CGI becomes reality
March 14, 2012 8:49 PM Subscribe
Pipe Dream [2001, CGI (previously)]
Pipe Dream [2011, physical (another video) (yet another)] [MLYT]
Pipe Dream [2011, physical (another video) (yet another)] [MLYT]
It's a shame that the air jets in the physical version are so loud that they interfere with the music. Pretty awesome otherwise.
posted by Pope Guilty at 9:03 PM on March 14, 2012 [1 favorite]
posted by Pope Guilty at 9:03 PM on March 14, 2012 [1 favorite]
Amusing enough, but one wonders how much better and cheaper their processors would be without spending resources on such projects, just for some geek cred.
AMD 4 Lyfe, yo!
posted by Samizdata at 9:15 PM on March 14, 2012
AMD 4 Lyfe, yo!
posted by Samizdata at 9:15 PM on March 14, 2012
It's a shame that the air jets in the physical version are so loud that they interfere with the music.
False. The air jets make it more awesome. Especially during the marimba bits. It sounds exactly like a robot band should sound like: Beautiful composition, with an overtone of machine gun.
posted by Popular Ethics at 9:27 PM on March 14, 2012 [5 favorites]
False. The air jets make it more awesome. Especially during the marimba bits. It sounds exactly like a robot band should sound like: Beautiful composition, with an overtone of machine gun.
posted by Popular Ethics at 9:27 PM on March 14, 2012 [5 favorites]
I have to say, this is some pretty cool corporate art.
posted by Strass at 9:30 PM on March 14, 2012 [2 favorites]
posted by Strass at 9:30 PM on March 14, 2012 [2 favorites]
So is this Intel just coming out and admitting that their GMA chips are so lackluster that you may as well just build a physical model of whatever you're trying to render instead?
posted by Uncle Ira at 9:32 PM on March 14, 2012 [13 favorites]
posted by Uncle Ira at 9:32 PM on March 14, 2012 [13 favorites]
P.S. If anyone's wondering about the discrepancy between the date in my post for the CGI version (2001) and the date visible on the Intel animation at the beginning of the first linked physical video (2004), IMDB lists 2001 as the original release date (with a re-release in 2004) so I went with that.
posted by DevilsAdvocate at 9:32 PM on March 14, 2012 [3 favorites]
posted by DevilsAdvocate at 9:32 PM on March 14, 2012 [3 favorites]
I can't believe there are actually microprocessor fanboys.
posted by crunchland at 9:36 PM on March 14, 2012 [4 favorites]
posted by crunchland at 9:36 PM on March 14, 2012 [4 favorites]
crunchland: "I can't believe there are actually microprocessor fanboys."
Yeah.
And?
I can cut ya, ya know...
posted by Samizdata at 9:41 PM on March 14, 2012 [2 favorites]
Yeah.
And?
I can cut ya, ya know...
posted by Samizdata at 9:41 PM on March 14, 2012 [2 favorites]
one wonders how much better and cheaper their processors would be without spending resources on such projects, just for some geek cred
Well, if they didn't do any marketing, people wouldn't buy as much of their stuff, which would mean less money for the engineering department. Add to that the fact that necessity, as imposed by the need to complete a project, is often the mother of invention and may yield useful feedback about the limitations of the existing hardware.
Lest I be accused of fanboyism, my main machine is an AMD quad-core. I tend to alternate between the two manufacturer.
posted by anigbrowl at 10:41 PM on March 14, 2012 [1 favorite]
Well, if they didn't do any marketing, people wouldn't buy as much of their stuff, which would mean less money for the engineering department. Add to that the fact that necessity, as imposed by the need to complete a project, is often the mother of invention and may yield useful feedback about the limitations of the existing hardware.
