EEVblog #783 - Dumpster Dive Power Macs

[Godzil]

Lower noise floors and designed for good audio in the on board sound system, plus good enough software to record and play back with the editing. You do not really have much advantages in the new software, MP3 and WAV audio is pretty much the same since at least the last 20 years, and once it is digital and sufficiently good bit depth and sample rate wise using the uncompressed date in another format on another machine does not need the original low noise and distortion front end.

That's solved today with high quality outboard USB connected ADC's and DACs. No more internal soundcards for audio work. 
Back in the hayday of these G5 Macs they were the only thing that could reliably do A/V editing (and they had all the best software tools too).  But todays's Intel PC's have become fast enough now, and there's much more software available, some of it even exclusive to PC (sadly)...

I'm no Mac fan boy, nor a Wintel lover.... I'm a Linux nerd, and the limited A/V work I do is covered by the available open source tools.

USB for audio?
It's almost impossible to have a steady bandwidth with USB, so realtime audio with USB...
That why professional audio tool use at least something like the IEEE1394 or now the Thunderbolt connectors that have a band witch warranty.

USB is just the worst thing for realtime audio need, that's also why I laugh everything that I see "Audiophile" product that use USB for delivering audio via USB to their marvelous DAC to get the best jitter/precisoin/bitcolor/whatever that audiophiles are so fond of.

(And there is not a lot more of Windows only audio tool today than 10 years ago, at least on the professional side)

[apis]

Nice find! 

I've taken apart a few lower end macs a friend (who unlike me is an apple fanboy) has thrown out over the years and they have all been as nicely engineered on the inside as the outside. Well, almost, because if you lift the skirt on some of the hard to get parts (like the psu) you might be in for a surprise. One imac I looked at that was really nice had a horrible third-party psu unit which most likely was the point of failure (cheap caps). It would have been interesting to see the insides of the parts that wasn't intended to be user serviced on these high end macs as well, although I suspect they didn't cut corners in these units that target professional users, but who knows.

Mac's before Jobs returned was a joke, when Jobs got back he made a lot of good changes for apple: the new BSD based OS, nice perfectionist hardware designs, etc. I wonder if things will deteriorate again now that he is gone.

[Pillager]

You could make an awesome looking bench supply or something like that  O0

[Someone]

Something about the fans being accessible in operation without tools, and without proper warning labels.

They have to have fan guards if they can be accessed without tools. Apple were retiring it anyway so didn't bother complying.

The thermal design was okay for the time. A lot of fans, far more than necessary if properly designed. That made them dust magnets and a little noisy. Other aspects were quite well done though.

I know someone who bought one instead of a car. He was rather upset when the X86 machines launched a few months later.

Its baloney, desirable features of a computer are easy tooless servicing, and quiet operation. But consumers are forced to have one or the other (fan guards add either noise or take lots of space) to protect the stupid who operate the computer open and poke their fingers into the fans?

Lower noise floors and designed for good audio in the on board sound system, plus good enough software to record and play back with the editing. You do not really have much advantages in the new software, MP3 and WAV audio is pretty much the same since at least the last 20 years, and once it is digital and sufficiently good bit depth and sample rate wise using the uncompressed date in another format on another machine does not need the original low noise and distortion front end.

That's solved today with high quality outboard USB connected ADC's and DACs. No more internal soundcards for audio work. 
Back in the hayday of these G5 Macs they were the only thing that could reliably do A/V editing (and they had all the best software tools too).  But todays's Intel PC's have become fast enough now, and there's much more software available, some of it even exclusive to PC (sadly)...

I'm no Mac fan boy, nor a Wintel lover.... I'm a Linux nerd, and the limited A/V work I do is covered by the available open source tools.

Look at some of the older mac pro systems being sold on Ebay, you can get a kitted out box with a multi IO professional audio interface and licensed software for just a few hundred. Having the professional workflow and tools is really a big step up, and the older professional interfaces are great value once they're no longer supported by the current versions/drivers/OS.

[Rasz]

I bought a complete G5 setup for £97 a few years ago, purely for making music.  It's far better for audio work than even any modern Windows PC I've tried.

What applications? Can't see how OS or hardware can affect "making music"

latency for starters

[kc8apf]

Now this brought back some memories.  I worked in the hardware performance group at Apple starting just before the G5 launched.  When the "world's fastest personal computer" announcement came out, there was a lot of contention about the claim.  I ended up sleeping in our lab that night to rerun a bunch of benchmarks with various tweaks that were claimed to make the PCs faster.

