jwatte:

Bitterman: If all we wanted to do was Kameo-style platformers, then XACT would indeed be sufficient

You misunderstand my point slightly. I'm not saying XACT doesn't need any more features, or that there is no game that will ever need more than XACT offers. Quite the opposite as it happens - I'm in absolute agreement that more features and better ease of use for XACT would be a good thing, and the XACT team will probably be able to tell you I was asking for exactly that pretty much constantly from about 2003 to 2007 (when I left Rare). (Hi guys, if you're reading!)

My comment about Kameo (and a number of other games I worked on or know of that used XACT, but Kameo was the one of those that seemed to be best received in terms of audio in reviews) was that even though it didn't have HRTF (shock!) it still managed to get 9s and 10s for its audio in reviews.

So, the claim that keeps being made in this thread and others that HRTF is absolutely 100% essential to avoid your audio sounding like pure crap, and that 5.1 surround is insufficient, sounds rubbish, not 3D and makes people's ears bleed through sheer badness is simply not true. (Or, possibly, it is true but no-one cares, which is the same thing in the end). All the research I've seen about people's listening habits is that fewer than 1% of them play games with headphones. Most use stereo speakers or 5.1, and - in terms of quality - XACT is absolutely fine in those conditions.

I'd also agree with Shawn's comment that "if HRTF is so great and non-experimental, how come no one uses it" - someone mentioned it was invented in 1881 or something unexpected like that - so why doesn't every console, every audio, every game, every movie and TV program and radio show support it already, and been doing so for that last 100 years? Because the benefits it brings are minor compared to the number of people who'd actually listen to it.

Lord Ikon:

If I remember correctly some floating point calculations take more than 7-8 times as long.

Take a look at the first post in that 3rd link that you didn't look at. :)

Those posts were pre-2.0 it seems, are they still applicable?

The XBox CLR wasn't updated for 2.0, and it sounds like it may not even be updated much for 3.0.

It is frustrating building a high-performance (relatively) game engine through XNA and knowing that it likely won't run well on the 360 until (if ever) the floating point performance is improved.

There are always performance improvements to be made, like LOD, less polys here or there, math optimizations. But there is nothing you can do to get around slow floating point calculations, they must be done.

Shawn said in the past that even as slow as the performance of the 360 is for XNA right now, it is still quite a bit faster than the most optimized code is on the original XBox, but unfortunately when making a new game engine you cannot have a target of 8 year old hardware and expect to impress anyone.

Machaira:

Lord Ikon:

If I remember correctly some floating point calculations take more than 7-8 times as long.

Take a look at the first post in that 3rd link that you didn't look at. :)

Those posts were pre-2.0 it seems, are they still applicable?

This thread is getting a little off-topic here, but in some cases I observe an order of magnitude, if not more, difference between Intel and Xbox performance.  I have a suite of benchmarks that I have written to analyze CLR floating-point performance, and I measured no improvement going from 1.0 Refresh to 2.0.  In fact, the elapsed time averages for the calculations came out to be around 1-2% slower across the board, for whatever reason.  Though, as with Jon, I feel as though I'm beating a dead horse bringing this up yet again.  I do understand the business perspective of what features are allocated time, but I am put off by the fact that Zune support got added when there are many areas of XNA Game Studio that need attention.

There are architectural differences that will make the performance of any code vary across platforms.  However, an order of magnitude is a bit much for two relatively similar platforms, in terms of peak floating-point through-put per core.  I do not know any Xbox CLR implementation details, but my guess is that this is a side-effect of using the Compact Framework CLR.  When XNA Game Studio first came out, I read that hardware floating-point support was a new feature for the Compact Framework, instead of the software floating-point implementation used on most of the Compact Framework's targets.  While this greatly improved performance, I believe the floating-point code generator is still young and immature.  The problem may also lie with the instruction scheduler.  Unlike Intel platforms, the PowerPC chip used in the Xbox is an in-order processor and compiler instruction scheduling is a lot more important than it is on Intel platforms.  It's impossible to say for sure without CPU profilers, but I wouldn't be surprised to find avoidable pipeline bubbles all over the place.  This may be another area where the Compact Framework falls short of generating optimized code.

Machaira:

How poor, just for the record?

About a factor 10, when compared to a 2.0 GHz Core Duo Mobile (note: not a Core 2!)
I haven't timed it precisely, but it's bad enough that things that run 60 Hz on my laptop won't work at all on the Xbox (< 10 fps)
(Meanwhile, the Xbox has a lot more fill rate than my laptop ;-)

Shawn Hargreaves:

I would count something as experimental if it has been implemented a long time ago, and the implementation has been widely available, yet it has not recieved widespread uptake. ...  As to why binaural audio has never really taken off, I don't have much clue.

Clearly we all have different definitions for "experimental", so there's no point in debating definitions when we all agree on the facts :).  We're all aware of the history of binaural audio now, so the important question is why hasn't binaural audio caught on if it is -- at least theoretically -- so incredibly cool?

1. Most people don't use headphones for games.  Binaural audio is meant for headphones.  It will work with speakers, but only if you're in just the right spot.

