Menu
What's new
By:
Filters

Mac Studio for Composers - Tested!

Very helpful, indeed! What I still wonder though comparing different SSDs (and I appreciate that fact you tested a cheap SATA ssd and nvme and the internal Apple one tied to the processor), is there a benefit in how low you might be able to set the buffer size of Sine player or Kontakt? In other words: could you go for a lower RAM option by storing the libraries used in the template on the internal SSD or doesn’t that make all that much of a difference?
 
blaggins said:
Fantastic video @David Kudell. It was very interesting to see how little of a difference disk speed makes to sample and project load times. Kind of goes to show you how unoptimized Kontakt and Sine are (and let's be honest probably every other sample player as well, not just those two) for super fast disks. I suppose it's good news since we don't have to worry too much about using internal storage for samples. I wonder which player will be the first to re-architect their I/O access model and thread pooling to take advantage of 5000+ MB/s transfer speeds.
Click to expand...
That would be very welcome. Microsoft rearchitected its storage stack with direct storage for the xbox and Windows. The games/Software that take advantage of it show tremendous improvements in load times.
 
David Kudell said:
Hopefully this video will be useful if you‘re considering a Mac, but even if you’re not, the tests comparing running samples on fast internal and Thunderbolt SSDs to standard USB SSDs were quite surprising.

Click to expand...

Great video David. Thank you soooo much for taking the time (must have taken days) to do this.

Interestingly, I have experimented with different sample sets on internal NVMe SSDs and the same on external Thunderbolt 2 NVME's in a Sonnet chassis, and came up with quite similar results - in that the differences were similar, though project loading times were slower due to the older / slower technology - though not that much slower. Note that this is on a dual Intel Xeon machine on a Supermicro X10DAi board which is PCIe Gen3. The main difference of my older machines to the M2 Ultra Strudio is that it can only run half that number of tracks in Nuendo12 before audio drop out ensues. That is probably because the Xeon chips can only boost to about 3.6GHz, despite each CPU having 36 CPU threads.

Anyway, thanks :2thumbs:
 
Hansu Heya said:
Very helpful, indeed! What I still wonder though comparing different SSDs (and I appreciate that fact you tested a cheap SATA ssd and nvme and the internal Apple one tied to the processor), is there a benefit in how low you might be able to set the buffer size of Sine player or Kontakt? In other words: could you go for a lower RAM option by storing the libraries used in the template on the internal SSD or doesn’t that make all that much of a difference?
Click to expand...
You can use the smallest streaming buffer size in Kontakt once you use an SSD to deliver data (at full speed). Of course, the more instances of Kontakt, the more likely it becomes that a single SSD cannot deliver the required data in time. A limiting factor can also be the SSD enclosure because if that limits the throughput, you cannot do much.
I would advise against saving on RAM. Yes, it's expensive, but it also makes your system future-proof. As for Kontakt-related project loading times, a small streaming buffer seems to result in more performance bump than an extra fast SSD. With Sine, thinking about a faster SSD (NVME, for instance) seems more beneficial.
 
  • Like
Reactions: bdr
babylonwaves said:
You can use the smallest streaming buffer size in Kontakt once you use an SSD to deliver data (at full speed). Of course, the more instances of Kontakt, the more likely it becomes that a single SSD cannot deliver the required data in time. A limiting factor can also be the SSD enclosure because if that limits the throughput, you cannot do much.
I would advise against saving on RAM. Yes, it's expensive, but it also makes your system future-proof. As for Kontakt-related project loading times, a small streaming buffer seems to result in more performance bump than an extra fast SSD. With Sine, thinking about a faster SSD (NVME, for instance) seems more beneficial.
Click to expand...
My expectation is that saving on space of the internal SSD is way more feasible than saving on RAM, as well. I just haven’t seen any test on exactly that question and with the release of the m1 there were a lot of people claiming that these systems could use a lot less ram with the internal SSD. I would really like to put some numbers on those claims, as this is the last question I have regarding my configuration for a m1/m2 Mac Studio.
 
Very interesting tests. One thing not mentioned is the use of ASIO guard here unless I missed it, in which case I apologize. I'm guessing you had it shut off for the tests? I wonder how Opus would behave? For now I'm keeping most samples on the internal and my almost never-used ones on the external.
 
babylonwaves said:
Of course, the more instances of Kontakt, the more likely it becomes that a single SSD cannot deliver the required data in time.
Click to expand...
True. I've had problems with that (when using the 6k buffer in Kontakt) also from my internal drive – which reads samples at more than 2300 mb/second.
 
IFM said:
Very interesting tests. One thing not mentioned is the use of ASIO guard here unless I missed it, in which case I apologize. I'm guessing you had it shut off for the tests? I wonder how Opus would behave? For now I'm keeping most samples on the internal and my almost never-used ones on the external.
Click to expand...
I have ASIO guard off for the VEPro plugin (which is Vienna's recommendation) and on for Kontakt.
 
Thanks so much for doing this! I really appreciate the time you must have put into it. It's very helpful. I've been trying to figure this stuff out too. I was curious about a few things, did you adjust the buffer for the spitfire player too when you were testing? If so, what were the settings? And have you ever tested your drives random 4k read speed? I know amorphous disc mark measures that. I used that to measure my current drives, my M2 air internal (1GB) got around 3200 MB/s sequential read/write, and around 1100 MB/s random 4k read, write was around 100 MB/s. My sandisk extreme 4tb got around 800/900 read/write MB/s sequential, and around 100 MB/s random 4k. I have a samsung T7 2TB that got around 700 read and 500 write MB/s sequential, and also around 100 MB/s random 4k. There were some fluctuations when I tested them a few times, but this ballpark felt ok, assuming I didn't mess it up.