Lest I be accused of fanboyism, my main machine is an AMD quad-core. I tend to alternate between the two manufacturer.
posted by anigbrowl at 10:41 PM on March 14, 2012 [1 favorite]
I call bullshit. These kinds of trajectories are notoriously sensitive to wind currents and other factors, but the video didn't show any mistakes. I think if you look carefully, ball 54 actually missed its mark, but the lighted bar still fired. Also, balls 267 through 293 were out of synch with the soundtrack. The balls are lob-synching.
posted by twoleftfeet at 10:53 PM on March 14, 2012 [4 favorites]
posted by twoleftfeet at 10:53 PM on March 14, 2012 [4 favorites]
anigbrowl: "one wonders how much better and cheaper their processors would be without spending resources on such projects, just for some geek cred
Well, if they didn't do any marketing, people wouldn't buy as much of their stuff, which would mean less money for the engineering department. Add to that the fact that necessity, as imposed by the need to complete a project, is often the mother of invention and may yield useful feedback about the limitations of the existing hardware.
Lest I be accused of fanboyism, my main machine is an AMD quad-core. I tend to alternate between the two manufacturer."
Okay, then, I will give you that. I will also give you that I am an AMD fan boy and have been for years, truthfully.
posted by Samizdata at 11:43 PM on March 14, 2012
Well, if they didn't do any marketing, people wouldn't buy as much of their stuff, which would mean less money for the engineering department. Add to that the fact that necessity, as imposed by the need to complete a project, is often the mother of invention and may yield useful feedback about the limitations of the existing hardware.
Lest I be accused of fanboyism, my main machine is an AMD quad-core. I tend to alternate between the two manufacturer."
Okay, then, I will give you that. I will also give you that I am an AMD fan boy and have been for years, truthfully.
posted by Samizdata at 11:43 PM on March 14, 2012
Oh, come on. The instruments aren't real; they're just plastic disks with LEDs and (presumably) pressure sensors. I mean hooray for rhythmically spitting plastic balls at plastic targets, but if you're just going to synthesize the music anyway, why bother?
tl;dr: Needs more cowbell.
posted by erniepan at 2:17 AM on March 15, 2012 [2 favorites]
tl;dr: Needs more cowbell.
posted by erniepan at 2:17 AM on March 15, 2012 [2 favorites]
if you're just going to synthesize the music anyway, why bother?
Have you used the Internet? Do we really need to explain the directly proportional relationship between the pointlessness of doing something and how awesome it is to do it? Steampunk wine-openers? The ketchup robot? Manned spaceflight?
posted by bicyclefish at 2:37 AM on March 15, 2012 [1 favorite]
Have you used the Internet? Do we really need to explain the directly proportional relationship between the pointlessness of doing something and how awesome it is to do it? Steampunk wine-openers? The ketchup robot? Manned spaceflight?
posted by bicyclefish at 2:37 AM on March 15, 2012 [1 favorite]
Do we really need to explain the directly proportional relationship between the pointlessness of doing something and how awesome it is to do it? Steampunk wine-openers? The ketchup robot? Manned spaceflight?
I think you proved erniepan's point (which you seem to have missed) rather than countered it.
posted by DU at 4:43 AM on March 15, 2012 [2 favorites]
I think you proved erniepan's point (which you seem to have missed) rather than countered it.
posted by DU at 4:43 AM on March 15, 2012 [2 favorites]
I'll guess they considered making it work with regular instruments but the cost would have gone way, way up. So they cut corners. A lot of corners. Corners that were, at least for me, the entire point of the thing. But that's okay. Some people got paid to make a cool thing which could be called art, albeit fairly tepid and watered down art. And I'm on board with that!
posted by seanmpuckett at 4:49 AM on March 15, 2012 [1 favorite]
posted by seanmpuckett at 4:49 AM on March 15, 2012 [1 favorite]
Was it just me, or did it seem like they were only using sensors to detect when the balls hit the panels, rather then actually having something that would resonate. If they were actually hitting all their targets it's an impressive use of robotics in order to fire those balls so precisely - I think building a physical device like that would actually be a lot harder then you would first think.
But still, from the video it's impossible to tell if it's actually working, the balls don't show up very well. Maybe they should have used black-lights so that the balls would glow and you could see them on video. Of course this was really about showing off to people at a conference, not for video.
I was really into reading about PC hardware advances when I was in high-school in the 90s. It's gotten kind of boring since then though, CPU speeds totally stagnated, we've gone from 2.5 to maybe 3.6* or so in 10 years. On the other hand '90-2000 saw a hundred fold increase, from 25mhz to 2.5ghz
(We have more cores now, but in the 90s the number of transistors was also going up as well, and more transistors let you do more per clock. Early x86 chips took several cycles to do one instruction, while modern ones can crunch through several in one)
The irony is that the actual top speed of CPUs was always still going up, it's just that the energy output per chip was so high at room temperature that the only way you can get to those speeds was to use liquid nitrogen.
