9:38
bob sherbin:
matlab, he says, is the excel of engineering. Tuesday September 21, 2010 9:38 bob sherbin
9:39
bob sherbin:
Next announcement: Multi-GPU AMBER11. Tuesday September 21, 2010 9:39 bob sherbin
9:42
bob sherbin:
in computational biology, some of the simulations are enormously complex. like walking, gene sequencing. researchers are simulating molecules at the atomic level using quantum chemistry. AMBER11 is among the most popular molecular simulation packages. UCSD recently made this GPU capable. Tuesday September 21, 2010 9:42 bob sherbin
9:44
bob sherbin:
really cool slide now of the Kraken supercomputer, it costs tens of millions and is the fastest amber machine. and there's a picture of jhh standing in front of it with a gpu-computer the size of a college dorm refrigerator that has the same amount of power. this makes supercomputing much more accessible. Tuesday September 21, 2010 9:44 bob sherbin
9:45
bob sherbin:
what other fields, jhh asks, would benefit from extreme acceleration in hpc? Tuesday September 21, 2010 9:45 bob sherbin
9:46
bob sherbin:
jensen says one is ansys, a simulator for engineers. Tuesday September 21, 2010 9:46 bob sherbin
9:47
bob sherbin:
using algorithems, ansys can predict the behavior of product design. it's used by 97 of world's top 100 industrial companies. there are 35K customers in 40 countries. it was used, among other things, to design michael phelps olympics swimsuit Tuesday September 21, 2010 9:47 bob sherbin
9:48
bob sherbin:
to talk about ansys, jensen brings on stage dr. s. subbiah, an ansys veteran exec Tuesday September 21, 2010 9:48 bob sherbin
9:48
bob sherbin:
dr. s is a cool looking guy, pewter-colored hair, open collar dark suit Tuesday September 21, 2010 9:48 bob sherbin
9:49
bob sherbin:
dr. s Tuesday September 21, 2010 9:49 bob sherbin
9:51
bob sherbin:
....is talking about ansys being used on a 747's brake. if a 747 needs to brake for on an abortive takeoff, it requires stopping a huge plane going 200 mph within 5000 feet. sounds pretty complex to me. Tuesday September 21, 2010 9:51 bob sherbin
9:52
bob sherbin:
dr. s is describing the phenomenally complex problems that his company's customers are trying to figure out by running them on clusters, which can take days per calculation. Tuesday September 21, 2010 9:52 bob sherbin
9:53
bob sherbin:
on a quadcore chip, a speed up on a xeon machine could be more than 2x. Tuesday September 21, 2010 9:53 bob sherbin
9:55
bob sherbin:
jhh asks, now that ansys will be using gpu's, where do you see simuation-driven product development playing out? Dr S: it will be multi-disciplinary. Tuesday September 21, 2010 9:55 bob sherbin
9:55
bob sherbin:
another trend is that design is more democratically achievable. it's not just big companies that can figure out this stuff. someone in a garage can. Tuesday September 21, 2010 9:55 bob sherbin
9:56
bob sherbin:
dr. s leaves the stage. Tuesday September 21, 2010 9:56 bob sherbin
9:58
bob sherbin:
3ds Max is used by just about all game developers, and increasingly architecture and design firms. JHH says nvidia and autodesk have been collaborating for years. today, a giant step forward: a capability never seen before. to help introduce that, ken pimntel from auto desk and michael kaplan from nvidia's iray unit come on stage. Tuesday September 21, 2010 9:58 bob sherbin
10:00
bob sherbin:
3ds Max comes on a screen. there a 3d rendering of an artist's view of an indoor glass table with eight scattered chairs. Tuesday September 21, 2010 10:00 bob sherbin
10:00
bob sherbin:
i've just been told by another pimentel that i misseplled the gent's last name!
