HDX card = 3 to 7 x Accel card? Avid needs to explain.

[ggegan]

Previously I had been told by a sales person that the processing power of an HDX card would be the equivalent of a TDM HD5, however, I now see advertisements that state the processing power would be between 3 and 7 times that of an Accel card, which makes 5x an average estimate.

I don't understand why the processing power has such a broad range. What type of configuration would result in 3x as opposed to 7x? If it turns out to be closer to 3x, then buying 2 HDX cards would mean that you would have 512 voices, yet only the processing power of a TDM HD6. From my experience, a TDM HD6 can easily be pushed to the limit for film mixing with only 192 voices, so this would be a woefully underpowered system. If the power is closer to 7x, then that seems to be a better matchup for 256 voices, but 7x is the high end of the range.

What is reasonable to expect and what are the factors that determine where a system would fall within the processing power range? I need a comprehensive official explanation from an Avid rep before I can even consider an investment in HDX.

[John_Toolbox]

This reminds me of the Mix card vs DSP farm, and Accel card vs HD process card comparisons. In both instances, the cards were generally supposed to be equal to roughly 3x their predecessor. On the older card transitions, like DSP farm to Mix, there were plugins that used an entire DSP chip on a DSP farm, that only used 1/3 of a chip on mix cards.... but you needed to use 3 of them for it to make efficient use of the chip. There were also plugins that only worked on a specific type of chip, so in some cases the DSP farm was much closer to the mix card in power depending on what plugins you used.

I believe that the HDX cards have only one type of DSP, so the results should be more consistent, especially since all of the plugins are effectively brand new and not based on old code like many TDM plugins were. If there really is a real world difference in power of an HDX being somewhere between HD3 and HD7, that is pretty severe and definitely needs some more explanation.

Voice count is also something that could be a potential bottleneck, From what I've read on this forum, the voice stealing issue of mixing native and DSP plugins still exists on HDX, however unlike HD you can increase voice count by adding more cards. i think the issue will be based on how many 3rd party plugin companies decide to make a DSP version of their plugin. If everyone only releases AAX native versions of their plugins, voice count could be an issue. But since the difference in work required to make a dsp version of a native plugin is much less, we may see a lot more dsp aax plugins.

[Emcha_audio]

As john stated it depends on your plugins but also your host processor power, as the hdx card is an hybrid system that can benefit from the host processor power.

[ggegan]

As I read it, the 3x to 7x statement is not referring to native processing. To my eye the comparison is clearly between HDX and TDM DSP processing, since that is what is being sold by Avid. It may be possible that the computer system configuration affects the efficiency of the card, I don't know, but whatever the case, there needs to be a clear explanation so we understand what exactly we would be getting.

[Craig F]

I think it is also dependent on sample rate/bit depth of the session
ie: 16/48 is 7x were 192/32 is 3x

hopefully this is cleared up when the HDx cards do ship

[matt-o-]

But what's about the efficiency of the recoding? I remember screenshots of PT10 with AAX Revibe using 54% of a DSP, with a 2nd assigned using 54% of another DSP. Will PT use the remaining 46% for other plug-ins? The TDM Revibe used straight 50%, making it possible to use 2 per chip. I think, there are still a lot of questions and details to verify, before making any decision... but anyway, no need to hurry...

[John_Toolbox]

Originally Posted by Craig F

bit death

Best typo I've seen this year

[John_Toolbox]

I posted a link to this thread on the transition to aax thread, hopefully DaveTremblay might be able to shed some light on this:

http://duc.avid.com/showpost.php?p=1...&postcount=131

[reichman]

This is just speculation, but if you look at the videos from AES, lots of plug-in developers were talking about getting AAX running, and then optimizing their code and having it run twice as efficiently after optimization. The nice English engineer from Avid on the night of the PT10 announcement was talking about wildly variable efficiencies depending on the code. I don't expect drawing an exact comparison will be possible.

[Danny Caccavo]

Originally Posted by reichman

This is just speculation, but if you look at the videos from AES, lots of plug-in developers were talking about getting AAX running, and then optimizing their code and having it run twice as efficiently after optimization. The nice English engineer from Avid on the night of the PT10 announcement was talking about wildly variable efficiencies depending on the code. I don't expect drawing an exact comparison will be possible.

What he said.