I had a second hand A500, then an A1200 before I switched over to PCs (when I started studying.. first one was a second hand 486sx33.. running Linux).
but yeah, the Amiga add-on HW was insanely expensive. As a pennyless teenager I couldn't afford anything but a 512kb expansion card, then an Action Replay, and later a 120mb HD for the 1200
The technological advancements, especially between the mid 90ies and late 00s, have really been nothing but astonishing (!).
During that time I always bought last-gen equipment, knowing that the HW would become obsolete after a rather short while anyway.
Anyways.. today I put on my lab coat and did some preliminary benchmarks with my (at the moment highly experimental, and heavily multi-threaded) VST host:
Here are some of my notes (grain of salt and all..):
// (note) all benchmarks w/o dropouts, all init-presets
// (note) CPU load according to task manager
// (note) 16 cores used for audio processing
//
@2600Mhz / 44.1kHz / 1024 samples:
// (note) 16 * 118 instances of Dexed => ~68% CPU load (1888 instances!!)
// (note) 10 instances of Diva -> 58% CPU load (11 cause crackles)
// (note) 16 * 4 instances of Pianoteq6 => ~84% CPU load (64 instances)
// (note) 16 * 17 instances of ZebraHZ => ~78% CPU load (272 instances)
//
@2600Mhz / 44.1kHz / 128 samples:
// (note) 16 * 7 instances of ZebraHZ => ~36% CPU load (112 instances)
//
@2600Mhz / 44.1kHz / 256 samples:
// (note) 16 * 15 instances of ZebraHZ => ~71% CPU load (240 instances)
// (in Reason: 5 * 14 => 2 DSP bars (70 instances))
// (note) 16 * 3 instances of Pianoteq6 => ~74% CPU load (48 instances)
// (in Reason: 5 * 3 => 2 DSP bars => ~34% CPU load (15 instances), BUT: heavy CPU spikes + crackles around every 2 seconds)
// (note) 16 * 32 instances of MicroShift => ~80% CPU load (512 instances)
// (note) 16 * 2 instances of Adaptiverb => ~66% CPU load (32 instances)
// (note) 16 * 130 instances of SparkVerb => ~73% CPU load (2080 instances)
// (note) 16 * 40 instances of Valhalla Plate => ~77% CPU load (640 instances)
// (note) 16 * 10 instances of Soundtoys LittlePlate => ~66% CPU load (160 instances)
// (note) 16 * 3 instances of Outer Space => ~52% CPU load (48 instances)
// (note) 16 * 2 instances of Satin => ~81% CPU load (32 instances)
// (note) 16 * 48 instances of Saturn => ~70% CPU load (768 instances)
// (note) 16 * 20 instances of Oddity2 => ~42% CPU load (320 instances)
All VSTs were processed in 64 sample frame chunks (just like in Reason).
Why 64 frames ? Since these are all VST2s, the host needs a chance to insert note ons/offs, program+ctl changes, even when using relatively large ASIO buffers.
In VST3, it is possible to pass a queue to the plugin, but I seriously dislike the VST3 API (who doesn't..
).
Well, I did these benchmarks just to get a rough idea of what's technically possible on this i9 CPU.
Keep in mind that all plugins ran their init patches, i.e. the real-world performance with real-world patches can differ dramatically (especially with modular-like synths like Zebra).
I also did not really try to find the absolute maximum (some plugins need over a second to initialize - for example Little Plate - but most of these were rather fast (20..30 milliseconds)).
Ok, don't wanna turn this into my personal blog, just thought some of you might find this interesting, even it is only remotely Reason-related.