Honestly, reading this article, it was a great fit for my "what's the fastest video compression workstation I can build that I can also play games on." Seemed to meet those needs quite nicely. I'm running a dual Athlon 2800+ MP system right now, but some next gen gear is PCI-X only, so it's getting near time to update.
One big issue in getting a real workstation from Dell or HP or whatever is that they tend to be configured with high end 3D and SCSI. I'm not doing 3D animation or running a database. I want good gamer 3D, wickedly fast 2D (for playing back HD video), and S-ATA RAID (video requires high bandwidth, but isn't very sensitive to latency). The SCSI premium is still way too high, especially compared to the nearly-as-good S-ATA.
So, I'm quite likely to be building a box quite similar to this, and found the article generally helpful. Honestly, if anything it was a little low end. The audio system (professional 7.1 channel) for testing HD and DVD content costs way more than the workstation will.
And why should that matter? PCM is PCM. They'd likely have different latency, but I can't imagine what would cause different frequency response in playback.
While these G5 systems are expensive for consumers, that really isn't your market.
Your parents should look at the consumer lines. The eMac starts at just $799. I'd probably toss an extra 256 MB stick in there, but you'll still have a complete internet/productivity machine, with monitor, for ~$850.
Sure, you can get the cheapest Wintel machine for less than that, but if you're providing technical support, it's probably worth a little of their money for them to have a machine without BIOS messages and effectively immune to viruses.
I'm always startled to see people getting excited about three-digit price differences between base machines for things meant for the video market.
I'm building a HD editing/compression system. When you add in storage, displays, audio, etcetera, even a cheap setup makes a $500 difference pretty tiny.
For my needs, I'm shaving THOUSANDS off by going Mac. Why? Xserve RAID. I can get 2.5 TB for about $12K, that's fast enough for 1920x1080i60 at 10-bit uncompressed capture. I haven't been able to find anything that's close to that price performance on Windows.
This is an edge case, granted. But for anyone who bills by the hour, a few hundred bucks in a system that's going to be making you money for a couple of years is nothing - like a quarter a day. Downtime for one tech support incident could eat up the entire differential.
Well, I'm typing this on the new 1.33 GHz G4 17" PowerBook. I've got the same RAM and graphics card. You win on drive speed (I've got a single 5400 rpm). On a large but not complete set of tasks, your faster processor and memory bus will make it faster. Price was roughly comparable (a few extra $100 on my end).
But I run rings around you in Final Cut Pro:). And can use it in my lap without getting scalded. And when I open the screen, the machine is fully awake before I can get the lid all the way open. And the battery can last for 3+ hours. And its' certainly a lot lighter and thinner than yours.
And I can't tell you how long it takes to boot, since I never have to reboot it. Once for the 10.2.8 update.
The big problem with the dual and higher Xeons is that they're limited to a shared 533 MHz bus. For the wide class of memory-bound problems, the G5 will sweep the floor with them, since each processor has its own 1000 MHz bus.
For my Lamy, I just got an optional cartridge that lets me refill it from standard bottled ink, instead of having to use the disposable cartridges. But considering how ink-stained my hands got back in those days, that might not have been the right choice.
I'm thankful that I almost never have to do handwriting anymore! I only started using fountain pens for everything because I would get horrible hand cramps writing with anything that required any pressure. It always through off my physics professors when I'd turn in assignments written in italic calligraphy on unlined paper.
Actually, there are a lot of things a paid service can offer that pirate P2P can't:
Guaranteed quality Much better browsing No spoofing or hoax downloads
legal P2P can offer those, but you're still left with:
Download speed
Most P2P systems don't offer very reliable download speeds. I suppose a BitTorrent system could work, but I question how well it could scale up to hundreds or thousands of different songs for each user. The number of users who are logged in at any given time who have the correct song might not be that high. Also, most users has asymmetric download.
If you're doing a paid service for relatively small files like music, it seems to make more sense to just own your own servers and pay for your own bandwidth. Much more straightforward to users, and not that expensive in context.
Sure, SIMD optimization is certainly useful. But the code isn't documented in a way that would make porting it to a different SIMD architecture at all easy (it's been looked at a number of times).
PNG/MNG are natively RGB, and so aren't well suited towards video.