I also have some more general questions that I've been pondering for a while related to all this. You and many others here have your large templates and much of the focus I've seen on here regarding drive speed and such I eventually figured out was focused on how long it takes to load those templates, but are there other benefits/drawbacks? Does lowering the buffer stress the cpu more? When I got my M2 air and an external sandisc usb I lowered the buffer on kontakt and spitfire, when I got clicks and pops with only a few tracks. I increased the buffer a bit, no more clicks but was that the processor or the drive?

I would think that if you divided your samples among several external drives you would have more bandwidth to work with, but is it possible to saturate that bandwidth with sample players? And do we know the numbers on the many different players? I read that kontakt maxes at 150 MB/s (if I'm remembering correctly), but if that's bandwidth, how does that correlate to the random read numbers? It seems the other players (Sine, OPUS, Spitfire, VSL VI and Synchron players, Falcon/UVI player, Superior Drummer/EZ Keys/etc, Ableton Simpler/Sampler.....) are probably faster than kontakt, but it would be great to know by how much?

Also, BH photo video is having a pretty good sale on Mac's and drives and stuff today, if anyone is looking.
 
Hey, this is a tangent, but can someone else who has Logic, Kontakt 7, and VE Pro running on an Apple Silicon Mac confirm what I found?

Because it it's confirmed, it would seem to be a more fundamental issue than what drives to buy and how much memory you need.

I just tried it again, and yes, loading the same Rhodes in Kontakt 7 inside Logic is fine at a 64-sample buffer; loading it in VE Pro gives me clicks and pops (although it's fine at a 128-sample buffer).

Latest versions of everything - macOS, Logic, VE Pro, Kontakt 7.

My thought is that it could be a Rosetta issue? (I have Logic in Native mode.)
 
^ I used the Blue Ballad program in the Scarbee Mk 1 that's included with Komplete, but you'd think that any Rhodes with some chorusing and reverb would produce the same result.
 
For all those who don't use VEPro, I wonder how relevant the comparisons between internal/external drives are – since without VEPro (and a not too low Kontakt buffer), it's possible to use quite low buffer settings in DAWs (at least in Logic). Would this mean that without VEPro, we'll get both a more stable (no clicks from VEPro when pressing play) system and lower latency (due to using lower buffer settings)?
 
dohm said:
I also upgraded a month ago from a 2019 iMac to a maxed out Mac Studio m2 ultra. Thanks for the video and great info!

I love this computer! My main reason for also having a fast nvme external drive is that it makes any transferring of files, sample libraries, backups, etc., really fast and that gives me more flexibility in moving things around anytime I want. I set my buffers size in logic at 64 and have not had a single pop or glitch since using the computer. Computer has handled everything with ease. I do everything on the computer, no VEP.
Click to expand...
Just out of interest, What's the largest project you have run and what instruments did you use? As this is the kind of test I'm interested in as I don't have VE Pro so am interested in non VE Pro performance of these Macs. Also if anyone has a maxed out M2 Max MBP I'd be interested to know the numbers as well.
Vik said:
For all those who don't use VEPro, I wonder how relevant the comparisons between internal/external drives are – since without VEPro (and a not too low Kontakt buffer), it's possible to use quite low buffer settings in DAWs (at least in Logic). Would this mean that without VEPro, we'll get both a more stable (no clicks from VEPro when pressing play) system and lower latency (due to using lower buffer settings)?
Click to expand...
This sort of plays into what I'm looking into and hoping will get answered.
 
  • Like
Reactions: Vik
TheNashJMusic said:
This sort of plays into what I'm looking into and hoping will get answered.
Click to expand...
IMO Apple already seems to have a Mac which is both portable and has the power I need: the newest 16" MacBook Pro. And for those who need more – it's only a question of time before 99% of all VI users can buy an even more powerful (M3 based) Mac. Such Macs are most likely available some time in 2024 – or earlier.

Based on many talks with other DAW users on various forums and in RL), I also doubt that many are willing to deal with roundtrip latency in the 7-12 ms range – or higher.

That's why I'm (still) very close to ordering a MBP with 96 gb RAM and 4 tb storage. Maybe I simply should do it and perform those VI related tests – not involving VEPro – that are hard or impossible to find.
 
This has been such a well-timed video for me (thanks David you're a star!...that was a lot of work right there!). I'm just about to push the button on a Mac Studio. I'm wondering .. what are people (particularly media composers) using as back-up protocols? Not sure I can run to NAS drives for sample back ups etc. though I WILL be going down the raided route for project back ups. Looking for some lower cost (but still effcetive) alternatives.
 
This has really made me consider whether I need to house samples on the Mac Studio. Seems external will work pretty nicely - which is way more flexible I think.
 
Minsky said:
This has really made me consider whether I need to house samples on the Mac Studio. Seems external will work pretty nicely - which is way more flexible I think.
Click to expand...
I see no good reason to house samples on the Mac Studio internal drive other than not having the cabling and getting the external drives out of sight. You get a bit of a speed bump, but it seems to really only affect load times, and according to David's experiments, not that much. At the same time, if you put your samples on the system drive, you will be using the system drive quite a lot more, and even if you aren't necessarily writing to it that often, more stuff on it affects performance, at least it does with the older Intel machine architectures.
 
jbuhler said:
I see no good reason to house samples on the Mac Studio internal drive other than not having the cabling and getting the external drives out of sight. You get a bit of a speed bump, but it seems to really only affect load times, and according to David's experiments, not that much. At the same time, if you put your samples on the system drive, you will be using the system drive quite a lot more, and even if you aren't necessarily writing to it that often, more stuff on it affects performance, at least it does with the older Intel machine architectures.
Click to expand...
Yep, makes sense!
 
You must log in or register to reply here.
Top Bottom