Dorico performance benchmarks?

Wow. Really looking forward to the 15" or 16" MBP sometime! M2 by then, perhaps… :crossed_fingers:t3:

(Not to mention the native version of Dorico!)

2 Likes

I can’t wait for the new iMacs (although the laptops appear so powerful perhaps an iMac wouldn’t be “necessary”). My main limiting factor is the ram; I need a bare minimum of 32, but would prefer 64gb. I weirdly suspect this will be the limiting factor for a while.

My other half’s multitrack cello stuff is all Logic and Final Cut, and of course those apps already have native M1 versions. She’s finding that those programs absolutely fly on an M1 MBP (with 16GB RAM) vs my i9 MBP (with 32GB RAM). It may be that you can get away with less RAM than you think.

Presumably the next iteration will allow more RAM, either way.

I would love if that were indeed an option but alas, I run pipe organ sample sets in hauptwerk and one of my sample sets takes 30gb ram compressed. To fully load it would take almost 50gb.

It doesn’t support buffering and loading only the first few KB of each sample?

EDIT:

Never mind. I went and read their documentation and answered my question. Really surprising!

"Hauptwerk deliberately does not stream samples from hard-disk/SSD in order to achieve the high polyphony necessary for a pipe organ, so it’s important that you have sufficient free memory (RAM) in order to be able to load each organ that you intend to use within Hauptwerk fully into RAM. "

Yes— it’s literally an entire pipe organ sampled note by note (and stop by stop!) with multiple samples per note, sometimes 4-8 seconds long, that you then play in real time for exquisite realism.

If Apple haven’t spotted the opportunity for a Maxi - midway between a Mini and the Pro - I’d be amazed.

1 Like

Look up the ‘mythical xMac’. People have been saying that for years. Clearly, Apple have run the numbers, and it’s not worth their while. The market for ‘a bit faster but not too fast’ is not that large. And extra components are all handled by Thunderbolt peripherals now.

Hi Ben. Thanks, I wasn’t aware of the mythical Xmac… I think I told you a while back I was for buying a Mac Mini - still am, but the RAM limitation on the new M1 is a deal breaker for me. It made me wonder whether they’d deliberately cut the price and limited the RAM to be able to introduce a second model in the same range later on. I said Maxi, but really meant a Mini Pro - really taking it back to where it was before.

ETA: Just read an article saying that the M1 can’t have its memory enhanced.

1 Like

These first M models are all at the ‘low-end’ of Apple’s product line-up. I’m sure the RAM limit was some trade-off in design or production, which they thought was acceptable for these base models. It’s all on the chip, don’t forget. Without doubt, the next wave will have more memory on offer.

But it may also be the case that the ARM chip uses RAM in a different or more efficient way, and old notions of ‘enough’ are not entirely comparable.

Apple still sells the Intel Mini, which is still a very capable machine: I have the middle i5 6-core CPU, 32Gb RAM, and I can’t get close to its limits - Logic, Dorico, graphics apps, etc, etc. I’m certainly not going to get rid of that for a good few years yet. But I’m certainly tempted by a 16" AS laptop.

I think having it on the same dye is the issue. Also, I’ve seen multiple reports that fir “regular workflows” the machine can eek by with less ram and still function well. This obviously doesn’t apply to people like us who use large, ram-heavy libraries.

So I realized I didn’t have XMP enabled when I ran those tests. It’s now enabled. I also disabled HPET on a whim. So here are the old numbers:

5338 ms enabling condensing (1000 bars)
4616 ms disabling condensing (1000 bars)
850 ms enabling condensing (100 bars)
764 ms disabling condensing (100 bars)
75625 ms XML export

And the new:
5024 ms enabling condensing (1000 bars)
4354 ms disabling condensing (1000 bars)
776ms enabling condensing (100 bars)
676 ms disabling condensing (100 bars)
67087 ms XML export

That’s roughly a 10% improvement in some areas. Check your XMP! (And maybe HPET).

We are able to build a native Apple Silicon version of Dorico now, with a bunch of caveats that I won’t get into just at the moment, but I thought you’d find it interesting to see how it performs on the tests described in this thread:

2529 ms enabling condensing (1000 bars)
2270 ms disabling condensing (1000 bars)
47176 ms XML export (1000 bars)
300 ms enabling condensing (100 bars)
376 ms disabling condensing (100 bars)
22553 ms XML export (100 bars)

The machine in question is an M1-powered MacBook Pro 13" with 16GB RAM. It’s all pretty unscientific, and of course this is a development build of the next version of Dorico rather than the current Dorico 3.5 release, but it’s very promising indeed for the performance of Dorico on these new ARM-powered machines.

(@stefan.podell has pretty much the same kind of Intel MacBook Pro as I use as my daily work machine, and that costs upwards of €4000 in the configuration listed there. The M1-based MacBook Pro with 16GB RAM costs around €1900, and once we have a native Apple Silicon version of Dorico promises to provide performance that’s roughly twice as good for half the price. And it does it with battery life more than twice as good, and it is – to all intents and purposes – silent and cool in operation. Pretty amazing!)

18 Likes

I approve of this message :slight_smile:

1 Like

This all sounds so very promising (and exciting) to this mac user. :exploding_head:

2 Likes

Is this the incoming Dorico 3.5 maintenance update? or Dorico 4?

Unless things change very dramatically, there will not be a native Apple Silicon or Universal version of Dorico 3.5. We expect the next major version of Dorico to be a Universal application, however.

3 Likes

Thanks for your response.
This means that I won’t get a benchmark like this if I use Dorico 3.5 through Rosetta 2, right?

(Prior to asking you about this this time, I was thinking of selling my 2018 mac mini and getting a M1 mac mini instead. However, if there’s no significant performance boost, I may hesitate to upgrade my mac now.)

Shiki,
Daniel’s results were using the development build. There will be no native universal binary version of 3.5 released; so to fully benefit from the M1’s performance you’d need to upgrade to D4 whenever that arrives. That said, everyone who has tested the M1 machines indicates that they still outperform even very high-end macs for many, if not most, tasks, so I feel comfortable in stating that you’d likely stand to gain performance either way.

1 Like

I posted M1 benchmarks for Dorico 3.5 further up the thread. Certainly an improvement over my i9 MacBook Pro…

4 Likes