Another reason why I support GUIs switching to a fully accelerated model. back in 1993, I saw a Matrox card that did HW accelerated TrueType. Given the fact that we have a 100 times as many transistors to work with these days, why can't the make a card that accelerates all of the Postscript primatives? I'd buy one, and the QuarkXPress guys would kill for one.
True, it does depend on the reference system. However, in astronomy the reference point is the common center of mass. Since that center of mass is about a thousand miles under the surface of the earth, you can in all truth say the moon orbits about the earth.
Or just use OS X. Quartz should be perfect for scalable GUIs (when, of course, we get HW acceleration for Postscript primitives.) I dream of a day when the user can specify a scaling size for their desktop, and program designers can work in a virtual coordinate space without having to worry about resolution.
Do you realize how few people know that 200dpi==40,000 pixels per square inch? You're talking about a country where 1/5th of people don't know whether the earth goes around the moon or vice versa. (Which, BTW compares favorably to the 1/4-1/3 in parts of Europe.)
Actually, you wouldn't be pusing 36MB x 60fps to do Quake. Remember, 3D is vector graphics. Aside from textures, the data rate should be the same wether you're rendering at 1600x1200 or 16,000x12,000. Also, 50fps video at that res would require about 1.8GB/sec of bandwidth. While that's not within the realm of AGP 4X, it is within the 3-5GB/sec of bandwidth that even mid-range SGIs have.
You could NOT break down a wall and rebuild it farther for $3000. Also, a lot of people don't like a big monitor hogging their desk. If you're working in a cozy study or in the corner of the living room, then a flat panel display is ideal. Additionally, if you're a minimalist, then a standard CRT probably doesn't go with the "look" of your house. Not to mention he might just be tired of frying his brain with cancer causing emissions.
Of course, 10MB/sec is quite pessimistic. Given the huge density of the device, and the fact that it will eventually spin at CD (12x) speeds (probably) you'll get a good deal closer to the 25-30MB/sec of fast harddrives.
I don't know about that. This is still a spinning medium and access time with non-solid state devices will always suck. Given the density of this, and the spin rate(12x (eventual) max for CD-based technologies), you could persume some massive bandwidth (probably harddrive speeds) However, RAM is a moving target, and things like 1T-SRAM (SRAM that only uses one transistor) are offering SRAM-type speed at densities approaching regular RAM (128mbit chips) I don't think we'll ever get rid of RAM, it simply makes too much sense.
They're both lossy. MPEG2 I frames are stored at a compression similar to JPEG, and in the process of getting the motion data, some information is discarded.
Hardware stuff is only part of it. If you look at the general quality of driver, you'll notice too kinds. Crappy ones, and solid ones. Those with the solid/featureful drivers are (rightfully) reluctant to give an advantage to those with crappy drivers.
HDTV is higher res than SVGA, but who uses SVGA anymore? The maximum horizontal resolution for HDTV is something like 1200-something, while on a monitor it is 1840 (for highest quality monitors.) A 1280x1024 capable monitor should be easily able to display HDTV.
Well, on a laptop, you're probably not running programs that max-out the CPU. Try running a couple of compiles in the background and a 3D Studio render and watch the fliker-show.
Hello? We're talking about AOL here? The same group of brillient who made AOL 5.0 feel like it was written in Java? If Mozilla is this bloated now, I absolutely tremble at the though of AOL getting its grubby hands on it.
Of course that kind of "being optimized" applies. Additionally, a 10GHz chip better have a hell of a lot more bus bandwidth than a 1 GHz chip. You said something along the lines of "I hope the P4 isn't optimized for 5Ghz because then it wouldn't work efficiently at 2GHz." The P4 is optimized for high clock-speed in general. It takes the 20 stage pipeline just to reach 2GHz and the chip is capable of reaching 5-8 GHz or so. That's about the same range as the 60Mhz Pentiums reaching 233 or the 233MHz PII's reaching 1000Mhz (with the PIII) I thought you were implying that clocking a chip meant for 5Ghz at 2Ghz (given the same architure) would somehow decreased the performance more drastically than the decrease in clock speed would account for. My mistake if I underestimated you.
I've always wondered. Does Linus use CVS? I can understand him wanting to be the central relay station for any and all patches to the kernel, but how does HE manage the source? Does he have thousands of source archives of different patches on his harddrive? Does he manually diff and check in patches? Does he have only the one working copy of the kernel?! Maybe it would be easier on him if he used CVS himself, but made sure only a trusted few (him and Alan probably) had commit privleges so he could still be the grand poobah, but he'd have access to the other features (rollback to any state, automatic merging, source management, etc) of CVS.
