Fermi development on a laptop and the GTX 480M Experiences developing for Fermi on laptops

[anthonyfmorse]

Apple MacBook Pro
NVIDIA GeForce GT 330M with 256 MB (thats Fermi)
[url=“Buy MacBook Pro - Apple (UK)”]http://store.apple.com/uk/browse/home/shop...mco=MTAyNTQzMzk[/url]

only the one graphics card though so you can’t dedicate it to HPC only

[avidday]

Apple MacBook Pro

NVIDIA GeForce GT 330M with 256 MB (thats Fermi)

No it isn’t. The GT330M is a GT200 derived part, only compute 1.2 capability and DX10.1 support.

[avidday]

Apple MacBook Pro

NVIDIA GeForce GT 330M with 256 MB (thats Fermi)

No it isn’t. The GT330M is a GT200 derived part, only compute 1.2 capability and DX10.1 support.

[anthonyfmorse]

No it isn’t. The GT330M is a GT200 derived part, only compute 1.2 capability and DX10.1 support.

Oh yea, sorry I had thought it was fermi but it seems it’s not

[anthonyfmorse]

No it isn’t. The GT330M is a GT200 derived part, only compute 1.2 capability and DX10.1 support.

Oh yea, sorry I had thought it was fermi but it seems it’s not

[headbubba]

Thanks everyone for the replies - all useful.

The definite verdict I’ve got here is that it’s too early for Fermi on a laptop right now. Any laptop that can deal with it would be just too big, heavy and hot and nobody’s piped up to say that they’re doing it.

The breakout box idea looks like a reasonable compromise.

Cheers,

Rob

I am a mobile developer, sharing time between two cities. My company wants to target the Fermi architecture. We’ll mostly use desktops for this, but I would like my next laptop to have native Fermi capability so that I can do development and testing on it. I would like the Fermi card to be dedicated (or dedicatable) to running of HPC stuff. Hence, I need some other mechanism for running the display, like on board video, or another video card. Ideal would be to have two of these cards dedicated to HPC and some other mechanism for the display, but that is probably too much to ask for…

It appears as though the GTX 480M is Nvidia’s first, and currently the only, Fermi card for laptops. Is that correct?

I would like to hear from anyone who has a laptop with the ability to natively run on a Fermi board (a GTX 480M I guess). What is the make of your laptop? What is the size, weight, screen size? What CPU configs are possible? What’s the heat like? Do you use another video subsystem or card to handle the display? If so, how is this done? Is there a docking station for your laptop? Does it have two internal HD bays?

Any sharing of experiences would be much appreciated.

Cheers,

Rob

[headbubba]

Thanks everyone for the replies - all useful.

The definite verdict I’ve got here is that it’s too early for Fermi on a laptop right now. Any laptop that can deal with it would be just too big, heavy and hot and nobody’s piped up to say that they’re doing it.

The breakout box idea looks like a reasonable compromise.

Cheers,

Rob

I am a mobile developer, sharing time between two cities. My company wants to target the Fermi architecture. We’ll mostly use desktops for this, but I would like my next laptop to have native Fermi capability so that I can do development and testing on it. I would like the Fermi card to be dedicated (or dedicatable) to running of HPC stuff. Hence, I need some other mechanism for running the display, like on board video, or another video card. Ideal would be to have two of these cards dedicated to HPC and some other mechanism for the display, but that is probably too much to ask for…

It appears as though the GTX 480M is Nvidia’s first, and currently the only, Fermi card for laptops. Is that correct?

I would like to hear from anyone who has a laptop with the ability to natively run on a Fermi board (a GTX 480M I guess). What is the make of your laptop? What is the size, weight, screen size? What CPU configs are possible? What’s the heat like? Do you use another video subsystem or card to handle the display? If so, how is this done? Is there a docking station for your laptop? Does it have two internal HD bays?

Any sharing of experiences would be much appreciated.

Cheers,

Rob

[laughingrice]

The chips that are supposed to be available soon (I’ve seen laptops advertised already, not sure about availability) are the gtx 480m and quadro 5000m. The few laptops I’ve seen are 17" monsters. Note that this chip takes 100W of power, so I’m not sure how laptopy these laptops are.

I believe that there are solutions by clevo.
This appears on the nvidia website as well
[url=“http://www.v3gamingpc.com/buy/avid_p55m.html”]http://www.v3gamingpc.com/buy/avid_p55m.html[/url]
and there is an announcement about the EliteBook 8740w.
http://www.maximumpc.com/article/news/hp_s…-inch_elitebook

You may also want to look at the solutions by next computing
[url=“http://www.nextcomputing.com/products/mobile-workstation-computer-radius.php”]http://www.nextcomputing.com/products/mobi...uter-radius.php[/url]
not cheap but cool

[laughingrice]

The chips that are supposed to be available soon (I’ve seen laptops advertised already, not sure about availability) are the gtx 480m and quadro 5000m. The few laptops I’ve seen are 17" monsters. Note that this chip takes 100W of power, so I’m not sure how laptopy these laptops are.

I believe that there are solutions by clevo.
This appears on the nvidia website as well
[url=“http://www.v3gamingpc.com/buy/avid_p55m.html”]http://www.v3gamingpc.com/buy/avid_p55m.html[/url]
and there is an announcement about the EliteBook 8740w.
http://www.maximumpc.com/article/news/hp_s…-inch_elitebook

You may also want to look at the solutions by next computing
[url=“http://www.nextcomputing.com/products/mobile-workstation-computer-radius.php”]http://www.nextcomputing.com/products/mobi...uter-radius.php[/url]
not cheap but cool

[neoideo]

you could wait for the alienwares to come with fermi, within these 2 months,

alienware m11x is coming with fermi, i think is better to have a minimal fermi system for developing and fast-testing, then benchmark it on a ral desktop system, thats one way to go

[neoideo]

you could wait for the alienwares to come with fermi, within these 2 months,

alienware m11x is coming with fermi, i think is better to have a minimal fermi system for developing and fast-testing, then benchmark it on a ral desktop system, thats one way to go

[jgoffeney]

The chips that are supposed to be available soon (I’ve seen laptops advertised already, not sure about availability) are the gtx 480m and quadro 5000m. The few laptops I’ve seen are 17" monsters. Note that this chip takes 100W of power, so I’m not sure how laptopy these laptops are.

