![\"Fairmont_out\"](\"http:\/\/i1.wp.com\/alienbabeltech.com\/main\/wp-content\/uploads\/2009\/10\/fairmont_out-jpg.jpg\" "\\"Fairmont_out\\"")
<\/a> ![\"2ndFloorplanFairmount\"](\"http:\/\/i1.wp.com\/alienbabeltech.com\/main\/wp-content\/uploads\/2009\/10\/2ndfloorplanfairmount-jpg.jpg\" "\\"2ndFloorplanFairmount\\"")
<\/a> ![\"floorplanfairmount-002\"](\"http:\/\/i0.wp.com\/alienbabeltech.com\/main\/wp-content\/uploads\/2009\/10\/floorplanfairmount002-jpg.jpg\" "\\"floorplanfairmount-002\\"")
<\/a><\/p>\nIf you want to check out the jam-packed schedule, or read anything and everything to do with the GTC or the just-released Fermi GPU architecture, Nvidia\u2019s own site<\/a> has a wealth of information available from the conference.<\/p>\n The first installment of this short series covering Nvidia\u2019s GTC, Day One was published here<\/a>.\u00a0 Their three-in-one GPU technology conference used an entire floor at the impressive luxury Fairmont San Jose Hotel<\/a> in beautiful downtown San Jose, California for 3 days \u2013 September 30 through October 2.\u00a0 We are continuing to give you our impressions of this event for day 2 (now) and day 3 (tomorrow). For the final part of this series (this week), you can expect a more polished summary of Fermi architecture when we review all of our notes, transcribe all of our audio and the forty-plus gigabytes of full HD-HR video that we shot, and finally make sense of everything that we observed.\u00a0 We will also post our own video and links to full discussions of what we covered at Nvidia\u2019s GTC.\u00a0 We also had a couple of semi-private press conferences with Nvidia officials that we will cover for you.\u00a0 We thank them for their time and their extraordinary attention to our questions as we help our readers to understand their new revolution in GPU computing.<\/p>\n Nvidia\u2019s GTC was not about hardware.\u00a0 We only saw a single production mock-up of the coming Fermi (Tesla division) in Jensen\u2019s hand at the general press conference yesterday.\u00a0 The Fermi production silicon is so new, it is still hooked up to wires in Nvidia\u2019s box \u2013 completely functional as we saw demonstrations actually using it \u2013 but not yet ready for public presentation.<\/p>\n The Nvidia officials that we spoke with, all insisted that Fermi will ship this year \u2013 and besides excelling at GPU computing, it will also be the single fastest GPU \u2013 period!\u00a0 Rumors floating around the GTC whispered that Fermi would be 1.6 to 1.8 times faster in general gaming than GT 200; but those are just rumors that are neither substantiated nor confirmed by those who really know.\u00a0 Nvidia is not yet ready to talk about gaming performance of the GeForce Fermi F100 GTX other than to say, \u201cit will be the fastest\u201d.<\/p>\n In the above picture (left), the production mock-up is of the new Fermi Tesla \u2013 their GPU professional computing card. We did get confirmed that this is the first time that one of Nvidia\u2019s code names has actually become the marketing name for it.\u00a0 Previously, \u201cGT 200\u201d was Nvidia\u2019s internal codename for GPU Tesla ver. 2.0<\/em>.\u00a0 What was unusual was that the codename for the Tesla architecture then crossed over to become the name of their newest third brand, Tesla.\u00a0 Nvidia\u2019s Tesla products are intended for the high performance computing market. Their lack of ability to output images to a display is the primary difference between Tesla products and Quadro workstation and GeForce consumer video cards. The main function of Tesla products are to help with simulations, large scale floating-point calculations and image generation for professionals, using Nvidia\u2019s GPU programming language, CUDA.<\/p>\n Fermi features several big innovations over GT 200 (Tesla) architecture that are essential to greatly improving GPU computing \u2013 up to eight times in some cases!: However, this GPU Technology Conference is all about software and the companies and the individuals that will be using it to program for a much expanded general computing architecture in Fermi.\u00a0 We saw languages such as C++ and Fortran newly riding on top of CUDA \u2013 Nvidia\u2019s own open source language which is the GPU programming equivalent of x86 instructions for the CPU.\u00a0 It was invented by Ian Buck \u2013 the genius programmer who wrote Brook for his Master\u2019s thesis.\u00a0 He also sat down with ABT in a small meeting and shared information about programming for Fermi.