It's not hard, and it has been done. It's just that if I really want a.mp4, I'll use a stand-alone MPEG-4 encoder like Squeeze or Compression Master, and skip those extra steps.
WMV9 is a tremendously good codec, and beats out next generation MPEG-4 for high definition tasks.
Head on over to dv.com and read the article I wrote in the current issue about HD delivery codecs. Microsoft is working hard to win this battle on technical merits.
FWIW, they've also been submitting the technology to various standards boards, including SMPTE.
Xvid is a MPEG-4 implementation, so commercial products based on it will need to pay the MPEG-4 license fee, FWIW. It's cheap.
Theora is free as in every kind of speech, beer, or anything else you could imagine. However, they haven't locked down their bitstream yet, so it's hard to say how good it will be as a codec.
Huffyuv is open source, but full of x86 assembly, so it isn't usefully portable. I'd love to see an equivalent technology that'd be more portable, and LGPL so it could be used more widely.
I think you meant something other than DirectX. DirectShow does have third party MPEG-2 decoder filters available, but it's hardly the dominant format used with DS.
As for DRM, the iTunes Music Store has shown that many people don't object to DRM per se, they mainly show that people reject DRM models which don't allow them to use the content in the way they want to use them.
It's not like you hear about people boycotting Macrovision encoded DVD that often:).
Divx 4.x and 5.x are themselves MPEG-4 Video codecs (pt. 2 video, not the new pt. 10/AVC/JVT/H.264 stuff). They aren't the best, but they're far better than the worst. The bitstream itself is compatible with ISMA compliant decoders, but...
Divx files themselves use the AVI file format instead of the MPEG-4 file format. This is for historical reasons, and the biggest problem I personally have with Divx, since it is incompatible with stock MPEG-4 tools, but not in a way that adds any user value. This is a legacy of how Divx was originally a hack to use a proprietary Microsoft codec in AVI files.
Divx files also use all kinds of audio codecs, which are rarely MPEG-4 compatible. AAC-LC is a great audio codec, and it's ISMA compatible.
So, I really wish Divx would get their tools support exporting to.mp4 with aac-lc audio. They've done 98% of the hard work to interoperate well. It's just that last 2% I'm waiting for. I don't mind if they maintain legacy support, but I want support for ISO standard MPEG-4 before I'm likely to use their stuff for much in practice.
Also, you are correct, Windows Media can't play MPEG-4 by default. There is a plugin available from Envivio for WMP that will enable this, though.
Yeah, does WINE give you ColorSync or equivalent? I worry about color management.
As for "far more attractive" I don't see that in the numbers. A G5 is quite reasonably priced for a dual processor workstation. Sure, you can buy cheap x86 boxes, and Apple doesn't make anything that low end, but animators aren't using eMachines boxes!
After I got laid off 2.5 years ago, I decided to go the contract route, and it has been pretty good. The key is to have existing clients, and unique skill sets. I was the lead consultant for my former company, but a lot of the potential customers for that kind of service had competing products. So once I was on my own, a lot of my initial clients made products that competed with my former employer's stuff. My billable hours actually went UP.
Glad I never signed that non-compete!
Anyway, it's important to note that the business didn't really take off until I decided it really was a business, not just what I was doing until I found another job. Once I decided this is what I wanted to be doing for the next ten years, I was motivated to go out, take out some loans, spend the capital to get some marketing, and that kind of thing. The average service business like this isn't really profitable for the first two years. If you accept that and plan accordingly, that's not a big problem.
The mistake I've seen others make is to blow their whole nest egg early in the process, not leaving enough to live on as things start to get rolling. What a business is evolves a lot depending on what clients you actually land, so you need to keep enough money in reserve to be able to keep adjust the plan mid-stream.
Based on the article, it seems like Apple's new G5 systems would almost be a better fix for Linux for a lot of this stuff. SGI was too slow, and Windows too different an API to port. Mac OS X lets you run all the Adobe stuff natively, a real UNIX underneath for porting, and with the G5, a processor fast enough to get in the ballpark of x86 for price/performance. As the article mentioned, Shake on Mac is so much cheaper, you basically get the Mac for free!
The new Photoshop CS announced yesterday definitely has the deep paint (and non square pixel support!) required to make it a great film/video app for a lot of stuff.
