AMD's never been really strong in the mobile area. IMHO, it's because they're not a platform developer-supplier like Intel is, so they have to deal with inferior VIA chipsets with fairly crappy onboard video. At least that's always been my reasons for not even considering AMD a viable x86 mobile alternative to intel. ATI's recent jump into the chipset market changes the picture a bit, but if you want low power, obviously, the Athlon's not what you want in your laptop.
For the most part that's true, but it also comes down to the question of what platform runs your $20k per seat license workstation software fastest AND most reliably. In which case, you might decide an entire platform and just go with the standard IT OEM buy. In this case, Intel does have some nice control because they sell the platform, not just the processor. After all, if you're an OEM and you want to sell intel chips, do you offer them on VIA or Sis chipset motherboards if you're, say, Dell or even IBM?
Now if my software ran in 10 minutes on an AMD box vs 30 minutes on Intel box and they asked what they could do to improve my performance, I'd ask them to direct me to an OEM building around their reference systems--after all, the performance is already there, I just want reliability, and no company should be sending out crappy machines as reference boxes.
I'm no lawyer, but I think that only applies in criminal cases. Since this is a civil matter, they still have to prove you're guilty, but you also have to prove you're innocent.
Makes you wish you could find his collection of audio and video stuff and set it on fire and when he screams about it, you tell him he didn't need backups, right?
LOL Actually I really wish I could do this, but I'd probably enjoy it too much.
The Enterprise is a museum piece that was used to test the structure of the shuttle in atmosphere. I don't think it ever even glided in anywhere. Anyway, it became a museum piece a little before Columbia was built and launched, IIRC
Xeon's going to end up being their immediate competition, but in that market, brand has more say then performance or platform cost (nevermind the software). Pick up in research and universities would be a nice plus for AMD and associated system builders.
Must be why the server side using Linux is still getting released on time. Hrm, maybe I should pick up an opteron in the summer JUST to run Linux on it.
I believe unix oses behave similarly. At least Solaris 9 (? or was it 8) did that. Of course, we generally ran out of RAM before we ran out of CPU time.
Actually I think it has more to do with Altivec's permute function allowing bit shifts and bit rotations all in one cycle. There was a paper about this on ars-technica.
This was about Rambus' underhanded tactics with JEDEC which required all members to put patents related to what they were pushing as far as SDR and DDR SDRAM on the table BEFORE they made the specs final. Rambus, however, didn't do that. Instead, they held onto their patents while helping all the ram manufacturers settle on the SDRAM and DDR SDRAM specifications. Later, when the market wasn't adopting RDRAM fast enough--and the memory manufacturers saw no need to get RDRAM licenses (only samsung was making them at the time), Rambus went out and started telling the same ram manufacturers they worked with that they had to pay licensing fees.
Technologically RDRAM's fine with some weaknesses early on as far as the desktop platform (embedded was another matter). Had Rambus acted honestly from the start, this whole matter could have been avoided.
Hrm, I could be wrong, but I thought civil engineers also had to go through mentor programs and get professionally certified before they could go out building bridges and the like?
RISC instructions must retire in 1 cycle after the pipeline's been filled (assume no bubbles), this is part of the nature of RISC in the pipelined world. RISC is also fixed instruction length.
You are correct about SOME CISC instructionS as it may tell the processor to do something in a very SIMD fashion (fetch all data between these 2 memory address and do something to them then retire).
RISC was the move to load/store architectures (you fetch from memory to load, but you do real work in registers) and fixed length instructions that could be implimented in pipeline fashion--which means you're supposed to be able to arbitrarily push up clock speed. Pipelining, however, has penalties (hence some companies build very wide, very shallow processors--see ars-technica, Hannibal is VERY good at explaining this in detail).
It's all part of the decode stage, simple instructions (the set that follow the RISC model in the first place) Can generally hit the lookup table immediately, the more complex DMA style instructions (string handling, etc.) need to be broken down and the microps ordered correctly. Much of this takes less time then you think, although if intel wanted pure performance, the entire process would be done without a lookup table (like alpha-all logic, no lookup). BTW all processors have an instruction decode stage, how it is broken up and how it works, as you can expect varies.
