I was lucky enough to spend some time working with digital video at SGI (Silicon Graphics) earlier this year. Quite an education. One thing that I had overlooked was that compressed video has to come from somewhere... and SGI had that somewhere on Octane2 workstations. I belive 1080i was between 124 MByte/sec and 248 MByte/sec depending on the bitdepth (2, 3, or 4 bytes per pixel -- the film industry loves 4 byte / 48 bit color). To even play back uncompressed 1080i video required three channels of fibrechannel disk arrays attached to the workstation. Overkill, perhaps, but the machine handled the huge uncompressed video like a modern PC can handle a 320x240@15fps AVI.
MPEG2 compresses digital video very well and depending on the settings used, 1080 can be done with not much more thruput (bandwidth) than 720 or even 480. But belive me, an uncompressed digital version of a film digitized to 1080p/24Hz is a sight to be seen, especially on a Sony HD studio monitor. Without compression, "HDTV" really shines but the ungodly amounts of disk space required don't make it worthwhile with current consumer technology. Uncompressed video is beautiful, compressed video (when done right) is still *very* good.
I don't know about the thruput of streaming digital video, but based on my limited knowledge of MPEG2 and the various flavors of HD digital video would lead me to belive that 1080i at 20Mbit/sec would certainly be obtainable. I'm not sure about the audio, though, especially if it's to be played on a high-quality 5 or 6 channel system.
As far as the term "HDTV", you won't hear it used much outside of marketing circles. It's sort of like saying "computer" rather than "1.4 GHz Athlon running Linux" or "833 MHz Alpha running Tru64". HDTV is a vague consumer term.
The folks in that Palo Alto neighborhood had better upgrade their TVs/monitors/projectors! Very cool stuff indeed.
...I was using an Apple IIe that 'just worked'. No silly OS to worry about, no silly drivers or incompatibilities. Adding an 80-column card or a Grappler serial card never brought about problems. I wrote all of my undergrad papers on that thing and not once did AppleWorks fail to do its job. Even the BeagleBrothers TimeOut addons worked without a hitch. My games -- granted they were on different floppies with thier own OSes -- worked fine and no amount of gaming weakened the stability of my system.
I wish modern computers were even 1/4 as reliable and stable as that old Apple IIe. If there was an easy way to open Word and Excel files as well as interface the thing to my color inkjet printer and use different fonts, I would probably continue to use it. Over the 7 years I heavily used that machine, the only problem I ever encoundered was the lousey spacebar, I had to reseat the little metal support about once a year.
A heardy thank you to the Woz for creating such a delightful series of machines.
How do you figure 20Mbps/channel without the picture looking like ass?
HDTV (1080i) = 1920*1080*16bitcolor*30fps = 9.4x10^8 bits/sec = 940 Mbit/sec uncompressed for the video alone. Ignoring audio, that's going to require 47:1 compression. Even DiVX won't be able to compress that well and still make the image look good, at least not on an HDTV. Still need to figure in the thruput required to transport the dolby digital or dts sound as well, I don't know how well they compress.
Sure, it's a tradeoff, but if I bought an HDTV or high-res projector for my PC, I'm not going to want to stream crap video through it, regardless of the cost.
Any idea what their upstream connections are? With a *neighborhood* of gigE, I would imagine they would have at least a pair of OC3s (from two different providers), probably a pair of OC12s. Unless it's paid for by some grant, that can't be cheap, even if they had 1000 customers. More details would sure be interesting.
http://www.gigabitgarage.net/images/lix_arch.jpg
Doesn't really cut it. I can't make heads or tales out of that, especially when the text on the labels is 1 pixel tall. I would really like to see a good diagram or read some detailed specs.
Being a Mac user (as well as a NeXTSTEP/Openstep, Solaris, IRIX, OpenBSD, Linux, and Windows user), I have indeed run across far too many "typical mac users". Half of which are addicted to the platform and refuse to even look at anything else... the other half not knowing the difference between USB and HBO (or IEEE1394 and the IRS 1040).
