GF 100 Fermi X2 in the works from Nvidia

cfanboi · **Joined:** Sun Dec 13, 2009 4:54 am **Posts:** 9

apoppin wrote:

When i was at GTC, the emphasis was on giving scientists the option to work in their OWN language instead of being forced to learn CUDA.

http://alienbabeltech.com/main/?p=11911

Code: Select all

 We saw Nvidia announce new support for traditional programming languages like C++ and Fortran that ride on top of CUDA to make it much easier for high performance users to write in their own domain languages.  We also saw Nvidia release new GPU/CPU debugging tools and even make Microsoft Visual Studio completely GPU friendly

.

You are comparing CUDA to languages like C++ and Fortran - you missed the point.
CUDA is NOT programing language. It is, let's say it in simple words, an addition to programing
language. Programmer still writes program in a language like C/C++/Fortran. When it comes to
pass execution to the Nvidia card for GPU acceleration, programmer writes some CUDA instructions, instructing the program on what and how needs to be passed to the GPU.

New support for C++ and Fortran means that programing in these languages will be even easier
then it was with older versions of CUDA. And even more, Fermi will have hardware acceleration for some features of those languages!

apoppin · Posted: Tue Dec 15, 2009 3:37 am

You don't know what i am comparing it to

i met Ian Buck at GTC. CUDA is basically instructions for the GPU as x86 is a language set for the CPU.

All of these other domain languages will now be able to ride on top of CUDA; be debugged easily, and allow the researchers to do their work without worrying about CUDA - or knowing anything about it at all. And Fermi is designed to have native support for other languages.

cfanboi · **Joined:** Sun Dec 13, 2009 4:54 am **Posts:** 9

There is fundamental difference in writing CPU-only, sequential processing program,
to writing GPU accelerated parallel program. I cannot imagine how one can do it without knowing anything about the technology below. At last what has to know, is how to tell the program which data needs to be processed on CUDA cores - in parallel. Program cannot just guess it. CUDA can simplify it, programing tools like Visual Studio debugger can simplify it, but not need to know anything about it is a science fiction.
It's the same thing like saying that game developers would not need to know anything about DirectX.

But again, CUDA is GREAT technology, it simplifies extremely complex task to the maximum possible level of simplification.

apoppin · Posted: Tue Dec 15, 2009 11:18 am

cfanboi wrote:

There is fundamental difference in writing CPU-only, sequential processing program,
to writing GPU accelerated parallel program. I cannot imagine how one can do it without knowing anything about the technology below. At last what has to know, is how to tell the program which data needs to be processed on CUDA cores - in parallel. Program cannot just guess it. CUDA can simplify it, programing tools like Visual Studio debugger can simplify it, but not need to know anything about it is a science fiction.
It's the same thing like saying that game developers would not need to know anything about DirectX.

But again, CUDA is GREAT technology, it simplifies extremely complex task to the maximum possible level of simplification.

Well, that is what Nvidia is attempting to do; simplify the researcher's lives by allowing them to program in their own language. They do teach how to make the best advantage of CUDA at their seminars; i attended a couple. And one definitely does not have to be extremely proficient in CUDA (anymore) to debug with Virtual studio; just to know the basics.

Actually, eventually, one will not need to know directX inside and out to make a game.

Yes, i am quite impressed with CUDA.

cfanboi · **Joined:** Sun Dec 13, 2009 4:54 am **Posts:** 9

apoppin wrote:

cfanboi wrote:

And one definitely does not have to be extremely proficient in CUDA (anymore) to debug with Virtual studio; just to know the basics.

I agree with that. The basics of CUDA is what one must know in order to use it.
That's exactly what makes is so great and simple.

What Nvidia created with CUDA and GPU computing is something that
happens every 20,30,40 years, it's a revolution. It's comparable to the
invention of the first microprocessors by Intel and Motorola. Next few years will show, what is possible to achieve with GPU acceleration.
Now it's turn to software developers, Nvidia gave as great machines and great tool to use them, we have to make the software.

apoppin · Posted: Tue Dec 15, 2009 12:40 pm

cfanboi wrote:

apoppin wrote:

cfanboi wrote:

And one definitely does not have to be extremely proficient in CUDA (anymore) to debug with Virtual studio; just to know the basics.

I agree with that. The basics of CUDA is what one must know in order to use it.
That's exactly what makes is so great and simple.

What Nvidia created with CUDA and GPU computing is something that
happens every 20,30,40 years, it's a revolution. It's comparable to the
invention of the first microprocessors by Intel and Motorola. Next few years will show, what is possible to achieve with GPU acceleration.
Now it's turn to software developers, Nvidia gave as great machines and great tool to use them, we have to make the software.

Damn, you have been listening to Jensen, haven't you? :tease:

- imo CUDA for GPU is the equivalent of x86 for CPU.

i don't understand the resistance to CUDA; there is nothing better (period!)
- strange that ATi fans bitch about it being "proprietary" and that they hate it
(but otoh many of them claim to love Brook; which were both developed by Ian Buck)
:rolleyes:

Video from : www.youtube.com

cfanboi · **Joined:** Sun Dec 13, 2009 4:54 am **Posts:** 9

apoppin wrote:

Damn, you have been listening to Jensen, haven't you? :tease:

- imo CUDA for GPU is the equivalent of x86 for CPU.

i don't understand the resistance to CUDA; there is nothing better (period!)
- strange that ATi fans bitch about it being "proprietary" and that they hate it
(but otoh many of them claim to love Brook; which were both developed by Ian Buck)
:rolleyes:

I must had read it somewhere.
But it's not hard to guess it's a revolution, when people are gaining something like
100X performance boosts after few weeks of modifying their legacy applications.

Fans are fans. They are driven by emotions, not by logic. They decide to love or hate something no mater
what the facts are.
But after all, competition is good. ATI is miles behind Nvidia in GPU computing, but if some day in the distant future
they manage to give us even greater technology...wow, looking forward to see it!

apoppin · Posted: Wed Dec 16, 2009 11:43 am

if you qualify, you might like Nvidia's webinar on Fermi Tesla GPU computing

unfortunately it is today at 10 AM and i have to go to my other job

https://www2.gotomeeting.com/register/606754650

Quote:

NVIDIA’s next generation CUDA architecture, code named “Fermi” is the most advanced GPU computing architecture ever built. Join us for a live webinar to learn about the new Tesla GPU Compute solutions built on Fermi and the dramatic performance capabilities they offer customers who are tackling the most difficult, compute-intensive problems. In addition you will learn about our limited time offer, the Mad Science Promotion, whereby you may qualify for a promotional upgrade to a new NVIDIA Fermi-based Tesla product when you purchase a NVIDIA® Tesla™ C1060 GPU Computing Processor or a S1070 1U GPU Computing System today.

Who Should Join: Researchers, Scientists and other Professionals looking to accelerate compute-intensive applications with the GPU

Presenter: Sumit Gupta, Sr Product Manager, NVIDIA GPU Compute Solutions

Wed, Dec 16, 2009 10:00 AM - 11:00 AM PST

if someone does attend, i'd sure appreciate a recording of it

neliz · **Joined:** Fri Dec 25, 2009 8:12 am **Posts:** 6

You know why we have the "x2" Fermi rumours? because they need it to become the "fastest" card

gstanford · Posted: Fri Dec 25, 2009 8:17 am

neliz wrote:

You know why we have the "x2" Fermi rumours? because they need it to become the "fastest" card

Shouldn't you be busy kissing ATi's butt?

GF 100 Fermi X2 in the works from Nvidia

Who is online