Well, if the review mags like CNet/ZDnet start pointing out that aftermarket refills don't work, and marking down those printers for it, that'll have an effect.
This seems like typical MS bundling. They have a nice PVR app and maybe a codec. Instead of releasing it as the "MS PVR" app or an add-on to Windows Media Player, like anyone else would, they *bundle it into the OS*. WTF? This is exactly what they got reamed for doing with IE, media player, WMA, DRM codecs, etc. etc. And they're still doing the same thing! (Yeah, that DoJ slap on the wrist really hurt...)
Don't fall for this marketing crap. Make them release it as a simple app. I don't care if it will only install if you have the XYZ tuner card or whatever, but it is NOT a new version of the OS! It is NOT "Windows MCE" whatever they tell you! It's just Windows XP with a simple (and quite nice, from what I can tell) full-screen PVR app.
Once they succeed in convincing everyone that the PVR app is "part of the OS" the market for 3rd party PVR apps will dry up, because of course who would pay money for something that comes for free? And then they can cripple it any way they want in future versions (no recording pay-per-view, no skipping commercials, no transcoding, etc.)
Don't buy this line. Anand is starting to behave like a PR flack; it's too bad. Think for yourselves, people! There are a few PVR Windows apps out there, and some of them are getting quite decent. Vote with your wallets.
-- Tristero
Re:Musings on CPU and UI Performance
on
Is Mac OS X Slow?
·
· Score: 0
Go to Apple's benchmark page. The dual 1.25GHz Mac is 1.9x the speed of the single 2.53GHz P4. That means that even on their best Altivec-optimized benchmark, the fastest single Mac CPU is 0.95x a mid-range P4 (2.8GHz would be a fairer comparison: hottest CPUs in each class).
My own highly optimized image-processing software bears out these numbers. It's not OSX that's slow, it's the CPU. Come on, Jobs, stop fooling around and get OSX on Intel/AMD!!!
Minolta. Updated the Dimage7 camera firmware to nearly double the autofocus speed, add direct manual focus (autofocus trim), new color modes, real-time histogram, and lots more. Over a year after product introduction too. Fairly amazing firmware upgrade actually.
Yes, but USB serial adapters are more like $35. I need at least two, one for my Garmin GPS and one for my PalmV. Adding $70 to a motherboard's price just for the cool "legacy-free" label just doesn't cut it for me.
I guess my life just isn't legacy-free yet.
I presume you're kidding about the video editing part. SGI's largest OEM outside the government is, I believe, Discreet, probably the most popular high-end nonlinear editing/compositing/effects packages in the industry.
Pretty much any FX movie you watch anywhere was at least partially made on a Discreet, and thus on IRIX.
Now as for watching your DVDs, well... Octanes are not really desktop machines.:-)
The idea of radial context menus is they are self-documenting. You can start by looking at the menu, and then after a while the ones you use become second nature. You don't even really have to "learn" them, as opposed to gestures, which, while obviously not so difficult, still have to be looked up somewhere until you get used to them.
And if there's a rarely used gesture, it's utterly useless. In a radial menu, you can at least wait for the menu to show up and then follow its cues.
If a radial menu is well-designed, it becomes pretty similar to a set of gestures. Unfortunately the Mozilla radial menus use tiny hard-to-read icons and so are much too slow to actually use. But in other systems with text menus, they're quite fast to use and learn.
Our high-end image processing software runs cross-platform on Windows, IRIX, Linux, and OSX. And I can tell you that my dual 1GHz Athlon RedHat box already smokes a dual 1.2GHz G4, by about double. And you can now get a dual 2.8 P4. It's not even funny anymore. (By the way, our Windows version is even faster because the Intel compiler is just amazing. Can't wait to use it on Linux.)
Just look at the Photoshop benchmarks on Apple's site: they disingenuously compare a dual 1.25GHz G4 with a single 2.53 P4, and say it's 90% faster. That means against a dual 2.53 P4, it would almost keep up, even with all the full-on Altivec optimization they did for Photoshop. Oh joy!