I use a number of G4s at work as they've been powerful, stable, wonderful machines to work with for DV video editing and compositing (under Mac OS 9.1). The G4 case was a deam to work with when adding ram and a second hard drive to each machine. We currently have two SGIs and three PCs running Maya but will certainly try Maya under OS X later this summer. At home and for some of the organizations I work with, I use a variety of x86 and workstation platforms. I do my gaming and game serving on two homebrew Windows PCs. No need to bash a platform or try to convert the dissimilar. Use what you enjoy and what works for you.
I use a bit of everything and while I only own one mac these days, we use almost nothing but G4s at work (FinalCutPro 2 with DVcam and DVCPro gear). At home I spend most of my time on a few old SGI workstations but slowly using more and more Mac OS X on my PBG3. As soon as Apple finishes (or starts, heh) full hardware acceleration for the Aqua widgets, I'll probably go fullbore OS X and never look back. There's something very nice about using Photoshop, postgresql, apache, and iMovie all on the same machine under the same os. I do admit that OS X hasn't been perfect, but our Mac OS 9.1 machines at work have been rock solid.
The crossbar architecture reminds me a lot of my Silicon Graphics Octane... built in January 1997. As far as wide memory bitpaths, I think the Sun Blade 1000 is the highest for less than $12,000 --288-bit ram path (upgrade four dimms at a time).
At any rate, I think I'll wait for the SiS735.
onboard graphics performance?
on
nVidia nForce
·
· Score: 2
The reference board has only two PCI slots and no onboard Ethernet. Ouch. I'll stick with my Mac for now.
Do you think this stuff could really compete with mac anyway?
Compete is a relative term, depends on the task (say, Photoshop vs Quake). I'm interested to learn how the onboard gfx of nForce compares to say, a GF3 in a Mac or PC. GF3 may be a bit faster, but is limited by AGP 4x/6x/8x and the system ram thruput. I'd also like to see if NVIDIA is working on some sort of unified color correction API for Windows (a la Apple's ColorSync... something that is universal, not just one part of a kluge of components).
Seriously. DC was about as much fun and welcome as a neighborhood bully. Everything about the company -- from their business plan to attitude towards geeks to the way they tried to strongarm the UPC industry -- stunk. Not only did they prove the limitations of free-as-in-beer, they set over a dozen exampled of how not to run a business, how not to conduct PR, and how not to treat your userbase. DC was a closed minded facist group that didn't open its eyes and ears until it was far too late. It's hard, if not impossible, to feel sorry for them.
Golly, it was only about two years ago that the most common Linux tagline was "and you don't even need the latest hardware to run it... a 486 is fine and a P233 is a dream come true". Shucks, these days a guy almost needs a PII/400 just to run the latest windowmanagers, desktop environments, and web browsers. *sigh*
Here in rural North Dakota we only get the Simpsons via over-the-air broadcast about three times a week... and I'm itching to learn if Simpsons has a 'laugh track'. I can't for the life of me recall if there is one or not...
I've had very good luck with the drivers for my various standalone PCI video capture cards on my PCs. It's been the graphics card drivers that have given me nothing but trouble (especially early Matrox and NVIDIA drivers).
One of my biggest complaints about Linux is that there isn't a modern open source Fortran compiler for it.
I'm sure you probably mean Linux for IA32/x86, but this may be of interest anyway... seems SGI is hard at work on a "commercial-grade" freeware set of compilers for Linux IA64/Itanium. Not only will their Pro64 compilers include F90/F95 support, but they will also be bringing over their SCSL libraries gained from Cray. Neat stuff.
My university has two large SGI systems that crank away on compiled fortran code day and night. Our analog EE's, biowarfare research group, and meteorology folks keep both machines at about 95% load around the clock. One machine is a shiny new Origin 3000 (96x R14K/500) and the other is an Origin 2000 (64x R10K/250). The O2K has a stack of R14K/500 CPU nodeboards and an additional 64 GB ram ready to replace its aging R10Ks, but the users keep telling the admin staff to "wait another day, my big batch job is almost done".
The decision to switch from a Cray C90 to SGI had a lot to do with SGI's on-going fortran compiler development, optimization, and obscure/rare bug hunts. Their short term Cray R&D ownership also brought about some updated scientific libraries and optimized routines. Lots of happy folks here.