As for instruction reordering for performance, as stated before, the compiler's supposed to do the work, and intel's compiler writers are incredible at their job. In a lot of ways, the P4 was perfectly engineered to leverage intel's strengths all the way through the design and production stage, and the power of their compiler FOR THEIR processor is one of the examples of this.
Actually he's not delusional, I guess you missed the slashdotted article about the RIAA report that only saw 10% of people NOT buying CD's AND downloading. I would assume that 90% of the people surveyed were either not buying CDs or downloading, or they were downloading AND buying CDs.
I don't want to JUST preview music, I want to buy it, albeit, one track at a time, if they sold me an mp3 in 320kbps for $1, I'd buy it--even before napster, they were lucky to get one CD out of me per year. It's all about selling a product with a value people will pay for.
I'm not saying people don't like getting stuff for free, they do, but you're even more of a pessimist than me if you believe people will steal from you at EVERY opportunity. People are generally willing to pay a fair price for a fair product. Of course, if you try to fleece people out of their money, they'll strike back, and they don't need to do so in a legal manner. Remember, if you're too busy watching for the knife in your back, you'll just find it in your chest one day.
It sounded more like a settlement to me, I'd rather see them lose and lose big. I'm talking, no mercy, we're taking you to the cleaners so you better bring lube for the reaming you're gonna get when we get there.
It'd be funnier if you tried to get them patented and the USPTO actually gave you a patent for 'discovering' your own genes. Even I have to think patent examiners aren't THAT idiotic--although I'm afraid they might not be that far off.
As far as a consumer is concerned, usage is a factor in the value of an item. I get many more hours of entertainment out of a DVD then I EVER have out of a CD, after all, how long can you listen to the best 3 to 5 songs on a CD 'fore you're done.
At least with a DVD, I get ~2 hours of enjoyment everytime I open it up and watch it, or flip through the specials, or read a translator's notes and see if I catch any new meaning behind the movie from the notes.
Even when buying a game, I consider hours of gameplay and replayability. There are some games where you just want to beat them once and that's that, there are others that you can replay over and over and the enjoyment doesn't quite go away. Sure it diminishes over time, but I've always gone back to my favorites in 6 months to a year to feed the addiction, as light as it is.
Slashdot Mirror
Holy Cow man!!! Where do you live?!? The 9500 Pro should be more like $25 to $50 more then the Ti4200.
AMD's never been really strong in the mobile area. IMHO, it's because they're not a platform developer-supplier like Intel is, so they have to deal with inferior VIA chipsets with fairly crappy onboard video. At least that's always been my reasons for not even considering AMD a viable x86 mobile alternative to intel. ATI's recent jump into the chipset market changes the picture a bit, but if you want low power, obviously, the Athlon's not what you want in your laptop.
For the most part that's true, but it also comes down to the question of what platform runs your $20k per seat license workstation software fastest AND most reliably. In which case, you might decide an entire platform and just go with the standard IT OEM buy. In this case, Intel does have some nice control because they sell the platform, not just the processor. After all, if you're an OEM and you want to sell intel chips, do you offer them on VIA or Sis chipset motherboards if you're, say, Dell or even IBM?
Now if my software ran in 10 minutes on an AMD box vs 30 minutes on Intel box and they asked what they could do to improve my performance, I'd ask them to direct me to an OEM building around their reference systems--after all, the performance is already there, I just want reliability, and no company should be sending out crappy machines as reference boxes.
I'm no lawyer, but I think that only applies in criminal cases. Since this is a civil matter, they still have to prove you're guilty, but you also have to prove you're innocent.
Makes you wish you could find his collection of audio and video stuff and set it on fire and when he screams about it, you tell him he didn't need backups, right?
LOL Actually I really wish I could do this, but I'd probably enjoy it too much.
The Endeavor, you mean?
The Enterprise is a museum piece that was used to test the structure of the shuttle in atmosphere. I don't think it ever even glided in anywhere. Anyway, it became a museum piece a little before Columbia was built and launched, IIRC
*I* said nothing of the sort ;-). I think you meant the other guy.
Xeon's going to end up being their immediate competition, but in that market, brand has more say then performance or platform cost (nevermind the software). Pick up in research and universities would be a nice plus for AMD and associated system builders.
32-bits maxes out at 4 GB. Doesn't it?