Intel (or AMD, whatever) has *got* to be the way to go for Apple. The PPC is losing ground fast. You can't keep up with the tremendous consumer-demand-driven processors from Intel. Jump that sinking ship, Jobs!
New York had this for almost a century. You could get a complete breakfast, lunch and/or dinner 24 hours a day from Horn & Hardart's Automat. It was completely unstaffed storefront, with a wall of vending machines in the back, and cafe tables in front. I think the first one opened in 1912. The last one closed in 1991.
The problem with the proprietary hardware is specifically the Motorola CPUs. THEY ARE SLOWER. If you just run, say, gcc3 compiling a large app on an OSX box vs. a comparable-vintage x86 Linux box, the Linux box will blow it away. Or try the Gimp. Or an mp3 encoder. Or just about anything.
Yes, photoshop is fast because it's highly optimized for the PPC's simd/vector instructions, and takes advantage of both CPUs. So's my own software, and I can tell you, x86 creams the PowerPC.
Clock-for-clock, or Mhz-for-Mhz, on regular code without handcoded SIMD (for either CPU), the PPC is a bit faster. But Apple's just shipping the 1.25 GHz PPCs now, and the P4s are at 2.5, pushing 2.8 shortly. Sorry, it's just no contest anymore.
Wake up and smell the coffee, Apple. I love OSX, but real people have real work to do and they need the speed. The PPC is a boutique part. It has no volume, the price will always be exorbitant, and Motorola has next to no other customers demanding high-performance PPCs. Get with the program.
I've been developing on it for about a month now (since C85) and I can say it is REALLY NICE. Perl, python, emacs (latest version), zsh (latest version), and so on -- all included. My Cons-based builds just work, no fuss, no muss, no bother.
One problem that continually bites me though, is the line-ending convention. Some of the system header files use CR-only line-endings. Just try grepping for something in those; grep thinks they're all one long line!
Someone needs to fix the gnu text tools to recognize a lone CR as a line ending, and then I'll be totally happy.
(OK, the NetInfo stuff is a little weird too; I'm pretty used to regular Linux/FreeBSD sys admin stuff, and it seems like they gratuitously replaced a lot of that. But no big deal.)
This was all decided on Usenet in the mid-80s, back when Usenet was a real community.
And yes, over the top is the right answer, unless you are a cat owner.
My great-uncle had one; it was wonderful!
on
The Bulova Accutron
·
· Score: 1
He loved it. He worked at the National Bureau of Standards in Boulder, and he used to hang his Accutron on the bedpost upside-down each night. If he hung it right-side up, it would lose or gain a half-second in the night, or so he said. But hung upside-down, and during the day, it kept perfect time.
Check out the new Abit MAX series. No legacy ports at all. 10 USBs, 3 firewire, 100bT, RAID, etc. This is the wave of the future. I think some other mobo vendors are starting to do the same thing.
However, I'm not getting one because my GPS and PalmV both need serial ports, and a USBSerial adapter is an extra $45 or so! Why can't someone make a super-cheap $10 USBserial adapter?
[Single-user doesn't require Apache.] That's cool. I wonder if it wouldn't be possible to just use samba and nfs to access a repository database without needing any server protocol. Locking might be an issue, perhaps, but maybe Berkeley DB can handle that? Well anyway, I guess it's better to just use Apache as intended.
Several folks pointed out that WebDAV is a protocol that comes with Apache. Sorry, my mistake -- I didn't read the docs carefully enough. Well, I'll put my money where my mouth is: I'm trying it right now. It's taken two hours so far.
APR was dead simple. (No RPM available, but not needed.)
Autoconf: I had a version of autoconf in/usr/bin, and the version subversion needs autoconfig'ed into/usr/local (of course). I fiddled with that, OK. (RPM available.)
libtool 1.4: no problem. (RPM available.)
I downloaded neon, and subversion built it automatically. (RPM available, but I didn't use it.)
