Laptop specs for EEG data processing and pattern matching
Hi all,
If I were wanting to do quite fast and intensive EEG data processing and pattern matching (for example the data that comes out of the OpenBCI boards) what kind of specs would you recommend for a laptop? If there is anyone already doing this kind of activity and has some advise that would be great.
Thanks in advance.
Regards,
Sam
Comments
I can't believe this is the first post about this! Such an important topic. The minimum requirements are so easy to meet. I develop the OpenBCI_GUI application and OpenBCIHub on a macOS 1.6 GHz Intel Core i5, 8 GB 1600 MHz DDR3 from 2015. That being said, running the GUI with 16 channels, FFT, and Accelerometer at 1000Hz is tough for the OpenBCI_GUI, the OpenBCIHub will continue to collect data though. The headplot is a CPU killer too, that's why by default we don't show that.
The most important thing you need is at least a dual core processor. I noticed that when the processing GUI was tasked with collecting data, a lot of problems occurred with transfer from low level driver to the GUI. Therefore I moved the data acquisition to the OpenBCIHub application, which is it's own process, and can dominate a CPU core, truth be told the OpenBCIHub does't beat on the CPU too bad. The OpenBCI GUI can then focus on processing and drawing the data.
@wjcroft Hello.What laptop CPU/RAM/GPU are suitable for designing and creating OpenBCI derived feedback signals as used, for example, in LENS and Cygnet?
Merged your question into this existing thread. See previous comments.
@Scipio, hi. I already answered your question, by merging it with this previous thread. No need to make another thread.
@wjcroft @pushtheworld Is the 'headplot' mentioned above in this thread millivolt measurements of.all electrodes at various points on the head using the 10/20 system of scalp electrode placement?
Yes, the head plot is 10-20 based. And position to channel mapping can be altered.
https://www.google.com/search?as_q=change+head+plot&as_sitesearch=openbci.com
@wjcroft @pushtheworld @s.cheshire @Greenleafs Does this YouTube video
not suggest that a laptop (https://www.newegg.com/rog-metallic-asus-gl502vm-db71-gaming-entertainment/p/N82E16834234299) with a dedicated 6 GB graphics card such as the NVIDIA GeForce GTX 1060 and 16 GB of Memory (RAM) would provide suitable power for processing with the OpenBCI_GUI application?
You don't need a fancy graphics card. Certainly do not need a so called "Gaming PC". Those are generally expensive. Price on that model (now likely dated, from 2016) is listed at $1700. Whew. Save your money. See previous comments in this thread for suggestions on what elements are more relevant.
https://www.amazon.com/GL502VM-G-SYNC-Gaming-Laptop-i7-7700HQ/dp/B01MS14CQH
Almost all mid range laptops have decent 3D graphics support. Not 'gamer' amped up specs, but adequate. And 3D would only likely be useful if you are running some type of VR / AR animation that runs in parallel with your (neurofeedback or BCI) app.
@wjcroft I think I will follow the advice you gave earlier and try some benchmarking with my lowly 2 gb RAM netbook although, as you suggested, it would provide a poor development environment. My netbook may simply be adequate for viewing OpenBCI EEG data processed by the 8-channel Cyton and stored in the OpenBCIHub.
@wjcroft In fact using the laptop in the YouTube video above as a hypothetical benchmark, I was contemplating purchasing a formidable mobile workstation for EEG signal processing:-
https://www.ebay.com/itm/NEW-HP-ZBook-15-G6-Workstation-15-FHD-E-2286M-32GB-512GB-RTX-3000/193430424090
But, following your advice, I think that I would be better off using that kind of money in purchasing an OpenBCI headband kit and Cyton board. This makes more sense.
Note that GUI v5 no longer uses the Hub. Instead all device I/O is through the BrainFlow libraries. This results in a much slimmer and faster GUI experience. Still, Windows 10 on a 2GB netbook, would likely be a miserable experience. Minimum RAM recommendation for Windows 10 decent performance would be at least 4 and better 8GB or more.
https://www.zdnet.com/article/how-much-ram-does-your-windows-10-pc-need-2019-edition/
@wjcroft Yes. My experience with a 2gb netbook and Windows 10 was miserable, recently leading to the Netbook crashing and forcing me inadvertently to downgrade back to its former Windows 7 Starter which is much better suited to 2gb RAM. Unfortunately, Windows 7 Starter would appear to me to be incompatible with the GUI v5, apart from no longer receiving support from Microsoft in terms of system updates.
There are two possible issues with the old machines called 'netbooks'. I'm not sure they are made anymore, they were only popular some years ago.
https://en.wikipedia.org/wiki/Netbook
One concern would be if the machine is 32 bit vs 64 bit. My understanding is that 32 bit GUI is no longer supported.
https://openbci.com/index.php/downloads
There may be older GUI releases that have 32 bit binaries, but of course none of these support BrainFlow. And instead use the older 'Hub' process, which would further slow down on old machines.
The second consideration, is that the Processsing Language (Java underneath), in which the GUI is written, requires up to date OpenGL graphics drivers. Which may be impossible to find, since the netbooks are so old.
@wjcroft @pushtheworld I was reluctant to start a new thread to pose the following enquiry which is that judging from the foregoing, would it not be fair to say that a budget laptop computer with 4gb RAM, a 1.6 to 2GHz processor and two cores would be adequate simply for the purpose of data acquisition on a microSD card (not processing) and streaming/viewing EEG waveforms in real-time (using the latest OpenBCI GUI)?
Most budget laptops, including the one you cite, will be fine for running the GUI and streaming. 4GB is livable as long as you are not running too many apps at once. For example Google Chrome browser is a notorious memory hog; you can watch your memory and cpu usage with such tools as Process Explorer. I'd suggest purchasing through a site such as Amazon that has a 30 day return policy if you are not happy with performance.
https://docs.microsoft.com/en-us/sysinternals/downloads/process-explorer
The processor speed you mention is also viable. The GUI and many other EEG apps are not excessive users of CPU.
Don't bother mentioning AJ Keller (pushtheworld), his email registered to that username is now defunct. His company is now Neurosity:
https://neurosity.co/
Most new laptops have SSD instead of mechanical hard drives. Further improving performance over old machines, and reducing impact of less than optimal RAM or slightly slower processor.