The one big barrier I see is a complete port of Maya (Maya Complete not being complete, of course).
*NIX to OS X porting will get a lot easier with 10.3 (out within 30 days?), which has X11 built-in, and a lot more Linux-like API stuff.
Okay, you've got strong opinions. I hear a lot of this kind of comment, and since we're on Slashdot, I can ask you a question back! Let's see what kind of policy you're imagining.
So, specifically, which programs would you cut and by how much?
What policies would deal with the potential gaps between what private charity can provide and the needs? The last time the USA didn't have a significant welfare state, private charity was manifestly not up to the task of providing adequate services for all (although it did a good job for some).
What proportion of income do you imagine the average citizen will put into private charity a year? How much are you willing to personally commit to giving?
How much deficit are you willing to accept?
Based on those savings, and target deficit, how would you distribute the resulting tax increases or decreases? You don't have to confine yourself to income taxes - look at all federal taxes, including payroll, social security, capital gains, and corporate.
If you could run things, what do you see as a better distribution?
I know people love the GIMP, but man, is that UI confusing to a new user. It can do everything, but is a lot harder to use than Elements for Elements' sweet spot of consumer photo retouching and manipulation.
That said, there is a ton of stuff that GIMP can't do that isn't possible in Elements. It really depends on the need.
Well, my customers wouldn't pay me so much if they thought I was milking the job. $300/hour is the price for "works fast with obsessive quality on brutal deadlines applying unique skills."
That said, I wind up doing most of my work these days with per-project billing.
I did the math a while ago, and figured that I'd only get half as much work if I increased by rate from $150/hour to $300/hour. Working half as much for the same money seemed like a pretty good deal:).
Slashdot Mirror
Honestly, reading this article, it was a great fit for my "what's the fastest video compression workstation I can build that I can also play games on." Seemed to meet those needs quite nicely. I'm running a dual Athlon 2800+ MP system right now, but some next gen gear is PCI-X only, so it's getting near time to update.
One big issue in getting a real workstation from Dell or HP or whatever is that they tend to be configured with high end 3D and SCSI. I'm not doing 3D animation or running a database. I want good gamer 3D, wickedly fast 2D (for playing back HD video), and S-ATA RAID (video requires high bandwidth, but isn't very sensitive to latency). The SCSI premium is still way too high, especially compared to the nearly-as-good S-ATA.
So, I'm quite likely to be building a box quite similar to this, and found the article generally helpful. Honestly, if anything it was a little low end. The audio system (professional 7.1 channel) for testing HD and DVD content costs way more than the workstation will.
And why should that matter? PCM is PCM. They'd likely have different latency, but I can't imagine what would cause different frequency response in playback.
While these G5 systems are expensive for consumers, that really isn't your market.
Your parents should look at the consumer lines. The eMac starts at just $799. I'd probably toss an extra 256 MB stick in there, but you'll still have a complete internet/productivity machine, with monitor, for ~$850.
Sure, you can get the cheapest Wintel machine for less than that, but if you're providing technical support, it's probably worth a little of their money for them to have a machine without BIOS messages and effectively immune to viruses.
I'm always startled to see people getting excited about three-digit price differences between base machines for things meant for the video market.
I'm building a HD editing/compression system. When you add in storage, displays, audio, etcetera, even a cheap setup makes a $500 difference pretty tiny.
For my needs, I'm shaving THOUSANDS off by going Mac. Why? Xserve RAID. I can get 2.5 TB for about $12K, that's fast enough for 1920x1080i60 at 10-bit uncompressed capture. I haven't been able to find anything that's close to that price performance on Windows.
This is an edge case, granted. But for anyone who bills by the hour, a few hundred bucks in a system that's going to be making you money for a couple of years is nothing - like a quarter a day. Downtime for one tech support incident could eat up the entire differential.
Good points.
A big part of getting the TCO for paid music better than free is lowering the price.
Well, I'm typing this on the new 1.33 GHz G4 17" PowerBook. I've got the same RAM and graphics card. You win on drive speed (I've got a single 5400 rpm). On a large but not complete set of tasks, your faster processor and memory bus will make it faster. Price was roughly comparable (a few extra $100 on my end).
