CUDA/GPU on Notebook?

Apologies if this post is a) dumb; or B) not in the right place, but the forum search function is broken at the moment, and I did see a couple of somewhat related posts under this section…

I’m going to be migrating some software to CUDA/GPU. The problem is that I do essentially all my development work on a notebook. (Lenovo T61 running SuSE Linux, if it’s relevant.) I don’t really want to spend the time needed to either set up a desktop especially for CUDA work, or get a new notebook.

So is there a good way to add a GPU card to this notebook? It would only need to be used for CUDA: since I don’t really do any extensive graphics work (and am not a gamer or anything), the current on-board Intel GMA video is perfectly satisfactory.) It does not have to be powerful, since it would only be used for development work, and for power consumption would be better if it was at the low end.

I did see an ExpressCard adapter mentioned in this thread: [url=“The Official NVIDIA Forums | NVIDIA”]The Official NVIDIA Forums | NVIDIA (The HarmonicInversion link about halfway down the thread.) Anyone have thoughts/experience on whether this would be a workable solution?

Thanks,
James

The T61 has a soldered graphics card… it’s not upgradable.
I myself have a T61p which DOES have a Quadro and runs CUDA.

You can run the CUDA emulator in toolkit 2.3, or Ocelot with no graphics card at all.

But the best answer: upgrade to a new laptop. I know you said you don’t want to spend the time to do it, but it’s easier than fancy external hackery.
It’s easy to get a $500 laptop with a quite reasonable CUDA GPU these days.
Even cheap netbooks running Atom chips have CUDA GPUs.

The T61 has a soldered graphics card… it’s not upgradable.
I myself have a T61p which DOES have a Quadro and runs CUDA.

You can run the CUDA emulator in toolkit 2.3, or Ocelot with no graphics card at all.

But the best answer: upgrade to a new laptop. I know you said you don’t want to spend the time to do it, but it’s easier than fancy external hackery.
It’s easy to get a $500 laptop with a quite reasonable CUDA GPU these days.
Even cheap netbooks running Atom chips have CUDA GPUs.

Another alternative would be something like an Asus eee box or Zotac ZBOX, which used the Ion platform and are perfectly capable development machines.

Another alternative would be something like an Asus eee box or Zotac ZBOX, which used the Ion platform and are perfectly capable development machines.

I personally like these: [url=“ASUS Eee PC 1201N-PU17-BK Black 12.1" WXGA NetBook - Newegg.com”]http://www.newegg.com/Product/Product.aspx...N82E16834220659[/url]

Drop an SSD in it, and you get a lightweight CUDA capable system that is pretty quick for anything other than median encoding or high end gaming.

This one only has a single core CPU, but does a little better in terms of battery life: [url=“ASUS Eee PC Seashell 1201PN-PU17-BK Black 12.1" WXGA NetBook - Newegg.com”]http://www.newegg.com/Product/Product.aspx...N82E16834220760[/url]

I personally like these: [url=“ASUS Eee PC 1201N-PU17-BK Black 12.1" WXGA NetBook - Newegg.com”]http://www.newegg.com/Product/Product.aspx...N82E16834220659[/url]

Drop an SSD in it, and you get a lightweight CUDA capable system that is pretty quick for anything other than median encoding or high end gaming.

This one only has a single core CPU, but does a little better in terms of battery life: [url=“ASUS Eee PC Seashell 1201PN-PU17-BK Black 12.1" WXGA NetBook - Newegg.com”]http://www.newegg.com/Product/Product.aspx...N82E16834220760[/url]

What exactly does "NVIDIA ION graphics " mean? how do I know what card/capabilities it has??

thanks

eyal

What exactly does "NVIDIA ION graphics " mean? how do I know what card/capabilities it has??

thanks

eyal

It is fairly low-end (I think that it is similar to a 9400M). The appeal is the ability to be able to run and debug your code from a coffee shop, definitely not to benchmark anything…

It is fairly low-end (I think that it is similar to a 9400M). The appeal is the ability to be able to run and debug your code from a coffee shop, definitely not to benchmark anything…

That one (“first generation”) is an MCP79a north bridge, 16 cores, compute 1.1 + zero copy memory support. There are new Ion graphics for the Intel Pineview series Atom processors which are effectively a discrete compute 1.2 part hanging off a couple of PCI-e lanes.

That one (“first generation”) is an MCP79a north bridge, 16 cores, compute 1.1 + zero copy memory support. There are new Ion graphics for the Intel Pineview series Atom processors which are effectively a discrete compute 1.2 part hanging off a couple of PCI-e lanes.

Yes, I realize that. But as I said, I don’t care at all about the graphics capability of the GPU (absurd as that may seem in writing :-)), I just want to use it as a compute engine. So, since the notebook has all these external device slots (like the expresscard), I was wondering if something would fit.

But is it easy to get a $500 laptop (or ANY laptop - money isn’t that big an issue) with a decent display resolution (1600x1200 or 1920x1200), that runs silently in normal use? (That is, when I’m just writing/debugging code, which is about 95% of the time.) Last time I had to go through 3 or 4 machines & returns to customer service to get rid of fan noise problems, then all the time needed to figure out kernel options &c, all of which is non-productive.

James

Yes, I realize that. But as I said, I don’t care at all about the graphics capability of the GPU (absurd as that may seem in writing :-)), I just want to use it as a compute engine. So, since the notebook has all these external device slots (like the expresscard), I was wondering if something would fit.

But is it easy to get a $500 laptop (or ANY laptop - money isn’t that big an issue) with a decent display resolution (1600x1200 or 1920x1200), that runs silently in normal use? (That is, when I’m just writing/debugging code, which is about 95% of the time.) Last time I had to go through 3 or 4 machines & returns to customer service to get rid of fan noise problems, then all the time needed to figure out kernel options &c, all of which is non-productive.

James