\u00a0 He is a very personable and yet down-to-earth individual who obviously loves his work as do all of the Nvidia employees that we met.\u00a0 One cannot fake the pride, enthusiasm and excitement they all show for their company and its newly released GPU and new programming and debugging tools.\u00a0 And the attendees and presenters were just as enthusiastic at the new software releases. It was contagious.<\/p>\n ABT’s attendance at this conference has turned into a short series on Nvidia\u2019s GPU computing and what this conference is really all about.\u00a0 It is really a computing revolution that started as part of Nvidia\u2019s vision a relatively few short years ago.\u00a0 It is Nvidia\u2019s revolution to make the GPU \u201call purpose\u201d and just as important as the CPU in computing.\u00a0 We saw this genesis begin with last year\u2019s Nvision08 and expand and grow into the GTC, only one short year later.<\/p>\n As we can see from our images, the conference was physically large and absolutely jam-packed with multiple sessions occurring at the same time.\u00a0 It reminded me of college, with everyone hurrying to the next lecture.\u00a0 All along the second floor were over 100 posters that showed what Nvidia\u2019s GPU revolution had already accomplished and what was being planned.\u00a0 Everywhere we were reminded of the extreme advantages of the GPU\u2019s parallel processing contrasted with the CPU\u2019s serial processing; advantages in cost, energy-saving and time.\u00a0 Of course, the first poster belonged to Nvidia and reminded everyone to conserve and to think \u201cgreen\u201d (in every way).<\/p>\n Nvidia also knows how to treat their press very well.\u00a0 They gave out a very nice press kit and were continually helpful to this reporter.\u00a0 The food supplied at the conference was very good with several choices available \u2013 including vegan.<\/p>\n <\/p>\n <\/p>\n Day Two \u2013 focusing on research and the emerging companies<\/strong><\/span><\/p>\n It was warmer in San Jose the next two days and no jacket was required any longer as the conference also got into full swing.\u00a0 We noted that the first day\u2019s events was followed by a nighttime pub crawl through downtown San Jose to raise money for charity.\u00a0 Quite a few people were quite tired \u2013 but happy \u2013 even the morning after and no enthusiasm appears to have been lost for the GTC whatsoever.<\/p>\n Day Two Keynote: Research<\/strong><\/p>\n Harvard Professor Hanspeter Pfister began a fascinating 90 minute talk at 9 AM on how GPUs are being used to answer some of science\u2019s thorniest problems \u2013 like how the brain is wired, how the human visual system works, as well as how the universe began.\u00a0 He also discussed the very pragmatic subject of preventing heart attacks with real-time advanced MRI-type scanners that will likely become hand-held miniature diagnostic devices that should evolve into the \u201ctri-corder\u201d of the future.\u00a0 What was unusual is that he was able to make the subject of \u2018matter interacting at the quantum level\u2019<\/em>, interesting to experts as well as to non-technical viewers in the audience.\u00a0 That is a rare gift.<\/p>\n The Connectome Pipeline<\/strong><\/p>\n Let\u2019s cover some of the subjects that he presented that demonstrate the need for GPU parallel computing and its incredible advantages over CPU supercomputers. One project has to do with finding out how the brain is wired and it is massive, rivaling the Human Genome project in every way.\u00a0 It is called finding the Connectome pipeline and this diagram is the simplest representation of what they want to achieve as they map the synaptic connections.<\/p>\n They are using rat brains in their research and the idea is to slice the brain tissue into extremely thin but long slices \u2013 much as one uses an apple peeler or a lathe.<\/p>\n Next it is imaged with electron microscopes.\u00a0 Also, notice the 40 GB of data that is required to image such tiny sections and the fact that they need many samples to make a 3D image of their overall sample that will put a massive strain on the fastest supercomputer.\u00a0 We are talking about 1 cubic MM requiring 1.5 PB of data to manipulate!<\/p>\n They certainly aren\u2019t done yet.\u00a0 After they have properly imaged the section of rat brain sample they took, they next need to reconstruct it and map it.\u00a0 Here we see reconstruction and active brain ribbons.