The current DRI Radeon drivers are fine for 3D, providing pretty much the full power of the machine to the Linux world -- but the 2D performance is awful. Why, you might ask? Because when ATI contracted with VA Linux to create the drivers, they only funded development of 3D, and didn't specify that the 2D performance should be accelerated.
This turns out to be a real problem in the visual effects community -- a lot of our work depends on having good, fast 2D. Film frames, after all, are still just 2D images.
I hope that this new driver they speak of, and future drivers, recognize this. We'll see. At this point, we have to by nVidia boards, even while the drivers are closed-source, they provide pretty good 2D support.
I ask because Linux is a very bad choice, if that's what you're planning on using. How many professional 3D packages are available for Linux?
The two most important high-end 3D packages are Maya and Houdini. Both are available, shipping now, in complete supported implementations, under Linux.
Every single high-end visual effects and animation facility is using Linux for render farms, and almost all of them are pushing it very strongly on the desktop. For example, Pacific Data Images (you know, the people that did Shrek) have completely moved over to Linux and abandoned their previous platforms.
The way that 'Massive' works is ingenious. A statement of the battle simulation problem is "How do you get all of these simulated actors to respond correctly to their individual environment?"
The solution employed at WETA is to use state-of-the-art graphics hardware to render the view from the eyes of each simulated actor. He then plans his actions based on this view, correctly placing his feet on the changing terrain, correctly attacking bad guys while avoiding attacking good guys, and so on.
Of course, this is only one part of the system. There are a few thousand motion-captured full-body gestures used for each simulated actor to implement these desired goals; for example.
I have been extremely impressed, stunned even, by the astonishing pace of development on this movie. They've come up with shockingly great solutions to very hard problems, on a time scale that I wouldn't have thought possible. If you read the article, John Labrie says that he wishes that he had more time to plan, and didn't have to just react -- but I can't believe that there is any better way to make a movie. When the director-artist-programmer team really starts to hum, you can get amazing progress in a very short time.
One of the most chilling things that my father-in-law said about flying F-4s over Vietnam is that they figured out pretty quickly that being shot down and killed is not the worst thing that could happen to you.
Ouch.
ObDisclaimer -- my company did effects for a dozen shots in Behind Enemy Lines; stuff that you'll never know was effects, though.
At Hammerhead, we have been approached to do a character like this as well, call her Desiree [name changed to protect, well, me:)]. This group of producers wanted to create a pop star that didn't have all of the baggage that real people have; they could then choose the actual voice behind the character strictly on the basis of singing ability. As with these Japanese characters, Desiree would have a biography, would appear on radio interview shows, would endorse products, etc. With the recent advances in skin rendering, one could make a far more compelling image than the plastic-like characters to date.
The producers wanted the character to be 'racy' and 'revealing', like Britney -- something that I consider a tragic mistake. Perhaps as her career evolved over ten years or so I would think that she shoul d go that way, but I feel that flashy but modest clothes would be far more appealing over time than the same old skin. Clothes are adornment, not just censorship.
Personally, I don't see this as very different than Britney Spears. Britney is almost as synthetic as Desiree -- and at least Desiree would lip-synch competently.
This project drifted along for a while and finally died, as do 95% of all proposed projects. Still, it will definitely happen, the economics work. Desiree need only 'live' a few dozen minutes a year; and those appearances could be funded at a pretty reasonable rate.
This contrasts to what the poster above commented about porn stars. He commented that this would be an obvious venue, as you could build perfect bodies that would do anything. It seems to me that porn stars already have next-to-perfect bodies, and from what I can tell, there is very little that they won't do. More importantly, they are cheap. The most expensive full-length porn movies don't cost nearly as much as a synthetic music video would cost, say, $500,000. From what I've read, porn stars make most of their money outside of the films by performing live -- they treat the films as advertising for the live shows. Needless to say, this is beyond the capabilities of synthetic characters to this point.
Re:Why should the GPL be a problem?
on
Behind the Scenes
·
· Score: 2
I know that at Disney, for instance, using anything GPL'd is a problem. Disney is a rabid protector of its intellectual property*, and Disney's lawyers do not feel that they can guarantee with complete certainty that using GPL'd tools will not expose them to some threat to the intellectual property created with those tools at some time in the future.
I was surprised when I heard this, but I heard it directly from the developers at Disney and The Secret Lab (TSL is formerly DreamQuest, which Disney bought a few years ago to be their visual effects branch. TSL is scheduled to be shut down shortly as it has proven to be non-viable).
