• Welcome to the General Discussion forum for UAD users!

    Please note that this forum is user-run, although we're thrilled to have so much contribution from Drew, Will, and other UA folks!

    Feel free to discuss both UAD and non-UAD related subjects!

    1) Please do not post technical issues here. Please use our UAD Support Forums instead.

    2) Please do not post complaints here. Use the Unrest Forum instead. They have no place in the the General Discussion forum.

    Threads posted in the wrong forum will be moved, so if you don't see your thread here anymore, please look in the correct forum.

    Lastly, please be respectful.

A \"different\" type of 64-bit question ...

billybk1

Shareholder
All internal processsing is done @ 64bit already. It is then \"bit-reduced\" using the UAD-1's hardware UltraDither algorithm to match the resolution of the output signal. It is all done in the UAD-1's DSP.
 

Akis

Sadly, left this world before his time.
Moderator
Maybe because the I/O of the card is 32-bit with the current software. Perhaps they'll change that for the 64-bit release; maybe they could save some DSP cycles, too, by not using the hardware dither?
 

Tony Ostinato

Active Member
yep, uad-1 is 64-bit internally, they maybe shoulda hyped that more as people seem way gullible for it.


and yet we settle for 16 bit 44.1k as our delivery medium. :roll:
 

toader2

Member
Akis said:
Maybe because the I/O of the card is 32-bit with the current software. Perhaps they'll change that for the 64-bit release; maybe they could save some DSP cycles, too, by not using the hardware dither?
Hopefully 64-bit is not too far off considering Windows Vista is coming up really soon now. Also, a chainer would be nice... to save trips to and from the PCI bus - maybe it would help the whole system operate more efficiently.
 

Cabbage

Active Member
AFAIK the VST interface is 32-bit. I don't know if this has changed in VST3, though.

Petter
 

ambrose

Member
The current vst standard (2.4) supports double precision 64 bit floats as i/o between the host and the plugin.

Vst 3.0 will contunie to do this. I think personally that the sonic damage comes when converting data types all the time (float->double), (float-> int24), etc, so it would make a lot of sense for plugins to use 64-bits.

It isn't actually much (if any) slower on modern cpu's.

Plus, check out the amazingly good sounding Reaper (http://www.reaper.fm) if you want to hear just how excellent a 64-bit start-to-end sample pipeline can be. The compressors are worth a serious look too, as is the 64-bit convolving reverb.
 

robi

Member
...so, it isn't necessary to port the actual vsts to 64bit but only the driver for the uad-1 itself? Is that correct?
 

ambrose

Member
robi,

unfortunately, developers need to make use of the vst 2.4 standard fully for 64-bit communication between host and plugin. All Sonalksis stuff now does, for example, and voxengo has some new 64-bit vst 2.4 plugs.

However, many vst do not (probably because many native plugs still process at 32bits). Unless developers update to the newer vst standards, and choose to use the 64bit i/o options, there is no other way to get 64bit i/o out of your plugins. Sorry.

I guess UAD is using 32bit communication because it means fewer transfers between card and host.

On a modern native processeor, processing 64bits is less problematic due to the amazing memory bandwidths available. Most FPUs have been able to do 64bit operations in very few cycles for quite a while now!

Hope that helps
 

Tony Ostinato

Active Member
How is it not a waste to mix in 64 bits when your final delivery medium is 16bit?


(or less).


why is everyone so complacent about CD? DVD-A and SACD have pretty much been dead in the water.


are we fooling ourselves?
 

bulls hit

Active Member
If the final delivery medium was 64 bits, you'd need to mix in 128 or 256 bit numbers.


You're always going to need more bits than the final resolution to allow for the accumulation of errors in each mathematical operation within the audio engine
 

cAPSLOCK

Active Member
Tony Ostinato said:
How is it not a waste to mix in 64 bits when your final delivery medium is 16bit?


(or less).


why is everyone so complacent about CD? DVD-A and SACD have pretty much been dead in the water.


are we fooling ourselves?
Only if we are still mixing in 16 bit.

It's rounding errors, headroom and noisefloor Tony. And it is one of the easiest to hear 'esoteric' improvements you can make.

How easy to hear 64bit to 32/24? That's another question. But any improvement over 16 will amke a difference in most audio, and a HUGE difference in dynamic audio.

