is just communism. If you wrote a book for the sole fact that that is what you do, you're an author and you got compensation from the state (which is the people after all) in the form of food, clothing, entertainment and lodging stipends you'd be set.:-)
I kid. Stick with what we know best. Capitalism with a side of cheating.
As a soon to be published author who is making little money on the deal I don't see the big incentive for me to get all upset about my work being distributed. I mean I don't because I want my publisher to print the books but financially I have nothing riding on the books...
And perhaps with media like that that's the way it should be...
I'm not saying "net on tv" is bad. I'm saying it won't make it any better and as it stands now our network designs are not meant to handle it.
It takes a lot of hardware to make things symmetric. Think of a 8 lane divided highway. If you make 7 lanes go the same direction you get a lot of traffic going in that direction. Now split it evenly. Even though you may not be using the other direction you've lost it. Networking is the same. If I have to send you smaller packets [to give you a chance to send something my way] it means I have to read more ACK packets and what not. It makes the process less efficient.
I'd rather the net be more symmetric as it gives everyone a chance to participate. I hate DSL modems that have a 2Mbit/down and 128Kbit/up (Sympatico used to be that way) ratios, etc, etc.
Just because you can apply technology to something doesn't make the story any better. Like right now after braveheart the show "VIP" came on. VIP == teh stupid. It's a product of "me too" ism. If anything, random ondemand TV will just make that worse. Everyone will be a TV producer and the quality of the entertainment and news will suffer more than it already does.
Also the internet is not meant for broadcast. 80 million people watched Friends each week. That's totally asynchronous. The net is not meant to be so heavily lodsided.
Sure maybe when we can all simultaneously sustain 100mbit/sec from our homes to the net it may be practical but right now it's nowhere near practical enough.
Think about it. At $5 per GB a 4Mbit/sec stream costs you $201 per day. Now suppose get a deal and pay only $0.50 per GB. That's still $20.11 per day per stream. At a minimum they would have to charge you $0.84 per hour. Now look at the average digital package at say 60$ [say you have movies] per month. That's roughly $0.083/hr of viewing.
So right now it's nearly ten times more expensive to watch something over the net. Not to mention how it's not entirely a good use of broadcast resources.
Also TV is not 25mbit/sec. normal TV is more like 2-4mbit [at most]. What NTSC digital subscribers get is more like 2mbit/sec MPEG-1 (it's compariable to what I can encode with tools locally in quality).
HD on the otherhand is ~19-20Mbit/sec.
Anynways, is Raymond or survivor in quad-fi super-HD really that much better?
Though to be honest I don't see of the appeal of HD over the net. It's the same bullshit video tape of a monkey falling out of a tree or something, just now it's got 16 times the pixels.
I really question if people are fundamentally more efficient as people with all their extra doo-dahs and thingy-ma-jobs. I may have the worlds coolest phone but I still need to be at [the very least] a terminal to code properly. Nothing says impractical as typing C source code on a 3x4 numpad!
I think most business types just carry that shit around for status. I mean in all my travels I see people having conference calls and shit on them but it's like, if the meeting was that important you wouldn't be sitting in an airport gate waiting to board...
Also hates ringtones. I swear to god if I hear that stupid Razr flipfone default tone one more time I'll start killing people...
The drivers aren't GPL though and they don't include GPL code. They merely are compatible with GPL code. This is like saying my source files are GPLed because GCC can parse them. Or this webpage is MPLed because Mozilla can read it, etc...
Just because the kernel can load your module doesn't mean your modules is GPLed. The way I understand the GPL is anything you derive from GPL code must be open source and what not. The drivers are proprietary and just happen to be compatible with GPL code.
I don't see the whole holdup on this. I give out free software that is possibly worth millions [depending on who is doing the sales] and yet I routinely do contract and other random work for next to nothing.
There is nothing wrong with copyright when it's applied fairly. It's when you take away the ability for people to derive things from your work product that it becomes a problem. If I write a book and then you copy it verbatim [or largely verbatim] why should that be allowed? Now suppose you write a book based on what I was talking about but you add your own contributions?
DRM strives to remove your ability as a member of society to grow off the work of another which I think is fundamental to the evolvement of humanity [everyone gets ideas from someone else].
I think we can safely separate the original goals of copyrights and that of DRM.
You're saying because some people abuse property rights nobody should have them?
Please give me your home address. I think your right ot own property is "fairly stupid" and I should be free to take your stuff. Give me that freedom you hateful bastard!!!
I've been saying for a long time that passwords should just be stored on a friggin swipe card. It isn't like a reader is advanced technology. While it's not strictly two factor authenication it is better than having the user either just write down the password or use something that is easy to memorize.
