Bump.

One thing one could do is shoot in jpg + raw and use the jpg on the notebook and process them there just to have a first indication, Use the notebook during the night to send the raws to a bigger machine at home for processing remotely or by someone else... It sort of works ;-)

Track wrote:

gkaefer wrote:

Track wrote:

Bump.

what is a 640M LE?
ok... Googles....
what do you wanna do with your Notebook with this mobile nvidia chip?
doing 6images Fisheye panos being on vacation: yes than 2GB is better than 1GB
doing gigapixels or even 500-1000MP panos: 1GB: verrrrry slow to 2GB: verry slow

PS: utilize it... yes if you enable the GPU Option in apg than the preview (and only the preview) will benefit from it. Rendering can only be speeded with more RAM and "better" IO on harddisc side (so scratch/temp/source/Destination Folders on SSD)

Georg

2GB of video memory on the graphics card is verry slow..?!

yes. because this Memory is not used to render the pano. so you can have 20GB vRAM and Rendering will not be faster.
and - i did not check it with the 640M LE - but onboard nvidia gpus on Notebooks often use shared memeory, no dedicated Memory. So if your note/net-book does have 16GB RAM and 2 GB dedicated GPU Memory (preview window does Profit from GPU memory) available and you use a SSD to hold temp dirs and you OS is 64bit, than you too can do bigger panos.
but doing gigapixels on a Notebook waiting 1-3 days to finish - yes this is slow.

Last edited by gkaefer (2013-02-24 21:23:02)

Track wrote:

I don't understand your logic.

If the vRAM is not used to render the pano, then why would 1GB be slower than 2GB? This is what you said before.

vRam is used for the PREVIEW-Feature. So if you load your Image and detect the pano and directly render it, than it makes no difference havin 1,5,10GB vRAM.
but if you want to optimize your pano after detection in any way (Setting CPs, zoomin into the pano to Review if there are stitching Errors in preview) than more vRAM massively will Speed up your workflow.

Last edited by gkaefer (2013-02-25 01:14:42)

Track wrote:

gkaefer wrote:

Track wrote:

I don't understand your logic.

If the vRAM is not used to render the pano, then why would 1GB be slower than 2GB? This is what you said before.

vRam is used for the PREVIEW-Feature. So if you load your Image and detect the pano and directly render it, than it makes no difference havin 1,5,10GB vRAM.
but if you want to optimize your pano after detection in any way (Setting CPs, zoomin into the pano to Review if there are stitching Errors in preview) than more vRAM massively will Speed up your workflow.

So, when it comes to optimizing before rendering, the vRAM amount counts more than the rendering ability?

As in, a GTX 680 with 2GB of RAM will optimize as fast as a GT 640 LE with 2GB of RAM?

Wouldn't 2GB be wasted on a GT 640 LE?

ad Q1: if you click render directly after you detected your pano than no vram is used. If you wanna check the Images overlapping by deepzooming in the Editor or using the Preview Option for "live" Rendering than GPU usage will Speed up your workflow, so here more Memory for your GPU will Speed things up.
so again: after you detected your pano you can press 2 Buttons:
edit button: GPU is used, more vRAM for the GPU Speeds things up.
Render Panorama->Render Button: here GPU is not used

ad Q2: (I dont talk here about optimization step after render-button is clicked, its about the preview in edit mode etc...) I rather think that a GTX680 is faster than a GT640LE because:
GTX680: GDDR5 Memory with 256bit, GT640LE: DDR3 Ram with 128bit.
CUDA cores are not relevant, apg does not use cuda (nore SLI) - so it does not make sense to use 2 GTX680 on a desktop.

Georg

Last edited by gkaefer (2013-02-25 13:46:27)

if pano fits in 4 GB then no use having more then 4 GB. If you open multiple panos in editor it will use more memory and having more then 4 GB can be usefull.
I have no knowledge of a GPU with 32 GB of ram, do you have a link?