Berkeley DB was pretty simple, except that the whole subdir of docs/ that explains the build process was missing in my download -- I found the instructions on the web. This also created errors when doing the 'make install', had to use make -k to work around them. (No RPM for this version available.)
However, when I went to configure subversion, it didn't recognize my Berkeley DB install (in the default place,/usr/local/BerkeleyDB.4.0), so I just copied it into the subversion build tree and it built it OK.
As for apache, the first thing is subversion requires 2.0.40, but the latest I can find is 2.0.39. OK, so I figured out that means I need the latest CVS version. Also, I already have an older 1.3.xx version running because RT requires it, so now I have two versions running simultaneously! A bit of an admin hassle, but again, not a showstopper. (No RPM available for this version.)
(I also installed python 2.0, which required a bunch of other stuff, but I gather that was optional.)
I guess, from the comments I've seen, that I'm the only one who thinks it's weird to require people to install a web server just to do source control (with more than one machine). HOWEVER: now that I've started to try it out, I have to say it's really a pretty cool idea. It might even be useful to my company!
I expect all this will get much simpler, but for now it's not for the faint of heart. Still, from what I can tell I think subversion will be the best post-CVS CVS!
We use CVS here, and like everyone else I'm fed up with the lack of rename support and branching. But looking at the install requirements of Subversion is very intimidating! It requires:
- Berkeley DB, a particular version (this makes sense)
- Apache 2.x
- WebDAV
- Neon
and a bunch of other stuff, IIRC. (Their site is/.ed, so I can't check, sorry) All we need at my company is a server to run on one Linux machine and clients for all the others (MacOSX/WinXP/Linux/IRIX), all within our firewall.
Doesn't all the above stuff, especially the Apache/WebDAV/Neon stuff, seem like overkill just to implement a network protocol for a version control system? Setting up a CVS server is certainly not this complicated, and it seems like with a little more effort on the developers' part, much end-user time and pain could be saved. Does Apache/WebDAV/Neon really buy enough so it's worth the install&admin overhead?
I'm not trying to rag on the Subversion developers; it looks like a really cool system, once you get it up & running. It also looks like they've really done a great job of meeting their goals. I'm definitely looking forward to checking it out -- as soon as I have enough time.
Actually, it *is* about quality. As you say, a pristine 1st gen film print on a clean projector is better than the best digital projection (better colors, finer and less regular grain, wider exposure ratio). After a week or two of constant projection, though, the film gets scratched, dirty, and the colors wash out due to the lamp intensity. And normal film projectors at your suburban multiplex also have sprocket jitter, weave, and other problems. So yes, unless you plan to see everything on opening night at Grauman's Chinese Theater, the quality of a digital print is often likely to be better.
1080p wouldn't fit into the 6MHz bandwidth limit. The others do. 480p is 720x480, 30 full frames per second. 1080i is 1920x1080, 60 interlaced fields per second. 1080p would need twice as much bandwidth.
It all still has to fit into one TV channel, which is 6 MHz wide.
Now nobody knows what will happen. OSX is pretty nice, and our Sapphire plug-ins run pretty nicely on Shake on a dual G4, but still, there are a lot of Windows users out there. Apple is not doing anybody a service by being so close-mouthed about things. People in the post-production business are used to preannouncements of upcoming products, and if Apple won't do that they'll lose out to other companies (Discreet, Quantel, Avid, Sony) who will.
At my company GenArts, we write image processing plug-ins for high end film and video processing systems. Performance is absolutely key. If it takes an hour to render a sequence, even a 5% gain is important. 30-40% is amazing, and our customers REALLY CARE. We're starting to port our stuff to Linux, e.g. for Shake and the fact that the Intel compiler smokes both MSVC and GCC is important for us. We couldn't care less about compile time; we use CL/GCC for debugging builds, then switch to icl toward the end of the cycle.
GCC isn't going to go away any time soon, but anything we can do to get more high-performance real-world apps (like audio and video editors) into Linux (even if only on x86) is good news for the Linux market.