At least if you keep the swipe card and your other factor isolated (e.g. on in your pocket the other in your bag or whatever) a compromise of one is not of both...
Swipe cards are also easy to reprogram and revoke [unlike say biometrics]. Storing a random 16-char string on a swipecard is entirely doable and much better than "password" as a password.
"Not true. The memory controller onboard HURTS MP performance, it adds extra latency (beyond 2~4-way) because MP systems have their own complex memory subsystems. Not a smart move from the current king of server chips."... flamebait...
This is total bullshit. Go learn what NUMA is. Software written for NUMA can be much more efficient than that on a traditional FSB setup. Also the FSB approach doesn't scale. As you raise nodes the shared FSB has to get slower and further divided. Memory bandwidth on a 4-way Xeon is MUCH lower than that on a 4-way Opteron.
"they could have more and are just tight lipped"
Changes to the core don't happen overnight. Opteron has gone from 1.6Ghz 130nm cores to 2.6ghz [and higher] dual core 90nm cores. That alone is a huge process change and upgrade.
I can't really say anything about plans but just suppose that AMD isn't totally asleep at the wheel. If you were halfway objective you'd learn what makes the Opteron approach efficient instead of just spouting ignorance.
as far as I know core is based on the Pentium M with a larger cache, etc. It doesn't help that Intel doesn't really have any tech docs on their website.
It's next years MCW series that are going to be more promissing.
I'd still take my Opteron 2P workstation over Core2DuoSuperDuper any date.:-)
DDR2 support will amount to more when DDR2-800 comes about.
And even though next years cores are looking better from Intel they still lack HT links and independent memory controllers which are critical for any sort of MP setup.
Also... you don't think AMD doesn't have platform changes in the woodworks?
Intel is hedging their bets. That's why they still produce so many Netburst based cores while researching a completely new core, etc, etc.
The reason for sticking with x86 is only because of legacy software. If it cost you 5% of your die space to get a fast decoder and the rest is cache/execution parts then its worth it to both the producer and customer.
Think about it. For your 150$ processor 5% of the cost is $7. For that you get an efficient processor that will run all of your existing software without having to rebuild it.
Granted in the OSS world that's moot. So long as gcc has been ported to the ISA one could port Gentoo to it.
Or perhaps they're concerned with more than video game performance?
HT links for example give more than 40% boost to most SMP setups compared to a shared FSB setup.
Dual channel memory, etc, etc, etc.
I think you'd find that a 50Mhz K8 would achieve a higher IPC than a similarly clocked 80486 core. Heck I'd bet you could pit a 100Mhz 486 against a 50Mhz K8 and still have the K8 win, specially with FPU intense code.
I would take it more seriously if they stopped using video games as the benchmark for processors. Video games are more GPU and memory intense than anything else.
Who knows, maybe the ATI or Nvidia drivers are just poorly optimized for Intel or AMD.
It means JACK SHIT.
Why not use real benchmarks using tools people need to do work since really for "funputers" people will buy whatever they think is whizbang cool.
Real developers care about things like IPC, memory throughput, NUMA, etc..
cache is cheap? Hahaha. You know that 1000 dollar Opteron? Look at how much is cache:-)
As for RISC vs CISC that argument is long since dead. After the decoders the processor is effectively RISC. While decoders are not trivial the ALU, AGU, FPU, LSU and other units account for more logic than the decoder. Look at things like PPC or ARM. They're both RISC yet their IPC is LOWER than the K8 core. Even on PPCs which are superscalar.
Granted there is no reason why you couldn't map most of the macro-op control onto the PPC instruction set the market is just not there to spend that much money on it.
In the pre-686 days the processors were entirely CISC. This made for short pipelines and slow execute units that couldn't ramp in frequency. But that pretty much went to the wayside when decoders came about in the x86 world.
Keep in mind this represents their NEW products not current. If they get to retail first they'll enjoy that success [e.g. lower watts/mips] but they're still a ways out.
Slashdot Mirror
is just communism. If you wrote a book for the sole fact that that is what you do, you're an author and you got compensation from the state (which is the people after all) in the form of food, clothing, entertainment and lodging stipends you'd be set. :-)
I kid. Stick with what we know best. Capitalism with a side of cheating.
As a soon to be published author who is making little money on the deal I don't see the big incentive for me to get all upset about my work being distributed. I mean I don't because I want my publisher to print the books but financially I have nothing riding on the books...
And perhaps with media like that that's the way it should be...
Tom
I'm not saying "net on tv" is bad. I'm saying it won't make it any better and as it stands now our network designs are not meant to handle it.
It takes a lot of hardware to make things symmetric. Think of a 8 lane divided highway. If you make 7 lanes go the same direction you get a lot of traffic going in that direction. Now split it evenly. Even though you may not be using the other direction you've lost it. Networking is the same. If I have to send you smaller packets [to give you a chance to send something my way] it means I have to read more ACK packets and what not. It makes the process less efficient.