<\/p>\n Again, this takes massive computational power as they do an elliptical approximation of ribbons with a compression ratio of 1000 to 1.\u00a0 Here we see the active ribbons processing on the GPU contrasted with the CPU supercomputer.\u00a0 A twenty-three times speedup is far beyond extra-ordinary.\u00a0 It completely changes the scientist\u2019s work for the better.<\/p>\n They already have some incredible results with these multi-scale images.\u00a0 What is making it even more interesting is that the test rats are genetically bred so that their tissue now show up in colors under ultraviolet light so that their brains can be visualized and mapped easier.\u00a0 They call it the \u201cBrainbow\u201d.<\/p>\n One of the slides half-humorously compared the brain\u2019s wiring to that of a PC:<\/p>\n The future is wide open to this kind of GPU parallel computing that is so much faster than what they had before using only the CPU.<\/p>\n <\/p>\n <\/p>\n Day Two Keynote (continued):<\/strong> The Universe\u2019s Beginning<\/strong><\/p>\n This section will been even briefer yet it will demonstrate GPU processing meeting the needs of some scientific communities that are doing important research in Australia\u2019s outback by using radio astronomy to look at the beginning of the universe \u2013 the first 300,000 to 1,000,000 years of its existence that is not at all well understood yet.<\/p>\n To do this they must use many small linked antennas over 1 square kilometer using red shift and a 3D projection of the sky.<\/p>\n They must be far from civilization\u2019s radio interference to do this work.<\/p>\n This means they have special needs for low power and high performance per watt. All of their supercomputing must be done with generators and this means an extremely limited power supply for their project.\u00a0 Even though the antennas are miniature, the sheer amount of data that is require to be processed would require 20KW of power using traditional CPU clusters.\u00a0 In a remote situation, that is impossible so they turned to the GPU.<\/p>\n They also have extreme computing needs with mass amounts of data.\u00a0 Using the GPU parallel processor has allowed them to accomplish this with a savings of 10 to 1 performance to watt over using just the CPU.\u00a0 Success with GPU computing and a speedup of 20 times over the CPU is an added bonus far beyond the power saving.\u00a0 What does this lead to and what kind of data needs will they have in ten short years?\u00a0 Well, let the following slides presented at the conference answer it.<\/p>\n We can see practical benefits for all of us coming from this research.\u00a0 Besides understanding the universe’s beginning, being able to predict solar flares accurately can help us to safeguard our communication systems and our power grid.<\/p>\n These are exciting times and the GPU\u2019s parallel processing makes the formerly impossible work.<\/p>\n Simulations<\/strong><\/p>\n Now lets us look at a few of the challenges of GPU computing in tackling simulations.<\/p>\n Professor Pfister went on to address the many challenges of simulations including a specific example of determining the organic properties of molecules using the drug Taxon as a quantum multi-body problem.\u00a0 Since the researchers are simulating the molecule’s bouncing electrons, massive computational power is necessary.\u00a0 Here the GPU absolutely left CPU computing in the dust in every way.\u00a0 The GPU will allow for simulation of small molecules as well as large ones now.\u00a0 The future of this kind of research may also lead to improved organic voltaic materials from copying plant properties.<\/p>\n Next up, he presented the challenge of determining how the brain performs object recognition.\u00a0 They are determining models and parameters of the human visual system which require intense computing.\u00a0 Again, the speed differences between using the GPU to perform these tasks and the CPU were shown to be similar to the difference between using an airplane verses a car.<\/p>\n Predicting & Preventing Heart Attacks<\/strong><\/p>\n Here absolutely massive amounts of computing power is needed \u2013 think MRI in real time.\u00a0 Again, the GPU makes possible in emergency situations what the CPU simply cannot do it in a timely fashion.\u00a0 Scientists are using multiscale hemodynamics to predict plaques eruptions which release dangerous artery clogging fat.\u00a0 They look at the circulation in real time and determine where there are potential problems.