Jump to content

Optimising For Large Orchestral Software Sample sessions


SamRedfern91

Recommended Posts

Hi folks,

I’m trying to understand how to optimise my computer performance when running large orchestral sessions that feature plugins such as Spitfire BBC Symphony Orchestra, Spitfire Solo Strings, Spitfire Hans Zimmer percussion, Spitfire Albion Legacy. I’m experiencing sample drop outs, and I wonder how configuring the settings within Logic Pro X, Kontakt, and the Spitfire Plugins can help.

 

Here are the specs:

 

Computer:

Mac mini (2018)

3.2 GHz Intel Core i7

64 GB 2667 MHz DDR4

Macintosh HD

Intel UHD Graphics 630 1536 MB

 

Audio Interface:

Focusrite Scarlette 8i6 3rd Gen.

 

Here’s some videos of the problem happening:

 

 

Steps taken so far:

- I’ve been on the phone with logic, I screenshared my session, and they can see no performance issues regarding the CPU metering in both 1024 buffer size and 512 buffer size.

 

- Within the BBC Symphony Orchestra Plugin, I’ve messed around with the settings “preload size” and “Stream Buffer Size” but I wasn’t exactly sure what I was doing, so I could use some clarity: What I noticed was that halving the Stream Buffer Size from its default fixed the issue of sample drop outs, but I want to know exactly why that is, and therefore, with that understanding, what I can do to further optimise.

 

- BBC Symphony Orchestra plugin: I also doubled the preload size from default, but I noticed that in my Activity monitor I was getting closer to the 64GB RAM limit, and if I doubled the preload size again, my Macintosh HD Flash Storage would begin to get used. Does increasing the preload buffer size mitigate sample drop out issues, if so, why?

 

- BBC Symphony Orchestra Plugin: To save on RAM, I am purging articulation that I don’t need.

 

- Within Kontakt, on a Mac mini 2011, I used to purge samples and reset markers, and limit the available voices, but for a system like this, I don’t expect a step like that to be necessary.

 

- In Kontakt > Settings > Memory, this is a screenshot of my settings:

1992090520_Screenshot2019-11-15at16_58_33.png.364e075640789d3730d22b9ee12c4cd2.png

 

- In Kontakt, I am purging articulations that I don’t need

 

Thanks in advance!

Link to comment
Share on other sites

Why 240kb prebuffer? I use 6 or 12 when running samples off SSDs. :)

I guess that provided drive is not super fast, i stream samples from NVMe drive. (6x the speed of SATA3 SSD). I would dare streaming samples directly from this drive, bypassing RAM altogether, its so fast.(2.5GB/s)

 

But this sounds like drive isn't keeping up to me.

 

edit: with prebuffer set to 6kb, i can run 150 of berlin strings violins and SSD is at 2% and CPU at 10%

 

have you tried batch resaving the library?

Link to comment
Share on other sites

Make sure you have a dummy track selected any time you aren’t actually needing to play your midi controller for any reason in live mode. If any one track is selected then the instrument in question will all be handled on a single core. If you use a multi timbral setup with say 16 tracks feeding a single multi timbral instrument, for example; then if you select any one of those 16 tracks, all 16 will be running through a single core.

 

Also consider vepro. But I don’t think you should be having such problems with your relatively new Mac mini.

Link to comment
Share on other sites

I didn't see the videos before my last post. I don't see the CPU meter peaking out, but would be helpful if you show us the CPU meter that shows all you're cores. Also notice the HD meter is barely moving, but I don't know if that meter catches HD use by plugins.

 

I also concur with what logicben said. The dropouts are happening at exactly the same place in both videos..which in my mind cries out something is amiss with the midi programming.

Link to comment
Share on other sites

I guess that provided drive is not super fast, i stream samples from NVMe drive. (6x the speed of SATA3 SSD). I would dare streaming samples directly from this drive, bypassing RAM altogether, its so fast.(2.5GB/s)

 

I wanted to ask you about this. What NVMe are you using and in which Mac?

 

Typically when I have researched whether to upgrade to NVMe, I have found overwhelming response from people saying that while the benchmarking of those drives will reveal amazing speeds, the real world difference in booting up, loading apps, even loading sample projects....is about the same......perhaps because of lack of parallelism and other bottlenecks. I am wondering what kind of comparison tests have you done between NVMe and SSD beyond benchmarking, but actual time to load LPX projects, for example? It makes sense that NVMe should theoretically be able to stream better then SSD, though I I don't know how to confirm for sure whether other bottlenecks aren't reducing the difference. Just kind of curious to learn more about what you have found out..not meaning to disagree here. I probably will not spend the bucks to upgrade my 5,1 MacPro to NVMe at this point anyway for now...but I'd like to hear more specifics about it... The firmware in my 5,1 MacPro finally got updated to support booting from NVMe, I think with the Mojave update they did that. but still, I have a thousand dollars worth of SSD's in it now that would be expensive to convert to NVMe.

 

I also tried a hardware raid setup, that basically produces 2-3x the performance in benchmarks compared to a single SSD. But it made absolutely zero difference to project load times, app load times, boot up time, etc.. However, I suspect that a NVMe has faster seek times and bigger queue depth compared to SSD raid. not sure though.

Link to comment
Share on other sites

I use an external Thunderbolt 3 drive, AliExpress enclosure fitted with Samsung EVO 970. I don't boot from it tho, i run my projects and stuff from it.

it's faster than SATA3 SSD via USB3.0 i have. Samesies as the internal one.

 

RAID0 inherently adds some latency so seek time will probably be worse, if i recall correctly

Link to comment
Share on other sites

I think you're right for sure about RAID0, but also I believe NVMe has not only lower seek times then SSD, but its the queue depth, which is superior in some way. But its not clear whether DAW's take advantage of that enough to make a difference.

 

Here i am. 5GB project, 1700 regions, 540 audio files of very various sizes.

load time on SATA3 USB3.0 SSD: 13s

load time on NVMe TB3 SSD: 4s

 

quite a difference. I'll try to generate a long files with a lot of transients or find a drum recording, but i remember difference in transient analysis (and transient editing in the audio editor!) was much much faster on SSD as opposed to HDD.

 

i'm going to be getting usb3.1 gen2 (10gbps) NVMe enclosure soonish with a cheaper drive, will test that as well when i get it.

Link to comment
Share on other sites

That is a huge difference. Unfortunately it’s also hard to tell for sure how much of the reason is attributable to ssd vs NVMe or USB3 vs TB3.

 

Also would be a bit interesting if it was loading a few hundred sample instruments vs many regions, as regions are generally not preloaded into ram they are streamed by the daw. The load times in both cases are quite low. I have projects that can take minutes to load 100 sampler instruments

Link to comment
Share on other sites

I'll try to see if i have something sample heavy, or just make a dummy project and compare it.

There's probably USB overhead there too yeah, i imagine a SATA3 Thunderbolt3 would perform better, although the bandwidth of TB3 is really wasted on a SATA3 drive.

 

But loading Kontakt samples from NVMe (batch re-saved) compared to SATA3 is already noticeably faster. Also BFD3. It just slams them right up, can't even see the progress bar

Link to comment
Share on other sites

In my 5,1 MacPro, it actually only has SATA2 built in sockets for drives. However I have a third party SATA3 board that I can place SSD drives. I have compared SSD's on SATA2 vs PCIe-based SATA3. The benchmarks are a bit faster, but I got zero improvement in project load times. zero.

 

I have not been able to test NVMe, that is what I'm interested in, but of course to really find out if there is a benefit, the other factors need to be made the same for a true comparison.

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Restore formatting

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...