-- SilentTristero
ps: to compiler folks, "kernel" means a small bit of highly computational code, usually with no I/O; not an OS kernel. Kernels to them are things like FFT, convolution, and so on.
I agree with you that they are a fraud, however your argument is incomplete. Most real data is not random. If you can figure out a more effective way to exploit nonrandomness in a data set, you can compress it.
Unfortunately Mr. Shannon has some bad news for these boys with his Sampling Theorem.
Some articles say 20 feet long, Monterey Bay site (mbari.com) says 4 to 5 meters, but nobody says whether this includes the tentacles or not, which makes a big difference.
I've done this with some of the freeware Windows scope programs out there (I'm a former VLSI hardware designer), but even at 96kHz the latency vs. buffer size issue gets to be a problem when probing around in a circuit. It's marginally acceptable for tuning up tape deck heads and so on, but don't try it for any serious design or repair work.
The other problem is of course sound card inputs are AC, you really want DC coupling. And high impedance.
Bottom line: you're better off with a scope board from the back pages of one of the magazines.
Slashdot Mirror
Well, if the review mags like CNet/ZDnet start pointing out that aftermarket refills don't work, and marking down those printers for it, that'll have an effect.
- Tristero
I think you mean googol.
This seems like typical MS bundling. They have a nice PVR app and maybe a codec. Instead of releasing it as the "MS PVR" app or an add-on to Windows Media Player, like anyone else would, they *bundle it into the OS*. WTF? This is exactly what they got reamed for doing with IE, media player, WMA, DRM codecs, etc. etc. And they're still doing the same thing! (Yeah, that DoJ slap on the wrist really hurt...)
Don't fall for this marketing crap. Make them release it as a simple app. I don't care if it will only install if you have the XYZ tuner card or whatever, but it is NOT a new version of the OS! It is NOT "Windows MCE" whatever they tell you! It's just Windows XP with a simple (and quite nice, from what I can tell) full-screen PVR app.
Once they succeed in convincing everyone that the PVR app is "part of the OS" the market for 3rd party PVR apps will dry up, because of course who would pay money for something that comes for free? And then they can cripple it any way they want in future versions (no recording pay-per-view, no skipping commercials, no transcoding, etc.)
Don't buy this line. Anand is starting to behave like a PR flack; it's too bad. Think for yourselves, people! There are a few PVR Windows apps out there, and some of them are getting quite decent. Vote with your wallets.
-- Tristero
My own highly optimized image-processing software bears out these numbers. It's not OSX that's slow, it's the CPU. Come on, Jobs, stop fooling around and get OSX on Intel/AMD!!!
-- Tristero
Minolta. Updated the Dimage7 camera firmware to nearly double the autofocus speed, add direct manual focus (autofocus trim), new color modes, real-time histogram, and lots more. Over a year after product introduction too. Fairly amazing firmware upgrade actually.
Yes, but USB serial adapters are more like $35. I need at least two, one for my Garmin GPS and one for my PalmV. Adding $70 to a motherboard's price just for the cool "legacy-free" label just doesn't cut it for me. I guess my life just isn't legacy-free yet.
Pretty much any FX movie you watch anywhere was at least partially made on a Discreet, and thus on IRIX.
Now as for watching your DVDs, well... Octanes are not really desktop machines. :-)
The idea of radial context menus is they are self-documenting. You can start by looking at the menu, and then after a while the ones you use become second nature. You don't even really have to "learn" them, as opposed to gestures, which, while obviously not so difficult, still have to be looked up somewhere until you get used to them.
And if there's a rarely used gesture, it's utterly useless. In a radial menu, you can at least wait for the menu to show up and then follow its cues.
If a radial menu is well-designed, it becomes pretty similar to a set of gestures. Unfortunately the Mozilla radial menus use tiny hard-to-read icons and so are much too slow to actually use. But in other systems with text menus, they're quite fast to use and learn.