Tuesday September 21, 2010 10:00 bob sherbin
10:03
bob sherbin:
so, back to 3ds Max.....Ken P talks about how an iray product that does ray-traced rendering. mentral ray, when it was introduced, would take hours to fake reality. it can now be done in real time. Tuesday September 21, 2010 10:03 bob sherbin
10:04
bob sherbin:
to use mental ray, you need to be able to enter a jillion parameters that not many designers want to learn. with iray there's photorealistic rendering, and it will now be part of the 3ds Max package. Tuesday September 21, 2010 10:04 bob sherbin
10:04
bob sherbin:
Ken P says the simulation is actually physically simulating every photon of reality. This is so complex it's making my head hurt! Tuesday September 21, 2010 10:04 bob sherbin
10:06
bob sherbin:
So, iray traces every impact on every single photon in a scene -- different materials they bounce off of, atmospheric issues, colors. Tuesday September 21, 2010 10:06 bob sherbin
10:09
bob sherbin:
So, with a lengthy introduction, michael is getting ready to show how iray works. in the old world, it took an hour just to do pre-procesing. now it takes just a few moments to simulate with gpus. all the light in the room is indirect. it's coming into a room through double glass doors, it bounces off a shiny floor, it reflects off various surfaces. Tuesday September 21, 2010 10:09 bob sherbin
10:10
bob sherbin:
and needless to say, you see the architect's rendering come to life. it looks like a room you just walked into. Tuesday September 21, 2010 10:10 bob sherbin
10:11
bob sherbin:
now they're showing another scene, an outdoor view of a post-modern building that an architect rendered. it's now being shown as a photo-realistic image.i bet it took 15 seconds for the photo-realistic version to appear. Tuesday September 21, 2010 10:11 bob sherbin
10:13
bob sherbin:
wow, they're now talking about 3ds Max being available on the cloud, so an architect could turn up in a client's office, without schlepping a workstation with him, and show something that's been published already. you can walk around the room real time. it's fully interactive photorealistc rendering. Tuesday September 21, 2010 10:13 bob sherbin
10:14
bob sherbin:
these architecturally rendered images always look better than the spaces i tend to hang out in... Tuesday September 21, 2010 10:14 bob sherbin
10:15
bob sherbin:
michael's taking a rendered scene and adding objects to it, like a glass vase that's actually reflecting back scenes in the room. now he's changing the time of day, all in real time. Tuesday September 21, 2010 10:15 bob sherbin
10:16
bob sherbin:
jhh thanks ken p and michael and they head off, feeling pretty good about the audience's reception, which included some spontaneous applause. Tuesday September 21, 2010 10:16 bob sherbin
10:17
bob sherbin:
JHH is shifting gears. He announces three new OEMs with Tesla systems in production: Tuesday September 21, 2010 10:17 bob sherbin
10:18
bob sherbin:
the first is IBM, with its IBM BladCenter, which now incluldes Tesla. Another is Cray, with its T-Platforms TB2; a third is Cray, with its XE6. Tuesday September 21, 2010 10:18 bob sherbin
10:19
bob sherbin:
Tesla's rapidly reaching a broad footprint, he says. Tuesday September 21, 2010 10:19 bob sherbin
10:19
bob sherbin:
So, what's next? Another gear shift....Over to Adobe, which is democratizing digital photography and editing. The stuff they're working on will blow your mind, in jhh's phrase. Tuesday September 21, 2010 10:19 bob sherbin
10:20
bob sherbin:
Two Adobe heavy hitters come on stagej, david salesin, its' senior principal scientist, and todor georgiev, a senior research scientist. Tuesday September 21, 2010 10:20 bob sherbin
10:21
bob sherbin:
David is recalling trips he took in the past and apparently wasn't such a good photographer. he was frustrated by the poor images of his pics, which included lots that were out of focus. Tuesday September 21, 2010 10:21 bob sherbin
10:22
bob sherbin:
there's a lens they're showing, a plenoptic lens, which captures multiple views of a scene from different view points. each is high res. and with software you can stitch these together into one high res image. Tuesday September 21, 2010 10:22 bob sherbin
10:24
bob sherbin:
so, instead of taking one image of a subject, you have little lenses that take lots of smaller pics -- it creates a 4d image that allows computation to take the place of optics. after the fact, you can change the angle you're viewing and even the focus. Tuesday September 21, 2010 10:24 bob sherbin
10:24
bob sherbin:
gpus can stitch these together 500x faster than cpus. Tuesday September 21, 2010 10:24 bob sherbin
10:26
bob sherbin:
it used to take an hour to distill a photo from this and now it's exceptionally fast. Tuesday September 21, 2010 10:26 bob sherbin
10:27
bob sherbin:
he's showing a photo of a girl taking a picture. the focus on the picture was wronog, it was on some trash cans, not on the cute little girl. they're changing the focus to the girl. now to the branches near her. this is the triumph of algorithems! Tuesday September 21, 2010 10:27 bob sherbin
10:28
bob sherbin:
whoah, the adobe guys are talking about this getting done on video, which would give huge flexibility Tuesday September 21, 2010 10:28 bob sherbin
10:29
bob sherbin:
jhh: so far, digital photography has been about digitizing analogue facts. now, with plenoptic lenses, you can gather far more information and do enormously more. this moves optics into the computer! Tuesday September 21, 2010 10:29 bob sherbin