The later models with the C19 plugs used a new system ASIC.  I and another guy needed to run large numbers of simulations of the memory controller design to evaluation various concepts but we didn't have enough machines to do so.  I found a whole rack of prototype Xserves being thrown away by another lab, fixed them all, put them in our lab, and ran simulations on them 24x7 for a month.  We ended up with a design that improved overall performance quite a bit.

Then it was all scrapped just after release to switch to Intel.

[jazz]

USB for audio?
It's almost impossible to have a steady bandwidth with USB, so realtime audio with USB...

I'm pretty sure that USB allows you to reserve bandwidth, which is then guaranteed. Only bulk transfers have no guarantees and use otherwise unused bandwidth. A 96k 24bit 2ch stream only needs about 4.6Mbit/s anyway, which is almost nothing compared to the 480Mbit/s that are availabe since USB 2.0. Latency is the same as IEEE1394, if you use High Speed transfers.

But maybe I am missing something here?

[nitro2k01]

I bought a complete G5 setup for £97 a few years ago, purely for making music.  It's far better for audio work than even any modern Windows PC I've tried.

What applications? Can't see how OS or hardware can affect "making music"

The way the OS handles realtime scheduling and other functionality related to audio and MIDI does affect applications for music making. Things like how low you can set the latency (ie audio buffer length) without getting underruns or whether plugging a new MIDI device requires restarting the music application.

[station240]

The fault with the big one could simply be, a duff hard drive, or dust in the memory slots.
Given how the PSU connections are designed, should be easy to check the power rails as well.

If you do want to make something out of the case, how about a high power dummy load ?
Silicon Chip had a design decades ago that involved banks of electric jug elements to create a 1000w 2/4/8 ohm load.

[Godzil]

USB for audio?
It's almost impossible to have a steady bandwidth with USB, so realtime audio with USB...

I'm pretty sure that USB allows you to reserve bandwidth, which is then guaranteed. Only bulk transfers have no guarantees and use otherwise unused bandwidth. A 96k 24bit 2ch stream only needs about 4.6Mbit/s anyway, which is almost nothing compared to the 480Mbit/s that are availabe since USB 2.0. Latency is the same as IEEE1394, if you use High Speed transfers.

But maybe I am missing something here?

Yes you are missing the fact that bandwidth is not the only important metric.

USB work in burst, latency is unpredictable as the bus is shared between different devices, the bus is "probed" directly by the software.

USB hasn't been made for continuous transfer, unlike IEEE1394 (and other buses)
Thats why, USB2 vs 400M IEEE1394, for external disc operation, for large file copy, IEEE1394 is faster than USB, just because USB can't sustain long transfert without bursting.

Also, described by some people as a possible security probleme, the IEEE1394 slave device have the ability to directly do DMA transfert without the need of the CPU.
For a crude comparison, IEE1394 work like SCSI, USB like PATA, the first is a "smart" bus where there is not a single master, and everything else is slaves, but two peripheral could transfer from one to the other, without telling the third about that.

I will quote some info from wikipedia:

The system was commonly used to connect data storage devices and DV (digital video) cameras, but was also popular in industrial systems for machine vision and professional audio systems. Many users preferred it over the more common USB 2.0 for its then greater effective speed and power distribution capabilities. Perhaps more important, FireWire uses all SCSI capabilities and has high sustained data transfer rates, important for audio and video editors. Benchmarks show that the sustained data transfer rates are higher for FireWire than for USB 2.0, but lower than USB 3.0.

The FireWire host interface supports DMA and memory-mapped devices, allowing data transfers to happen without loading the host CPU with interrupts and buffer-copy operations. Additionally, FireWire features two data buses for each segment of the bus network whereas, until USB 3.0, USB featured only one. This means that FireWire can have communication in both directions at the same time (full-duplex), whereas USB communication prior to 3.0 can only occur in one direction at any one time (half-duplex).

If all DV camera used IEEE1394 for realtime video, there where a reason, it was far better suited than USB for that, and not only for bandwidth reason.
Isochronous transfert on USB prior to USB 3 was merely a joke

[Fungus]

USB is just the worst thing for realtime audio need, that's also why I laugh everything that I see "Audiophile" product that use USB for delivering audio via USB to their marvelous DAC to get the best jitter/precisoin/bitcolor/whatever that audiophiles are so fond of.