They are not particularly laptopy. My company got one for demos and it meets a minimum portability requirement and runs quite well but is also big, heavy and hot.

[jgoffeney]

The chips that are supposed to be available soon (I’ve seen laptops advertised already, not sure about availability) are the gtx 480m and quadro 5000m. The few laptops I’ve seen are 17" monsters. Note that this chip takes 100W of power, so I’m not sure how laptopy these laptops are.

They are not particularly laptopy. My company got one for demos and it meets a minimum portability requirement and runs quite well but is also big, heavy and hot.

[den3b]

Hey everybody,

I was pretty much thinking about buying a laptop with a GT420M, so that I can program and test the software on the laptop and then use it on my workstation (which has a GTX470).

Do you think it’s a wise idea??

Thankx
den3b

[den3b]

Hey everybody,

I was pretty much thinking about buying a laptop with a GT420M, so that I can program and test the software on the laptop and then use it on my workstation (which has a GTX470).

Do you think it’s a wise idea??

Thankx
den3b

[SPWorley]

I was pretty much thinking about buying a laptop with a GT420M, so that I can program and test the software on the laptop and then use it on my workstation (which has a GTX470).

In terms of capability, you do want a 4 series mobile chip just because it can execute anything a GTX470 can, including double precision.

So any 4 series GPU will work for compiling and quick testing, but remember the 420M has a lot less horsepower than the GTX470…roughly only 15-20% of the speed.

If you’re a serious developer, you may want to invest in a 460M which is a quite nice mobile chip, about 50% the speed of the GTX470. It usually needs a physically bigger laptop (usually targeted at gamers, but that doesn’t mean it’s not a workstation for us CUDA developers!)

One really really nice laptop with a 460M is only about $1450.

[SPWorley]

I was pretty much thinking about buying a laptop with a GT420M, so that I can program and test the software on the laptop and then use it on my workstation (which has a GTX470).

In terms of capability, you do want a 4 series mobile chip just because it can execute anything a GTX470 can, including double precision.

So any 4 series GPU will work for compiling and quick testing, but remember the 420M has a lot less horsepower than the GTX470…roughly only 15-20% of the speed.

If you’re a serious developer, you may want to invest in a 460M which is a quite nice mobile chip, about 50% the speed of the GTX470. It usually needs a physically bigger laptop (usually targeted at gamers, but that doesn’t mean it’s not a workstation for us CUDA developers!)

One really really nice laptop with a 460M is only about $1450.

[den3b]

In terms of capability, you do want a 4 series mobile chip just because it can execute anything a GTX470 can, including double precision.

So any 4 series GPU will work for compiling and quick testing, but remember the 420M has a lot less horsepower than the GTX470…roughly only 15-20% of the speed.

If you’re a serious developer, you may want to invest in a 460M which is a quite nice mobile chip, about 50% the speed of the GTX470. It usually needs a physically bigger laptop (usually targeted at gamers, but that doesn’t mean it’s not a workstation for us CUDA developers!)

One really really nice laptop with a 460M is only about $1450.

Actually for what I do, I can easily downsize my problem just to see if everything runs fine on the laptop. After that, I just change few variables on my algortihms and I can let it run on my desktop.

Besides, I want to buy the laptop for other uses… and I want it to be smaller.

I think an Acer with GT420/425 for less than $800 would do fine, what do you think?

Thanks a lot!

den3b

[den3b]

In terms of capability, you do want a 4 series mobile chip just because it can execute anything a GTX470 can, including double precision.

So any 4 series GPU will work for compiling and quick testing, but remember the 420M has a lot less horsepower than the GTX470…roughly only 15-20% of the speed.

If you’re a serious developer, you may want to invest in a 460M which is a quite nice mobile chip, about 50% the speed of the GTX470. It usually needs a physically bigger laptop (usually targeted at gamers, but that doesn’t mean it’s not a workstation for us CUDA developers!)

One really really nice laptop with a 460M is only about $1450.

Actually for what I do, I can easily downsize my problem just to see if everything runs fine on the laptop. After that, I just change few variables on my algortihms and I can let it run on my desktop.

Besides, I want to buy the laptop for other uses… and I want it to be smaller.

I think an Acer with GT420/425 for less than $800 would do fine, what do you think?

Thanks a lot!

den3b

[rjl]

Actually for what I do, I can easily downsize my problem just to see if everything runs fine on the laptop. After that, I just change few variables on my algortihms and I can let it run on my desktop.

Besides, I want to buy the laptop for other uses… and I want it to be smaller.

I think an Acer with GT420/425 for less than $800 would do fine, what do you think?

Thanks a lot!

den3b

While not Fermi, I purchased an ASUS 1215N for mobile development about a month ago. It has both the intel and GT218 graphics, so you can dedicate the GPU to CUDA. $450, 6 hour battery life, 11 inch screen. A CUDA netbook with two GPUs. I have even gotten NSight to work on this. I believe compute capability is 1.2. I don’t use double precision (not sure which compute capability provides that), but I have had to do some preprocessor directives because I use __ballot and some other 2.0 functions in my code. It was just easy, so I am happy.