The storage world is almost as bad as the audio world... a 6 TB SAN is the equivalant of a $500 all-in-one Sony stereo from Circuit City. An IDE/ATA RAID is a used, first-generation Rio MP3 player with a pair of $3 headphones. Get yourself a (4 - 512 CPU) SGI Origin 3000 or a Cray SV1ex and (25 - 500 TB) storage from EMC. Don't forget a 1.5 PB robotic tape silo from StorageTek for backup.
Haven't you seen the television commercial with the nifty animations and the "Chiropractor" chick? If you buy their "micromineature circuitry-based" Internal Antenna, they'll throw in The Wave Scrambler! According to the animation, this little sticker deflects the harmful waves away from your head when applied to the speaker of a cell phone!
Why would you need a console monitor on a server? Run a serial cable from your laptop or pda into the server's serial port / alternate console port and chug away. Here in my office all of my servers are headless and have their serial ports plugged into a portmaster.
Linus is great, but lets get our facts straight...
Just because an application uses a cluster does not automatically mean that it's running on a stack of comodity PCs running Linux in a beowulf-style cluster interconnected via GigE or Myrinet. It also doesn't automatically mean that the application isn't capible of running well on a single large machine.
For example, NOAA recently put together a cluster for computing weather models for the upcoming hurricane season. Their cluster is actually 8 machines, each a 128 CPU SGI Origin 3800 runing IRIX 6.5. The 8 machines are interconnected through a thick mesh of GSN (gigabyte system network, a modern version of HiPPI that can transfer 800 megabytes/sec per link). The messaging protocols used are a mixture of shmem, OpenMP, and MPI.
Linux is great and all, but ASCI Red uses Intel's Paragon OS, a derrivative of Unix.
Slashdot Mirror
Interesting thread.
I was lucky enough to spend some time working with digital video at SGI (Silicon Graphics) earlier this year. Quite an education. One thing that I had overlooked was that compressed video has to come from somewhere... and SGI had that somewhere on Octane2 workstations. I belive 1080i was between 124 MByte/sec and 248 MByte/sec depending on the bitdepth (2, 3, or 4 bytes per pixel -- the film industry loves 4 byte / 48 bit color). To even play back uncompressed 1080i video required three channels of fibrechannel disk arrays attached to the workstation. Overkill, perhaps, but the machine handled the huge uncompressed video like a modern PC can handle a 320x240@15fps AVI.
MPEG2 compresses digital video very well and depending on the settings used, 1080 can be done with not much more thruput (bandwidth) than 720 or even 480. But belive me, an uncompressed digital version of a film digitized to 1080p/24Hz is a sight to be seen, especially on a Sony HD studio monitor. Without compression, "HDTV" really shines but the ungodly amounts of disk space required don't make it worthwhile with current consumer technology. Uncompressed video is beautiful, compressed video (when done right) is still *very* good.
I don't know about the thruput of streaming digital video, but based on my limited knowledge of MPEG2 and the various flavors of HD digital video would lead me to belive that 1080i at 20Mbit/sec would certainly be obtainable. I'm not sure about the audio, though, especially if it's to be played on a high-quality 5 or 6 channel system.
As far as the term "HDTV", you won't hear it used much outside of marketing circles. It's sort of like saying "computer" rather than "1.4 GHz Athlon running Linux" or "833 MHz Alpha running Tru64". HDTV is a vague consumer term.
The folks in that Palo Alto neighborhood had better upgrade their TVs/monitors/projectors! Very cool stuff indeed.
...I was using an Apple IIe that 'just worked'. No silly OS to worry about, no silly drivers or incompatibilities. Adding an 80-column card or a Grappler serial card never brought about problems. I wrote all of my undergrad papers on that thing and not once did AppleWorks fail to do its job. Even the BeagleBrothers TimeOut addons worked without a hitch. My games -- granted they were on different floppies with thier own OSes -- worked fine and no amount of gaming weakened the stability of my system.
I wish modern computers were even 1/4 as reliable and stable as that old Apple IIe. If there was an easy way to open Word and Excel files as well as interface the thing to my color inkjet printer and use different fonts, I would probably continue to use it. Over the 7 years I heavily used that machine, the only problem I ever encoundered was the lousey spacebar, I had to reseat the little metal support about once a year.
A heardy thank you to the Woz for creating such a delightful series of machines.
