Windows 10 support for RAM compression

[braincell]

An August 10, 2015 "Windows Insider Preview" update for Windows 10 added support for RAM compression.


Not out yet for the general public but this should be good news for DAW users. As I understand it, RAM will be compressed when you run out of RAM thus freeing up more RAM. It's not a new technology but new to Windows. Probably Mac OS has this already. I don't know.

[stonberg]

It only compresses data in RAM which would otherwise have been swapped out (and slaps it into the system process's address space); overall this means more RAM use (as the compressed data is in RAM, not written to the swap file), but faster application switching as the applications' data can now be read out of the compressed RAM store, not the swap file.

[Nestor]

If you were to have 32GB of RAM or more, you would benefit from it so?

[braincell]

I have 32 gigs and I use all of it sometimes.

[Nestor]

Wow...., I expected to be fine with 16GB. What do you do to use 32GB, can you describe a project using so much?

[ronnie]

Has anybody tried using a RAM disk? I used to use them in DOS and Windows in the good old days and it made quite a difference. I haven't thought about it until now and there are some programs available. This one is free and has some good specs and reviews.

SoftPerfect Ramdisk

[jksuperstar]

Creating a project to max out your PC is nice and all, but for centuries people have been more creative with their samples to work around such obscene limits and end up with a much thicker coat of hair on their chest as a result. With no wave compression at all (even avoiding the most basic of sampler tricks), that provides 99.5 FUCKING HOURS of stereo wave data. You can't write a song with 99.5 hours of samples?

[ronnie]

Yup. That's what was so great about the early days of MIDI. You needed basically a calculator with I/O operations which was the DOS CPU with an MPU-401, or better an OP-4001 card (ISA). Add a JL Cooper PPS-1 for sync with a multi-track (even cassette) and route your hardware MIDI modules (CZ-101, DX-or TX, M1, DSM-1, EMU, Romplers, etc.,) through a mixer with cheap effects (MicroVerb, Digitech, BBE). Latency? What latency? Grit maybe, but latency no. And then there was Steinberg and the native/CPU/OS cat and mouse games began. And then there was, uh, SCOPE and it still is. The mouse that roars. Course there's more detail to it but it doesn't really change things much.

[kensuguro]

I do think a lot of it is just because sample libs just use up resource like air. Even with disk streaming, if you have crazy layering, multiple mic positions.. it all adds up very quickly. From my perspective, it's very little to do with the ability to compose.. surely, one can write the melodies and the parts with something more basic, but in creating the recording.. man, these libraries are just insane. In certain cases, there are work arounds, like just using 1 mic position, etc.. but all that's up to whether the library or vsti works that way. I do feel like the lavishness of these resource hog instruments artificially inflate system requirements, with very little in the way of developer side effort to make instruments more resource efficient.

[dawman]

I stick to Kontakt since the NCW upgrade a while back.
You can convert to NCW too from what I have seen, although I just only use developers
who use NCW from the get go.
Loads are instant, entire instances can be loaded during a song if you have M.2 SSDs/ASRock boards.
Also Omnisphere in dual live mode use to take a few seconds.
I could only use 1 multi per song. M.2s + a good MIDI Controller (Physis k4) make it so I can load 4 per song.
If you have the cash the U.2 adapter that fits into an M.2 slot is for a 400GB Intel 750 using NVMe.
Windows 8.1 is fine this way and 330,000+ IOps can load so much so fast, there's no limit.
Hollywood Strings and other PLAY instruments load much quicker, but those hogs need to loaded once and left alone.

RAM disk caching is what sampler apps already use by placing buffers/targets for streaming into RAM.
So no benefit there.
Compressed sounds cool though.
When I decide to fuck up my perfectly working rig and get Windows 10 I'll check into it further.

[braincell]

I do think a lot of it is just because sample libs just use up resource like air. Even with disk streaming, if you have crazy layering, multiple mic positions.. it all adds up very quickly. From my perspective, it's very little to do with the ability to compose.. surely, one can write the melodies and the parts with something more basic, but in creating the recording.. man, these libraries are just insane. In certain cases, there are work arounds, like just using 1 mic position, etc.. but all that's up to whether the library or vsti works that way. I do feel like the lavishness of these resource hog instruments artificially inflate system requirements, with very little in the way of developer side effort to make instruments more resource efficient.

The thing is, when you get creative, you always are loading more samples than you actually end up using. You are not going to use every sample in a patch. This is the nice thing about having a lot of RAM. The samples and articulations can be there in case you want to use them. Maybe you want to change the key etc.. Eventually you do use up your RAM and most software does not have the option to automatically dump the samples you are not using in your sequence. Manually removing samples you are not using slows down your creative flow.

For me acoustic samples offer a richer palette than synthesizers do and I like to make my own a lot. I bought a saxophone and cello just to make samples. I'm not a great player but you don't have to be very good to make interesting samples.

[jksuperstar]

I read my post and realized the sarcasm and jest probably didn't come through as clearly as they would have in speech.

When the PC is your one instrument, in fact, many instruments, I get it...it takes a lot of resources to "be the band".

[ronnie]

I upgraded my entry-level ASUS laptop (PDF music reader, MIDI and MP3 player and web browsing/streaming/YouTubing) to Windows 10. Basically let it do it by itself. Took about an hour. Went really smooth except that Kaspersky disappeared! Known problem and after 30 seconds of research and a 5 minute re-install, it's back. To tell the truth, I don't see any real advantage over Win 8.1 after 8.1 is setup for a "classic" GUI. Could be the OS is better for audio but I'm not going to upgrade my working audio machines with Win 8.1 for native stuff and my Win 7 for SCOPE PCI. I don't want any surprises down the road.

[braincell]

Creating a project to max out your PC is nice and all, but for centuries people have been more creative with their samples to work around such obscene limits and end up with a much thicker coat of hair on their chest as a result. With no wave compression at all (even avoiding the most basic of sampler tricks), that provides 99.5 FUCKING HOURS of stereo wave data. You can't write a song with 99.5 hours of samples?

HAHAHA. No comment!

[jksuperstar]

Thanks for laughing, it was more a poke to the creators of software and instruments these days, who see "resource limits" more in terms of development time and profit margins rather than CPU, memory, latency, and quality.

[Nestor]

I do think a lot of it is just because sample libs just use up resource like air. Even with disk streaming, if you have crazy layering, multiple mic positions.. it all adds up very quickly. From my perspective, it's very little to do with the ability to compose.. surely, one can write the melodies and the parts with something more basic, but in creating the recording.. man, these libraries are just insane. In certain cases, there are work arounds, like just using 1 mic position, etc.. but all that's up to whether the library or vsti works that way. I do feel like the lavishness of these resource hog instruments artificially inflate system requirements, with very little in the way of developer side effort to make instruments more resource efficient.

The thing is, when you get creative, you always are loading more samples than you actually end up using. You are not going to use every sample in a patch. This is the nice thing about having a lot of RAM. The samples and articulations can be there in case you want to use them. Maybe you want to change the key etc.. Eventually you do use up your RAM and most software does not have the option to automatically dump the samples you are not using in your sequence. Manually removing samples you are not using slows down your creative flow.

For me acoustic samples offer a richer palette than synthesizers do and I like to make my own a lot. I bought a saxophone and cello just to make samples. I'm not a great player but you don't have to be very good to make interesting samples.

Thanks for explaining it Brain, it makes total sense to me what you say... If I had this extra money, I would get 32GB ram to, as I believe and would like to work in the same way, particularly when I have been working into pieces, just because of lack of power.