I did my interviews with Disney on this subject about a year and a half ago, and perhaps they've changed their minds since then, as Linux has become far more mainstream. But at least at the time, they were dead set against using Linux, GIMP, or any GPL'd product.
* I used the term 'intellectual property' holding my nose, only because that's how they say it.
Re:Once again Quake to the rescue
on
Behind the Scenes
·
· Score: 4, Insightful
What Quake really did was single-handedly get OpenGL through a critical couple of years. If it were not for Quake, and Carmack in particular, OpenGL would have been relegated to a few very expensive but low performance (because of low volume) cards; and Microsoft would have a monopoly on yet another field of computing.
I develop software for a visual effects company, and I give thanks every day to Carmack, Daryll Strauss, and the Fahrenheit gang for contributing in their own vital ways in keeping OpenGL viable.
John, I find the service that you provide as Cryptome to be essential. You remind me strongly of the title character in Vonnegut's short story
Report on the Barnhouse Effect. Your reporting keeps the entire world somewhat more honest; and I can't think that it's possible that governments are more careful knowing that someone is watching.
The end of the story, is, of course, of the passing of the torch to Barnhouse's apprentice. I am worried that there's nobody with the combination of integrity, fearlessness, and intelligence to carry on with your work, when your time to perform it is over. Do you worry about that, and are there people to carry the load?
This article by Michael Zyda is a remarkable document, and well worth reading. It is the single document from which this Insititute of Creative Technologies was created from.
I was part of the pitch that UCLA made to attempt to get this institute located at their campus. The idea was that UCLA's contacts with the visual effects community would be helpful in building these simulation games. Unfortunately for UCLA, their perennial cross-town rival USC made a better presentation.
It was weird to see a group of relatively high-level military people sitting around the conference table talking about writing next-generation video games; in dead earnest.
You can cluster IDE drives trivially, cheaply, and powerfully (but only for a limited time!) using 3ware's Escalade controller cards. This article on a cheap terabyte file server describes the card.
Basically, the Escalade cards make a bunch of cheap IDE drives look like a big SCSI drive. What could be better? You get the intelligence of SCSI, the protection of a RAID, at the price of IDE. With just a few IDE drives, you get scalding performance that more the beats the most expensive SCSI drives.
Sadly, 3ware has decided to get out of the controller card business. I've bought a couple of cards that I'm going to keep until I need to build some more file servers; they say that they are going to keep selling the cards until December, but only until then.
I am certain that there are tremendous efforts underway to find and interpret the data on the flight recorders of the airplanes. Rarely are suicide bombers recorded during their attacks, but these people should have been.
I cannot bring myself to believe that the captains of the planes were alive during these attacks, but I also don't understand how the planes could have been flown so precisely by amateurs -- unless the purchase of the jetliner simulator by that Saudi prince is not just the toy that it was taken to be. And, if the terrorists were schooled well in the airplanes, they would have known how to disable the flight recorders, unfortunately.
Actually, how I read this is that they were worried that somebody would try to use this software for production, and something would horribly fail, causing extreme duress. In the inevitable ensuing lawsuits, the Broadcast 2000 people would end up being sued into oblivion. It's not a nice way to go.
No amount of disclaimers and click-through agreements can keep these lawsuits from getting started, and once started they are incredible money-sinks.
Exactly this kind of thing happened to Burt Rutan, the designer of almost every interesting airplane over the last 20 years. His VariEze, and follow-on LongEZ were spectacular designs, but a few people built them poorly, died, and Burt was sued. He defended four of five of these lawsuits, and won every one, but decided that there were better ways to spend one's life, and pulled the plans off the market. In something parallel to what will happen here; there are xeroxed versions of the LongEZ plans out there if you really want them, in a samizdat kind of operation. Burt's current company, Scaled Composits continues to build exciting airplanes, but only for the corporate market.
In the early 90's, a huge percentage of the leading people in computer graphics research went to Microsoft Research. Some of the people involved were Jim Kajiya, Steve Gabriel, Andrew Glassner, and many others. They selected people who were prolific writers, wide ranging in their interests, and locked them away.
They made people rediculous offers to lure them away from their universities and other companies.
One example recruitment I heard went like this.
ring ring
Hello?
Hello, this is Microsoft Research, we'd like you to come work with us
Why should I? I'd never work for the great Satan. [thinking that this would make the caller hang up. But, what would Satan say? You got it...]
