Originally Posted by DaveTremblay
I talk to a lot of plug-in developers including Sonnox, McDSP, and Audioease. I have never received feedback that AAX-DSP is more difficult than TDM development. Quite the contrary. We've already talked about the reasons for limited instance counts of McDSP and Sonnox, so no need to rehash it here. That limitation was our issue, not the plug-in developers. As to Audioease, I imagine the effort to create a TDM version of Altiverb was herculean. I can't speak for them, but I imagine the AAX-DSP version will be quite a bit simpler.
And Marsdy, I know I've talked about a lot of these other points you're making as well. I think I haven't, or we haven't, done the best job of explaining why we would continue to make DSP accelerated hardware. Platform transitions are difficult, there is no doubt about that, but we think it's worth investing in the future. We could no longer make TDM systems due to variety of hardware issues, but we still strongly feel that there is value to embedded DSP resources in a DAW environment. Let me talk a bit about why...
1) Integrated DSP Systems have lower latency. No matter how much power is in the Intel cores, they still can't run effectively below a 64 sample buffer, and even that is pushing the limits of reliability. With converter latency, that probably ends up around 7-10mS of latency. For some people, that is plenty low enough, but certainly not for everyone. No matter how powerful those Intel chips get, you likely won't be able to push this much lower due to the OS. With a dedicated real-time OS, you can get lower, but still not as low as our HD and HDX cards which mix in a single sample buffer. You simply can't get better than that. And the "simplicity" of the TI DSP chips lends itself well to this type of situation. Intel does not.
2) Integrated DSP Systems have higher determinism. Again, the "simplicity" of the TI DSP chips is an advantage here. The chips run our own real-time OS and have near perfect predictability for processing. This means that we know, with absolute certainty, that the engine on the DSPs can operate without hiccups. We also know, in advance, whether a DSP based effect can fit on a processor or not. Compare that with Native, where you add another effect, it throws errors and you either remove it or go into the hardware settings dialogs to tweak buffer size. You absolutely cannot get this with a Native DAW. Even if you installed a hard real-time OS, you wouldn't have the same level of predictability. You're playing the odds. For some customers, this is a fine tradeoff, for many it is not.
No matter how powerful the Intel processors are, they two reasons above make the DSP resources unique. And that uniqueness is what makes them interesting. Sure, they do add power to your system which many customers appreciate, but don't overlook these other advantages.
The way I see it, it's our job to do whatever it takes to give you guys the resources to create great music and films. And in my opinion, we are capable of doing that like no other company in the world. We can give you every advantage of a fully Native DAW (run AAX Native plug-ins), and in addition, we can provide you with additional resources that have strengths where native DAWs have weaknesses. In the end, you can combine all of the positive attributes of all of these different resources to make a highly reliable, low latency, high performance DAW.
The ideal DAW combines high-power/higher-latency resources with lower-latency/higher-reliability resources in a SEAMLESS way that gives you the advantages of both.
Dave
|