I love those those two emachines boxes under that monitor...
G
http://www.fiberhood.net/first/images/DSC00111.JP
How do you figure 20Mbps/channel without the picture looking like ass?
HDTV (1080i) = 1920*1080*16bitcolor*30fps = 9.4x10^8 bits/sec = 940 Mbit/sec uncompressed for the video alone. Ignoring audio, that's going to require 47:1 compression. Even DiVX won't be able to compress that well and still make the image look good, at least not on an HDTV. Still need to figure in the thruput required to transport the dolby digital or dts sound as well, I don't know how well they compress.
Sure, it's a tradeoff, but if I bought an HDTV or high-res projector for my PC, I'm not going to want to stream crap video through it, regardless of the cost.
Any idea what their upstream connections are? With a *neighborhood* of gigE, I would imagine they would have at least a pair of OC3s (from two different providers), probably a pair of OC12s. Unless it's paid for by some grant, that can't be cheap, even if they had 1000 customers. More details would sure be interesting.
http://www.gigabitgarage.net/images/lix_arch.jpg
Doesn't really cut it. I can't make heads or tales out of that, especially when the text on the labels is 1 pixel tall. I would really like to see a good diagram or read some detailed specs.
Being a Mac user (as well as a NeXTSTEP/Openstep, Solaris, IRIX, OpenBSD, Linux, and Windows user), I have indeed run across far too many "typical mac users". Half of which are addicted to the platform and refuse to even look at anything else... the other half not knowing the difference between USB and HBO (or IEEE1394 and the IRS 1040).
I use a number of G4s at work as they've been powerful, stable, wonderful machines to work with for DV video editing and compositing (under Mac OS 9.1). The G4 case was a deam to work with when adding ram and a second hard drive to each machine. We currently have two SGIs and three PCs running Maya but will certainly try Maya under OS X later this summer. At home and for some of the organizations I work with, I use a variety of x86 and workstation platforms. I do my gaming and game serving on two homebrew Windows PCs. No need to bash a platform or try to convert the dissimilar. Use what you enjoy and what works for you.
I use a bit of everything and while I only own one mac these days, we use almost nothing but G4s at work (FinalCutPro 2 with DVcam and DVCPro gear). At home I spend most of my time on a few old SGI workstations but slowly using more and more Mac OS X on my PBG3. As soon as Apple finishes (or starts, heh) full hardware acceleration for the Aqua widgets, I'll probably go fullbore OS X and never look back. There's something very nice about using Photoshop, postgresql, apache, and iMovie all on the same machine under the same os. I do admit that OS X hasn't been perfect, but our Mac OS 9.1 machines at work have been rock solid.
The crossbar architecture reminds me a lot of my Silicon Graphics Octane... built in January 1997. As far as wide memory bitpaths, I think the Sun Blade 1000 is the highest for less than $12,000 --288-bit ram path (upgrade four dimms at a time).
At any rate, I think I'll wait for the SiS735.
The reference board has only two PCI slots and no onboard Ethernet. Ouch. I'll stick with my Mac for now.
Do you think this stuff could really compete with mac anyway?
Compete is a relative term, depends on the task (say, Photoshop vs Quake). I'm interested to learn how the onboard gfx of nForce compares to say, a GF3 in a Mac or PC. GF3 may be a bit faster, but is limited by AGP 4x/6x/8x and the system ram thruput. I'd also like to see if NVIDIA is working on some sort of unified color correction API for Windows (a la Apple's ColorSync... something that is universal, not just one part of a kluge of components).
Seriously. DC was about as much fun and welcome as a neighborhood bully. Everything about the company -- from their business plan to attitude towards geeks to the way they tried to strongarm the UPC industry -- stunk. Not only did they prove the limitations of free-as-in-beer, they set over a dozen exampled of how not to run a business, how not to conduct PR, and how not to treat your userbase. DC was a closed minded facist group that didn't open its eyes and ears until it was far too late. It's hard, if not impossible, to feel sorry for them.
Golly, it was only about two years ago that the most common Linux tagline was "and you don't even need the latest hardware to run it... a 486 is fine and a P233 is a dream come true". Shucks, these days a guy almost needs a PII/400 just to run the latest windowmanagers, desktop environments, and web browsers. *sigh*
Here in rural North Dakota we only get the Simpsons via over-the-air broadcast about three times a week... and I'm itching to learn if Simpsons has a 'laugh track'. I can't for the life of me recall if there is one or not...
