3gb vs 6gb Videocard memory? will it make a difference?

[rick1]

Hello all,

I want to try processing datasets with 1000-5000 photos.

Will Photoscan processing (alignment & dense cloud) stages using high (or highest) settings be affected by video card GPU memory?

ie will a GTX 1060 3gb be slower vs a GTX 1060 6gb?  3Gb vs 6gb performance & requirements..

And follow up question, if I get a DUAL GTX 1060 (3gb or 6gb) card setup, will they be faster vs a Single GTX 1070 TI card ?

Thanks ahead of time for any input/feedback

[rick1]

Alexey, yours thoughts and expertise on this question?

Sorry to be persistent, but I am trying to make a decision today whether  6gb GPU memory is needed for the higher settings within alignment and dense cloud generation instead of 3gb. And if a Dual gtx1060 setup will outperform a single gtx1070ti card.

Please let me know.

Note I tried a friend’s gtx1070 card, non “ti” version, and during the reconstructing depth stage, I noticed it used 5.3gb of 8gb gpu memory. I dont know if it would have failed with 3gb or significant time difference.

Thanks

[Alexey Pasumansky]

Hello rick1,

Two GTX 1060 would likely outperform single GTX 1070 Ti due to the higher number of CUDA cores.

The benefit of 6 GB card over 3 GB shouldn't be significant. PhotoScan is adaptive to available GPU memory, so it will just split the work into parts, but wouldn't fail.

[rick1]

Hello Alexey,

Thanks for the info, it’s greatly appreciated

I will base my purchase decision based on your input.

Thanks again and have a nice evening:)

[SAV]

Hi rick1,

Dual GPU is the way to go. Two GTX1060 will outperform a single GTX1070 Ti.

This might be an interesting read for you:
https://www.pugetsystems.com/labs/articles/Agisoft-PhotoScan-GPU-Acceleration-710/

All the best.

Regards,
SAV