:). And can use it in my lap without getting scalded. And when I open the screen, the machine is fully awake before I can get the lid all the way open. And the battery can last for 3+ hours. And its' certainly a lot lighter and thinner than yours.
But I run rings around you in Final Cut Pro
And I can't tell you how long it takes to boot, since I never have to reboot it. Once for the 10.2.8 update.
The big problem with the dual and higher Xeons is that they're limited to a shared 533 MHz bus. For the wide class of memory-bound problems, the G5 will sweep the floor with them, since each processor has its own 1000 MHz bus.
For my Lamy, I just got an optional cartridge that lets me refill it from standard bottled ink, instead of having to use the disposable cartridges. But considering how ink-stained my hands got back in those days, that might not have been the right choice.
I'm thankful that I almost never have to do handwriting anymore! I only started using fountain pens for everything because I would get horrible hand cramps writing with anything that required any pressure. It always through off my physics professors when I'd turn in assignments written in italic calligraphy on unlined paper.
Actually, there are a lot of things a paid service can offer that pirate P2P can't:
Guaranteed quality
Much better browsing
No spoofing or hoax downloads
legal P2P can offer those, but you're still left with:
Download speed
Most P2P systems don't offer very reliable download speeds. I suppose a BitTorrent system could work, but I question how well it could scale up to hundreds or thousands of different songs for each user. The number of users who are logged in at any given time who have the correct song might not be that high. Also, most users has asymmetric download.
If you're doing a paid service for relatively small files like music, it seems to make more sense to just own your own servers and pay for your own bandwidth. Much more straightforward to users, and not that expensive in context.
Sure, SIMD optimization is certainly useful. But the code isn't documented in a way that would make porting it to a different SIMD architecture at all easy (it's been looked at a number of times).
PNG/MNG are natively RGB, and so aren't well suited towards video.
It's not hard, and it has been done. It's just that if I really want a .mp4, I'll use a stand-alone MPEG-4 encoder like Squeeze or Compression Master, and skip those extra steps.
Nope, that's right.
File format=avi
video codec=MPEG-4
audio codec=one of many
That's a pretty bogus analysis.
WMV9 is a tremendously good codec, and beats out next generation MPEG-4 for high definition tasks.
Head on over to dv.com and read the article I wrote in the current issue about HD delivery codecs. Microsoft is working hard to win this battle on technical merits.
FWIW, they've also been submitting the technology to various standards boards, including SMPTE.
Xvid is a MPEG-4 implementation, so commercial products based on it will need to pay the MPEG-4 license fee, FWIW. It's cheap.
Theora is free as in every kind of speech, beer, or anything else you could imagine. However, they haven't locked down their bitstream yet, so it's hard to say how good it will be as a codec.
Huffyuv is open source, but full of x86 assembly, so it isn't usefully portable. I'd love to see an equivalent technology that'd be more portable, and LGPL so it could be used more widely.
I think you meant something other than DirectX. DirectShow does have third party MPEG-2 decoder filters available, but it's hardly the dominant format used with DS.
:).
As for DRM, the iTunes Music Store has shown that many people don't object to DRM per se, they mainly show that people reject DRM models which don't allow them to use the content in the way they want to use them.
It's not like you hear about people boycotting Macrovision encoded DVD that often
Divx's video codec is exactly MPEG-4, which they're public about, and they are paid up on their license fees.
The technology did start out as an illegal hack, but it's 100% legit know as far as I know.
This all gets kind of confusing :).
.mp4 with aac-lc audio. They've done 98% of the hard work to interoperate well. It's just that last 2% I'm waiting for. I don't mind if they maintain legacy support, but I want support for ISO standard MPEG-4 before I'm likely to use their stuff for much in practice.
Divx 4.x and 5.x are themselves MPEG-4 Video codecs (pt. 2 video, not the new pt. 10/AVC/JVT/H.264 stuff). They aren't the best, but they're far better than the worst. The bitstream itself is compatible with ISMA compliant decoders, but...
Divx files themselves use the AVI file format instead of the MPEG-4 file format. This is for historical reasons, and the biggest problem I personally have with Divx, since it is incompatible with stock MPEG-4 tools, but not in a way that adds any user value. This is a legacy of how Divx was originally a hack to use a proprietary Microsoft codec in AVI files.