I have a tape with 28 full episodes!
That can't be VHS... what are you using? DLT or DDS4 with the episodes in DiVX format?
I've had very good luck with the drivers for my various standalone PCI video capture cards on my PCs. It's been the graphics card drivers that have given me nothing but trouble (especially early Matrox and NVIDIA drivers).
One of my biggest complaints about Linux is that there isn't a modern open source Fortran compiler for it.
I'm sure you probably mean Linux for IA32/x86, but this may be of interest anyway... seems SGI is hard at work on a "commercial-grade" freeware set of compilers for Linux IA64/Itanium. Not only will their Pro64 compilers include F90/F95 support, but they will also be bringing over their SCSL libraries gained from Cray. Neat stuff.
http://oss.sgi.com/projects/Pro64/
My university has two large SGI systems that crank away on compiled fortran code day and night. Our analog EE's, biowarfare research group, and meteorology folks keep both machines at about 95% load around the clock. One machine is a shiny new Origin 3000 (96x R14K/500) and the other is an Origin 2000 (64x R10K/250). The O2K has a stack of R14K/500 CPU nodeboards and an additional 64 GB ram ready to replace its aging R10Ks, but the users keep telling the admin staff to "wait another day, my big batch job is almost done".
f ortran.html a po.html m ips73features.html
The decision to switch from a Cray C90 to SGI had a lot to do with SGI's on-going fortran compiler development, optimization, and obscure/rare bug hunts. Their short term Cray R&D ownership also brought about some updated scientific libraries and optimized routines. Lots of happy folks here.
http://www.sgi.com/software/scsl.html
http://www.sgi.com/developers/devtools/languages/
http://www.sgi.com/developers/devtools/languages/
http://www.sgi.com/developers/devtools/languages/
The storage world is almost as bad as the audio world... a 6 TB SAN is the equivalant of a $500 all-in-one Sony stereo from Circuit City. An IDE/ATA RAID is a used, first-generation Rio MP3 player with a pair of $3 headphones. Get yourself a (4 - 512 CPU) SGI Origin 3000 or a Cray SV1ex and (25 - 500 TB) storage from EMC. Don't forget a 1.5 PB robotic tape silo from StorageTek for backup.
Haven't you seen the television commercial with the nifty animations and the "Chiropractor" chick? If you buy their "micromineature circuitry-based" Internal Antenna, they'll throw in The Wave Scrambler! According to the animation, this little sticker deflects the harmful waves away from your head when applied to the speaker of a cell phone!
Why would you need a console monitor on a server? Run a serial cable from your laptop or pda into the server's serial port / alternate console port and chug away. Here in my office all of my servers are headless and have their serial ports plugged into a portmaster.
Heh, I can just see a good trivia question...
Who is the "Cray woman on the upper right-hand side of most cray.com pages?"
http://www.cray.com/products/index.html
Linux isn't running on the second-fastest supercomputer either, Paragon OS is.
My mistake.
Cray, Inc. is much more alive than their former owner, SGI...
Lots of new products and they're even making a profit.
http://www.cray.com/products/systems.
Nice varitey of systems, from their own SV1/SV1ex/SV2 machines, to Linux clusters, to maspar Alphas, to NEC vector-based machines, and more.
Linus is great, but lets get our facts straight...
Just because an application uses a cluster does not automatically mean that it's running on a stack of comodity PCs running Linux in a beowulf-style cluster interconnected via GigE or Myrinet. It also doesn't automatically mean that the application isn't capible of running well on a single large machine.
For example, NOAA recently put together a cluster for computing weather models for the upcoming hurricane season. Their cluster is actually 8 machines, each a 128 CPU SGI Origin 3800 runing IRIX 6.5. The 8 machines are interconnected through a thick mesh of GSN (gigabyte system network, a modern version of HiPPI that can transfer 800 megabytes/sec per link). The messaging protocols used are a mixture of shmem, OpenMP, and MPI.
Linux is great and all, but ASCI Red uses Intel's Paragon OS, a derrivative of Unix.