Slashdot Mirror
Creative's X-Fi Audio Chip Reviewed
theraindog writes "The Tech Report has posted an in-depth review of Creative's new X-Fi audio processor. The 51-million transistor chip employs a unique audio ring architecture that pushes an apparent 10,000 MIPS, supports up to 128 hardware-accelerated voices for 3D audio, and can upsample and upmix stereo 16-bit/44.1kHz audio to multichannel 24-bit/96kHz. Creative says that the X-Fi's upsampling and upmixing capabilities can make MP3s sound better than the original CD, and although that claim isn't validated by listening tests, the X-Fi does sound better than other consumer-level audio cards. It also performs better in games, in part because precious few sound cards feature hardware acceleration for 3D audio."
Ah, good old MIPS. I always love this number. The first thing they tell you in computer architecture classes is, "This is the MIPS value. People used to use it, but it's very much a bullshit number."
Missing link?- x-fi/index.x?pg=1
http://techreport.com/reviews/2005q4/soundblaster
...But this guy sounds a lot better, regardless.
OK, actually, it sounds a lot better when it's connected to a Home Theater receiver/amplifier. Whatever. It's a far better way to spend your $100.
-- I wanna decide who lives and who dies - Crow T. Robot, MST3K
Reads:
The Tech Report has posted an in-depth review
Should read:
The Tech Report has posted an in-depth review
(Thanks to synthparadox for the link)
Try playing Half Life 2 with poor sound. It is one of the few FPS I play where gameplay depends on what you hear and in what direction.
I also run a mini theatre with my PC: DVD Player/DVD Recorder(Device), DVD Audio(PC), DivX(PC), Mpeg(PC), Avi(PC), DirectX Applications(PC), Flash(PC)
And a mini studio: Fruity Loops, Vegas, Acid, Reason, Midi In/Midi Out
----- You know you have ego issues when you register a domain in your name.
-AT
Working in a DevOps shop is like playing in a band made up entirely of keytarists.
Exactly. The statement might be revised to "Precious few sound cards feature full hardware acceleration for 3D audio", since some cards like the SB Live! do 3D acceleration, just not on enough 3D channels to cover it all in hardware.
This used to be a big issue in games where 3D (surround) sound was used. These days with faster processors, it isn't such a big issue any more. In fact, many modern games (Half-Life 2's Source engine, and DooM 3) both do all sound processing exclusively in software (though Creative later blackmailed Id into adding hardware support for DooM 3). It was decided for this current generation of games that CPUs were fast enough to do the sound processing in hardware, and that it was the best way to provide a consistent presentation no matter what sound card is used. Both games do all their 3D mixing, and their post-processing (reverb for example) entirely in software.
Does doing it in hardware still provide a CPU benefit? Yes. Is it that important anymore? No, unless you're going nuts for framerates.
I seem to recall a benchmark done years ago on an Athlon 1.4 that showed 40% CPU usage exclusively for 3D sound on the SB-Live, and something like 5% on the Audigy. Now, with current high-end CPUs at something like 3x faster than that, spending 15% of a game's CPU budget on sound is fine. Multithreading in games (to support multicore processors) will further reduce this, since you'll be able to offload sound processing to the second core.
They bought E-MU who was a synth manufacturer, and started releasing some very high quality stuff under the E-MU brand. I point to the 1820M which has unbelivable specs which have all been verified by independant tests. This sound card is a low end *MASTERING* grade unit for about 550$.
Religion is a gateway psychosis. -- Dave Foley
HL2 does ALL of it's sound processing (3D, effects, etc) entirely in software. The only benefit a higher priced Creative card can provide is a better SNR, which isn't the end-all-be-all.
Using a generic onboard card with surround support will not be much of a different experience than using an X-Fi. You'll notice cleaner sound due to the better SNR, but that's it.
Valve did this because CPUs have advanced to the point where sound processing can be done in software without too much of a processing time investment, and it ensures that everybody gets to hear the same soundscape/quality no matter what soundcard they are using. No longer does environmental audio depend on what version of EAX your soundcard supports.
Umm... cubic curve interpolation (or even more simply, a capacitor) is probably not new since you studied DSP.
Indeed, this sounds like market speak for "We add some nice sounding junk to your audio, but it'll be nowhere near the undistorted CD audio.".
Under normal circumstances, there is no room to improve a 44kHz 16bit signal intended for end user audio (I'm not talking about mixing & stuff). Most humans can't hear above 16kHz (20 if you're lucky) and 44.1kHz can represent signals up to 22kHz. The only reason to use higher sampling frequencies, is to make the design of digital low-pass filters easier. 16bit is enough to represent a dynamic range of 90dB, which is far more than what you'll hear in the overcompressed crap that populates the charts today.
The only way to 'improve' the sound, is to modify it somehow, using some filters or effects. In other words, it will be distorted.
Maybe they found out what typical MP3 distortion sounds like, and invented a filter that undoes it. But this will also distort other sounds that accidentally have the same characteristic as MP3 distortion. Or maybe they just add some harmonics to the sound. Whatever it is, High Fidelity it isn't. Maybe that's why the name is 'X-Fi' and not 'Hi-Fi'...
The _usage_ of the "MIPS" here might be mostly marketing bullshit (as it doesnt make the sound "better"), its not compareable to PMPO or co.
Its just a bit decieving, because getting mips in audio chips is _REALLY_ easy. You are mostly dealing with 16 or 24bit integer values, in neat streams. You can build a whole function unit for a few 1000 transistors...
