[Eeglablist] I7 vs i9

[Seri, Stefano]

hadar.levi at mail.huji.ac.il

Dear Hadar,
If ICA is your main computational load, i can add to the advice from Arno to consider CUDAICA using a NVIDIA card with as many CUDA cores you can afford, that of sourcing an HP Z-series workstation on ebay. There are quite a few around the world being decommissioned from corporation or medium size academic labs.
We have acquired a Z-620 with dual xeon v2 (total 16 cores) and a NVIDIA K6000 + 64 GB RAM for around USD 800.00. An ICA decomposition on 128 channels sampled at 2048 Hz takes 8minutes for 30' EEG. Not bad.
If you want a More versatile computer but still have ICa as your main load, i would Save on the processor (i7 ) and spend a bit more on a card with at least 2000 CUDa cores. A GTX 1080 ti could be a good choice in terms of value for money

Bw

Sent from my iPad




Prof. Stefano Seri MD, FRCP
Professor of Clinical Neurophysiology and
Clinical Director Aston Brain Centre
Aston University, Birmingham UK

Consultant Clinical Neurophysiologist
Birmingham Women’s and Children’s Hospital NHS Foundation Trust




Contact details:
Hospital: +44(0)121-3339260<tel:+44121-3339260>
NHS Email:  s.seri at nhs.net<mailto:s.seri at nhs.net>


University: +44(0)121-2044103<tel:+44121-2044103>
Email: s.seri at aston.ac.uk<mailto:s.seri at aston.ac.uk>