Dorico performance benchmarks?

Thanks to PianoLeo’s benchmark results.


I also wonder whether using 4k monitor or 1080p monitor can bring significant differences towards the benchmark results (including “add / remove instruments”, " reordering instruments", “condensing change”, etc.).

If you want to suggest a specific set of tasks (on a specific project?) I’m more than happy to compare them across a 2019 16" i9 Macbook Pro and a 13" 2020 M1 Macbook Pro, with and/or without the same 27" 4K monitor. I’m off to bed right now, though :wink:

2 Likes

Then we can use this project to do the test:

I don’t have any specific tests in mind but am considering buying a new Mac mini with M1 processor and wonder how much the maximum of 16 GB Ram would limit its Dorico performance with orchestral library samples. I suspect your 2019 i9 Macbook pro 16" has even more than 16 GB of Ram and wonder if you notice a significant reduction in handling orchestral scores when you switch to the smaller Macbook Pro with the M1 processor (and 16 GB ram?).

I’m afraid I’m not one for using large orchestral libraries myself, and in any case the 512GB storage limitation would make that impractical probably sooner than the 16GB RAM limitation. Anecdotally what I’ve heard from people in the Mac developer community is that the unified memory architecture of the M1 allows it to page things in and out of memory much more quickly than on other architectures, and that might mean that virtual memory performance is also much better on M1 than on Intel. But whether or not that extends to paging huge samples in and out of memory, I couldn’t say.

1 Like

Do set me some tasks and I’ll try them out on both machines.

1 Like

I gave this a go in both Dorico and Logic Pro, and I had to tweak the settings a bit, but found it could keep up reasonably well on this front. Better than I expected. It definitely did hit some issues when trying to run with, say, all of an orchestra piece loaded up with Spitfire BBC Symphony Orchestra and Spitfire Percussion. That being said, my iMac with 64 GB of RAM definitely performed much better on this front. That is actually one of the few things holding me back for just going all in on this new MacBook Air as my only machine. I will also note that export was totally fine, and much faster than on that iMac – it was just live playback that was slightly challenging.

A friend whose job is basically “processing data as fast as possible” says of the M1 Mini:

"I have just 8GB RAM in my M1 mini, and I regularly run it 7/24 with 2 or 3 GB swapped all the time.

However, even with this amount of swap space being used, I find there’s no paging going on. That is, swap-outs are not occurring. There is some low level swap-in activity going on at times, but the performance seems to be unaware of this.

I have purposely run workloads on the M1 mini to severely over subscribe its RAM and the mini does not bat an eyelid. It simply did not care about me abusing it in this manner. The Activity Monitor ‘Memory’ tab did show a few yellow spikes in it ‘memory pressure’ graph, but that was it.

The M1 mini is incredibly robust, and deals efficiently with my workload. "

So, while the second wave of new Macs may have higher memory specs, I think that assuming similar memory usage/requirements as on Intel Macs is a false comparison.

1 Like

Excuse me. Could you please tell me where these logs are located?
I am gonna do another test (regarding OpenType-PostScript and OpenType-TrueType) and I need the location of the log files.

On Windows it’s in %APPDATA%\Steinberg\Dorico 3.5 and on Mac it’s in /Users/your-username/Library/Application Support/Steinberg/Dorico 3.5 (aka ~/Library/Application Support/Dorico 3.5)

There will usually be 10 numbered logs plus one unnumbered log. The unnumbered one is the most recent one.

1 Like

Here’s some more benchmarks. I believe the .12 update since the earlier posts may have brought some slight performance improvements.

Mac Mini, 2018, 6-core 3GHz:
Enabling Condensing 1000 bars: 6218 ms
Disabling Condensing 1000 bars: 5554 ms

100 bars:
Enabling 894 ms
Disabling: 779 ms

Exporting XML (Uncompressed, all layouts) was 52092 ms.

My venerable 2014 MBP (i7, 2.2Ghz) was, weirdly, faster at Condensing 1000 bars: 5314, 5242 ms.
Though slower for 100 bars: 1039, 954 ms.
Exporting XML was 60930 ms.

The old MPB does have 8 threads, whereas the Mini has only 6. Also, interestingly, I was able to shave over a second off the Condensing time on the Mini by repeating it:

2021-04-25 11:29:43.250 : notifyPostCommandExecute: Edit.ToggleLayoutCondensing?Set=true (6218 ms)
2021-04-25 11:31:07.164 : notifyPostCommandExecute: Edit.ToggleLayoutCondensing?Set=false (5554 ms)
2021-04-25 11:31:12.913 : notifyPostCommandExecute: Edit.ToggleLayoutCondensing?Set=true (5739 ms)
2021-04-25 11:31:51.587 : notifyPostCommandExecute: Edit.ToggleLayoutCondensing?Set=false (4905 ms)
2021-04-25 11:31:59.841 : notifyPostCommandExecute: Edit.ToggleLayoutCondensing?Set=true (5082 ms)

.

And indeed, even adding more music (4 whole notes for the Oboes), the time was faster than the original: 5500, 5011 ms.

Argh: Actually, those numbers are using my own Playback Template. Using the HSSE SE Halion template, I get a much faster 4038 / 3760 ms on the Mini, and 5266 / 4912 ms on the MBP. That makes more sense.

So interestingly, Playback VST configuration is a factor in Condensing!

1 Like

Just revisiting this thread after a new build. All times except the Export are 3 times averaged together, although there honestly wasn’t much variance. PC is Ryzen 5950x, 64 GB. I haven’t really played around with BIOS settings yet other than a few simple things using Gigabyte’s EasyTune app so perhaps there are improvements to be had once I have it all configured.

1000 bars
Enabling condensing: 3857 ms
Disabling condensing: 3649 ms

100 bars
Enabling: 727 ms
Disabling: 670 ms
Exporting XML (Uncompressed, all layouts): 41072 ms

Ah, I had missed this. I just purchased a 5900x, so I guess I can expect basically the same results.

It’s not an extravagance. Most hymnal engraving is thousands of little nudges. Left hand on the Stream Deck, right hand on the trackball. And when I zoom closer than 400%, there’s a noticeable lag… enough so that it messes up my clicking. Going to a Ryzen 2700 definitely helped, but the 5900x should crush it completely.

1 Like

I just installed the 5900x, and it’s ridiculously fast. The lag at high zoom is gone, which was the single reason I upgraded. I’m lovin’ it…

3 Likes

I’d love to see some v3.5 to v4 comparisons in here!