*Actually, I just looked online and the fasted chip you can buy at this point is actually 4.2ghz, which is actually a pretty big jump over just a couple months ago - I suppose it has to do with the reduced power consumption that we've seen lately.
posted by delmoi at 5:01 AM on March 15, 2012
But still, from the video it's impossible to tell if it's actually working, the balls don't show up very well. Maybe they should have used black-lights so that the balls would glow and you could see them on video. Of course this was really about showing off to people at a conference, not for video.
Amusing enough, but one wonders how much better and cheaper their processors would be without spending resources on such projects, just for some geek cred.You mean Holding events where they overlcock their chips with liquid helium [2] wasn't done for the geek cred!!? (this is the most recent one, 8.42ghz)
AMD 4 Lyfe, yo!
I was really into reading about PC hardware advances when I was in high-school in the 90s. It's gotten kind of boring since then though, CPU speeds totally stagnated, we've gone from 2.5 to maybe 3.6* or so in 10 years. On the other hand '90-2000 saw a hundred fold increase, from 25mhz to 2.5ghz
(We have more cores now, but in the 90s the number of transistors was also going up as well, and more transistors let you do more per clock. Early x86 chips took several cycles to do one instruction, while modern ones can crunch through several in one)
The irony is that the actual top speed of CPUs was always still going up, it's just that the energy output per chip was so high at room temperature that the only way you can get to those speeds was to use liquid nitrogen.
*Actually, I just looked online and the fasted chip you can buy at this point is actually 4.2ghz, which is actually a pretty big jump over just a couple months ago - I suppose it has to do with the reduced power consumption that we've seen lately.
posted by delmoi at 5:01 AM on March 15, 2012
Is it just me, or did it seem like they were only using sensors to detect when the balls hit the panels, rather then actually having something that would resonate.
I don't even think they were doing that. I was paying attention to the cowbell, and it didn't seem to have anything to do with the sound at all. I suspect it might simply be recorded music playing alongside independently scripted ball firing alongside independently scripted light flashing.
It would be more believable if there were any mistakes or misfires.
posted by CaseyB at 6:13 AM on March 15, 2012 [3 favorites]
I don't even think they were doing that. I was paying attention to the cowbell, and it didn't seem to have anything to do with the sound at all. I suspect it might simply be recorded music playing alongside independently scripted ball firing alongside independently scripted light flashing.
It would be more believable if there were any mistakes or misfires.
posted by CaseyB at 6:13 AM on March 15, 2012 [3 favorites]
yeah, I thought at first that it was all just drum triggers as well. There is no way that a plastic marimba sounds that close to a midi guitar sound. Then I noticed how poorly the cowbell assembly was spinning, and realize that this is essentially a lip sync. I feel kinda bad for whatever technology art company got roped into making this, and then couldn't actually get it under budget.
A similar thing could be done with humans dressed as robots throwing ping pong balls at pie plates while the soundtrack for the original plays in the background. Would probably be slightly cheaper, and for me, a more valuable video.
posted by jonbro at 6:19 AM on March 15, 2012 [1 favorite]
A similar thing could be done with humans dressed as robots throwing ping pong balls at pie plates while the soundtrack for the original plays in the background. Would probably be slightly cheaper, and for me, a more valuable video.
posted by jonbro at 6:19 AM on March 15, 2012 [1 favorite]
Has enough cowbell.
posted by obscurator at 7:12 AM on March 15, 2012 [1 favorite]
posted by obscurator at 7:12 AM on March 15, 2012 [1 favorite]
and the fastest chip you can buy at this point is actually 4.2ghz,
Nearly all of the 2500Ks can be overclocked in Turbo Mode to 4.4Ghz, and about half will go even further. My BIOS claims that this is only for limited periods of time, but in fairly extended testing on my 2600K (same thing, just with hyperthreading), I don't see where the Turbo Mode ever turns itself off, and the chips are incredibly fast.