Um, audiophiles don't need "real time audio"'.

And thanks to chips like the PCM2704 you can get amazing quality sound from your PC for under $12.

eg. http://www.ebay.com/itm/310799247879

(Of course no true audiophile would spend so little. This is for the rest of us...)

[Godzil]

Audiophiles dont like jitters, latencies and anything like that.

And originally I wasn't speaking about audiophiles, but the "real" pro in the audio domain (aka people that works in audio stuff).

[Fungus]

Audiophiles dont like jitters, latencies and anything like that.

If you bother reading the TI datasheet you'll see that the TI chips eliminate jitter with "SpAct™ architecture, TI's unique system that recovers the audio clock from USB packet data."

Latency makes no difference when you're listening to music. None. Zero. Nada.

And originally I wasn't speaking about audiophiles, but the "real" pro in the audio domain (aka people that works in audio stuff).

Your claim was that USB is useless to audiophiles.

[Godzil]

Are you sure to know what an Audiophile is? Not sure from what you say.

An Audiophile is someone that use black magic for getting better audio than he really have. Not using rational EE things for audio.

The best fear of an Audiophile is Jitter, you know if the left channel is 1 ns slower than the right channel, this is a big problem, and when I see that lots of audiophile product nowadays use USB for data transfer, it just make me laugh, when you see all the black magic the manufacturer put in their product, but they use something as unreliable (for an audiophile point of view) as USB.

I never say that for normal audio listening usb is bad, it's maybe not the best, but it does it's job. What I says is that for PRO environment USB is not well suited for realtime work.

Sorry the author is french, and speak french only, but just skim in this video to see what is inside the "Manta" from the brand "Esprit" which cost a lots of money, and is typical of what audiophile like to buy:



(Jipihorn is somewhat the french equivalent of Dave)

[Fungus]

Are you sure to know what an Audiophile is? Not sure from what you say.

"Audio" = audio
"Phile" = lover

The best fear of an Audiophile is Jitter, you know if the left channel is 1 ns slower than the right channel, this is a big problem,

That's not what jitter is.

https://en.wikipedia.org/wiki/Jitter

PS: 1ns delay between left/right channels is about 0.0003 millimeters at the speed of sound. You'd better adjust your headphones very carefully when you wear them or your worst nightmare will happen(!)

And don't even think about listening to speakers - the sound image will go to hell if you move your head even by 0.5 mm.

[rollatorwieltje]

A pc is not a realtime system. If you need a realtime system, get something purpose built (like a DSP), or deal with the latency. Even an internal soundcard on a PCI bus has several milliseconds latency and only in ASIO mode, in shared mode there's about 50ms latency in OS buffer alone. OSX seems better in this regard, manufacturers like RME claim that USB 2.0 performance under OSX is on par with PCI (round-trip time of about 5ms, 32 samples buffering).
Buffering has absolutely no effect on sound quality, it just introduces a delay, which is only really an issue when you're doing live recording. A buffer underrun will cause very noticeable glitches, not some minor jitter or synchronization errors.

My Scarlett Solo USB 2.0 sound card has a loopback feature to deal with the latency, when you're recording you can hear your own voice or instrument directly without going through the PC. Obviously without any processing (by the way, how much latency do those effects introduce?), but that's just the limitation of using a general purpose computer instead of a proper DSP. Either learn to work with the unprocessed signal or learn to compensate for the latency in your recording.

[Godzil]

@Fungus: so you don't know what Audiophile really means. An Audiophile is like creationist, or 911 conspirationist, but on the Audio side. Believe in black magic to make audio sound better.
I met someone one day that told me that the the audio quality you have was depending on the calibration of the laser lens for an Audio CD, because the "colour" of the bit changes because of that, and the DAC does not reproduce the same if the laser is not correctly calibrated.

That's an Audiophile. And Latin/Greek will not help you to understand what this word mean.

And about jitter, yes you are right, now goes to audiophile website and look at what they say. They tend to use words that they don't understand, in situation that does not mean anything.
They are people that buy digital cables 10x the normal price just because they are gold platted, and have some black magic box on the cable that make the sound better. Yes on a digital cable, like HDMI.
I saw too ferrite on optical cable, and golden platen TOSLINK.

Rollatorwieltje: No a PC is not a realtime system for a nuclear power plant that's for sure, but for audio it's more than enough and we can call it realtime. And between a port like USB or IEEE1394/Firewire, that make a noticeable difference when working in professional audio environement.