Divx files also use all kinds of audio codecs, which are rarely MPEG-4 compatible. AAC-LC is a great audio codec, and it's ISMA compatible.
So, I really wish Divx would get their tools support exporting to
Also, you are correct, Windows Media can't play MPEG-4 by default. There is a plugin available from Envivio for WMP that will enable this, though.
Yeah, he had it completely wrong.
RealOne is the new player, RealPlayer was the old player. They both play RealMedia, and they can both use the current set of codecs and all.
Real is actually doing a nice job of reconnecting to the *NIX community.
Anyone know what compression format this thing is going to use?
MPEG-4 would seem like the obvious choice for a set top box, but I haven't heard any indication of what it really is going to be.
Yeah, does WINE give you ColorSync or equivalent? I worry about color management.
As for "far more attractive" I don't see that in the numbers. A G5 is quite reasonably priced for a dual processor workstation. Sure, you can buy cheap x86 boxes, and Apple doesn't make anything that low end, but animators aren't using eMachines boxes!
After I got laid off 2.5 years ago, I decided to go the contract route, and it has been pretty good. The key is to have existing clients, and unique skill sets. I was the lead consultant for my former company, but a lot of the potential customers for that kind of service had competing products. So once I was on my own, a lot of my initial clients made products that competed with my former employer's stuff. My billable hours actually went UP.
Glad I never signed that non-compete!
Anyway, it's important to note that the business didn't really take off until I decided it really was a business, not just what I was doing until I found another job. Once I decided this is what I wanted to be doing for the next ten years, I was motivated to go out, take out some loans, spend the capital to get some marketing, and that kind of thing. The average service business like this isn't really profitable for the first two years. If you accept that and plan accordingly, that's not a big problem.
The mistake I've seen others make is to blow their whole nest egg early in the process, not leaving enough to live on as things start to get rolling. What a business is evolves a lot depending on what clients you actually land, so you need to keep enough money in reserve to be able to keep adjust the plan mid-stream.
Based on the article, it seems like Apple's new G5 systems would almost be a better fix for Linux for a lot of this stuff. SGI was too slow, and Windows too different an API to port. Mac OS X lets you run all the Adobe stuff natively, a real UNIX underneath for porting, and with the G5, a processor fast enough to get in the ballpark of x86 for price/performance. As the article mentioned, Shake on Mac is so much cheaper, you basically get the Mac for free!
The new Photoshop CS announced yesterday definitely has the deep paint (and non square pixel support!) required to make it a great film/video app for a lot of stuff.
The one big barrier I see is a complete port of Maya (Maya Complete not being complete, of course).
*NIX to OS X porting will get a lot easier with 10.3 (out within 30 days?), which has X11 built-in, and a lot more Linux-like API stuff.
Okay, you've got strong opinions. I hear a lot of this kind of comment, and since we're on Slashdot, I can ask you a question back! Let's see what kind of policy you're imagining.
So, specifically, which programs would you cut and by how much?
What policies would deal with the potential gaps between what private charity can provide and the needs? The last time the USA didn't have a significant welfare state, private charity was manifestly not up to the task of providing adequate services for all (although it did a good job for some).
What proportion of income do you imagine the average citizen will put into private charity a year? How much are you willing to personally commit to giving?
How much deficit are you willing to accept?
Based on those savings, and target deficit, how would you distribute the resulting tax increases or decreases? You don't have to confine yourself to income taxes - look at all federal taxes, including payroll, social security, capital gains, and corporate.
If you could run things, what do you see as a better distribution?
I know people love the GIMP, but man, is that UI confusing to a new user. It can do everything, but is a lot harder to use than Elements for Elements' sweet spot of consumer photo retouching and manipulation.
That said, there is a ton of stuff that GIMP can't do that isn't possible in Elements. It really depends on the need.
Well, my customers wouldn't pay me so much if they thought I was milking the job. $300/hour is the price for "works fast with obsessive quality on brutal deadlines applying unique skills."
:).
That said, I wind up doing most of my work these days with per-project billing.
I did the math a while ago, and figured that I'd only get half as much work if I increased by rate from $150/hour to $300/hour. Working half as much for the same money seemed like a pretty good deal