I'd rather the net be more symmetric as it gives everyone a chance to participate. I hate DSL modems that have a 2Mbit/down and 128Kbit/up (Sympatico used to be that way) ratios, etc, etc.
Tom
And you mark my words ... it'll still suck.
Just because you can apply technology to something doesn't make the story any better. Like right now after braveheart the show "VIP" came on. VIP == teh stupid. It's a product of "me too" ism. If anything, random ondemand TV will just make that worse. Everyone will be a TV producer and the quality of the entertainment and news will suffer more than it already does.
Also the internet is not meant for broadcast. 80 million people watched Friends each week. That's totally asynchronous. The net is not meant to be so heavily lodsided.
Sure maybe when we can all simultaneously sustain 100mbit/sec from our homes to the net it may be practical but right now it's nowhere near practical enough.
Think about it. At $5 per GB a 4Mbit/sec stream costs you $201 per day. Now suppose get a deal and pay only $0.50 per GB. That's still $20.11 per day per stream. At a minimum they would have to charge you $0.84 per hour. Now look at the average digital package at say 60$ [say you have movies] per month. That's roughly $0.083/hr of viewing.
So right now it's nearly ten times more expensive to watch something over the net. Not to mention how it's not entirely a good use of broadcast resources.
Tom
25mbit x 50-100 users on a local hub == lots.
Also TV is not 25mbit/sec. normal TV is more like 2-4mbit [at most]. What NTSC digital subscribers get is more like 2mbit/sec MPEG-1 (it's compariable to what I can encode with tools locally in quality).
HD on the otherhand is ~19-20Mbit/sec.
Anynways, is Raymond or survivor in quad-fi super-HD really that much better?
Tom
"No Shit".
Though to be honest I don't see of the appeal of HD over the net. It's the same bullshit video tape of a monkey falling out of a tree or something, just now it's got 16 times the pixels.
ooooh boy.
Tom
I really question if people are fundamentally more efficient as people with all their extra doo-dahs and thingy-ma-jobs. I may have the worlds coolest phone but I still need to be at [the very least] a terminal to code properly. Nothing says impractical as typing C source code on a 3x4 numpad!
I think most business types just carry that shit around for status. I mean in all my travels I see people having conference calls and shit on them but it's like, if the meeting was that important you wouldn't be sitting in an airport gate waiting to board...
Also hates ringtones. I swear to god if I hear that stupid Razr flipfone default tone one more time I'll start killing people...
Tom
By this logic any tool that needs autoconf should also be GPLed?
Tom
I thought that only went inwards.
... the driver doesn't incorporate ANY GPLed code. It's merely compatible with it.
E.g. if glibc was GPL it means you couldn't write closed source binary only software. Therefore the glibc package is LGPL which allows this.
Now, going the otherway
So you give out a OSS package [and the source] and then include the drivers. The package won't be GPL even though components of it are.
I don't see the problem there.
Tom
It only taints the kernel if you load the module. The kernel itself [the bzImage] is entirely based on GPL code.
/me shakes head...
So don't autoload the drivers and the kernel will not load with a tainted status.
Tom
The drivers aren't GPL though and they don't include GPL code. They merely are compatible with GPL code. This is like saying my source files are GPLed because GCC can parse them. Or this webpage is MPLed because Mozilla can read it, etc...
Just because the kernel can load your module doesn't mean your modules is GPLed. The way I understand the GPL is anything you derive from GPL code must be open source and what not. The drivers are proprietary and just happen to be compatible with GPL code.
I don't see the whole holdup on this. I give out free software that is possibly worth millions [depending on who is doing the sales] and yet I routinely do contract and other random work for next to nothing.
There is nothing wrong with copyright when it's applied fairly. It's when you take away the ability for people to derive things from your work product that it becomes a problem. If I write a book and then you copy it verbatim [or largely verbatim] why should that be allowed? Now suppose you write a book based on what I was talking about but you add your own contributions?
DRM strives to remove your ability as a member of society to grow off the work of another which I think is fundamental to the evolvement of humanity [everyone gets ideas from someone else].
I think we can safely separate the original goals of copyrights and that of DRM.
Tom
Nice troll there.
You're saying because some people abuse property rights nobody should have them?
Please give me your home address. I think your right ot own property is "fairly stupid" and I should be free to take your stuff. Give me that freedom you hateful bastard!!!
Tom
I've been saying for a long time that passwords should just be stored on a friggin swipe card. It isn't like a reader is advanced technology. While it's not strictly two factor authenication it is better than having the user either just write down the password or use something that is easy to memorize.