Then that's not the smartest thing to do. There is a lot of stuff in a server that you don't need in a desktop OS, a lot of stuff that isn't that you do, and optimizations that work for one, not the other. For example, on the TCP/IP stack of a desktop OS, you probably want to optimize the stack for fast response on a couple of connections, while on a server you want to be able to handle many connections consistantly. The VM management also has to be different, (on a desktop, the forground process needs a lot more CPU than the rest of the system). The sheduling quanta has to be different (WinNT Server: 120ms fixed length, WinNT workstation: 20 background 20-60 forground variable length, Linux 2.2 100ms fixed, Linux 2.4 50ms fixed, BeOS 3ms fixed), the priority management has to be different (most computation time, or interactive speed?), disk buffer management has to be different, etc. If you try to do it all (and no I'm not going in and changing these values in the Linux source code!) then you get a sub-optimal experience for all.
In the test kernels, NetFilter (the official name) is not that much fun, but not terribly bad either. You have to use a program called iptables instead of ipchains (duh) which never seems to be compiled correctly for the kernel modules I have at the time. The syntax for iptables is a little different, but my is fairly basic so I can't help much there. However, you still have to option to use ipchains in the 2.4 kernel if you have extensive config stuff already written.
To tell the truth, it all depends on what you use it for. On my system with Detonator 3 NVIDIA drivers, Win2K has locked up at least a dozen times this month. Its not random instability, Win2K just doesn't like Windows Media Player 7 and OpenGL. As long as you avoid weird GL apps (not Quake but stuff like the demo programs you get on flipcode) and Media Player, then it is quite solid. Unfortunately, WMA and OpenGL are the only reasons I need Win2K so it is a little self-defeating. Methinks it needs some more elbow grease to work the kinks out.
Slashdot Mirror
Another reason why I support GUIs switching to a fully accelerated model. back in 1993, I saw a Matrox card that did HW accelerated TrueType. Given the fact that we have a 100 times as many transistors to work with these days, why can't the make a card that accelerates all of the Postscript primatives? I'd buy one, and the QuarkXPress guys would kill for one.
True, it does depend on the reference system. However, in astronomy the reference point is the common center of mass. Since that center of mass is about a thousand miles under the surface of the earth, you can in all truth say the moon orbits about the earth.
Or just use OS X. Quartz should be perfect for scalable GUIs (when, of course, we get HW acceleration for Postscript primitives.) I dream of a day when the user can specify a scaling size for their desktop, and program designers can work in a virtual coordinate space without having to worry about resolution.
Actually those or pretty loose guidelines. Some companys (like Micron) have much tighter standards for their displays.
Do you realize how few people know that 200dpi==40,000 pixels per square inch? You're talking about a country where 1/5th of people don't know whether the earth goes around the moon or vice versa. (Which, BTW compares favorably to the 1/4-1/3 in parts of Europe.)
Actually, you wouldn't be pusing 36MB x 60fps to do Quake. Remember, 3D is vector graphics. Aside from textures, the data rate should be the same wether you're rendering at 1600x1200 or 16,000x12,000. Also, 50fps video at that res would require about 1.8GB/sec of bandwidth. While that's not within the realm of AGP 4X, it is within the 3-5GB/sec of bandwidth that even mid-range SGIs have.
You could NOT break down a wall and rebuild it farther for $3000. Also, a lot of people don't like a big monitor hogging their desk. If you're working in a cozy study or in the corner of the living room, then a flat panel display is ideal. Additionally, if you're a minimalist, then a standard CRT probably doesn't go with the "look" of your house. Not to mention he might just be tired of frying his brain with cancer causing emissions.
9 million pixels X 4bpp= 36MB framebuffer! Does anyone know what kind of graphics cards they have on those super computer terminals?
Another large slow 40GB haddrive. $120 on pricewatch.
Of course, 10MB/sec is quite pessimistic. Given the huge density of the device, and the fact that it will eventually spin at CD (12x) speeds (probably) you'll get a good deal closer to the 25-30MB/sec of fast harddrives.
I don't know about that. This is still a spinning medium and access time with non-solid state devices will always suck. Given the density of this, and the spin rate(12x (eventual) max for CD-based technologies), you could persume some massive bandwidth (probably harddrive speeds) However, RAM is a moving target, and things like 1T-SRAM (SRAM that only uses one transistor) are offering SRAM-type speed at densities approaching regular RAM (128mbit chips) I don't think we'll ever get rid of RAM, it simply makes too much sense.
