Hi NV guys,
My supervisor plans to buy a cuda-gpued notebook. i told her to wait till cuda2.0 comes out, where we can shift between directx and cuda on vista (or integrate kernels and shaders in a same program?).
The problem is, ordering a notebook from university officials takes 2-3 months, so we can’t get newest card if we wait till it’s already on market. So, we hope to order a newest card before it comes to the market.
Can we? when will cuda2.0 card for notebook come out? (i assume cuda2.0 needs new cards, not only new sdk) Most importantly, what’s the rough trend of convergence of CUDA in this year? you know, we wish to buy a “relatively stable” version, say, G80 is soon outdated because of the lack of atomic operations, which is in G86.
Hi NV guys,
CUDA 2.0 will not need new cards.
new cards will take some months I think, so getting a notebook with a 8800M (or dual 8800M) seems to be your best option.
As a BTW, I’m leaning towards Zepto notebooks. They’re only 14" big, but have 8600M GT cards. Not the absolute most powerful notebook cards, but may be good for certain purposes.
Thanks for you guys!
my supervisor asks about FX 3600M, which cuda version does it support? (1.0/1.1) if 1.1, then it’s preferable than 8800 on the cuda version, right? A further question is, will we need a new card for newer cuda/directx10 functionalities before July? if not, we’ll buy; if not, we’ll not.
Thanks for answering these sharp questions:*)
take a look at this site
There you can find what cards support CUDA all G80 are CUDA 1.0 and I think all G80+ architecture is fully CUDA 1.1 and 2.0
Oh, I have Fujitsu using 8400 MG. The problem is that the Fujitsu is very slow in releasing driver, right now, the version of the latest Forceware is only 101.28.
Is it possible for me to use CUDA in my laptop ?
Thanks! so FX3600 is not cuda2.0. We’ll wait for G92-level’s “M” card and notebook:) Thanks!
I think there is some confusion in what you believe to be a card that is usable for a particular CUDA version. No such thing exists. All CUDA capable cards are capable of running with ANY version of CUDA.
There is something called “compute capability” which specifies different graphics chips or graphis chips versions. Cards of higher compute capability may support additional CUDA features. Some are introduced with new versions of CUDA however even cards that do not support those features will work with the particular CUDA version.
So do not base your purchasing decision on CUDA versions but on compute capability (if you really need the new features).
Don’t bother with the Fujitsu drivers, I say. Just download them directly from nvidia.com and see if they won’t install.
thanks! i misused the “version”. i meant “capability” and thought it’s equivalent to “version”.