Laptop for Cuda development

I’m doing two CUDA projects at my university and need a pc I can use for this.
I have a few questions in this regard:

I’m considering buying a ASUS N61JV JX074V laptop running Windows7.
It uses the NVIDIA GeForce GT 325M.

  1. Would this work with the lastest CUDA release?
  2. Since the laptop only has one graphics card, is there anything I should do differently when running CUDA programs on it?
    (I mean since Windows will also be using the card at the same time…)

I really dont know the exact answer. But I have seen some laptops stamped with “CUDA enabled” stickers…
+
It depends on the Windows version u r using… if you r using windows vista or windows 7, you can expect some quirks… but generally, you can run CUDA kernels on actively displaying video cards…

I think that GPU should be fine for your work. But, my 2 cents worth from using CUDA on my laptop. Consider whether you really need the portability. Your laptop’s going to be big (i.e. a horrible thing to carry between classes) and the laptops I’ve seen with high end specs inevitably have heat problems. You should probably get a notebook cooler for it.

I’d advise you to get a desktop instead. You can get a faster card, and generally faster system, but still save yourself a lot of money. If you need to program on the go, then too bad, this laptop will probably be a reasonable choice.

If you are set on using a laptop, it might be worthwhile to find one with two graphics cards. If your display uses the same graphics card as your CUDA apps, the screen will freeze for the duration of the CUDA kernel. This is because the graphics card only does CUDA or only draws graphics, not both at the same time. This also imposes time limits on how long your CUDA kernels can run. If they run for too long, the driver will terminate them, otherwise Windows will reset the driver because it will think that the driver has crashed since it isn’t responding to graphics operations. But, someone with more experience and money than me will have to comment on using laptops with 2 gpus.

I am curious how this works in practice with Optimus-enabled laptops, that can switch between the integrated Intel GPU and an NVIDIA GPU. That could make for a nice solution if the NVIDIA GPU can be used purely for CUDA while the Intel GPU manages the display.

Thanks for all your answers :-)

The laptop has Optimus tech, so I’m hoping this means I can use the NVIDIA card solely for CUDA applications, while
the display is refreshed using an Intel integrated chip. Does anyone have experience with this?
Should any laptop with Optimus tech be able to use the dedicated NVIDIA card solely for CUDA applications?

I don’t think it works like that. Based purely on what I know about other Intel IGP chipsets it is likely that the Intel GPU will be disabled when the NVIDIA GPU is in place. I think the only big breakthrough with Optimus is that the switching now works using something like PCI-e hotplugging, rather than the old BIOS switch that earlier IGP chipsets used.

It sounds like there might be problems only having one graphic card, assuming Optimus does not enable the Intel integrated chip, so how could I deal with this?

It sounds like the system would be useless for any kind of performance testing (running under Windows7). I can live with the screen being frozen for the duration af app execution, but do you know

the approx. timeouts for this?

I would like to have a laptop, so what if I bought a usb graphics card - would this work, so I could do performance testing on the NVIDIA card, while using the usb card with an external monitor?