Hell, DivX sucks up 70% of my 300MHz processor running BeOS no less.
They're both lossy. MPEG2 I frames are stored at a compression similar to JPEG, and in the process of getting the motion data, some information is discarded.
Hardware stuff is only part of it. If you look at the general quality of driver, you'll notice too kinds. Crappy ones, and solid ones. Those with the solid/featureful drivers are (rightfully) reluctant to give an advantage to those with crappy drivers.
HDTV is higher res than SVGA, but who uses SVGA anymore? The maximum horizontal resolution for HDTV is something like 1200-something, while on a monitor it is 1840 (for highest quality monitors.) A 1280x1024 capable monitor should be easily able to display HDTV.
Umm, if you have the money to burn, nothing beats watching HDTV brodcasts of friends while hacking code ;)
Good god, that damn dealer sold me tainted crack again. I ought to sue ;)
Well, on a laptop, you're probably not running programs that max-out the CPU. Try running a couple of compiles in the background and a 3D Studio render and watch the fliker-show.
Hello? We're talking about AOL here? The same group of brillient who made AOL 5.0 feel like it was written in Java? If Mozilla is this bloated now, I absolutely tremble at the though of AOL getting its grubby hands on it.
Of course that kind of "being optimized" applies. Additionally, a 10GHz chip better have a hell of a lot more bus bandwidth than a 1 GHz chip. You said something along the lines of "I hope the P4 isn't optimized for 5Ghz because then it wouldn't work efficiently at 2GHz." The P4 is optimized for high clock-speed in general. It takes the 20 stage pipeline just to reach 2GHz and the chip is capable of reaching 5-8 GHz or so. That's about the same range as the 60Mhz Pentiums reaching 233 or the 233MHz PII's reaching 1000Mhz (with the PIII) I thought you were implying that clocking a chip meant for 5Ghz at 2Ghz (given the same architure) would somehow decreased the performance more drastically than the decrease in clock speed would account for. My mistake if I underestimated you.
I've always wondered. Does Linus use CVS? I can understand him wanting to be the central relay station for any and all patches to the kernel, but how does HE manage the source? Does he have thousands of source archives of different patches on his harddrive? Does he manually diff and check in patches? Does he have only the one working copy of the kernel?! Maybe it would be easier on him if he used CVS himself, but made sure only a trusted few (him and Alan probably) had commit privleges so he could still be the grand poobah, but he'd have access to the other features (rollback to any state, automatic merging, source management, etc) of CVS.
Actually, the guys at BeDope are winning here. Check out this press release
Installation instructions can be found here.
PS> You knew you were asking for this.
Then that's not the smartest thing to do. There is a lot of stuff in a server that you don't need in a desktop OS, a lot of stuff that isn't that you do, and optimizations that work for one, not the other. For example, on the TCP/IP stack of a desktop OS, you probably want to optimize the stack for fast response on a couple of connections, while on a server you want to be able to handle many connections consistantly. The VM management also has to be different, (on a desktop, the forground process needs a lot more CPU than the rest of the system). The sheduling quanta has to be different (WinNT Server: 120ms fixed length, WinNT workstation: 20 background 20-60 forground variable length, Linux 2.2 100ms fixed, Linux 2.4 50ms fixed, BeOS 3ms fixed), the priority management has to be different (most computation time, or interactive speed?), disk buffer management has to be different, etc. If you try to do it all (and no I'm not going in and changing these values in the Linux source code!) then you get a sub-optimal experience for all.
In the test kernels, NetFilter (the official name) is not that much fun, but not terribly bad either. You have to use a program called iptables instead of ipchains (duh) which never seems to be compiled correctly for the kernel modules I have at the time. The syntax for iptables is a little different, but my is fairly basic so I can't help much there. However, you still have to option to use ipchains in the 2.4 kernel if you have extensive config stuff already written.
You can find some FAQs and HOWTO's (scroll down) at the NetFilter homesite.
To tell the truth, it all depends on what you use it for. On my system with Detonator 3 NVIDIA drivers, Win2K has locked up at least a dozen times this month. Its not random instability, Win2K just doesn't like Windows Media Player 7 and OpenGL. As long as you avoid weird GL apps (not Quake but stuff like the demo programs you get on flipcode) and Media Player, then it is quite solid. Unfortunately, WMA and OpenGL are the only reasons I need Win2K so it is a little self-defeating. Methinks it needs some more elbow grease to work the kinks out.