I don't understand what this 'waste of resources' is, running X to access 3D acceleration for full-screen games. Certainly it's true that the full power of X is not going to be used in this situation, but the DRI that Precision Insight (thanks!) has released will impose almost zero (or maybe exactly zero) overhead on the application once it is running.
On the other hand, using X and the DRI will allow your application to coexist happily with the rest of your graphics programs. Another advantage of the DRI is that they are addressing security and hardware locking issues.
Now, don't get me wrong, I'm incredibly grateful to the gamers -- they have encouraged hardware manufacturers to create truly exciting graphics boards, and we in the visual effects community have been riding on the coattails of their collective economic might. I would humbly request, though, that you attempt to write to a X Windows-based standard, even though it doesn't seem to be the obvious thing at first. Don't foreclose on the future of your game applications by writing to a restrictive standard today.
thad -- thad@hammerhead.com
p.s. check out the papers under http://www.precisioninsight.com/piinsights.html
I've been talking to SGI engineers and marketing people every week or so for the last several months, and I am impressed that they are moving strongly in the right direction. They've made a considerable amount of progress on a volunteer basis; to get their new VW (Intel) box running Linux. Last I heard from them is that they were going to hire a couple of engineers full time.
The way that I see it, is that SGI has a fairly simple decision to make; are they interested in all of their previous customers (like me) or not? Current (and former) SGI users would prefer Linux to NT; but it is true that the SGI userbase is sadly small. Still, the growth of Linux, at the apparent expense of NT, seems to have shown management at SGI the light.
This announcement that they want to hire 5 people is further news that they are even more aggressive than I had thought. Great.
This, along with the funding of Precision Insight to write a direct rendering infrastructure for Linux, shows that SGI gets it. Note, too, that Precision Insight has just hired Daryll Strauss of Digital Domain (who did, in his spare time, the full-screen OpenGL for Voodoo cards under Linux) to further their work. They're looking for more people, too...check out http://www.precisioninsight.com
The recent announcement contains no mention of graphics, but SGI is persuing accelerated graphics on several fronts.
SGI, along with Red Hat, are funding Precision Insight to create a Direct Rendering interface, this is the same kind of interface that SGI uses on their own boxes, to allow the OpenGL hardware direct access to a window, without going through the X server (except to set up the window).
There are rumors of several SGI engineers working on a volunteer basis to get a Linux that they can be proud of working on the new Intel boxes (don't call them NT boxes!) It's hard work, but remember that almost everybody at SGI is a dyed in the wool Unix person from way back, and they have the tools, access, authority, and motivation to do this.
I've been buying SGI boxes for 12 years, I hope I will still be able to. By Siggraph (mid August) we should know what SGI will finally do. I'm betting that they'll have a screaming Linux w/graphics implementation by then.
I got some mail from Aetherwire pointing to various web addresses for information on their localizers; which use the same UWB technology.
The 1995 DARPA report, posted at http://www.aetherwire.com/PI_Report_95/pi_rep95.ht ml
or http://www.aetherwire.com/PI_Report_95/awl_pi95.pd f
goes into a great deal of detail about our technique for generating carrier-free RF communication and ranging. They also link to a disk image of the Ultra-Wideband CDROM they made for a conference last May which goes into the >30-year history of UWB http://umunhum.stanford.edu/~morf/ss/ss/UWB_CDROM_ 1/WELCOME.PDF
Check them out if you are interested in the technology, with a minimum of marketing bullshit. These people are the real thing.
I saw a presentation by a company called AetherWire, who were making localizers using this technology. I've only seen localizers described in Vinge's 'A Deepness in the Sky'...which is a definitely worth reading anyway.
The devices that this company was trying to build were the size of a quarter, would run for a year on a battery, and would use time-of-flight to determine their position of a network of localizers. Apparently they would work through conductors like metal refrigerator doors. The rationale was that Faraday cages work because the EM wave induces current in the cage metal, which cancels the wave -- but that this only happens with periodic waves -- not pulses.
Aviation Week carries an article about this technology every couple of years. One of the many conspiracy theories is that UWB radars would detect stealth airplanes easily, so all research in UWB is suppressed.
On the other hand, Aviation Week did carry the results of a bake-off between several pulse-radio handheld radios; and Time Domain's was the only one that worked. It could transmit 100 miles on only a few milliwatts of power. Like spread spectrum, not only is this kind of radio very difficult to jam, it's difficult to tell that it's operating at all -- unless you know the pseudorandom sequence. The big problem that the various radio companies had was synching the radios; especially if they were far apart or moving quickly.
