Welcome Guest! To enable all features please Login or Register.

Notification

Icon
Error

Forum Jump  
Options
Go to last post
Posted on Tue, Jan 28 2020 10:57
by Isaihsan
Joined on Mon, Oct 01 2018, Posts 3
I'm troubleshooting audio distortion while using synchronized dimension strings in Logic and am wondering how disk read speed factors in. I have an internal nvme in my 2013 6 core mac pro with 48 gb of ram and it can only play 5-7 tracks or so without cutting, pops and crackles--the usual. This only happens with synchronized dimension strings...which I've come to love and want to use more of.

I've tested this with 2 different drives, the original that came with the mac and an xgp 8200pro, and both are read by VSL at between 17-20% of how fast the drives test in Blackmagic speed test: the original mac tests at 700mb/s and vsl reads it at 135; the adata tests at over 1400 and vsl reads it at 250. I know external drives are preffered for samples, but I'm limited to usb 3.0 since the tb2 ports have virtually no ssd support, esp. no nvme enclosures.

I've been discussing this with VSL reps to no avail as of yet, so I'd like to hear from the community to see if my drives are being read at appropriate speeds, or if there is something throttling performance.

How fast does VSL read your drives, and at what percentage of your drives' tested speeds?
Have you compared the speed when using them internally vs externally?
Does anyone have any thoughts as to the cause of my limited track count with synchronized dimension strings besides drive read speed? Could Logic be the cause?

Thank you!


*Update: I put some samples on a slow external HDD just for testing and discovered that the external drive is read in VSL at a much higher percentage--43% of the drive's tested max speed (50 mb/s out of a possible 117). So it seems an external drive is indeed faster just by virtue of pathway. But of the new macs (excluding Pro--way too rich for my blood), it seems the new mini has the fastest INTERNAL drive speed of ~2500mb/s, which would yield a 500mb/s read speed in VSL, unless the plugin is more efficient in the newer cpus. External nvme via TB3 is the winner it seems, pricier than usb c but the 2500mb/s read rate would mean 1K+ in VSL, which is the best available that I can see in a simple setup (not sure about slave pc or RAID), and would be awesome. I'd still love to hear about others' actual speeds and the macs/drives they use...wondering if I should upgrade now or hold out. Thanks again!
Posted on Tue, Jan 28 2020 13:28
by Paul
Joined on Sat, Aug 03 2002, Vienna, Posts 12112

Hi Isaihsan, 

SYNCHRON-ized Dimension Strings are great, but convolution on each player takes quite a toll on the performance, additionally to the streaming demands. 

My suggestion is to be careful with Velocity XFade (only use it when you really need it) and switch to the nonconvolution mixer presets we added with a Library Update, for your workflow. You can switch to the convolution settings for the mixdown. 

Also interested to hear what others are reporting. 

Best, 
Paul

Paul Kopf
Product Manager - Vienna Symphonic Library
Posted on Tue, Jan 28 2020 21:58
by Isaihsan
Joined on Mon, Oct 01 2018, Posts 3
Thanks, Paul. The above tests were done without any v. xfade. I did try the light convolution setting, which gave me a few more tracks, but I am trying to get the best performance without sacrifices right now to ascertain my cpu...I think I will upgrade to the new mini--seems the bugs were worked out--and then, if the new machine still can't handle 10 or so tracks of full Syn. Dim. Strings, I'll start minimizing convolution and such. I'm not sure if that is part of the update, might have been there before, so I'll look into it later on to see what's new.

-Isa
Posted on Wed, Jan 29 2020 08:15
by Paul
Joined on Sat, Aug 03 2002, Vienna, Posts 12112

Hi, 

I'd try the nonconvolution mixer presets instead of the light convolution... is that better?

Best,
Paul

Paul Kopf
Product Manager - Vienna Symphonic Library
Posted on Thu, Jan 30 2020 13:30
by Isaihsan
Joined on Mon, Oct 01 2018, Posts 3
Better? Well, if you mean it requires less cpu power, then yes. But it sounds worse, so the point is moot. The other settings, those with convolution, sound great, though. Oh well.

