Ramdrive sizing for image processing (Startools)

[Imme (Jon)]

Seems I'm on a never ending quest updating things and trying to tweak bits and pieces for better results!

This weeks focus is my processing computer. I'm running DSS and Startools as my main programs and don't see this changing anywhere in the near future.

I've recently upgraded my boot drive to a Crucial ssd (480gb) and have retired the hdd's (approx 950gb) to be storage only. I've seen a significant increase in performance by doing this....boot time has gone from around a minute to 20-25 seconds. Programs are loading faster and Startools appears to be working quicker as well.

I've got some new ram coming tomorrow which will take me from 8gb to 32gb (got a good deal on it).

From what i read a ramdrive is the way to go for processing, Ivo (Startools) is pretty clear it's the best avenue.

My question is size.....I'll have 32gb available so what size should i assign to a Ramdrive?

Also what program do people use/recommend to use to create the drive? (free would be ideal!)

[Imme (Jon)]

Little bit of testing tonite while I still have the 8 gig in.

Assigning 4 gig to a ramdrive and running everything from that made absolutely no difference doing 5 seperate tasks when compared to running it straight from the ssd with the 8 gig assigned to normal ram. (Tests were - open program, open file, auto dev, gradient wipe, noise reduction via tracking).

To be honest it was actually 4% quicker without the ramdrive (approx 0.4 seconds per 10 seconds of ‘tasking’)

Will try same test tomorrow night once the 32 gig is in and I’ll assign a larger portion to the ramdrive.....although I’m thinking maybe the way win 10 handles ram it won’t make a difference.

[Imme (Jon)]

So 32gb is in. Ran the same 5 tests using Startools both in a standard 32gig ram setup and assigning 16gig to a ramdrive with Startools loaded on and running from it.

No appreciable difference at all in any setup when comparing it to the stock standard 8gig i originally had (with ssd).

Watching task manager i did note that during some of the testing the ram use did go up to 12 so it must have at least used some above the standard 8 I had....but as far as processing time goes there wasn't a difference.

....guess on the bright side I can now multitask while processing!

[irwjager (Ivo)]

Hi Jon,


With a RAM drive and StarTools, you should be able to notice some speedups in between modules (e.g. when Tracking is doing the backward/forward propagation).

The speedup is quite dependent on the size of the image you are processing, how complex your workflow has been (this aspect is more noticeable in ST 1.6 which utilises a more optimised storage scheme for Tracking data), as well as single-core CPU speed. The latter is due to on-the-fly compression and decompression of the Tracking data. If the CPU becomes the bottleneck rather than the storage throughput, improvements in throughput will have less of an effect. It's one of those "YMMV" things.


If you run Linux, try mounting your tmp folder using tmpfs - it automatically uses up RAM for the RAM drive until a certain limit, then moves over to using swap.

[Imme (Jon)]

Originally Posted by irwjager

Hi Jon,


With a RAM drive and StarTools, you should be able to notice some speedups in between modules (e.g. when Tracking is doing the backward/forward propagation).

The speedup is quite dependent on the size of the image you are processing, how complex your workflow has been (this aspect is more noticeable in ST 1.6 which utilises a more optimised storage scheme for Tracking data), as well as single-core CPU speed. The latter is due to on-the-fly compression and decompression of the Tracking data. If the CPU becomes the bottleneck rather than the storage throughput, improvements in throughput will have less of an effect. It's one of those "YMMV" things.


If you run Linux, try mounting your tmp folder using tmpfs - it automatically uses up RAM for the RAM drive until a certain limit, then moves over to using swap.

Thanks for the reply Ivo....always good to have a software creator so available and willing to give advice (one of the reasons I purchased a licence)

After testing I've put the speed thing down to my processor....i5, 4 core, 4 thread running @ 3.7.

Still more than happy with the program and lets face it - we take hours upon hours to collect the data.... an extra few minutes processing isn't going to kill me!