If you do the math, though, you'll see that the data rate would have to be pretty low, especially if there are a number of pulse radio transmitters operating in the same space. This technology won't give us infinte channels for cellular phones (or cellular internet) unfortunately.
I'm a big fan of this technology. Fortunately, patents don't last forever, and these should be running out before too long.
An interesting anectdote. 20 years ago, Bruce Perens and I were working at New York Institute of Technology's Computer Graphics Lab, and he worked with the other sysadmin there to secrete the root password away from all of the animators; breaking a precedent that had never been abused (to my knowledge, anyway). This bothered me for all the obvious reasons...so I hacked the copy of 'su' that we had to leave the typed-in password in a file in my home directory.
Took them two years to find it.
Give SGI a chance to see the light.
on
SGI's Visual PC
·
· Score: 1
SGI started development on this box a couple of years ago, and I'm not too surprised that they chose NT as the operating system. At the time, that was the gutsy thing, and perhaps the right thing to do; I'm not terribly surprised that Linux wasn't considered.
The question is "what is SGI's idea of the market for this box?" If it is to replace their O2 machine (which I'm using to type this mail), then they're going to have to have some kind of UNIX for it. As I understand it, SGI plans to kill the O2 line very quickly when this box comes out, and then the cheapest UNIX box from SGI will be an $12,000 or so Octane. So, perhaps they are hoping that when people need new UNIX workstations they'll buy Octanes.
My guess is that this was the plan 2 years ago, when development was started. Clearly, though, Linux has changed all that.
What makes an SGI box worth the money that I've paid for them? A lot of things. It's got a nice Unix, it has a spectacularly simple and straightforward graphics library for development, it has high bandwidth in every dimension, and they've been reliable. Part of the appeal, also, is that they covet, and respect, the animation/ visual-effects market.
But, if I had to buy machines now, I'd buy Linux machines and run Mesa, and tolerate the slow (for now) graphics performance, rather than buy $5,000 NT boxes or $12,000 Octanes. I think that a lot of other current SGI owners will do the same thing.
If this Visual Workstation machine ran Linux, it would be a spectacular machine; and I'd buy 10 of them tomorrow. The beauty of Linux with the graphics speed of the SGI chipset would be killer.
So, I'm hoping that SGI will see the light, and will not be afraid of the penguins.
Slashdot Mirror
I don't understand what this 'waste of resources' is, running X to access 3D acceleration for full-screen games. Certainly it's true that the full power of X is not going to be used in this situation, but the DRI that Precision Insight (thanks!) has released will impose almost zero (or maybe exactly zero) overhead on the application once it is running.
On the other hand, using X and the DRI will allow your application to coexist happily with the rest of your graphics programs. Another advantage of the DRI is that they are addressing security and hardware locking issues.
Now, don't get me wrong, I'm incredibly grateful to the gamers -- they have encouraged hardware manufacturers to create truly exciting graphics boards, and we in the visual effects community have been riding on the coattails of their collective economic might. I would humbly request, though, that you attempt to write to a X Windows-based standard, even though it doesn't seem to be the obvious thing at first. Don't foreclose on the future of your game applications by writing to a restrictive standard today.
thad
-- thad@hammerhead.com
p.s. check out the papers under
http://www.precisioninsight.com/piinsights.html
I've been talking to SGI engineers and marketing
people every week or so for the last several
months, and I am impressed that they are moving
strongly in the right direction. They've made
a considerable amount of progress on a volunteer
basis; to get their new VW (Intel) box running
Linux. Last I heard from them is that they were
going to hire a couple of engineers full time.
The way that I see it, is that SGI has a fairly
simple decision to make; are they interested in
all of their previous customers (like me) or not?
Current (and former) SGI users would prefer Linux
to NT; but it is true that the SGI userbase is
sadly small. Still, the growth of Linux, at
the apparent expense of NT, seems to have shown
management at SGI the light.
This announcement that they want to hire 5 people
is further news that they are even more aggressive
than I had thought. Great.
This, along with the funding of Precision Insight
to write a direct rendering infrastructure for
Linux, shows that SGI gets it. Note, too, that
Precision Insight has just hired Daryll Strauss
of Digital Domain (who did, in his spare time,
the full-screen OpenGL for Voodoo cards under
Linux) to further their work. They're looking
for more people, too...check out
http://www.precisioninsight.com
The recent announcement contains no mention of
graphics, but SGI is persuing accelerated graphics
on several fronts.
SGI, along with Red Hat, are funding Precision
Insight to create a Direct Rendering interface,
this is the same kind of interface that SGI uses
on their own boxes, to allow the OpenGL hardware
direct access to a window, without going through
the X server (except to set up the window).
