đź–•
- Thread starter HeavyMusic3000
- Start date
Pinky
and The Brain...
I've tinkered with this... and with a system using a SSD I didn't see as much benefit compared to one running the OS/swap on a platter drive.
You'll need as much as you expect the PC to ever swap. I'd say double the RAM you would normally use [this is what I did], and of course keeping in mind when you max out your memory your OS will likely freeze. In the times I've tinkered with this configuration I never ran out of memory but always stayed well within the total memory I had.
Fwiw, I let windows manage my swap file now (Win 7 x64) and don't really find it to be an issue. System has been running this install for a couple years and is largely glitch free (music production is definitely glitch free), so with a SSD any 'loss' real or perceived in having windows use this caching doesn't seem to materialize in any performance degradation.
spitzer
New member
I don't use demanding effects/plugins on a DAW, so I'm not going to comment on that specifically. I do other demanding stuff (video stuff, a little bit of gaming) on this computer though. I have 8 GB of RAM and my paging file is locked at 4 GB. I experience crashes very rarely and for some reason that never happens when (because) I'm running the demanding apps, but instead when really nothing is running except for maybe one web browser window or something.
I'd say the suitable paging file size depends on how much RAM you have, first of all. The right number could be double the amount, exactly the amount you have or even zero (If you have a lot of RAM). Personally I would probably disable it if I had 16 GB or more, but... that's me.
I'd say the suitable paging file size depends on how much RAM you have, first of all. The right number could be double the amount, exactly the amount you have or even zero (If you have a lot of RAM). Personally I would probably disable it if I had 16 GB or more, but... that's me.
Mack Caster
Banned
the first test failed. I was getting clicks render for my vids. once I changed my settings and updated some drivers, it showed everything good. I will look into the page fault when I get back in town.
with an ssd it will be as fas as the controller can be serviced. Let's take an example of one of my machines, it has 16GB of ram. Do I really need windows to swap 1.6gb on the disk? When I turn swap off there is no bad side effects. Why have it on in the first place?
Now on my 8gb machine I noticed that it does degrade a little, but if I take and use a ram drive software solution, set up a 2gb partition in ram, and move the swap, tmp, and temp, and internet cache environments to the ram drive, I get a performance boost, and it keeps up with 16 @ 192Khz simultaneous tracking at a latency of 3.6ms (using a dante card) The Latency with a software ram solution is in the nano-seconds because its working at the ram frequeny and access times. Compared to a system that has a ssd as swap, its faster because of this.
Pinky
and The Brain...
The swaps aren't being done during VST processing, for example. That's why having a bunch of memory when you have VSTs and VSTi's is critical. They load themselves completely into RAM. Swapping (from the OS standpoint) is a legacy product, something certain limited applications do and Windows mostly uses to fast load previously opened applications/information, and dump items from RAM not being frequently accessed. An active project with VSTs wouldn't see any benefit.
It's possible some swapping is done with some DAWs but I've not looked to see what my drive I/O activity is regarding any paging going on with Sonar.
The reason to *not* do it has to do with inviting a small degree/risk of instability. This article [and the links to the test with/without a pagefile] explain why I stopped tinkering with the setting... my results were the same - no benefit, or only placebo (*). Depending on what you're doing perhaps you have found a reason it works or helps you, but that would seem to be more specific to your configuration.
Hopefully we're actually talking about the same thing. I suspect we are.
Do I personally object or would I dissuade someone from disabling their pagefile? Nope. But I don't bother.
[insert wink smiley here]
(*) - I had actually forgotten why until we discussed this and I went to google to verify what I suspected I knew. It's been a few years since I last experimented.