Well, what are your terms?
Ummm [trying to think of something completely unreasonable] How about $XXX.XXX [twice what he was getting then.]
Fine.
Ok, I want to work three months, then take a month off, work three months, take a month off...
We can't do that. How about this, you work for four years, then you get four years off at that same rate.
uhhhhhhh, well, ok.
When they set up the CG research group, they promised to have half the papers in Siggraph (the premier forum for computer graphics research) in a few years. This was a little scary, but not as scary as what really happened. What really happened is that these people pretty much stopped publishing at all; and stopped interacting with the rest of the graphics community.
I asked a few of the people there about it, and they seemed happy as clams, they weren't worried about it. To me, it appears that their world had shrunk to be just Microsoft. It's more than a pity, it's almost criminal.
The PrecisionInsight DRI drivers for the Radeon boards are great as far as they go -- which is to say they are great for 3D rendering. I've installed them on my Athlon box, and they perform as advertised (once I installed the MMTR bugfix for Linux 2.4.2).
Unfortunately, for 2D, the free linux drivers are terrible. I get around 13M pixels/sec glDrawPixels performance; while the closed source Xi drivers get ~80M pixels/sec; some 6 times as fast. The problem is that ATi didn't care to fund development of free high-performance 2D; so it didn't get done.
Perhaps it is surprising to some, but for many if not most visual effects applications, 2D performance is more important than 3D performance.
At this point, I would not recommend the ATi Radeon for visual effects applications for just this reason; and would recommend the nVidia cards which do have reasonably good free-driver 2D performance. I make this recommendation quite painfully, because I tremendously admire the work that the DRI team has done, it's just spectacular.
They started from a clean sheet of paper, and addressed all of the subtle issues involved in doing accelerated graphics in multiple windows, from context switching to security. Unfortunately, it's unclear whether that effort will lead to drivers that take full advantage of the cards. It is really quite sad.
Code Red II has a fighting chance of killing off
Code Red I, as it reboots machines that it finds.
So, it is partially a good thing; beyond the fact that it will probably convince a percentage of people to abandon Microsoft servers.
I have bought a couple of Dell laptops to run Linux over the last few months. They are truly wonderful machines, fast, reasonably rugged, breathtaking 1600x1200 screens.
I chose to get Windows ME instead of Dell's configuration of Linux, for two main reasons.
1) Configuration of linux is important, and not something that I'd give responsibility for to somebody else. Especially with laptops, there are dozens of little things that you'd like to configure yourself to get the machine to be 'just so'. Having somebody else do it the first time forces you to start from a somewhat unknown position. Better a completely blank machine than somebody else's configuration.
2) Windows ME is a good place to check out drivers and hardware. You can plug in your USB CompactFlash card reader, and see if the hardware is compatible (under Windows) before trying to dig up a driver (under Linux). I probably boot Windows at least once a month to do this; most recently when I was trying to get the Lucent winmodem working. (It does work, see http://www.linmodems.org )
I wouldn't be at all surprised if 95% of the Dell laptops running Linux were bought with Windows (it's no more expensive) for exactly these reasons.
This worm infected IIS servers, typical computers, relatively easily patchable. I would presume that this attack will lead to, among other things, mechanisms that automatically patch systems. These will just have to exist; as it is quite reasonable to expect that some worm-writers out there have systems ready to go just waiting for the next root-exploit. When the next one (and there will always be a next one!) is found, I expect that the exploit->attack interval will be remarkably smaller than a month.
This worm had a doubling time somewhere between 30 and 40 minutes, until it had compromised a significant proportion of the vulnerable machines. This time is proportional to the rate that a compromised machine can attack new ones, and to the ratio of vulnerable machines to the address space. I believe that the speed of attack cannot be increased much, this worm was remarkably efficient -- it will only be increased as the number of broadband-connected machines increases. The advent of six-octet IP addresses will, for a while, dramatically lower the percentage of vulnerable IPs compared to the address space, although that won't happen for a while.
What worries me is the advent of internet e-ppliances. These will probably not be patchable, and so, if infected, will remain so. Don't think that because machines don't have disk drives that they are invulnerable to worms, as this current worm didn't touch the disk at all, but stayed memory-resident. What's going to happen when a million e-refigerators start attacking root DNS servers, say?
I am in the process of building a fileserver for my little visual effects company. This article was extremely helpful and informative, although my solutions will be somewhat different from theirs.