10:30
bob sherbin:
we're now commaneded to put our 3D glasses on. and it turns on the picture that we were looking at is 3D. they're even changing the focus of 3d pics. Tuesday September 21, 2010 10:30 bob sherbin
10:30
bob sherbin:
"you guys are going to change the world," jensen tells them, as they scamper off. Tuesday September 21, 2010 10:30 bob sherbin
10:31
bob sherbin:
Nice segue here. jensen talks about a doctor changing the world. Tuesday September 21, 2010 10:31 bob sherbin
10:32
bob sherbin:
he introduces dr michael black of the california pacific medical center, where he's chief of the congenital heart program. he was bored, he said, and began to get into using robotics for surgery. Tuesday September 21, 2010 10:32 bob sherbin
10:33
bob sherbin:
the good doctor is recounting how he operated on a 22-week baby that weighed under a pound. Tuesday September 21, 2010 10:33 bob sherbin
10:34
bob sherbin:
back to 3D glasses, which everyone has donned. the video didn't work. ughh. Tuesday September 21, 2010 10:34 bob sherbin
10:34
bob sherbin:
now it shows up! Tuesday September 21, 2010 10:34 bob sherbin
10:36
bob sherbin:
they're talking about surgeons using small incisions and gpu imagery that effectively makes it appear that a beating heart has stopped. except it hasn't. thus, heart surgery can be done robotically on beating hearts. this would open up the door to procedures that aren't yet even possible. Tuesday September 21, 2010 10:36 bob sherbin
10:37
bob sherbin:
what's the implication of this, jhh asks? the good doctor says you can operate on a beating heart and sending a patient home the next day. or, you can perform surgery remotely on an astronaut. eventualy, little robots will work in the body. Tuesday September 21, 2010 10:37 bob sherbin
10:39
bob sherbin:
Very cool stuff. It can be used not just on beating hearts. You could operate within the brain, on very small fetuses, it's all dependent on visual information, and with gpus we can replicate this, the doc sez. Tuesday September 21, 2010 10:39 bob sherbin
10:41
bob sherbin:
the doctor heads down the stairs. Tuesday September 21, 2010 10:41 bob sherbin
10:41
bob sherbin:
Okay, I"m going to wrap up jhh says. Tuesday September 21, 2010 10:41 bob sherbin
10:43
bob sherbin:
we talked about how parallel computing is going to change the world. then, how momentum is accelerating, third how we're reaching an ever-larger footprint, and finally we gave a peek into the future.
but the question is, where are we in the advancment of computing Tuesday September 21, 2010 10:43 bob sherbin
10:43
bob sherbin:
for the first time-ever in NVIDIA's history, we're going to share with you our direction: Tuesday September 21, 2010 10:43 bob sherbin
10:44
bob sherbin:
today, jhh says, we're shipping fermi. kepler is coming in s2011 and maxwell in 2013. these are codenames Tuesday September 21, 2010 10:44 bob sherbin
10:45
bob sherbin:
fermi is just the beginning, he says. he shows a chart plott time and performance per watt. "perf per watt" equals perf. Fermi is 1.5 double precision gigaflops per watt. kepler, based on 28 nanometers, will be 3-4 times per watt of Fermi; we will invest a couple of billion in r&d on it. Tuesday September 21, 2010 10:45 bob sherbin
10:46
bob sherbin:
we're constaintly learning about barrier and walls to achieve the speed of light of parallel computing. Tuesday September 21, 2010 10:46 bob sherbin
10:47
bob sherbin:
the fanatical drive of perf per watt will take us way beyond kepler wtih maxwell. we believe we will continue to be on this ramp. with parallel computing transistors are free, power's not. Tuesday September 21, 2010 10:47 bob sherbin
10:47
bob sherbin:
maxwell will bring 16 gigaflops per watt. this will be 16x improvement over fermi for parallel computing. Tuesday September 21, 2010 10:47 bob sherbin
10:49
bob sherbin:
between now and maxwell, we'll introduce virtual memory, pre-emption, enhance the ability of gpu to autonomously block. these will take gpu computing to the next level, along with a very large speed up in performance. Tuesday September 21, 2010 10:49 bob sherbin
10:49
bob sherbin:
JHH winding down now. GTC is about you, he says, it's about the democratiziation of parallel computing. Thanks for support and have a great week. Tuesday September 21, 2010 10:49 bob sherbin
10:50
bob sherbin:
music goes back up, jensen moves off and now dan vivoli comes up on stage to talk about logistics. They both show off their tattoos. Tuesday September 21, 2010 10:50 bob sherbin
10:51
bob sherbin:
that's it for the live blogging. i had great fun and may stick with this gig!
LinkBack URL
About LinkBacks
Citar
y pa el 2013 se viene maxwell--- Doble OMg..
Visita la web donde laborowww.tu-nube.com
Iniciado por Sebastian Rojas
Wow, eso tiene tanto potencial... primero que CUDA se masificara mas rapido y facil ya que no necesitaran una VGA nvidia para crear y compilar CUDA, pero lo que realmente espero con esto es que una VGA de nvidia + un CPU poderoso multicore trabajen AMBOS cuda a la vez... se que es muy difícil hacer eso pero el desempeño del código CUDA se iría a las nubes... imaginen un Phenom x6 + Fermi, se vuela la barda.... tal vez nvidia pueda lograr la homogeneidad antes que intel o amd... SI, es un sueño, se vale soñar.
, colocale un SPOILER, esta brutal la cantidad de texto y no mas hace que el foro sea enorme y eterno el scroll para llegar abajo.
....
Permisos de publicación
