Q 1: What is CUDA?
- A parallel computing architecture for NVIDIA GPUs
- A computer language for GPUs
- A new way of writing graphics shaders
- A predatory fish
The correct answer is A. CUDA is our revolutionary parallel computing architecture for NVIDIA GPUs. CUDA enables programmers to extend the parallel processing power of GPUs to applications beyond pure graphics.
Q 2: How many CUDA-enabled GPUS have been shipped to date?
- Two million
- More than 100 million
The correct answer is ‘D’. The CUDA architecture was introduced with the GeForce 8 family of GPUs. All current desktop and notebook GPUs include cores with CUDA technology. Today we have shipped more than 100 million GeForce 8+ products, making CUDA the most widely available parallel computing architecture in the world. For a list of CUDA-enabled products, go to: http://www.nvidia.com/object/cuda_learn_products.html
Q 3: How many downloads of the CUDA 2.0 SDK have there been to date?
- More than 60,000
The correct answer is D. There have been over 70,000 downloads (71,814, in fact) of the CUDA SDK 2.0 and close to that many downloads of the CUDA 2.0 Toolkit. The CUDA SDK consists of code samples and docs. The CUDA Toolkit consists of the runtime libraries and docs as well as the toolchain for programming (compilers for various languages, debugger, and profiler).
Q 4: Which operating systems have support for the CUDA architecture?
- Mac OS
- All of the above
The correct answer is D. Software development kits for the CUDA architecture are available for all the major operating systems. This includes MAC OS 10.5.2 and later; popular 32/64bit Linux versions (including Redhat, SUSE, OpenSUSE, Ubuntu, Fedora); and Windows 32/64bit XP and VISTA
Q 5: Which of the following languages cannot be used to program the CUDA architecture today?
The correct answer is D. NVIDIA and third parties provide support for a variety of programming languages, including C, C++, Fortran, Java, Python, Perl and .NET. As of today, no third party has signed up to support Pascal on CUDA, but it’s possible it could happen in the future.
Q 6: Which emerging APIs (application programming interfaces) will be CUDA-accelerated?
- OpenCL and DX11 Compute
- OpenGL and DX10
- Open VCR
- None of the above
The correct answer is A. The CUDA architecture can be accessed directly via OpenCL (Open Computing Language) and DX11 Compute Shader from Microsoft. To see a video of the first OpenCL GPU demo (an “nbody” simulation, demonstrated by NVIDIA at Siggraph Asia 2008), go to: http://www.youtube.com/watch?v=r1sN1ELJfNo&feature=channel_page
Q 7: What is the relationship of CUDA to OpenCL?
- OpenCL runs on top of the CUDA architecture
- OpenCL competes with the CUDA architecture
- OpenCL has no relationship to the CUDA architecture
- They are identical
The correct answer is A. OpenCL runs on top of the CUDA architecture and uses the parallel cores of the GPU to give applications a big performance boost. OpenCL is good for GPU computing because it provides another way of unlocking the tremendous compute capabilities of the CUDA architecture. Background: The OpenCL API was submitted by Apple to the Khronos Group. The Khronos Group is an industry consortium creating open standards for parallel computing, graphics and dynamic media on a wide variety of platforms and devices. Neil Trevett, VP of Embedded Content at NVIDIA, is the Khronos Group president and OpenCL working group chair.
Q 8: Which of the following is an appropriate use of the word CUDA?
- “Developers around the world are programming to the CUDA architecture”
- “Tonight I am going to program in CUDA”
- “How CUDA you?”
- “CUDA is a compute API similar to OpenCL and DX11 Compute”
The correct answer is A. It is very important for us to communicate that CUDA is our parallel computing architecture, and not a language or an API.
Q 9: What is the range of the largest speedups currently achieved by scientific algorithms using the CUDA architecture?
- 2x to 10x
- 10x to 50x
- 50x to 100x
The correct answer is D. The CUDA architecture provides dramatic speedups of scientific algorithms – often greater than two orders of magnitude. For example, cone-beam computed tomography (CBCT) image reconstruction can be sped up by more than 300 times with the CUDA architecture. For more examples of CUDA speedups, see our database of user-submitted applications and research papers on the home page of CUDA Zone: www.nvidia.com/cuda
Q 10: The CUDA-based Tesla Personal Supercomputer provides:
- 4 teraflops of compute power via a standard power strip
- Supercomputing capability for under $10,000
- Programmability in C for Windows and Linux
- All of the above
The correct answer is D. The Tesla Personal Supercomputer is up to 250 times faster than traditional workstations, depending upon the application. It costs 1/100th the price of comparable performance CPU clusters and consumes about 1/20th the power. The introduction of the Personal Supercomputer brings the power of a traditional supercomputer to the developer?s desktop. For more info, go to: http://www.nvidia.com/object/personal_supercomputing.html
Q 11: How many universities have courses featuring the CUDA architecture?
- More than 100
The correct answer is D. Universities all over the world are starting to teach courses featuring the CUDA architecture. To see them on a Google world map, click here: http://tinyurl.com/CUDA-University-Map. If you know of professors/researchers (perhaps from your own alma mater) who could benefit from learning more about CUDA, please ask them to contact us at: UniversityPartnership@nvidia.com.
Q 12: Which of the following consumer applications are available with CUDA acceleration today?
- Badaboom (video transcoding)
- Power Director 7 (home video editing)
- TMPGEnc Xpress 4.0 (video filters)
- All of the above
The correct answer is D. Developers are creating new CUDA-accelerated consumer applications in the areas of games, video processing, photo/imaging, immersive web, and natural interface processing. See a presentation on NVIDIA and Badaboom, filmed at CES, here: http://tinyurl.com/nvidia-badaboom. See an example of great effects in a CUDA-accelerated game (Ghost Recon Advanced Warfighter) here: http://tinyurl.com/nvidia-graw
Q 13: What are some of the areas in which CUDA is making a difference?
- Life sciences
- Oil and gas exploration
- Financial modeling
- All of the above
The correct answer is D. The CUDA parallel programming architecture is being adopted across a broad range of industries. CUDA enables GPUs to accelerate many types of applications — from performing image processing of seismic data many times faster than before (SeismicCity) to creating fast, accurate ultrasounds to detect cancer (TechniScan). To watch TechniScan talk about why they like CUDA, go to: http://www.youtube.com/watch?v=OhmwF-py3PU . To see more success stories, go to: http://www.nvidia.com/object/tesla_testimonials.html
Q 14: Who said: CUDA “could save the world or my life someday”?
- Bill Gates, philanthropist
- John Markoff, reporter
- Rob Enderle, industry analyst
- Jack Bauer, high-tech consumer
The correct answer is C. In December, industry analyst Rob Enderle nominated his top favorite products for the year in an article titled: “The Most Magical, Excellent, Almost Perfect Products of 2008.” CUDA was included in the list. Enderle wrote: CUDA “…could save the world or my life someday…. What it is allowing is for people who otherwise either couldn’t afford, or couldn’t get access to, supercomputing resources the critical capability to get their work done. The kind of work ranges from medical to environmental research, and the result could be the safety of the human race. I’m big on living, so CUDA is a technology I expect great things from in 2009.” Read the story here: http://www.ecommercetimes.com/story/65564.html