So I just took delivery of a HD Native card yesterday. In talking to mixers overseas the general vibe was a HDNative card on a modern mac would be equal to a HD3 or greater. I was sceptical. I figured that many mixers (myself humbly included) are loathe to mix RTAS in amongst their TDM plugins so I figured they were comparing TDM only usage vs RTAS power and consequently seeing big disparities as they had never used their RTAS power before in any great way. There are delay comp issues going down that path, RTAS plugs have to be inserted before TDM plugs etc, all that.
So I decided to test my theory and get proper data once and for all. So I made a session on my HD3Accel rig called “MakeitCrash.ptx”
Yes, this test was all done on PT10 on Lion on an 8 core nehalem mac pro, 10gbs of Ram with the playback buffer at 512 and core usage at 10 cores, 99%.
So first I loaded up my tracks/voices to max, 192 of them. Then I switched on delay comp to long, 4k samples. Then I switched on all busses in the IO setup, 256 of them. Then I loaded as many auxes and sends and 5.1 outputs as I could before redlining the mixer. Now the 6 chip mixer was maxed out I proceeded onto inserts. I ran up TDM EQ3s on all the auxes in the TDM domain to avoid voice issues on audio tracks with consequent RTAS plugs (no voices left at this stage). That maxed my chips on all 3 TDM Accel cards, and then I went to RTAS on my audio tracks and started to insert RTAS EQ3s on 96 stereo tracks. Barely dented the CPU, so I started adding AAX native Channel strips.
The final session had most of the 96 audio tracks all running 5 instances of either EQ3 or channel strip, 2 sends and about 30 auxes with EQ3s on them as well and everything was redlining, but the session COULD play and was responsive without DAE errors.
Opening and closing the session – OMG, took forever. Opening it the second time I thought the machine had hung, it took about 8 minutes with no visual indication anything was happening.
So that was the TDM rig – I figured opening this session on native would come up with inactive plugs as it would be just using the equivalent RTAS power and would be missing the mixing/insert power of the 3 accel cards…what happened?
Firstly it opened snappy, within 5 seconds. Woah. Secondly everything was active. I pumped up the delay comp to 16K and I hit play….and got an error. So I went to my playback engine and bumped up the playback buffer from 512 to 1024 and cores from 10 to 12. Playback, no errors, all is good, snappy, wow…
How is this possible?!
Basically the HD Native 64 bit engine must be that much more optimised that it absorbed the extra processing power of a HD3Accel!! I was gobsmacked and grinning from ear to ear. Most of the plugins running were AAX channel strips but there were bucketloads of RTAS EQ3 and even a few RTAS TL Space in there from memory. Cant wait to see what usage I get when they are ported to the new and leaner AAX format!! There’s also the possibility that RTAS and AAX running in a TDM environment was just not optimised much at all, regardless, the results were conclusive!
So in switching from my HD3Accel to a HD Native card what have I lost and gained? I’ve lost a few TDM only plugins like Dolby surround tools, Ampfarm, Echofarm, Heat – that’s about it that I use. What have I gained? 4 times the delay comp, an extra 60 audio voices, amazingly snappy session open and close and re-routing and apparently if I switch up to 16 cores at 99% likely more grunt than I need for your average aussie feature mix! I’ve also gained my now redundant HD rig to use as a stem recorder and thanks to PT10 and Dcommand multi-mode I can access both from my console.
I am very, very, happy right now. Looking forwards to my first big mix and throwing around some Channel strips willy nilly!
The session can be downloaded here if you’re interested in running the test yourself on your own system.