-- Tristero
Our high-end image processing software runs cross-platform on Windows, IRIX, Linux, and OSX. And I can tell you that my dual 1GHz Athlon RedHat box already smokes a dual 1.2GHz G4, by about double. And you can now get a dual 2.8 P4. It's not even funny anymore. (By the way, our Windows version is even faster because the Intel compiler is just amazing. Can't wait to use it on Linux.)
Just look at the Photoshop benchmarks on Apple's site: they disingenuously compare a dual 1.25GHz G4 with a single 2.53 P4, and say it's 90% faster. That means against a dual 2.53 P4, it would almost keep up, even with all the full-on Altivec optimization they did for Photoshop. Oh joy!
Intel (or AMD, whatever) has *got* to be the way to go for Apple. The PPC is losing ground fast. You can't keep up with the tremendous consumer-demand-driven processors from Intel. Jump that sinking ship, Jobs!
-- Tristero
-- Tristero
The problem with the proprietary hardware is specifically the Motorola CPUs. THEY ARE SLOWER. If you just run, say, gcc3 compiling a large app on an OSX box vs. a comparable-vintage x86 Linux box, the Linux box will blow it away. Or try the Gimp. Or an mp3 encoder. Or just about anything.
Yes, photoshop is fast because it's highly optimized for the PPC's simd/vector instructions, and takes advantage of both CPUs. So's my own software, and I can tell you, x86 creams the PowerPC.
Clock-for-clock, or Mhz-for-Mhz, on regular code without handcoded SIMD (for either CPU), the PPC is a bit faster. But Apple's just shipping the 1.25 GHz PPCs now, and the P4s are at 2.5, pushing 2.8 shortly. Sorry, it's just no contest anymore.
Wake up and smell the coffee, Apple. I love OSX, but real people have real work to do and they need the speed. The PPC is a boutique part. It has no volume, the price will always be exorbitant, and Motorola has next to no other customers demanding high-performance PPCs. Get with the program.
-- Tristero
One problem that continually bites me though, is the line-ending convention. Some of the system header files use CR-only line-endings. Just try grepping for something in those; grep thinks they're all one long line!
Someone needs to fix the gnu text tools to recognize a lone CR as a line ending, and then I'll be totally happy.
(OK, the NetInfo stuff is a little weird too; I'm pretty used to regular Linux/FreeBSD sys admin stuff, and it seems like they gratuitously replaced a lot of that. But no big deal.)
This was all decided on Usenet in the mid-80s, back when Usenet was a real community.
And yes, over the top is the right answer, unless you are a cat owner.
He loved it. He worked at the National Bureau of Standards in Boulder, and he used to hang his Accutron on the bedpost upside-down each night. If he hung it right-side up, it would lose or gain a half-second in the night, or so he said. But hung upside-down, and during the day, it kept perfect time.
Thanks for the memory!
-- S.T.
However, I'm not getting one because my GPS and PalmV both need serial ports, and a USBSerial adapter is an extra $45 or so! Why can't someone make a super-cheap $10 USBserial adapter?
-- SilentTristero
[Single-user doesn't require Apache.] That's cool. I wonder if it wouldn't be possible to just use samba and nfs to access a repository database without needing any server protocol. Locking might be an issue, perhaps, but maybe Berkeley DB can handle that? Well anyway, I guess it's better to just use Apache as intended.
Several folks pointed out that WebDAV is a protocol that comes with Apache. Sorry, my mistake -- I didn't read the docs carefully enough. Well, I'll put my money where my mouth is: I'm trying it right now. It's taken two hours so far.
/usr/bin, and the version subversion needs autoconfig'ed into /usr/local (of course). I fiddled with that, OK. (RPM available.)
/usr/local/BerkeleyDB.4.0), so I just copied it into the subversion build tree and it built it OK.
APR was dead simple. (No RPM available, but not needed.)
Autoconf: I had a version of autoconf in
libtool 1.4: no problem. (RPM available.)
I downloaded neon, and subversion built it automatically. (RPM available, but I didn't use it.)