I am shelving the dimension strings. Waaaaay to many issues even when just 2-4 tracks are down, playback slows dramatically, I get sampling rate errors, and of course all the distortion. I think the VI are managable, but the synchronized are another beast.

You guys SERIOUSLY need to have a realistic conversarion anout the synchronized dimension strings and change the sys requirements for them. Intel Core 2 Duo and a 7200rpm HDD?? My machine is much higher spec'ed and can't support the samples properly.
Posted on Thu, Jan 30 2020 13:58
by MS
Joined on Wed, Feb 19 2003, Vienna, Austria, Posts 1748

Originally Posted by: Isaihsan Go to Quoted Post

I've tested this with 2 different drives, the original that came with the mac and an xgp 8200pro, and both are read by VSL at between 17-20% of how fast the drives test in Blackmagic speed test: the original mac tests at 700mb/s and vsl reads it at 135; the adata tests at over 1400 and vsl reads it at 250. I know external drives are preffered for samples, but I'm limited to usb 3.0 since the tb2 ports have virtually no ssd support, esp. no nvme enclosures.

The Synchron Player will test what is relevant for sample streaming - and that is random reads at small block sizes. Sequential performance, as measured by BlackMagic, is irrelevant for determining the proper precache size.

If a disk measures more than 600MB/s on random reads with 64k block size, the disk will also be tested using 4k block random reads. If the disk measures over 100MB/s at this block size, operating system file caching and read-ahead is bypassed as well. This leads to even lower memory/CPU overhead and increased system responsiveness. Currently there are very few drives except the Intel Optane series that handle this kind of speed.

The Intel 905p Optane drive I'm using measures 1084MB/s on 64kb blocks and 179.9MB/s on 4k blocks.

--
Martin Saleteg
Software Developer
Vienna Symphonic Library GmbH
Posted on Fri, Jan 31 2020 07:49
by stephen limbaugh
Joined on Tue, Feb 23 2016, Los Angeles, Posts 254

Slave Windows machine Synchron all of my drives says:

Speed 64k, 624MS/s

Speed 4k, 39.1MB/s

-----------------------------

Primary Mac machine Synchron on all my drives says:

Speed 64k, 1106MB/s

Speed 4k, 97.1MB/s

 

Right now I've got a 83 track-AU3 Multiport session that has some tutti sections where there are some pops and cracks at a 512 buffer in Logic X.  Dissappears when I raise that to 1024 buffer in Logic UNLESS I'm using 7 mic positions in Synchron Strings doubled/blended together with the Synchronized Dimension Strings on a velocity cross-faded trill... in which case the system takes a dookie.

I put the Synchron Pianos on my main machine, and if I'm playing a piece with a flurry of notes and more than 3 mic positions, the system also takes a dookie at all buffer sizes.  There is also a huge difference if I put the Synchron pianos in a local instance of VEPro vs a normal plugin in Logic... VEPro is 10x better on the pops/clicks.

My suspicion is that I don't have my threads and preload sizes optimized.  Though, experimenting with it, when I increase the preload size, I don't seem to get that much improvement in those big orchestra tutti sections, or in piano parts where it sounds like it's a piece for 4 hands.

I also don't really have a clear concept of what a "thread" is... other than it's a CPU thingy that is usually double the number of cores of your processor.   I've randomly moved the streaming/loading thread numbers around and don't hear much of a difference.

2019 MacBook Pro, 8 core i9, 32gb RAM. Heavy Digital Audio PC slave, 6 core Xeon E5-1650, 128gb RAM. Logic 10.4.8. Mojave & Windows 10.
You cannot post new threads in this forum.
You cannot reply to threads in this forum.
You cannot delete your posts in this forum.
You cannot edit your posts in this forum.
You cannot create polls in this forum.
You cannot vote in polls in this forum.