So just give the thing 50 adders, 50 mul-units, runn it at 100 Mhz and you get 10 billion possible instruction per second (which might be burned quite quickly if you want to do bigger effects on xx streams, but thats another matter).
HI O WISE PRINCE. WHT TOOK U SO DAM LONG?
Right.
So fair competition in your eyes involves malicious litigation knowing that the legal burden will drive your much smaller competititor under?
The only real audio card maker to have the balls to stand up to Creative was Diamond (remember them?). All the others wouldn't touch Aureal's tech while there was question marks over the legitimacy of Creative's claims which meant Aureal lost even more money. Later the courts would throw out every one of Creative's claims but by then it was too late.
The real injustice was the fact that Creative after losing the court case was allowed to pick over the remains of Aureal and acquire their IP. There is something seriously wrong with capitalism when companies are allowed to do this. Whatever the outcome, Creative was going to end up the winner while you, I and Aureal were most definitely the losers.
For all intensive porpoises your a bunch of rediculous loosers
You can't have aliasing from upsampling. Aliasing occurs from attempts to sample frequecnies that cycle at anything more than half the sampling rate. If you think about a wave, you have a peak and a trough. You need at least one sample on the peak and one on the trough to accurately represent that frequency. Any tone higher than one half of your sampling rate results in missed peaks and troughs, which the computer then represents as frequencies lower than what they originally were.
That said, you're right elsewhere, upsampling a crappy mp3 will only give you a more accurate representation of your crappy mp3.
WTFHEHO (who the fuck has ever heard of) "upmixing"?
This too, will end.
I always had bad experiences with stable machines becoming unstable after installing Creative's drivers, and never liked that you can't seem to just install what seems like a driver but have to screw up your system with what seems more like an application suite / (buggy) driver combo. What's worse is that despite the bloat Creative's stuff never has the features that I actually want in a sound card.
The only salvation for my SB cards has been the 'kx project' drivers:
http://kxproject.lugosoft.com/index.php?skip=1
(sorry I don't know to enter a URL here...)
If you are a musician these drivers have the features you actually want; WDM, ASIO, GSIF - other than the sound floor (on my SBLive) they make the card competative with a mid level music card. No bloat and I've found the driver to be solid, though the UI is rather yucky.
You can get optical out and decent sound for way cheap.
"Creative Labs, the worst thing to ever happen to sound card industry."
the runner up was
"Creative Labs, holding back soundcard innovation for over a decade"
If you wanna get rich, you know that payback is a bitch
I have bought the card last week, i have plugged it in today in no time, it take's up a psu cable (just as a hard drive) as a nice blue logo (nice addition to modded case) first thing i did was to hook up my jackson guitar in it to test one of the 2 preamplified 1/4 Jack that comes with it. The DAC makes its job well and the noise reduction is very good. Too good in fact that i had some trouble hooking up the hardware effects at first (before finding where to put on effects in the Xfi's pannel) , the noise reduction still give's me a hard time, it works too well. In a 30 minute trial to get to know the main reason to buy the Elite : Does the external module works to record guitar, voice and Bass -> Definitly. It comes packaged with Cubase and WaveLab and Ampli-tube, so everything is there to get started to rock, and do not forget that you can have 8 simultaneous Hardware effects that feels just like any pedal effects would (thrust me i heard worst distortion...metal zone anyone ?) So i definitly recommend this to home musicians who are gamers and want to be able to do anything, not just be able to record in a specialized software like protool cards without 3d accel. I bought the card directly on the site for 399$ US. Its not such a high price, because you would normally have to buy an ASIO sound card and a mini-console with 4 amplified I/O to have the same result.
My worst fears where that in the past Creative was well known to have horrendrous drivers and bloated install, this one has a package wich only install the tools for audio creation if you asks for it. Remember also that the elite pro comes with 64 MB of onboard RAM, wich frees up a bit of the main RAM for audio processing, tough I do not know if it makes a big difference.
For those that say its the speakers that makes a difference, try listening to AC98 onboard for 2-3 years, even with good speakers, you can still ear noise when nothing plays at high volume. Same speaker with this card feels like heaven.
A satisfied customer
As a former long-time Creative customer who has been burned many times and have seen many others burned, I'm no longer interested in Creative products. What I am interested in is this:
t goddl/
http://www.hidiaudio.com/products/mystique.html
http://www.bluegears.com/soundcard_xmystique.html
That's right, a card that can perform real-time Dolby Digital AC3 encoding (aka Dolby Digital Live, or DDL). The spiritual successor to the nVidia Soundstorm!
Turtle Beach has a card with the same chip, although their driver support is a bit lacking in comparison:
http://turtlebeach.com/site/products/soundcards/m
And this is the chip that drives them both:
http://cmedia.com.tw/product/CMI8768_plus.htm
The cards are pretty affordable - newegg has them both for under $100. Personally I'd rather go with the X-Mystique due to better driver support and on-board coax output (even though both cards come with optical cables, IIRC).
I guess Terratec has an Aureon 7.1 card that has DDL as well, but they don't market their cards to the U.S.
Arguing about vi versus Emacs is like arguing whether it's better to make fire by rubbing sticks or banging rocks.
You'd think so, but you'd be wrong. The shape of our heads and faces interact with natural sounds to change the way something sounds depending on its position other than on the left-right axis. Google for "head related transfer function" if you're interested.
In the context of digital signal processing MIPS refers to the number of multiply-accumulate operations per second, including incrementing buffer pointers. It is a well-defined number and comparing it is not meaningless even across different architectures.
Stop worrying about the risks of nuclear power and start worrying about the risks of not using nuclear power.