The only time I had a vst use the swap was the new waves plugins. I discussed optimizing solutions and they flat out told me if I was using their plugins @96K or higher its recommended that I get a waves soundgrid server. So I made that 8gb machine, and I loop my vst patches back to my other computer via dante. That a way, I can use my resource hungry plugins on a different machine. Granted, its more of a pain to set up at mix time, but it works. I also use a UAD box, but I wish it did more than just the UAD plugins. Personally, I still think going back to rack gear is the way to go even though I tried to embrace plugins. The recall and having presets is what I like about plugs, but I think real rack gear add things harmonically that the plugs don't. Another weird thing I noticed is that all of the modern music has no harmonics to them. I really notice this when a band plays live (even outdoors) there is so much harmonics going on, but when you pop in the cd and play it before or after the show, the cd very sterile.
Back on topic, the higher the sampling rate, the higher the sustained transfer rate needs to be on the hard drives.
Back on topic, the higher the sampling rate, the higher the sustained transfer rate needs to be on the hard drives.
jimmys69
MOODerator
Your LatencyMon results still don't look so good man. Running on my system for 30 minutes myself. Those pagefaults are disturbing.
I'll post a screenshot here in a bit.
Those page fault latencies are typical for a machine running a combination of software raid/hard drive controller, indexing, and malware/antivirus. Removing all anti-malware and virus scanning will clear it in most cases. But the problem with software hard drive controllers (weather they are in raid or not) is they take cpu time away to cue the file and output it to the cpu/memory on top of the time it transfers the data. You see, when software controllers do that, it takes an extra round trip of polling the irq to get the same data off the hard drive a real controller does in half the time. Even if you have an x99 chipset, if you use a software controller, it still going to be slow because of this. Unfortunately, you really have to look for real hard drive controllers that are integrated on the motherboards, and its getting more of the norm to install a card if you need hard drive performance of any real significance, and now these days it will be SAS based.
jimmys69
MOODerator
Ok, so I didn't have any 'page faults' until I opened the 'snipping tool'. Neither with Cubase running or not. The Highest measured interrupt was also peaked at this point when opening the snipping tool. That would be Windows taking over it seems. It is inconsistent as It never goes over 5 bars without opening the snipping tool. Cubase open/playing or not.
My system is built for running audio and 3 video monitors. Nothing else. Well, internet is connected but browser not opened. I only use it for occasional file transfer/export or software downloads/updates. Never for Windows updates. Here are the results:
My system is built for running audio and 3 video monitors. Nothing else. Well, internet is connected but browser not opened. I only use it for occasional file transfer/export or software downloads/updates. Never for Windows updates. Here are the results:
Attachments
Mack Caster
Banned
Thanks guys, next week I will print out your advice and hit up my local geek squad to see what I can do to implement your advice into actions. I have no desire for my PC to have to ride the short yellow bus! NYKTF
Last edited:
spitzer
New member
Mack, I'll just add/emphasize one thing: The "latency", aka the time it takes to resolve hard page faults is absolutely the most minor problem there. You shouldn't be getting those faults in the first place, or at least they should be rare (like in Jimmy's case). Experiencing them means either that you don't have enough RAM for what you're doing, or there's something wrong with your virtual memory settings.
You could see if they have a bios update. Sometimes they can effect the controller in a positive way.
But keep in mind your trying to multi-track at 192K which needs 73,728.00 Kilo bytes per second sustained transfer rate for every 16 channels. That is why 16 channels@192K takes up almost all the bandwidth of a MADI port. When you get into large track count at these higher sampling rates, you have to run another machine in parallel (with time code synchronized) for everything to run correctly.
Also, I think Geek Squad can't really help you because its going to be way out of their league, and they don't sell that class of hardware. You have to find a industrial computing/server specialist.
But keep in mind your trying to multi-track at 192K which needs 73,728.00 Kilo bytes per second sustained transfer rate for every 16 channels. That is why 16 channels@192K takes up almost all the bandwidth of a MADI port. When you get into large track count at these higher sampling rates, you have to run another machine in parallel (with time code synchronized) for everything to run correctly.
Also, I think Geek Squad can't really help you because its going to be way out of their league, and they don't sell that class of hardware. You have to find a industrial computing/server specialist.
Last edited: