VizWorld.com

NVidia has released the latest version of their popular NSight OpenGL & GLSL Debugging suite, boasting new support for OpenGL4.2, CUDA5, and new local-debugger features.

The final release of Nsight™ Visual Studio Edition 3.0 is available for download under Nsight™ Visual Studio Edition Registered Developer Program. This new release officially supports OpenGL frame debugging and profiling, GLSL GPU shader debugging, local single GPU shader debugging, the new Kepler™ GK110 architecture found in Tesla® K20 & GeForce GTX TITAN, and CUDA® 5.0.

Download it at the link below.

via Nsight Visual Studio Edition 3.0 New Features | NVIDIA Developer Zone.

Science cuda, nsight, nvidia, opengl

At NAB recently, NVidia has their usual presence showing off their Quadro products and various software technologies.  Many people may not have noticed, however, that NVidia was actually present in over 40 booths, covering everything from encoding technologies to color grading systems to rendering tools.  Many people thought that CUDA would fade away as the more open OpenCL took hold, but Nvidia is still reaping the benefits of the powerful software development pipeline they’ve built around CUDA as it continues to be integrated into more and more products.

While Nvidia admits the existence of OpenCL and offers support, the company says it is getting new customers for CUDA and it is not seeing a shift to OpenCL even though OpenCL gives developers a cross-platform approach. What Nvidia’s booth really demonstrates is that customers are seeing the benefits of optimizing for multi-core and GPU compute and they aren’t waiting for the OpenCL tools to evolve or for Intel to get tools out there for its multi-core MIC processor. There really was an impressive amount of Nvidia partners at NAB this year. The case has been made for multi-core, for GPU compute, and, for now, CUDA.

via GraphicSpeak » Nvidia preaches the GPU compute gospel at NAB.

Hardware cuda, nab, nvidia, opencl

There’s a new version of the glu3D fluid motion simulator for 3dsMax that brings it up to date with 3dsMax2012 and adds in CUDA computation for improved performance.

It speeds up intensive computations that glu3D performs to calculate particle fluid dynamics. With this glu3D edition it is possible to simulate faster and with more particles than before; it is possible to simulate your scene more times; it is almost interactive. Animated fluid scenes with more quality and level of detai

You can download their free Demo version at their site, or buy it for $760.

via glu3D GPU Edition for 3dsMax – NEW!.

Graphics autodesk, cuda, glu3d

Credit goes to InsideHPC for digging up a lecture by Rob Gillen at the recent CodeStock 2011 event on GPGPU Development.

Video: Introduction to GPGPU Development using CUDA | insideHPC.com.

Science conference, cuda, gpgpu

Click for Fullsize

InsideHPC dug up a nice paper from the ACM PLDI conference that discusses a prototype CPU-GPU Communication optimization tool called ‘CGCM’.  From their conclusions:

CGCM has two parts, a run-time library and an optimizing compiler.  The run-time library’s semantics allow the compiler to manage and optimize CPU-GPU communication without programmer annotations or heroic static analysis. The compiler breaks cyclic communication patterns by transferring data to the GPU early in the program and retrieving it only when necessary. CGCM outperforms inspector-executor systems on 24 programs and enables a whole program geomean speedup of 5.36x over best sequential CPU-only execution.

Impressive results to say the least.  Their examples seem based on CUDA code, not really OpenCL, although they do relate it to tools like OpenMP approaches and CUDA-lite.  It seems it becomes the programmer’s responsibility to mark data that’s to be shared between the GPU and CPU, and then some memory magic happens to move data between the two systems transparently.

It’s available as a PDF.

New Paper: Automatic CPU-GPU Communication Management and Optimization | insideHPC.com.

Science algorithm, cuda, gpgpu

NVidia has just announced the newest entrant into the ever-growing CUDA Center of Excellence program, Stanford University.  Stanford already has a CUDA architecture and parallel computing program, so adding them to the Center of Excellence program is really a mere formality.

“It’s vitally important that our faculty be at the forefront of computing technology so that we can continue developing state-of-the-art computational algorithms that drive innovation in the sciences and engineering,” said Margot Gerritsen, director, Institute for Computational & Mathematical Engineering, and associate professor, Department of Energy Resources Engineering, at Stanford University.  “This award allows us to broadly expand parallel computing education and research programs to large numbers of researchers and students from a wide variety of disciplines.”

The grants provided by NVidia as part of the program will be used to support some new research programs on mesh-based solvers for partial differential equations and probability and uncertainty quantification work.

Get the full release after the break.

Science cuda, nvidia, stanford

So, for the last week or so the internet has been abuzz with stories about “BitCoin”, the new all-digital currency that’s going to destabilize governments around the world and bring us to a new utopian society.  Well, yeah it’s a lot of hype.  But when I heard about the “mining” aspect of it, and how it’s almost entirely GPU based, I figured I would check it out.