A valid point about most people not using headphones.  Is it true? Maybe/probably.  But why?  Maybe they don't know what binaural audio is; I didn't until I had a need for it, and I've always suspected its existance (we only have 2 ears, right?), but now that I know, I MUST HAVE IT!!!  Also, consider that a lot of people turn off the music and/or sound effects as well and often overlay their own, so what does that tell you about the musical quality in games? 

2. Binaural technology, while theoretically amazing, requires personalized recording or personalized real-time synthesis.  Without a personalized HRTF, people with slightly or wickedly different shaped ears/head may not have the holophonic experience at all.

I would consider this the absolute, number one defining factor as to why binaural audio has not caught on.  For one, binarual SYNTHESIS has only recently (2007) been made available as a real-time API, although I'm not sure what that other guy is talking about it having been in sound cards for decades? and dropped from Vista?  For binaural audio to catch on, sound will have to be recorded with dozens of channels each with their own position, and sythesized in real time (via hardware or software) according to each person's personal HRTF data, which could catch on if it were carried around with them on their IPOD.  Games have an advantage over regular music, becuase their sounds are already in that format (all separate with position data), so games are perfect for exploiting binaural audio and personalized HRTF.  Binaural audio with personalized HRTFs could be the next Wii remote!

3. Surround speakers are nice bling for your house, and people who just don't quite get what binaural is about think they're better than 2-channel headphones.  Subjectively, given the difficulty of delivering personalized binaural audio, surround may be better from a practical perspective, but ultimately, binaraul in-ear audio is the theoretical optimum.  Of course, there will always be idiots who just prefer big-ass multi-speaker setups, just like there will always be people rolling around with huge spinning rims, a purple paint-job and deafening drop-bottom basses in the trunks of their gas-hogging SUVs. LOL.

Sad but true, I have no further comments on this point.  Hope that wasn't too offensive :)

I must be one of the idiots, then.

Indeed, at the risk of turning this into a pointless internet flame war, I'd go so far as to suggest that if having a really good set of 5.1 speakers makes one an idiot, I'd say an insistence on headphones and HRTF and such nonsense makes one pretentious.

Are you the same kind of person who spends £200 on a 2m audio cable because it's made out of gold and allegedly gives a better signal? (I know a few such people). You know there is no measurable difference between the signal on such £200 cables and an £8 cable from Dixons, right? But, no doubt, you'll insist that you can hear the difference and anyone who can't is an aural philistine.

Honestlly, that's fine. I don't mind being a philistine if it saves me £192 quid, and I don't mind being an idiot if it means I get a better audio experience that other people can hear at the same time and I don't have to sit in my living room wearing headphones. So really, we're all happy.

(I'm still laughing at the idea that HRTF can do for game audio what the Wii is doing for game controllers. Seriously? Prove it. Write the code yourself. Take over the world and make millions. But sometimes, there really is a reason that no-one's done it before - sometimes, they have, and failed so spectacularly you never even heard about it.

Andy Patrick:

I must be one of the idiots, then.

Indeed, at the risk of turning this into a pointless internet flame war, I'd go so far as to suggest that if having a really good set of 5.1 speakers makes one an idiot, I'd say an insistence on headphones and HRTF and such nonsense makes one pretentious.

Are you the same kind of person who spends £200 on a 2m audio cable because it's made out of gold and allegedly gives a better signal? (I know a few such people). You know there is no measurable difference between the signal on such £200 cables and an £8 cable from Dixons, right? But, no doubt, you'll insist that you can hear the difference and anyone who can't is an aural philistine.

Honestlly, that's fine. I don't mind being a philistine if it saves me £192 quid, and I don't mind being an idiot if it means I get a better audio experience that other people can hear at the same time and I don't have to sit in my living room wearing headphones. So really, we're all happy.

(I'm still laughing at the idea that HRTF can do for game audio what the Wii is doing for game controllers. Seriously? Prove it. Write the code yourself. Take over the world and make millions. But sometimes, there really is a reason that no-one's done it before - sometimes, they have, and failed so spectacularly you never even heard about it.

I agree whole heartedly with all of that. Glad somebody said it better than I could :).

You know there is no measurable difference between the signal on such £200 cables and an £8 cable from Dixons, right?

The difference is that there is lots of science that shows that snake oil hi-fi gear (as found at Audio Asylum) doesn't make any difference. There's plenty of science that shows that proper HRTFs work very well. That's why they've been used for 10-20 years in critical applications like air traffic control or intelligence dispatch.

Unfortunately, proper HRTFs require a personalized measurement for each listener. Without that, you will get inversion effects, lift effects, phase effects, and all other kinds of obviously bad sound. For some reason, my ears are either shaped differently than the researchers/developers working on "generic" HRTFs, or my hearing is more discerning -- I've never had a really good HRTF experience. The combination "needs headphones" and "works only on some people by default" pretty much kills the potential of HRTF in the market.

And, yes, I think that the idea of true, working 3D audio is great. Just like I think the idea of a time machine is great.