When you are working on realtime audio, buffering is not an option, when you apply a "realtime effect" or whatever buffering is absolutely not an option.
I have a few good friends that work on professional audio edition, live and things like this, they have tested lots of different solutions, and USB solution wasn't a good choice every time, they prefer solution like PCI card, IEEE1394, Midi etc.. but USB a part for high latency stuff and mouse/keyboard, they were never a real option.

[codeboy2k]

@Godzil... what you are describing is the acts of an Audiophool ... not an Audiophile.

An Audiophile is simply someone interested in High Quality audio, by whatever means.  When that becomes excessive, and garbled in nonsense such as ns jitter, carbon fiber cables, gold and silver intertwined jackets wrapped around poly-chromium foil encased in a ultra-flexible silicon jacket, ferrite on optical cables, gold plated TOSLINK, then it's Audiophoolery. ..

Someone with high-quality components that wants to get great sound is simply an Audiophile.  Most audiophiles know where reality is, anyone else is an audiophool.

[Fungus]

@Fungus: so you don't know what Audiophile really means.

Stop changing the subject.

You said: "USB is just the worst thing for realtime audio need, that's also why I laugh everything that I see "Audiophile" product that use USB for delivering audio via USB to their marvelous DAC to get the best jitter/precisoin/bitcolor/whatever that audiophiles are so fond of."

I pointed you to a perfectly good USB audio device for under $12. They exist.

USB is good for audio. QED.

[Rasz]

USB is good for audio. QED.

No, it is NOT. It might be good for listening to audio, but not making music.

[Jr460]

Making/recording music is so much different from listening to music.

If you are playing to existing tracks, then latency is a big problem to deal with.  Just to give you an example, having the kick drum be off about 10-20 ms can change the feel of the song from a groove to something that has no life.

You want to get real nasty, then deal with live audio mixing and latency.  It used to be not too much of a problem, but with performers mainly using in-ear monitors, a 10ms delay between them singing and what comes back in the in ears causes combo filter between the bone conduction path and the in ears.  The latency in a pro quality live board is on the order of 1.5ms (that would be a analog in, all processing, and back out to analog).  Use something like Dante (cool technology from AUS) to get your audio to and from stage and you are talking from 250us to a 1ms (configurable and guaranteed) each direction.  That would put you at 3.5ms worst case which the performers will not notice.

Trust me on this, I do live mixing, and have done a far amount of playing and singing live over the years.  I don't buy Audiofool things like $4k speaker cables, but some things do make a big difference.  Cheap AD/DA converters being the biggest problem.  The good designs take much better care of the signal on the analog side and also with power supplies even if they use the same converter chip as the cheap ones.

[nabeelr]

I'd love to see a teardown on one of these machines.

I used to work for Apple, and did a few repairs on them, they are really difficult to take apart and put together, but their boards are really beautifully laid out. (At least I thought)

I'd love to hear what your take is on the components and materials uses, as well as the layout.

[Fungus]

Making/recording music is so much different from listening to music.

Of course.

Cheap AD/DA converters being the biggest problem.  The good designs take much better care of the signal on the analog side and also with power supplies even if they use the same converter chip as the cheap ones.

Don't worry, phoolery is alive and well. There's a whole bunch of people out there swapping tips on how to upgrade the capacitors in those $12 ebay DACs.

FWIW, ELNA seems to be the brand of choice.

(Or you can pay a couple of dollars more on eBay to get an 'ELNA' model...)

[Godzil]

Making/recording music is so much different from listening to music.

Of course.

Cheap AD/DA converters being the biggest problem.  The good designs take much better care of the signal on the analog side and also with power supplies even if they use the same converter chip as the cheap ones.

Don't worry, phoolery is alive and well. There's a whole bunch of people out there swapping tips on how to upgrade the capacitors in those $12 ebay DACs.

FWIW, ELNA seems to be the brand of choice.

(Or you can pay a couple of dollars more on eBay to get an 'ELNA' model...)

$15 is just too cheap for them, it miss at least one zero.

[senso]

PCM270x is just a 16bits/48Khz interface, it's not really the best you can get, but Windows does not support USB Audio 2.0, so anything over 16 bits requires dedicated drivers, and you get into the land of either Xmos USB to I2S, or a couple chips for CMedia, Tenor and little else..
Even an Amanero or a Modi will set you back around 100$/€..