At least if you keep the swipe card and your other factor isolated (e.g. on in your pocket the other in your bag or whatever) a compromise of one is not of both...
Swipe cards are also easy to reprogram and revoke [unlike say biometrics]. Storing a random 16-char string on a swipecard is entirely doable and much better than "password" as a password.
Tom
Tell your work to invest in two-factor authenticators and don't store your own secrets on your work computer.
Tom
1. Communism is bad
2. WMD in Iraq
3. WMD in Iran
4. No WMD in Israel
5. "We're at war with terrorists" so it's ok to suspend your rights to make you safe.
Nuff said.
Tom [-- hates seeing neighbouring country being destroyed by lunatic security policies]
Two problems with your 1337n355.
1. You use windows.
2. You don't use two-factor authentication.
For two reasons alone you're just a paranoid twat who couldn't draw a threat model to save oneself.
Tom
"Not true. The memory controller onboard HURTS MP performance, it adds extra latency (beyond 2~4-way) because MP systems have their own complex memory subsystems. Not a smart move from the current king of server chips." ... flamebait ...
This is total bullshit. Go learn what NUMA is. Software written for NUMA can be much more efficient than that on a traditional FSB setup. Also the FSB approach doesn't scale. As you raise nodes the shared FSB has to get slower and further divided. Memory bandwidth on a 4-way Xeon is MUCH lower than that on a 4-way Opteron.
"they could have more and are just tight lipped"
Changes to the core don't happen overnight. Opteron has gone from 1.6Ghz 130nm cores to 2.6ghz [and higher] dual core 90nm cores. That alone is a huge process change and upgrade.
I can't really say anything about plans but just suppose that AMD isn't totally asleep at the wheel. If you were halfway objective you'd learn what makes the Opteron approach efficient instead of just spouting ignorance.
Tom
as far as I know core is based on the Pentium M with a larger cache, etc. It doesn't help that Intel doesn't really have any tech docs on their website.
:-)
It's next years MCW series that are going to be more promissing.
I'd still take my Opteron 2P workstation over Core2DuoSuperDuper any date.
Tom
DDR2 support will amount to more when DDR2-800 comes about.
... you don't think AMD doesn't have platform changes in the woodworks?
And even though next years cores are looking better from Intel they still lack HT links and independent memory controllers which are critical for any sort of MP setup.
Also
Tom
Intel is hedging their bets. That's why they still produce so many Netburst based cores while researching a completely new core, etc, etc.
The reason for sticking with x86 is only because of legacy software. If it cost you 5% of your die space to get a fast decoder and the rest is cache/execution parts then its worth it to both the producer and customer.
Think about it. For your 150$ processor 5% of the cost is $7. For that you get an efficient processor that will run all of your existing software without having to rebuild it.
Granted in the OSS world that's moot. So long as gcc has been ported to the ISA one could port Gentoo to it.
Tom
I know your joking but "heat" doesn't care about direction.
:-)
Also consider this, what's the temperature in a vacuum where there are *no* molecules to be moving at all?
Tom
Or perhaps they're concerned with more than video game performance?
HT links for example give more than 40% boost to most SMP setups compared to a shared FSB setup.
Dual channel memory, etc, etc, etc.
I think you'd find that a 50Mhz K8 would achieve a higher IPC than a similarly clocked 80486 core. Heck I'd bet you could pit a 100Mhz 486 against a 50Mhz K8 and still have the K8 win, specially with FPU intense code.
Tom
I would take it more seriously if they stopped using video games as the benchmark for processors. Video games are more GPU and memory intense than anything else.
Who knows, maybe the ATI or Nvidia drivers are just poorly optimized for Intel or AMD.
It means JACK SHIT.
Why not use real benchmarks using tools people need to do work since really for "funputers" people will buy whatever they think is whizbang cool.
Real developers care about things like IPC, memory throughput, NUMA, etc..
Tom
cache is cheap? Hahaha. You know that 1000 dollar Opteron? Look at how much is cache :-)
As for RISC vs CISC that argument is long since dead. After the decoders the processor is effectively RISC. While decoders are not trivial the ALU, AGU, FPU, LSU and other units account for more logic than the decoder. Look at things like PPC or ARM. They're both RISC yet their IPC is LOWER than the K8 core. Even on PPCs which are superscalar.
Granted there is no reason why you couldn't map most of the macro-op control onto the PPC instruction set the market is just not there to spend that much money on it.
In the pre-686 days the processors were entirely CISC. This made for short pipelines and slow execute units that couldn't ramp in frequency. But that pretty much went to the wayside when decoders came about in the x86 world.
Tom
Keep in mind this represents their NEW products not current. If they get to retail first they'll enjoy that success [e.g. lower watts/mips] but they're still a ways out.
Tom