\u00a0 One of the researchers actually prevented his own heart attack by participating in this project!<\/p>\n With an incredible speed-up of 120 times over the CPU, the near-future promises miniaturization and eventually portability.<\/p>\n W can see that the scaling of Nvidia\u2019s multi-GPU in the cluster is impressive.<\/p>\n There is no contest.\u00a0 The GPU computing wins by a huge margin and Dr Pfister has asked Nvidia for further expanded GPU capabilities so that we can have his future tools for the benefit of mankind. His first HTC lesson is this one:<\/p>\n He then asks for GPU HTC appliances similar to what other industries already have:<\/p>\n Finally, he says the scientists and researchers need to program in their own languages \u2013 they should not have to learn CUDA but rather be comfortable with what they already use.\u00a0 And we will find out shortly that Nvidia has listened to him.\u00a0 There are several domain-specific languages now available including C++ and Fortran with more on the way.\u00a0 Best of all, they released Visual Studio and also a full GPU debugger tool, Nexus \u2013 and much more software to go with their new Fermi and even their older GPUs.<\/p>\n The scientists say if they can get these tools, our future will have these things:<\/p>\n Well, that concluded just 90 minutes of one intense keynote presentation; and there were dozens of more presentations.\u00a0 However, it was time to race over to the Emerging Companies opening address – the kick off to the Emerging Companies Summit, featuring over 60 startups using Nvidia processors in a whole new range of ways.\u00a0 This particularly interested this editor as AlienBabeltech is<\/em> an emerging company that already plans to use GPU parallel computing in a brand new way so as to change the way humans interact on the Internet.\u00a0 So, we sadly had to forgo listening to Bill Dally, Nvidia\u2019s chief scientist, in the other lecture room.<\/p>\n <\/p>\n <\/p>\n Day Two: The Emerging Companies Summit<\/strong><\/p>\n Nvidia\u2019s GTC is three parts: first, we were introduced to research and soon we will look at programming with CUDA, C++ and other languages just now made available for the GPU.\u00a0 But now we are looking at a few really successful companies that have already leveraged GPU computing to produce extraordinary products.\u00a0 The main speaker was Nvidia\u2019s Vice President Jeff Herbst and he had 3 primary guests presenting their ongoing success stories to the 450 attendees in the audience.\u00a0 The audience had grown by 150 over last year.<\/p>\n Jeff Herbst stressed Nvidia\u2019s \u201cGPU ecosystem\u201d where they all work to support each other.\u00a0 His first guest was Laurent Gil, CEO of Viewdle<\/a>.\u00a0 This company is the first to be able to organize mass photos and especially video successfully by finding individual faces out of millions of images far more quickly and reliably than can be done with the CPU.\u00a0 The more successful a search is of a particular person, the easier it is to find more of that same person so as to even generate automatic slide shows.<\/p>\n How fast?\u00a0 Well, he said you can organize 2,000 photos in just a few minutes!\u00a0 He discussed his licensing plan to software houses and said some of his face recognition software will be free for some recognition and will be made available to consumers next year.\u00a0 He stressed that the Nvidia GPUs that consumers already own, will greatly speed up his program over just using the CPU.\u00a0 Of course, you can also expect this technology to be coming to an airport security camera near you.<\/p>\n He was followed by Tarek El Doktor, CEO of Edge 3 Technologies<\/a>, the premier gesture recognition company.\u00a0 They use a very cheap solution – two non-synchronized web cams that allow the computer to \u201csee\u201d and recognize gestures.\u00a0 The demonstration he gave was impressive.\u00a0 Again he stressed the importance of leveraging the GPU that is already in consumer\u2019s PCs to make his program far more effective and swift.<\/p>\n He further went on to say that although the program used 70,000 lines of code, its footprint was as small as an optical mouse’s and it only used 3% of the CPU \u2013 it depended on GPU processing.\u00a0 He speculated that 1,000 times more processing power would allow the computer to read facial gestures including emotions and that computers would be able to \u201csee\u201d within five years by mean of AI trained \u201cunsupervised learning\u201d.\u00a0 Exciting times indeed!