In particular, they claim to get 50 MB/sec transfer rates from their disk array. But, you can see that they also specify just a 10 MB/sec 100BaseTX ethernet; so 80% of that bandwidth is completely wasted.
I was curious what people's solutions to this are. Does one just get multiple 100BaseTX ethernet connections to a switch, or are people going Gigabit ethernet from the fileserver to the switch?
I plan to go the second route. There are many ethernet switches out there with one or two gigabit ports such as this 3COM switch. This should give 100MB/sec to the switch, and then the switch can distribute this bandwidth to all the client machines at 10MB/sec.
At least this is a reasonable stopgap until Gigabit Ethernet is more ubiquitous.
Windows XP is an incredible step forward for end users and partners, unlocking the possibilities of the digital world," said Jim Allchin, group vice president for platforms at Microsoft.
It's just insane to say something like this, Windows XP is just a small step in terms of usability. And it's more about locking in the possibilities [for microsoft] than unlocking them...
Joel Seigel once did a review of "Benji" on the air, panning it in every way he could. Then, as they were going to a commercial he said "and now, it's time for this message."
The movie poster quoted him "It's Time for this Message!"
One of the interesting things about movie-making is that you want people to see your work as high-quality, not that it actually be "right". Back at PDI we were doing a commercial for a Japanese snack food company, and we were asked to sync it to an English soundtrack. This was quite confusing to us; as the animation would only ever be shown on Japanese TV, in Japanese.
The clients reiterated that it should be sync-ed to English. When we pointed out that this wouldn't the Japanese dialogue, they said "Of course, all high-quality animation has lip-sync that doesn't match the audio". At least in the mind of these clients, high-quality animation meant Disney animation, and they specifically wanted the lip-sync not to work 'correctly'.
There are many other instances of this; for instance doing animation at 24 fps instead of 60 fps and adding film grain to pristine animation.
The first live-action color films used a similar
process; the Technicolor cameras had three strips of black-and-white film, with filters to separate the colors. These were huge cantankerous beasts, but they gave us color in film.
One extremely benificial aspect of this is that black-and-white film is extremely stable compared to color film. Black and white film uses metallic silver (expensive yes, but stable) and is an archival medium, whereas modern color films use dyes that are extremely quick to fade and degrade. Even a film as recent as the first Star Wars movie required extensive cleanup to restore to its original colors.
Some early computer graphic films used black-and-white images, step-printed (one frame of red, then green, then blue) with the colors combined later in an optical printer. My first computer graphics effects (Solar Crisis) was step-printed with the color image followed by the opacity image. These were again used on an optical printer to merge the CG with the live action.
It turns out, of course, that if the earth slows its rotation due to tidal drag (increased infinitesimally by tidal power stations) then the moon will move further out to preserve angular momentum.
Interestingly, we have slowed the rate of decrease of the earth's rotation measurably from hydropower. Hydropower dams are typically at reasonably high latitudes, and a lot of water has been stored there; closer to the axis of the earth than it would have been otherwise. Some calculations show that we'd have had dozens of more leap-seconds had this not been happening.
Global warming, though, will throw this all for a serious loop. If the icecaps continue to shrink and the sea level rises a few to ten meters; truly massive amounts of water will move further from the axis, slowing the earth by many seconds/year. I have no idea what the effect of this might be.
Slashdot Mirror
This turns out to be a real problem in the visual effects community -- a lot of our work depends on having good, fast 2D. Film frames, after all, are still just 2D images.
I hope that this new driver they speak of, and future drivers, recognize this. We'll see. At this point, we have to by nVidia boards, even while the drivers are closed-source, they provide pretty good 2D support.
thad
Wakko Warner sez:
I ask because Linux is a very bad choice, if that's what you're planning on using. How many professional 3D packages are available for Linux?
The two most important high-end 3D packages are Maya and Houdini. Both are available, shipping now, in complete supported implementations, under Linux.
Every single high-end visual effects and animation facility is using Linux for render farms, and almost all of them are pushing it very strongly on the desktop. For example, Pacific Data Images (you know, the people that did Shrek) have completely moved over to Linux and abandoned their previous platforms.
thad
The way that 'Massive' works is ingenious. A statement of the battle simulation problem is "How do you get all of these simulated actors to respond correctly to their individual environment?"
