DAVID KIRK’S LIVE CHAT ENCORE – YOUR QUESTIONS ANSWERED

If you had a chance to read or take part in our live chat with David Kirk, then you likely saw all the great questions we received. In fact, there were so many David didn’t have time to answer all of them during the one-hour chat. Not to worry, though, we were able to circle back with David and get his take on a few more of your questions. You can read his responses below.

Stay tuned to the blog for updates on when the next GPU Technology Conference (GTC) live chat will take place. And, just as a reminder, GTC is just around the corner so register now to benefit from the early-bird rates.

Question from lawless:
Do you think processor architectures like this are going to replace traditional CPU's?

David Kirk:
I think what you’ll see are more and more systems designed to use the GPU for parallel processing, while continuing to use a CPU type processor to run the operating system and manage basic serial functionality. This is a view shared by many in the high performance computing space, such as Jack Dongarra at the University of Tennessee who has been quoted as saying that future computing architectures will be hybrid systems with parallel-core GPUs working in tandem with multi-core CPUs.

Question from Benjamin:
Recently a new book called CUDA by Example was launched, could you talk about this book and how it could be a complement of your book?

    David Kirk:
    I see them as complementary. The book I wrote with Dr. Hwu, Programming Massively Parallel Processors: A Hands-on Approach, is designed for use in the classroom and as a reference for computing professionals. The new CUDA by Example book provides a great self-paced introduction to parallel programming, and CUDA API information using source code examples that help programmers who are just getting started learn by implementing examples for each concept covered in the book. So, both books can be effectively used together, or separately.

Question from Allan:
The question is whether NVIDIA sees their primary role as being the caretaker of the "PTX" opcodes and the CUDA platform with the market building on these technologies or will NVIDIA consider developing and promoting additional new GPU-leveraging software?

Similar Stories

  • Plamen Velikov

    What about OpenCL? Is it going to replace technologies like CUDA and ATI FireStreaming? Does NVIDIA implementation of OpenCL use all benefitst of CUDA performance?

  • http://profile.typepad.com/realpixelio Realpixelio

    DK, thanks for the answer. The “AXE” concept and branding are pretty clear.
    Hopefully there are more AXEs in the queue.
    -Allan

  • http://www.nvidia.com/cuda Will Ramey

    Hi Plamen,
    Great question!
    This boils down to personal preference. Developers will use the programming interface most comfortable to them, meaning one that supports a development environment, libraries and OS that they are accustomed to using.
    Since NVIDIA and the CUDA architecture supports language solutions like CUDA C/C++, CUDA Fortran (via PGI), etc. as well as device-level APIs like OpenCL and DirectCompute the choice sits with the developer. How a developer chooses their programming environment is based on a number of criteria: the language they use today, the operating systems they need to support, the codes they need to implement, the libraries they want to use, vendor technical support, legacy code they use, and so on.
    As always, developers can port their applications to use one (or more) of these options as needed since many of the parallel programming concepts are similar. No one forces a developer to use C, C++, C#, Java, etc. on the CPU today – developers get to choose the solution(s) that work best for them and, in the end, developers benefit from having a variety of choices.
    Regarding OpenCL specifically, NVIDIA has been a supporter of OpenCL from the very beginning. We support OpenCL on every CUDA-capable GPU (over 200 million shipped since 2006, before OpenCL was conceived!), we were the first company to publicly release OpenCL support for GPUs in September last year, and OpenCL support has been in every public driver release since then. You can read more about our industry-leading support for OpenCL at http://www.nvidia.com/object/cuda_opencl_new.html

  • Jesse Kaplan

    I’d like to ask a question related to your customer support services. I recently ran into a few problems using my video card in a home theatre set-up and all attempts at getting support from nVidia have been extremely frustrating. The fact that I have to take time off work because tech support is only available from 9-5 pacific time (despite being outsourced to the night-shift in India) makes the level of support available extremely frustrating. The customer support reps clearly have a set of questions and steps they go through regardless of the reported problem and then try to get rid of you as quickly as possible once those don’t help.
    Here are a few snippets from the online chat:
    Subhajit: In this case I suspect that the card is malfunctioning.
    Jesse Kaplan: why is that?
    Subhajit: This appears to be caused by a defective graphics card.
    Subhajit: Please contact the manufacturer of your graphics card to inquire whether it is still covered under their hardware warranty. If it is, please submit an RMA with your graphics card manufacturer to have your graphics card repaired or replaced.
    ..
    Jesse Kaplan: what sort of hardware defect could cause the contrast to go wild depending on the video playback?
    Subhajit: I am not sure, however I suspect it is due to failure of any hardware component of the card.
    Jesse Kaplan: why do you suspect that?
    Subhajit: We have already tried multiple driver version and issue still persists, in this case I doubt that the issue has caused due to display driver.

    After pointing out to him that the presence of multiple versions of the drivers indicates that maybe there can be bugs in drivers he finally agreed that it was possible that this wasn’t a hardware bug but insisted that I download 3dMark software and stress my video card for an hour before I could get any more help.
    Clearly the tech support tech’s goal was to get rid of me as quickly as possible once he runs out of trouble shooting tips. The “stress test” was an obvious attempt to ask me to do something that I couldn’t possibly do while on the phone. I’ve since made a few attempts at calling your support lines (the numbers were very hard to get and were always given with incorrect operating hours) and have been ditched by each tech even after promises from them to call/email back within a specified (and now past) time frame.
    Is there another way to get in-touch with a tech support rep that can do more than go through a script? For details on this issue (including the initial chat transcript) see [Incident: 100812-000068].
    Thank you,
    Jesse