HD2 vs 1xHDX, some thinking...

[soundworx]

Maybe I'm seeing this all wrong, but I'd like to hear from others - AVID in particular - because the silence about this issue is deafening and I want to make up my mind:

So HDX is supposed to be 3/5 times more powerful then an Accell card, I have two of the latter. I understand that it's hard to give an exact number because AAX-plugins might be more/less efficient than their RTAS/TDM equivalent. So let's go by some numbers from the AVID site:

An Accell card has 9x200 MHz processors (1800 MHz for argument sake)
A HDX card has 18x350 MHz processors (6300 MHz)

So that would account for 3.5 times the power. But I have an HD2 Accell set, it would then be 3600 vs 6300 MHz, so I would roughly double my processing power with an HDX upgrade, right? Or wrong...

I stumbled across a screenshot of PT10 HDX in action during the presentation ( http://www.gearslutz.com/board/post-...estion-me.html ), where you can see 5 or 7 processors of the 18 being used for the mixengine alone. That puts things in another perspective...

I guess the most important question I have at the moment is this:

Does the mixengine run on a separate FPGA chip (as mentioned by others, but not confirmed by AVID) or is it still eating up power from the "audio-processors" on the HDX card. Thanks for reading

[Pete Gates]

Interesting, the other thing I notice on that picture is that you can't even fit 2xRevibe IIs on one chip (and even D-Verb takes 43%!). More information/confirmation on this would be very useful for purchasing decisions!

Pete

[Frank Kruse]

Originally Posted by soundworx

Maybe I'm seeing this all wrong, but I'd like to hear from others - AVID in particular - because the silence about this issue is deafening and I want to make up my mind:

So HDX is supposed to be 3/5 times more powerful then an Accell card, I have two of the latter. I understand that it's hard to give an exact number because AAX-plugins might be more/less efficient than their RTAS/TDM equivalent. So let's go by some numbers from the AVID site:

An Accell card has 9x200 MHz processors (1800 MHz for argument sake)
A HDX card has 18x350 MHz processors (6300 MHz)

So that would account for 3.5 times the power. But I have an HD2 Accell set, it would then be 3600 vs 6300 MHz, so I would roughly double my processing power with an HDX upgrade, right? Or wrong...

I stumbled across a screenshot of PT10 HDX in action during the presentation ( http://www.gearslutz.com/board/post-...estion-me.html ), where you can see 5 or 7 processors of the 18 being used for the mixengine alone. That puts things in another perspective...

I guess the most important question I have at the moment is this:

Does the mixengine run on a separate FPGA chip (as mentioned by others, but not confirmed by AVID) or is it still eating up power from the "audio-processors" on the HDX card. Thanks for reading

For the reason you mentioned, IMO comparing HD to HDX does´t make sense since HD will be gone sooner than later anyway.
The more interesting question is how HDX1 compares to HDnative. Same voice count and features but how do they compare in terms of DSP?
Would be interesting to compare how the additional cost for HDX translates to dsp-power against HDnative.

HD2 against HDX1 probably looks best for AVID as a benchmark. But what about HDnative against HDX1? Latency left aside.

As far as your question goes: On HDX the mix-engine still eats chips just like under HDaccell. The main difference is more RAM on the card and better sharing, so you can squeeze more plugs into one chip so the power is used more effectively. A complex surround session will still eat lots of chips for the mixer alone (the mixing has to be done somewhere).
frank.

[Emcha_audio]

Originally Posted by Frank Kruse

The more interesting question is how HDX1 compares to HDnative. Same voice count and features but how do they compare in terms of DSP?
Would be interesting to compare how the additional cost for HDX translates to dsp-power against HDnative.

Well since HD|NAtive has no on board processing power that's not even a close comparison. And no it doesn't have the same voice count and track count, as HDX goes up to 768 vs 256 maximum for HD|NATIVE. You have to understand that HDX will -ALWAYS- offer more bang for the bucks compared to the native solution because you are adding computing power over your native computing power. The new AAX plugins format and the HDX card allows for a seamless match and use of the host processing power adding to it the dps power of the card. SO yes even in 6 years, your new shiny computer with 40 cores will still have more power with using the HDX card, than not.

[Frank Kruse]

Originally Posted by Emcha_audio

Well since HD|NAtive has no on board processing power that's not even a close comparison. And no it doesn't have the same voice count and track count, as HDX goes up to 768 vs 256 maximum for HD|NATIVE.

The title of this thread is HD2 against HDX1 not HDX3. HDX1 has 256 voices just like HDnative or even PT10+CPTK. So the comparison is pretty valid.

I know that HDnative doesn´t have "board processing" I´m talking about real world comparison as in: How many instances of given plugin can you open on HDX1 agaist natively on HDnative. Who cares about cards. I´m talking about results.

I don´t care if the cards do the processing or my grandmother using a pocket calculator. I would like to know how the result compare. If my grandmother is faster than HDX1 I will hire her.