The solution employed at WETA is to use state-of-the-art graphics hardware to render the view from the eyes of each simulated actor. He then plans his actions based on this view, correctly placing his feet on the changing terrain, correctly attacking bad guys while avoiding attacking good guys, and so on.
Of course, this is only one part of the system. There are a few thousand motion-captured full-body gestures used for each simulated actor to implement these desired goals; for example.
I have been extremely impressed, stunned even, by the astonishing pace of development on this movie. They've come up with shockingly great solutions to very hard problems, on a time scale that I wouldn't have thought possible. If you read the article, John Labrie says that he wishes that he had more time to plan, and didn't have to just react -- but I can't believe that there is any better way to make a movie. When the director-artist-programmer team really starts to hum, you can get amazing progress in a very short time.
thad
One of the most chilling things that my father-in-law said about flying F-4s over Vietnam is that they figured out pretty quickly that being shot down and killed is not the worst thing that could happen to you.
Ouch.
ObDisclaimer -- my company did effects for a dozen shots in Behind Enemy Lines; stuff that you'll never know was effects, though.
At Hammerhead, we have been approached to do a character like this as well, call her Desiree [name changed to protect, well, me :)]. This group of producers wanted to create a pop star that didn't have all of the baggage that real people have; they could then choose the actual voice behind the character strictly on the basis of singing ability. As with these Japanese characters, Desiree would have a biography, would appear on radio interview shows, would endorse products, etc. With the recent advances in skin rendering, one could make a far more compelling image than the plastic-like characters to date.
The producers wanted the character to be 'racy' and 'revealing', like Britney -- something that I consider a tragic mistake. Perhaps as her career evolved over ten years or so I would think that she shoul d go that way, but I feel that flashy but modest clothes would be far more appealing over time than the same old skin. Clothes are adornment, not just censorship.
Personally, I don't see this as very different than Britney Spears. Britney is almost as synthetic as Desiree -- and at least Desiree would lip-synch competently.
This project drifted along for a while and finally died, as do 95% of all proposed projects. Still, it will definitely happen, the economics work. Desiree need only 'live' a few dozen minutes a year; and those appearances could be funded at a pretty reasonable rate.
This contrasts to what the poster above commented about porn stars. He commented that this would be an obvious venue, as you could build perfect bodies that would do anything. It seems to me that porn stars already have next-to-perfect bodies, and from what I can tell, there is very little that they won't do. More importantly, they are cheap. The most expensive full-length porn movies don't cost nearly as much as a synthetic music video would cost, say, $500,000. From what I've read, porn stars make most of their money outside of the films by performing live -- they treat the films as advertising for the live shows. Needless to say, this is beyond the capabilities of synthetic characters to this point.
I know that at Disney, for instance, using anything GPL'd is a problem. Disney is a rabid protector of its intellectual property*, and Disney's lawyers do not feel that they can guarantee with complete certainty that using GPL'd tools will not expose them to some threat to the intellectual property created with those tools at some time in the future.
I was surprised when I heard this, but I heard it directly from the developers at Disney and The Secret Lab (TSL is formerly DreamQuest, which Disney bought a few years ago to be their visual effects branch. TSL is scheduled to be shut down shortly as it has proven to be non-viable).
I did my interviews with Disney on this subject about a year and a half ago, and perhaps they've changed their minds since then, as Linux has become far more mainstream. But at least at the time, they were dead set against using Linux, GIMP, or any GPL'd product.
* I used the term 'intellectual property' holding my nose, only because that's how they say it.
What Quake really did was single-handedly get OpenGL through a critical couple of years. If it were not for Quake, and Carmack in particular, OpenGL would have been relegated to a few very expensive but low performance (because of low volume) cards; and Microsoft would have a monopoly on yet another field of computing.
I develop software for a visual effects company, and I give thanks every day to Carmack, Daryll Strauss, and the Fahrenheit gang for contributing in their own vital ways in keeping OpenGL viable.
thad
John, I find the service that you provide as Cryptome to be essential. You remind me strongly of the title character in Vonnegut's short story
Report on the Barnhouse Effect. Your reporting keeps the entire world somewhat more honest; and I can't think that it's possible that governments are more careful knowing that someone is watching.
The end of the story, is, of course, of the passing of the torch to Barnhouse's apprentice. I am worried that there's nobody with the combination of integrity, fearlessness, and intelligence to carry on with your work, when your time to perform it is over. Do you worry about that, and are there people to carry the load?
thad
Plam sez: > ... which is fishy, because I'm
sitting here in my office on the sixth floor of
the Laboratory for Computer Science...