From what I can tell, the “mining” part is really just a brute-force hash attack, looking for specific numbers.  I ran some experiments with this once before as part of my Quadro5000 review using HashGPU.  Using the OpenCL bitcoin miner, I figured I could some up with some nice results.  I had a machine handy with two GeForce GTX285′s in it, and easily managed to eek out about 64Mh/s on each card, for a total around 128Mh/s. (Mh/s = Million Hashes per Second).  I also happen to have the Quadro 5000 card around, based on the Fermi Architecture, so I thought I’ld throw it in as it’s not currently in the Wiki Hardware Results.

I was very disappointed to find that my Quadro5000 could only manage about  58-59Mh/s, a startling 10% less than the GTX285.  This truly baffled me.  The Quadro 5000, which handily beats AMD cards in most benchmarks, falls waaay behind the ATI offerings which easily rake in 100+Mh/s, some hitting 300Mh/s.

All in all, I ran with two GTX285′s for about 4 days, and mined all of 2 BitCoins.  Presumably with a single AMD Radeon 6990, for $700 which claims to rake in over 650Mh/s , I could have make 5x that.  It’s interesting to see that as popular as CUDA is, there are still several problems where AMD’s “stream” design beats NVidia’s hands-down.

And of course, this wouldn’t be a BitCoin article if I didn’t include “If you liked this article, feel free to send some BitCoins to 1HXHDYeQux5BVzwTF5gEwMS2MgaKXwXeft“.

P.S. If anyone actually does send me any, shoot me an email with the amount for a Shout-Out here.

Update 7pm: Wow.. I just refreshed my wallet and thanks to the 3 people who actually sent me a total of 0.12 BTC, the equivalent of about $1 at current exchange rates.

Hardware, Science amd, bitcoin, cuda, gpgpu, nvidia, opencl

The Georgia Tech NVIDIA CUDA Center of Excellence is preparing a nice 2-day long tutorial on GPU programming and heterogeneous computing, including both CUDA and OpenCL.  The event will only cost you a $100 registration fee and the cost of your room and time, making it one of the best ways to get into GPU programming.

Hit the website for all the details and links.

Keeneland Workshop | Keeneland.

Science cuda, education, nvidia, opencl

Big news from Nvidia today as they announce the latest version of their GPGPU toolkit, CUDA 4.0.  This new version has all the usual performance enhancements and bugfixes, but also comes with 3 new features that I can guarantee all you CUDA developers are going to love.

• 

  GPUDirect2

  GPUDirect2.0 – The previous version of GPUDirect worked with clusters based on Mellanox Infiniband backbones, but this new version works with multiple GPU cards in a single machine.  Where previously the CPU was involved in memory transfers between cards, now you can DMA transfer directly between cards using MPI-style Send & Receive commands.

• Unified Virtual Addressing – Now, CPUs and GPUs all show up in a single uniform address space.  This makes moving memory between them much easier.
• Integrated Thrust Support – A great C++ library similar to LABLAS and CULAPACK, it adds in standard template constructs for all the popular data types and algorithms.  Thrust has a great following and active community, and boasts run-time selection of CPU vs GPU code, making the resulting code a bit more portable than previous CUDA.

All of this goes to further reinforce NVidia’s commitment to not just building nice graphics cards, but to continue to build and support a developer community around the computational capabilities of their hardware.  With renewed support in their Tegra line and the new ARM cores on the horizon, NVidia knows that having a wide community of developers ready to go on the new hardware is critical to mainstream market success.  Microsoft is already pumping up support for a future ARM-based Windows, and ARM already has wide support in many embedded applications like settop boxes, smartphones, and tablets.  Tools like Unified Virtual Addressing and GPUDirect2 further Nvidia’s attempts to tear down the barriers between the CPU and GPU, making future porting to ARM systems simpler.

Get the full details of the release in the Press Release after the break.

Read more…

Hardware, Science cuda, feature, nvidia

Another win for NVidia in the VFX and CUDA space comes from Double Negative, who found their proprietary fluid simulation system “Squirt” getting a nice 20x performance boost.

“Moving our fluid solver onto the GPU allows our artists to get the results of their simulations back much faster, without any impact to their workflow,” explained Dan Bailey, lead GPU developer, Double Negative. “By default, fluid simulations are now sent to a specialized GPU farm, affording the artists more time to iterate and ramp up the complexity of a shot to achieve a more believable result for the big screen.”

It took 6 months of hard work to complete the transition, and the 20x number they’re seeing isn’t even using a Quadro4000 (Fermi-based).  Just wait until they upgrade to see the numbers then.

via DNeg Accelerating Visual Effects with NVIDIA Quadro and CUDA « NVIDIA.

Graphics, Hardware cuda, doublenegative, nvidia, vfx