Berkeley DB was pretty simple, except that the whole subdir of docs/ that explains the build process was missing in my download -- I found the instructions on the web. This also created errors when doing the 'make install', had to use make -k to work around them. (No RPM for this version available.)
However, when I went to configure subversion, it didn't recognize my Berkeley DB install (in the default place,
As for apache, the first thing is subversion requires 2.0.40, but the latest I can find is 2.0.39. OK, so I figured out that means I need the latest CVS version. Also, I already have an older 1.3.xx version running because RT requires it, so now I have two versions running simultaneously! A bit of an admin hassle, but again, not a showstopper. (No RPM available for this version.)
(I also installed python 2.0, which required a bunch of other stuff, but I gather that was optional.)
I guess, from the comments I've seen, that I'm the only one who thinks it's weird to require people to install a web server just to do source control (with more than one machine). HOWEVER: now that I've started to try it out, I have to say it's really a pretty cool idea. It might even be useful to my company!
I expect all this will get much simpler, but for now it's not for the faint of heart. Still, from what I can tell I think subversion will be the best post-CVS CVS!
We use CVS here, and like everyone else I'm fed up with the lack of rename support and branching. But looking at the install requirements of Subversion is very intimidating! It requires: /.ed, so I can't check, sorry)
- Berkeley DB, a particular version (this makes sense)
- Apache 2.x
- WebDAV
- Neon
and a bunch of other stuff, IIRC. (Their site is
All we need at my company is a server to run on one Linux machine and clients for all the others (MacOSX/WinXP/Linux/IRIX), all within our firewall.
Doesn't all the above stuff, especially the Apache/WebDAV/Neon stuff, seem like overkill just to implement a network protocol for a version control system? Setting up a CVS server is certainly not this complicated, and it seems like with a little more effort on the developers' part, much end-user time and pain could be saved. Does Apache/WebDAV/Neon really buy enough so it's worth the install&admin overhead?
I'm not trying to rag on the Subversion developers; it looks like a really cool system, once you get it up & running. It also looks like they've really done a great job of meeting their goals. I'm definitely looking forward to checking it out -- as soon as I have enough time.
Actually, it *is* about quality. As you say, a pristine 1st gen film print on a clean projector is better than the best digital projection (better colors, finer and less regular grain, wider exposure ratio). After a week or two of constant projection, though, the film gets scratched, dirty, and the colors wash out due to the lamp intensity. And normal film projectors at your suburban multiplex also have sprocket jitter, weave, and other problems. So yes, unless you plan to see everything on opening night at Grauman's Chinese Theater, the quality of a digital print is often likely to be better.
1080p wouldn't fit into the 6MHz bandwidth limit. The others do. 480p is 720x480, 30 full frames per second. 1080i is 1920x1080, 60 interlaced fields per second. 1080p would need twice as much bandwidth.
It all still has to fit into one TV channel, which is 6 MHz wide.
-- SilentTristero
-- SilentTristero
ps: to compiler folks, "kernel" means a small bit of highly computational code, usually with no I/O; not an OS kernel. Kernels to them are things like FFT, convolution, and so on.
I agree with you that they are a fraud, however your argument is incomplete. Most real data is not random. If you can figure out a more effective way to exploit nonrandomness in a data set, you can compress it.
Unfortunately Mr. Shannon has some bad news for these boys with his Sampling Theorem.
Some articles say 20 feet long, Monterey Bay site (mbari.com) says 4 to 5 meters, but nobody says whether this includes the tentacles or not, which makes a big difference.
What's this thing's wingspan?
I've done this with some of the freeware Windows scope programs out there (I'm a former VLSI hardware designer), but even at 96kHz the latency vs. buffer size issue gets to be a problem when probing around in a circuit. It's marginally acceptable for tuning up tape deck heads and so on, but don't try it for any serious design or repair work.
The other problem is of course sound card inputs are AC, you really want DC coupling. And high impedance.
Bottom line: you're better off with a scope board from the back pages of one of the magazines.