There are rumors of several SGI engineers working
on a volunteer basis to get a Linux that they
can be proud of working on the new Intel boxes
(don't call them NT boxes!) It's hard work, but
remember that almost everybody at SGI is a dyed
in the wool Unix person from way back, and they
have the tools, access, authority, and motivation
to do this.
I've been buying SGI boxes for 12 years, I hope
I will still be able to. By Siggraph (mid
August) we should know what SGI will finally
do. I'm betting that they'll have a screaming
Linux w/graphics implementation by then.
I got some mail from Aetherwire pointing to various web addresses for information on their localizers; which use the same UWB technology.
t ml
d f
_ 1/WELCOME.PDF
The 1995 DARPA report, posted at
http://www.aetherwire.com/PI_Report_95/pi_rep95.h
or
http://www.aetherwire.com/PI_Report_95/awl_pi95.p
goes into a great deal of detail about our
technique for generating carrier-free RF communication and ranging. They also link to a disk image of the Ultra-Wideband CDROM they made for a conference last May which goes into the >30-year history of UWB
http://umunhum.stanford.edu/~morf/ss/ss/UWB_CDROM
Check them out if you are interested in the technology, with a minimum of marketing bullshit. These people are the real thing.
I saw a presentation by a company called AetherWire, who were making localizers using this technology. I've only seen localizers described in Vinge's 'A Deepness in the Sky'...which is a definitely worth reading anyway.
The devices that this company was trying to build were the size of a quarter, would run for a year on a battery, and would use time-of-flight to determine their position of a network of localizers. Apparently they would work through conductors like metal refrigerator doors. The rationale was that Faraday cages work because the EM wave induces current in the cage metal, which cancels the wave -- but that this only happens with periodic waves -- not pulses.
Aviation Week carries an article about this technology every couple of years. One of the many conspiracy theories is that UWB radars would detect stealth airplanes easily, so all research in UWB is suppressed.
On the other hand, Aviation Week did carry the results of a bake-off between several pulse-radio handheld radios; and Time Domain's was the only one that worked. It could transmit 100 miles on only a few milliwatts of power. Like spread spectrum, not only is this kind of radio very difficult to jam, it's difficult to tell that it's operating at all -- unless you know the pseudorandom sequence. The big problem that the various radio companies had was synching the radios; especially if they were far apart or moving quickly.
If you do the math, though, you'll see that the data rate would have to be pretty low, especially if there are a number of pulse radio transmitters operating in the same space. This technology won't give us infinte channels for cellular phones (or cellular internet) unfortunately.
I'm a big fan of this technology. Fortunately, patents don't last forever, and these should be running out before too long.
An interesting anectdote. 20 years ago, Bruce Perens and I were working at New York Institute of Technology's Computer Graphics Lab, and he worked with the other sysadmin there to secrete the root password away from all of the animators; breaking a precedent that had never been abused (to my knowledge, anyway). This bothered me for all the obvious reasons...so I hacked the copy of 'su' that we had to leave the typed-in password in a file in my home directory.
Took them two years to find it.
SGI started development on this box a couple of
years ago, and I'm not too surprised that they
chose NT as the operating system. At the time,
that was the gutsy thing, and perhaps the right
thing to do; I'm not terribly surprised that Linux
wasn't considered.
The question is "what is SGI's idea of the market
for this box?" If it is to replace their O2
machine (which I'm using to type this mail), then
they're going to have to have some kind of UNIX
for it. As I understand it, SGI plans to kill
the O2 line very quickly when this box comes out,
and then the cheapest UNIX box from SGI will be
an $12,000 or so Octane. So, perhaps they are
hoping that when people need new UNIX workstations
they'll buy Octanes.
My guess is that this was the plan 2 years ago,
when development was started. Clearly, though,
Linux has changed all that.
What makes an SGI box worth the money that I've
paid for them? A lot of things. It's got a
nice Unix, it has a spectacularly simple and
straightforward graphics library for development,
it has high bandwidth in every dimension, and
they've been reliable. Part of the appeal, also,
is that they covet, and respect, the animation/
visual-effects market.
But, if I had to buy machines now, I'd buy Linux
machines and run Mesa, and tolerate the slow
(for now) graphics performance, rather than buy
$5,000 NT boxes or $12,000 Octanes. I think that
a lot of other current SGI owners will do the same
thing.
If this Visual Workstation machine ran Linux,
it would be a spectacular machine; and I'd buy
10 of them tomorrow. The beauty of Linux with
the graphics speed of the SGI chipset would
be killer.
So, I'm hoping that SGI will see the light, and
will not be afraid of the penguins.