Perhaps this is fiscal April Fools Day? Those fiscal holidays are so hard to keep track of.
thad
I was part of the pitch that UCLA made to attempt to get this institute located at their campus. The idea was that UCLA's contacts with the visual effects community would be helpful in building these simulation games. Unfortunately for UCLA, their perennial cross-town rival USC made a better presentation.
It was weird to see a group of relatively high-level military people sitting around the conference table talking about writing next-generation video games; in dead earnest.
thad
My brother lives with his boyfriend in Berkeley. This is his favorite script, from when he worked at home:
ring ring
Hello, this is ABC company. Is Mr. Caner in?
[imagine his deep voice] No, Mr. Caner is not in.
Oh, then can I speak to Mrs. Caner?
Speaking
[caller gets perplexed, always hangs up]
thad
Basically, the Escalade cards make a bunch of cheap IDE drives look like a big SCSI drive. What could be better? You get the intelligence of SCSI, the protection of a RAID, at the price of IDE. With just a few IDE drives, you get scalding performance that more the beats the most expensive SCSI drives.
Sadly, 3ware has decided to get out of the controller card business. I've bought a couple of cards that I'm going to keep until I need to build some more file servers; they say that they are going to keep selling the cards until December, but only until then.
thad
I am certain that there are tremendous efforts underway to find and interpret the data on the flight recorders of the airplanes. Rarely are suicide bombers recorded during their attacks, but these people should have been.
I cannot bring myself to believe that the captains of the planes were alive during these attacks, but I also don't understand how the planes could have been flown so precisely by amateurs -- unless the purchase of the jetliner simulator by that Saudi prince is not just the toy that it was taken to be. And, if the terrorists were schooled well in the airplanes, they would have known how to disable the flight recorders, unfortunately.
We'll see.
thad
No amount of disclaimers and click-through agreements can keep these lawsuits from getting started, and once started they are incredible money-sinks.
Exactly this kind of thing happened to Burt Rutan, the designer of almost every interesting airplane over the last 20 years. His VariEze, and follow-on LongEZ were spectacular designs, but a few people built them poorly, died, and Burt was sued. He defended four of five of these lawsuits, and won every one, but decided that there were better ways to spend one's life, and pulled the plans off the market. In something parallel to what will happen here; there are xeroxed versions of the LongEZ plans out there if you really want them, in a samizdat kind of operation. Burt's current company, Scaled Composits continues to build exciting airplanes, but only for the corporate market.
thad
They made people rediculous offers to lure them away from their universities and other companies.
One example recruitment I heard went like this.
ring ring
Hello?
Hello, this is Microsoft Research, we'd like you to come work with us
Why should I? I'd never work for the great Satan. [thinking that this would make the caller hang up. But, what would Satan say? You got it...]
Well, what are your terms?
Ummm [trying to think of something completely unreasonable] How about $XXX.XXX [twice what he was getting then.]
Fine.
Ok, I want to work three months, then take a month off, work three months, take a month off...
We can't do that. How about this, you work for four years, then you get four years off at that same rate.
uhhhhhhh, well, ok.
When they set up the CG research group, they promised to have half the papers in Siggraph (the premier forum for computer graphics research) in a few years. This was a little scary, but not as scary as what really happened. What really happened is that these people pretty much stopped publishing at all; and stopped interacting with the rest of the graphics community.
I asked a few of the people there about it, and they seemed happy as clams, they weren't worried about it. To me, it appears that their world had shrunk to be just Microsoft. It's more than a pity, it's almost criminal.
thad
Unfortunately, for 2D, the free linux drivers are terrible. I get around 13M pixels/sec glDrawPixels performance; while the closed source Xi drivers get ~80M pixels/sec; some 6 times as fast. The problem is that ATi didn't care to fund development of free high-performance 2D; so it didn't get done.
Perhaps it is surprising to some, but for many if not most visual effects applications, 2D performance is more important than 3D performance.
At this point, I would not recommend the ATi Radeon for visual effects applications for just this reason; and would recommend the nVidia cards which do have reasonably good free-driver 2D performance. I make this recommendation quite painfully, because I tremendously admire the work that the DRI team has done, it's just spectacular. They started from a clean sheet of paper, and addressed all of the subtle issues involved in doing accelerated graphics in multiple windows, from context switching to security. Unfortunately, it's unclear whether that effort will lead to drivers that take full advantage of the cards. It is really quite sad.
thad
thad
I chose to get Windows ME instead of Dell's configuration of Linux, for two main reasons.