<\/p>\n Finally we had a talk by MirriAd\u2019s<\/a> CEO and then it was followed by a full panel discussion including the previous two guests, Steve Perlman of OnLive<\/a> (Games on Demand) and analyst Rob Enderle, CEO of Enderle Group, for another hour.\u00a0 MirriAd digitally embeds advertising into any video content, as though it was always there. This advertising allow content owners, whether film, TV or online \u2013 including social media like Facebook and YouTube – to effectively monetize their assets.\u00a0 It is incredibly high volume work requiring placement of hundreds of ads a day.\u00a0 This requires GPU processing and the real time demonstrations of placing ads into existing media seamlessly, was very natural and quite impressive.<\/p>\n Now it was a choice between attending a Motion DSP<\/a> presentation or hitting the exhibits.\u00a0 I picked viewing the exhibits and later got to talk with the MotionDSP people.\u00a0 They have a fascinating product that cleans up shaky video very quickly using CUDA-enabled GPUs (the fastest; they will also run on the CPU-only, albeit much more slowly).\u00a0 This editor wishes he had this product right now for many of the video and stills that cannot otherwise be used.\u00a0 Of course, we will get the MotionDSP press kit and review it thoroughly for you in a later review.<\/p>\n Unfortunately, this editor had a HW malfunction and had to run all over San Jose for a replacement \u2013 finding it finally at Fry\u2019s and then racing back through traffic to the conference for a 5 PM semi-private meeting with Nvidia\u2019s Drew Henry, the general manager of the GeForce business unit and with Jason Paul, the GeForce product manager.\u00a0 Here we discussed Fermi architecture, Ion, Tegra and Nvidia\u2019s plans for the future.\u00a0 We will detail this discussion in the last part of this series when we cover Fermi architecture and Nvidia\u2019s new direction.\u00a0 However, here is an image of the reverse side of new Fermi GPU:<\/p>\n The meeting began at 5 PM and was supposed to end in 1\/2 to one hour.\u00a0 It was actually over just before 6:30 PM and we appreciated the extra time that the Nvidia officials spent with us answering our questions.\u00a0 We will have those details for you in the next installments of our GTC coverage.\u00a0 Well, it was now getting dark and an exciting and jam-packed Day 2 of Nvidia\u2019s GTC had concluded.<\/p>\n We saw how Nvidia\u2019s revolution in GPU computing is exciting scientists, researchers and new companies. This left me looking forward to the last day of the GTC where we sit in on CUDA programming and learn to use Nvidia\u2019s new debugger tool for the GPU – Nexus –\u00a0 as well as to delve deeper into Microsoft’s Visual Studio, which now supports the GPU.\u00a0 For the last day, we see scheduled, Richard Kerris of Lucas Films and a \u201cfireside chat with Jensen and Peddie\u201d.\u00a0 We did not miss any of this, including visiting many of the emerging companies exhibits. We shall have it covered here for you tomorrow including the details of the new Fermi architecture<\/p>\n Before we go, as a tease, we would like to leave you with a couple of images.\u00a0 First of all, we found out from a very good source that Nvidia actually killed Dumbledore.\u00a0 Details are in the next installment.<\/p>\n And could this possibly be Jensen, Nvidia’s CEO and co-founder doing some kind of weird 3-D dance in the exhibit room?\u00a0 Yes indeed!\u00a0 We explain tomorrow.<\/p>\n Mark Poppin<\/p>\n ABT Senior Editor<\/p>\n <\/p>\n Please join us in our Forums<\/a><\/p>\n Become a Fan on Facebook<\/a><\/p>\n Follow us on Twitter<\/a><\/p>\n For the latest updates from ABT, please join our RSS News Feed<\/a><\/p>\n Join our Distributed Computing teams<\/p>\n<\/a> <\/a><\/p>\n
\n \u2022 512 CUDA cores (twice as many)
\n \u2022 NVIDIA Parallel DataCache technology
\n \u2022 NVIDIA GigaThread\u2122 engine
\n \u2022 ECC support<\/p>\n<\/a><\/p>\n<\/a> <\/a> <\/a> <\/a> <\/a> <\/a><\/p>\n<\/a><\/p>\n<\/a><\/p>\n<\/a><\/p>\n<\/a> <\/a><\/p>\n<\/a> <\/a><\/p>\n<\/a> <\/a> <\/a><\/p>\n<\/a><\/a><\/p>\n<\/a> <\/a> <\/a><\/p>\n<\/a><\/p>\n<\/a><\/p>\n<\/a><\/p>\n<\/a> <\/a><\/p>\n<\/a><\/p>\n<\/a><\/p>\n<\/a><\/p>\n<\/a><\/p>\n<\/a><\/p>\n<\/a> <\/a><\/p>\n<\/a><\/p>\n<\/a><\/p>\n<\/a><\/p>\n<\/a> <\/a><\/p>\n<\/a><\/p>\n<\/a> <\/a><\/p>\n<\/a> <\/a><\/p>\n<\/a><\/p>\n<\/a><\/p>\n<\/a><\/p>\n<\/a><\/p>\n<\/a><\/p>\n
<\/p>\n
<\/p>\n
<\/p>\n
<\/p>\n<\/a><\/p>\n
<\/p>\n
<\/p>\n
<\/p>\n
<\/p>\n\n
\n