I don't even have to add voltage to my chip to get 4.4Ghz. One core runs hotter than the others, so I presume it would fail first if I kept pushing, but at present, my idle temps are 30/40/30/30, and my extended load temps are around 68/75/68/68. And, again, that's with no extra voltage, just setting the Turbo Multiplier to 44, and setting the heat-limiting section of the BIOS up to 130W from the stock 95. (Many chips do require a little more voltage to comfortably hit 4.4, though, so that can add some complexity.)
You want an aftermarket cooler to do this, as the stock unit doesn't handle much extra heat, but you don't need anything exotic. A nice big hunk of aluminum or copper sitting on top of the CPU, with a fan, will do nicely. The Cooler Master Hyper 212s are like $25, and work fine. If you spend more, you can get a better fan -- the 212s come with a sleeve bearing fan, which fail sooner than ball bearings, or the high-tech fluid bearings.
You can, in other words, get an astonishingly fast chip on the cheap, if you're willing to spend some time fiddling with it... and the fiddling is easier than it used to be. Just make sure to get one of the K units, as most other Intel chips are multiplier-locked.
posted by Malor at 8:25 AM on March 15, 2012
Nearly all of the 2500Ks can be overclocked in Turbo Mode to 4.4Ghz, and about half will go even further. My BIOS claims that this is only for limited periods of time, but in fairly extended testing on my 2600K (same thing, just with hyperthreading), I don't see where the Turbo Mode ever turns itself off, and the chips are incredibly fast.
I don't even have to add voltage to my chip to get 4.4Ghz. One core runs hotter than the others, so I presume it would fail first if I kept pushing, but at present, my idle temps are 30/40/30/30, and my extended load temps are around 68/75/68/68. And, again, that's with no extra voltage, just setting the Turbo Multiplier to 44, and setting the heat-limiting section of the BIOS up to 130W from the stock 95. (Many chips do require a little more voltage to comfortably hit 4.4, though, so that can add some complexity.)
You want an aftermarket cooler to do this, as the stock unit doesn't handle much extra heat, but you don't need anything exotic. A nice big hunk of aluminum or copper sitting on top of the CPU, with a fan, will do nicely. The Cooler Master Hyper 212s are like $25, and work fine. If you spend more, you can get a better fan -- the 212s come with a sleeve bearing fan, which fail sooner than ball bearings, or the high-tech fluid bearings.
You can, in other words, get an astonishingly fast chip on the cheap, if you're willing to spend some time fiddling with it... and the fiddling is easier than it used to be. Just make sure to get one of the K units, as most other Intel chips are multiplier-locked.
posted by Malor at 8:25 AM on March 15, 2012
Oh, and re:AMD ... their current-gen Bulldozer chips are a disaster in all respects. They're just wrecked by Sandy Bridge in every category... absolute performance, absolute heat, price/performance, performance/heat, price/heat. There are precisely zero significant markets I'm aware of where a Bulldozer chip is a better bet than one of the Sandy Bridge variants. They're completely uncompetitive, which frustrates me, as Intel price-gouges whenever they're in the catbird seat.
AMD's last-gen Phenom IIs were actually a much better design -- it's not often that a company spends billions to go backwards, but that's precisely what AMD has done. And they don't make Phenom IIs anymore, so you can't actually get a decent AMD CPU. (I saw a claim that they're considering repackaging low-end Bulldozers as Phenom IIs to try to trick people into buying CPUs they think are good, so beware any Phenom II you see that's new.)
Bulldozers overclock like crazy, but they also suck down insane amounts of power to do it. You'll have to sink something like 300 watts into a BD to get really good performance out of it, and then deal with the cooling issues that implies. And then you can just counter by overclocking one of the Intel K chips, which will run substantially faster, while emitting less than half the heat to do it.
Their graphic cards aren't any better... they just started shipping a new process. Using 65% more transistors than the last generation, they're only 15%ish faster than their last-gen NVidia counterparts, and priced like they shipped in late 2009. I fully expect NVidia's new cards, when they ship, to issue a truly righteous beatdown.
