I have several SSDs laying around, but I also only use 21/64 gigs of my DDR4 PC3600 memory.
Would it be better to use a ram drive for my transcoding directory, or one of the SSDs?
(I already have my Plex Data Folder on another separate SSD all by itself, and Plex is installed on another separate OS SSD as well)
Why not leave it on the current SSD, are you trying to solve an actual problem, or just tweaking your install because you have unused sata ports and SSDs and “why not”
I left mine on the install after playing with other options. I found the RAM more useful as an OS file system cache (I have 64 as well) and found no notible improvement from moving it to another SSD, performance and wear wise.
are you on windows or linux?
on linux you can generally use /dev/shm which is tmpfs ram based drive.
I personally do that on linux, the main reason is not for speed, it is to avoid unnecessary writes to the ssd.
Indeed. I am aware that a number of people have prematurely worn out their SSD’s due to transcoding. I am on Windows, and although I have fast RAID10 Storage for my content, and Mirrored SSD’s for the OS / Plex Data, I also have more than enough RAM to be able to use 16GB for a RAM Drive. It’s fast as hell, and simply removes the possibility of premature wear on my SSD’s.
So if you have the RAM to spare, it’s definitely worth it.
Yea I’m on windows. Not trying to fix an issue, but like axemanuk666 said – tons of read/writes to an SSD is gonna kill it so wondering if just creating a ram drive is the smarter route, and if there would be any performance gain or if it’s kind of a wash. Gonna have to do some reading on how to create the ram drive
unless the transcoding drive is sooo slow that it is slower than the bitrate of what is being transcoded, then you would not get any speed benefit from it.
and if that were the case, then you have bigger problems.
the other thing you must consider, is the transcode buffer.
I would not recommend having less than 64 gig of ram before attempting to use a ram disk, and a minimum of 32 gig for the ram disk itself.
You should monitor the space used by your current transcode folder with your normal server load (ie how many transcodes/users does your server serve), to get an idea of how much temp space your transcodes currently use.
the higher the bitrate of the outgoing transcode, the more space each second/minute of buffer uses.
multiple that times the number of simultaneous users doing transcodes.
I use Starwind RAM Drive to create a 16GB Drive on bootup.
The only problem I find with it (or limitation) is it only gives the ability to create a FAT32 Drive on bootup, with no facility for NTFS, beyond manually Formatting it after creation. However after using this for around 3 years now, I have realized that there is no need for this partition to be NTFS, as the Transcode chunks are very small, and therefore I never run in to the 4GB file size limitation of FAT32.
I keep a close eye on it, however in that 3 years, there have been many occasions where I have had 4 users Transcoding, at once, and have so far never ran out of space.
Yes, of course bigger is always better, but so far in my experience, 16GB has always been enough. However I have never yet gone beyond 4 users Transcoding at the same time.
As far as I’m concerned, there is your benchmark for real world use.
The added benefit is not just reducing the wear on the SSD’s but there is also quite a performance gain to be had.
If you think about it, Transcoding on my setup goes like this…
Read the original File from RAID10 array → Transcode → Write the Transcoded chunks back to RAM Disk → Read the Transcoded chunks from RAM Disk to stream to user.
Now if you multiply this by 3 or 4 Transcodes at once, this method is actually very efficient.
As they say… Every little helps 
Oh I am not denying the efficiency of it, I am simply saying that any modern hard drive, that is not already overloaded, has plenty of iops to cover any plex transcoding tasks.
If a hard drive can not keep up with whatever any servers load of users/transcodes, then there are bigger problems at play.
Otherwise, the main bottlenecks will be either the internet bandwidth between server/client, and the server cpu (or gpu) in older/under powered servers (ie trying to serve too many users, or trying to decode HEVC).
Your OS, regardless of windows/linux, will have disk caching for the transcoder, which will automatically absorb the io of transcoding and other io loads.
There is a few old thread here on this forum which tested transcoding to ram extensively.
for starters
My plex app is installed on a RAID5 of 6 SSD, and I just left it at the default location based on previous testing. I have been keeping an eye on wear (still under 2% used) but will likely keep a closer eye on it 
on the other hand I have some otherwise useless 60-120 GB SSD, and an unused SATA port, maybe I should just create a volume dedicated to plex transcode … hmm
older small SSDs are worth what 20$? disposable at this point, hardly something that deserves protecting IMO.
This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.
