We’re working on specs for a Nikon Ti2 that will be dedicated to high-content screening. The scope will acquire transmitted and epifluorescence images from microplates, in the most extreme case 1536 images x 8 channels per experiment, which yields ~100 GB for our camera and bit depth.
One suggestion from Nikon is to max out the RAM on the acquisition computer to speed up acquisition (via NIS Elements), with the option to go up to 512 GB. I honestly do not see how this would speed anything up, since it’s well beyond the total file size of an experiment. Are there advantages to such a large amount of RAM that would speed acquisition in some way that is not clear to me?
I should mention this computer is to be used for acquisition only and not for analysis, which we perform outside of Elements with high-performance computing resources.
that sure does sounds like overkill to me. I really don’t think having more RAM just for the sake of having it can lead to noticeable performance gains, if you weren’t otherwise already paging out to to disk. And Elements will also be writing to disk as you go (so it’s not like you need to have enough RAM to contain the entire experiment).
It all of course depends on exactly what the theoretical bottleneck is. But I find it hard to imagine RAM being your limiting resource in the case of high content screening, where time required to move the stage is almost certainly longer than the time to write out the last image to disk. In other words: it’s not about total experiment size (GB), it’s more about the rate of acquisition (GB/s)
I think you’d be very justified in asking for them to substantiate that suggestion with some actual data
Thanks for the response. This agrees with my intuition, and I wasn’t convinced by the rep’s response that RAM would necessarily speed things up. Even on our current system with 16 GB of RAM where we run similar experiments, it’s literally the exposure time and filter wheel switching that eats up time, and I’ve never noticed meaningful latency on the file writing side. If we were doing live capture at insanely high frame rates or crunching data, maybe I could see it.
I just fired off an email to Nikon asking for a detailed explanation of why we need all this RAM.
Adding to Talley’s comment: do pay attention to the acquisition computer’s disk speed. Specifically make sure the disk can comfortably keep up with the average camera data rate or else you will need lots of RAM to keep the images until they can be written.