Think back to the VHS tape for a moment. If movies we gonna just end up there why not shoot on VHS in the first place? Did a film done in 35mm and then copied to VHS look better than one shot directly to VHS?

Same sort of thing.

cAPS
 

toader2

Member
Tony Ostinato said:
How is it not a waste to mix in 64 bits when your final delivery medium is 16bit?


(or less).


why is everyone so complacent about CD? DVD-A and SACD have pretty much been dead in the water.


are we fooling ourselves?

I don't know if 64-bit is a waste or not. What I CAN say though, is that I have done A/B/C comparisons with different DAWs. We had about 20 audio files - all 24-bit. We set the pan laws identical in Cubase, Samplitude, and Sonar 6 (in 64-bit mode). We set the master fader down the exact same amount in each DAW. We left all of the other faders at 0db.

I can say that hands down, Sonar 6 sounded the best - it was amazing. People say there are no differences in the sound of different DAW systems... I think those people either have not done the tests, or they have terrible ears. Seriously, it is incredibly obvious. I couldn't beleive it. I am a Nuendo user... I absolutely love and prefer the interface of Nuendo - it's much better than Sonar or Samplitude, but I have to admit, that the "sound" of Sonar was a lot better! I am not exaggerating. Do the test yourself if you know someone who has both softwares, or if you own them both yourself...

Is this difference because of 64-bit? I don't know, but there IS a difference... absolutely.
 
toader2 said:
A/B/C comparisons with different DAWs.
With the same converters and support hardware for the input/output?

32 bit float and 64 bit float should not sound so extremely different as you are saying here.

Also, the gain structures should probably be set by measurement, the same input having the same gain through to the output. At least this should be verified.
 

toader2

Member
John Lance said:
toader2 said:
A/B/C comparisons with different DAWs.
With the identical convertors and support hardware for the input/output?
Yep... exact same stuff... seriously. Give it a try some time if you have time to export all of the tracks, and to go through the process. It's kind of a hassle, but kind of enlightening... it's really strange, but they really do sound different. I do not know why... that is beyond me... but I know what I hear, and beleive me, there really is a difference - and I think it is enough to say that it IS significant... It is as significant as the difference between converters, or between preamps for example (which I have also done shoot-outs with as well).
 
Toader2, I'll have to wait for a bit to try this out. I am giving serious thought to adding another DAW software to my lineup. It probably won't happen until I have a dual-quad core processor machine along with the release version of Windows Vista. Vista: that could be a bag of worms there. Maybe XP 64 bit edition. At any rate my current machine is not quite cutting it.

You must have been right here in the forum while I was typing up that question! I sometimes \"commit\" and then think on it and go back and change things. That's my bad. I did a few edits since you looked at my post before I realized there was another post.
 
I am not fully familiar with the internal architecture of the UAD processor, but just because it is running software with calcs at 64 bits fp internally does not mean that it would be capable of more.

To have 64 bits fp in and out normally implies an internal data path that has more bit depth in order for the rounding errors to be minimized on the calcs before dithering down to the io bit depth.

That could be an architectural difficulty for the UAD1. Anyone here know?
 

toader2

Member
John Lance said:
To have 64 bits fp in and out normally implies an internal data path that has more bit depth in order for the rounding errors to be minimized on the calcs before dithering down to the io bit depth.
I don't know much about this kind of thing, but my guess is that if you are working inside a 64-bit application like Sonar 6, you wouldn't need to "dither down"... the UAD-1 could just pass 64-bit audio right into a 64-bit application. Any dithering would just be to protect against errors due to any truncation, but if there is no trunctation, no need for dither.
 
toader2 said:
I don't know much about this kind of thing, but my guess is that if you are working inside a 64-bit application like Sonar 6, you wouldn't need to "dither down"... the UAD-1 could just pass 64-bit audio right into a 64-bit application. Any dithering would just be to protect against errors due to any truncation, but if there is no trunctation, no need for dither.
The extra bit depth is needed to assist in keeping cumulative rounding errors to a minimum through the calculations. It is not so simple as you are suggesting. It would be so nice if it were. This is why you would see hardware with a 24 bit integer io and a 48 bit or more integer internal processing path. That is also at least one of the reasons you would see UAD currently at a 32 bit floating point io and a 64 bit floating point internal processing path.
 
UAD Bundle Month
Top