That´s why I´d find a HDnative against HDX1 comparison very interesting.
Maybe HDX1 is only 1.5xHDnative on a 12-core? WHo knows.

Frank.

[Emcha_audio]

Originally Posted by Frank Kruse

Maybe HDX1 is only 1.5xHDnative on a 12-core? WHo knows.

Frank.

No, it's your 12 core + the 5x time+ power of the old accel system all in one.

Well the real world result are everything that HD native can do you can addition on top of that what the HDX card can do. That's what you don't seem to get. HDX is able to combine the use of seamless integration of both native processing and dps processing. So what ever your native system can do, HDX will add to that even more.

It's a false comparison in the very basis of things because HD|NAtive doesn't give you any extra power, it's all your computer. You could have the same voices with pt 10 and cptk. HDX adds to that computer power, by combining both processing. So let's take the AES introduction of the HDX example, they had 2 HDX cards, about 100 tracks, only using half the voices, over 700 plugins (the cards were not even maxed out yet), so if you went by the simplest of logic (which by all account would be false, because it's all dependent on which plugins you are using) you'd get 350+ plugins per card, and that's not including what you would be able to instantiate on the native host.

But like I said, it depends on which plugins your using.

It's not like comparing native (computer) to the old accel cards, cause you couldn't mix match the plugins like you can now.

HDX cards is a mix of dsp fpga and host processing.

So back to comparing the HD2 to a HDX as the title of the thread implies. 192 voice counts which you lost a whole lots of voices by mix matching rtas and tdm plugs together. VS 256 voices, where you no longer loses voices by mix matching native and dsp plugs.

[Frank Kruse]

Originally Posted by Emcha_audio

Well the real world result are everything that HD native can do you can addition on top of that what the HDX card can do.

Well, that´s "what they say". In practice however you might know that on a card based system like HD and HDX the mixer eats lots of chips especially when you work in surround or lose voices for every AAX-native to AAX-dsp transition (just like under HDaccell). On a HD2 you can end up with your chips full by 60-80% without a single plugin open. This is the reason why you can open session that won´t run on an HD3 easily on a native PT. So what you say isn´t completely valid. If you spend most of your chips for the mixer HDX1 might not be so much better than HDnative. The question is how they compare in practice not theory.

So the question is if this is the case as well on HDX1 and if yes how many chips you lose with a given modern surround mixing environment.

For stereo it won´t matter....

So again: A comparison of HDX1 against native is very interesting, at least for me. Why would I want a card-system if 60-80% is eaten up by the mixer like on HD?

[soundworx]

Correct. I'm in post and only track 2 to 4 tracks at once. When I bought my HD2 set, it seemed to be the only reliable, setandforget PT system available. TDM gave me just that and I don't regret buying it.

I don't mind the cost of upgrading to HDX if there's significant benefits. But spending 7000 euro on the hardware alone, does make me wonder if I can do feature premixes on à native system (not eating up my voices), with loads of bussing, on a reliable platform. If that 's possible, investing in a HDX system might not be a wise thing to do for me. I'm a one man band, nice little studio, Nothing too complicated.

Originally Posted by Frank Kruse

The title of this thread is HD2 against HDX1 not HDX3. HDX1 has 256 voices just like HDnative or even PT10+CPTK. So the comparison is pretty valid.

I know that HDnative doesn´t have "board processing" I´m talking about real world comparison as in: How many instances of given plugin can you open on HDX1 agaist natively on HDnative. Who cares about cards. I´m talking about results.

I don´t care if the cards do the processing or my grandmother using a pocket calculator. I would like to know how the result compare. If my grandmother is faster than HDX1 I will hire her.

That´s why I´d find a HDnative against HDX1 comparison very interesting.
Maybe HDX1 is only 1.5xHDnative on a 12-core? WHo knows.

Frank.

[Emcha_audio]

Originally Posted by Frank Kruse

Well, that´s "what they say". In practice however you might know that on a card based system like HD and HDX the mixer eats lots of chips especially when you work in surround or lose voices for every AAX-native to AAX-dsp transition (just like under HDaccell)..

You are not losing voices from the AAX native to AAX-dsp transition. It's not like the TDM cards. And if I remember right what I read, the mixer is handled by the fpga chips, although I could be wrong on this one.

Yep I was right, the mix buss is handled by the FPGA chip

http://www.gigasonic.com/avid-pthdx-card.html

[Frank Kruse]

Originally Posted by Emcha_audio

You are not losing voices from the AAX native to AAX-dsp transition. It's not like the TDM cards. And if I remember right what I read, the mixer is handled by the fpga chips.

An avid-rep told me that it´s the same on HDX just like on HD. native and dsp in the same channel = lose voices. Just like on HD. If it´s not true AVID should really educate their local branches better.
Again: The mixer is NOT handled by the FPGA. The mixer needs chips just like on HDX1. You can see that clearly in the system usage window screenshots from HDX that are floating around.

frank.