AMD has stated now, on multiple occasions, that they are focusing primarily on using chip-design tools to make their processors, instead of humans. And their recent designs look like something a computer program would make -- fast on paper, but impractical in the real world because of heat issues. Basically, AMD is trying to outsource their chip design work, not to other countries, but to computer programs. Their management appears to be completely, absolutely inept, people who should not be allowed anywhere near a tech company. They are producing shit silicon, and unless something big changes, I expect them to be entirely dead, instead of mostly dead, within two or three years.
And I'm not anyone's fanboi, btw. I've run a lot of AMD chips over the years. But about the time Intel shipped the first Core processor, they started falling behind. The Phenom II was a good low-cost option, especially good if you wanted lots of features on your motherboard, but those are gone now. And Bulldozer is so awful that the only scenario I've been able to imagine where you might want one is if you A) want a bunch of threads, B) need a low up-front cost, and C) don't care about power consumption/run costs. I'm sure there must be at least a few people out there that fit this scenario, but that implies a strings-and-bailing-wire, low capitalization approach to computing needs, which in turn implies low volume, because they'll be too poor to afford more than a few chips. This strikes me as an exceptionally poor target market for a large chipmaker.
posted by Malor at 9:03 AM on March 15, 2012 [3 favorites]
AMD's last-gen Phenom IIs were actually a much better design -- it's not often that a company spends billions to go backwards, but that's precisely what AMD has done. And they don't make Phenom IIs anymore, so you can't actually get a decent AMD CPU. (I saw a claim that they're considering repackaging low-end Bulldozers as Phenom IIs to try to trick people into buying CPUs they think are good, so beware any Phenom II you see that's new.)
Bulldozers overclock like crazy, but they also suck down insane amounts of power to do it. You'll have to sink something like 300 watts into a BD to get really good performance out of it, and then deal with the cooling issues that implies. And then you can just counter by overclocking one of the Intel K chips, which will run substantially faster, while emitting less than half the heat to do it.
Their graphic cards aren't any better... they just started shipping a new process. Using 65% more transistors than the last generation, they're only 15%ish faster than their last-gen NVidia counterparts, and priced like they shipped in late 2009. I fully expect NVidia's new cards, when they ship, to issue a truly righteous beatdown.
AMD has stated now, on multiple occasions, that they are focusing primarily on using chip-design tools to make their processors, instead of humans. And their recent designs look like something a computer program would make -- fast on paper, but impractical in the real world because of heat issues. Basically, AMD is trying to outsource their chip design work, not to other countries, but to computer programs. Their management appears to be completely, absolutely inept, people who should not be allowed anywhere near a tech company. They are producing shit silicon, and unless something big changes, I expect them to be entirely dead, instead of mostly dead, within two or three years.
And I'm not anyone's fanboi, btw. I've run a lot of AMD chips over the years. But about the time Intel shipped the first Core processor, they started falling behind. The Phenom II was a good low-cost option, especially good if you wanted lots of features on your motherboard, but those are gone now. And Bulldozer is so awful that the only scenario I've been able to imagine where you might want one is if you A) want a bunch of threads, B) need a low up-front cost, and C) don't care about power consumption/run costs. I'm sure there must be at least a few people out there that fit this scenario, but that implies a strings-and-bailing-wire, low capitalization approach to computing needs, which in turn implies low volume, because they'll be too poor to afford more than a few chips. This strikes me as an exceptionally poor target market for a large chipmaker.
posted by Malor at 9:03 AM on March 15, 2012 [3 favorites]
Notably, this is by the same person who created this bit of nostalgia for the Canadian kids of the 90s
posted by DareTo at 5:18 PM on March 15, 2012 [1 favorite]
posted by DareTo at 5:18 PM on March 15, 2012 [1 favorite]
Metafilter: Where microprocessor fanboys meet.
posted by Chuffy at 6:08 PM on March 15, 2012 [1 favorite]
posted by Chuffy at 6:08 PM on March 15, 2012 [1 favorite]
« Older The Forrest Gump of Opera | RL in '78, TJ in '83 Newer »
This thread has been archived and is closed to new comments
I just forwarded him a link to the new video and I'm looking forward to his gloating vindication.
posted by crunchland at 9:01 PM on March 14, 2012 [6 favorites]