1) Configuration of linux is important, and not something that I'd give responsibility for to somebody else. Especially with laptops, there are dozens of little things that you'd like to configure yourself to get the machine to be 'just so'. Having somebody else do it the first time forces you to start from a somewhat unknown position. Better a completely blank machine than somebody else's configuration.
2) Windows ME is a good place to check out drivers and hardware. You can plug in your USB CompactFlash card reader, and see if the hardware is compatible (under Windows) before trying to dig up a driver (under Linux). I probably boot Windows at least once a month to do this; most recently when I was trying to get the Lucent winmodem working. (It does work, see http://www.linmodems.org )
I wouldn't be at all surprised if 95% of the Dell laptops running Linux were bought with Windows (it's no more expensive) for exactly these reasons.
thad
This worm had a doubling time somewhere between 30 and 40 minutes, until it had compromised a significant proportion of the vulnerable machines. This time is proportional to the rate that a compromised machine can attack new ones, and to the ratio of vulnerable machines to the address space. I believe that the speed of attack cannot be increased much, this worm was remarkably efficient -- it will only be increased as the number of broadband-connected machines increases. The advent of six-octet IP addresses will, for a while, dramatically lower the percentage of vulnerable IPs compared to the address space, although that won't happen for a while.
What worries me is the advent of internet e-ppliances. These will probably not be patchable, and so, if infected, will remain so. Don't think that because machines don't have disk drives that they are invulnerable to worms, as this current worm didn't touch the disk at all, but stayed memory-resident. What's going to happen when a million e-refigerators start attacking root DNS servers, say?
thad
In particular, they claim to get 50 MB/sec transfer rates from their disk array. But, you can see that they also specify just a 10 MB/sec 100BaseTX ethernet; so 80% of that bandwidth is completely wasted.
I was curious what people's solutions to this are. Does one just get multiple 100BaseTX ethernet connections to a switch, or are people going Gigabit ethernet from the fileserver to the switch?
I plan to go the second route. There are many ethernet switches out there with one or two gigabit ports such as this 3COM switch. This should give 100MB/sec to the switch, and then the switch can distribute this bandwidth to all the client machines at 10MB/sec.
At least this is a reasonable stopgap until Gigabit Ethernet is more ubiquitous.
Any comments?
thad
Windows XP is an incredible step forward for end users and partners, unlocking the possibilities of the digital world," said Jim Allchin, group vice president for platforms at Microsoft.
It's just insane to say something like this, Windows XP is just a small step in terms of usability. And it's more about locking in the possibilities [for microsoft] than unlocking them...
thad
The movie poster quoted him "It's Time for this Message!"
I'm not making this up.
thad
The clients reiterated that it should be sync-ed to English. When we pointed out that this wouldn't the Japanese dialogue, they said "Of course, all high-quality animation has lip-sync that doesn't match the audio". At least in the mind of these clients, high-quality animation meant Disney animation, and they specifically wanted the lip-sync not to work 'correctly'.
There are many other instances of this; for instance doing animation at 24 fps instead of 60 fps and adding film grain to pristine animation.
thad
One extremely benificial aspect of this is that black-and-white film is extremely stable compared to color film. Black and white film uses metallic silver (expensive yes, but stable) and is an archival medium, whereas modern color films use dyes that are extremely quick to fade and degrade. Even a film as recent as the first Star Wars movie required extensive cleanup to restore to its original colors.
Some early computer graphic films used black-and-white images, step-printed (one frame of red, then green, then blue) with the colors combined later in an optical printer. My first computer graphics effects (Solar Crisis) was step-printed with the color image followed by the opacity image. These were again used on an optical printer to merge the CG with the live action.
thad
Interestingly, we have slowed the rate of decrease of the earth's rotation measurably from hydropower. Hydropower dams are typically at reasonably high latitudes, and a lot of water has been stored there; closer to the axis of the earth than it would have been otherwise. Some calculations show that we'd have had dozens of more leap-seconds had this not been happening.
Global warming, though, will throw this all for a serious loop. If the icecaps continue to shrink and the sea level rises a few to ten meters; truly massive amounts of water will move further from the axis, slowing the earth by many seconds/year. I have no idea what the effect of this might be.
thad