Opteron's got a 128-bit path to RAM, the Athlon's just 64-bits to RAM. There'll probably some cache size differences too.
Must be why the server side using Linux is still getting released on time. Hrm, maybe I should pick up an opteron in the summer JUST to run Linux on it.
I believe unix oses behave similarly. At least Solaris 9 (? or was it 8) did that. Of course, we generally ran out of RAM before we ran out of CPU time.
you can find them on JC's cross comparison, they're old though
JC's Home Page
Scroll down, on the right, under Benchmarking.
Actually I think it has more to do with Altivec's permute function allowing bit shifts and bit rotations all in one cycle. There was a paper about this on ars-technica.
This was about Rambus' underhanded tactics with JEDEC which required all members to put patents related to what they were pushing as far as SDR and DDR SDRAM on the table BEFORE they made the specs final. Rambus, however, didn't do that. Instead, they held onto their patents while helping all the ram manufacturers settle on the SDRAM and DDR SDRAM specifications. Later, when the market wasn't adopting RDRAM fast enough--and the memory manufacturers saw no need to get RDRAM licenses (only samsung was making them at the time), Rambus went out and started telling the same ram manufacturers they worked with that they had to pay licensing fees.
Technologically RDRAM's fine with some weaknesses early on as far as the desktop platform (embedded was another matter). Had Rambus acted honestly from the start, this whole matter could have been avoided.
Hrm, I could be wrong, but I thought civil engineers also had to go through mentor programs and get professionally certified before they could go out building bridges and the like?
RISC instructions must retire in 1 cycle after the pipeline's been filled (assume no bubbles), this is part of the nature of RISC in the pipelined world. RISC is also fixed instruction length.
You are correct about SOME CISC instructionS as it may tell the processor to do something in a very SIMD fashion (fetch all data between these 2 memory address and do something to them then retire).
RISC was the move to load/store architectures (you fetch from memory to load, but you do real work in registers) and fixed length instructions that could be implimented in pipeline fashion--which means you're supposed to be able to arbitrarily push up clock speed. Pipelining, however, has penalties (hence some companies build very wide, very shallow processors--see ars-technica, Hannibal is VERY good at explaining this in detail).
It's all part of the decode stage, simple instructions (the set that follow the RISC model in the first place) Can generally hit the lookup table immediately, the more complex DMA style instructions (string handling, etc.) need to be broken down and the microps ordered correctly. Much of this takes less time then you think, although if intel wanted pure performance, the entire process would be done without a lookup table (like alpha-all logic, no lookup). BTW all processors have an instruction decode stage, how it is broken up and how it works, as you can expect varies.
As for instruction reordering for performance, as stated before, the compiler's supposed to do the work, and intel's compiler writers are incredible at their job. In a lot of ways, the P4 was perfectly engineered to leverage intel's strengths all the way through the design and production stage, and the power of their compiler FOR THEIR processor is one of the examples of this.
Even though the only real decision makers are the executives and board of directors?
Actually he's not delusional, I guess you missed the slashdotted article about the RIAA report that only saw 10% of people NOT buying CD's AND downloading. I would assume that 90% of the people surveyed were either not buying CDs or downloading, or they were downloading AND buying CDs.
I don't want to JUST preview music, I want to buy it, albeit, one track at a time, if they sold me an mp3 in 320kbps for $1, I'd buy it--even before napster, they were lucky to get one CD out of me per year. It's all about selling a product with a value people will pay for.
I'm not saying people don't like getting stuff for free, they do, but you're even more of a pessimist than me if you believe people will steal from you at EVERY opportunity. People are generally willing to pay a fair price for a fair product. Of course, if you try to fleece people out of their money, they'll strike back, and they don't need to do so in a legal manner. Remember, if you're too busy watching for the knife in your back, you'll just find it in your chest one day.
It sounded more like a settlement to me, I'd rather see them lose and lose big. I'm talking, no mercy, we're taking you to the cleaners so you better bring lube for the reaming you're gonna get when we get there.
So how much do you think an artist REALLY gets per CD--after the label recoups 'expenses'?
It'd be funnier if you tried to get them patented and the USPTO actually gave you a patent for 'discovering' your own genes. Even I have to think patent examiners aren't THAT idiotic--although I'm afraid they might not be that far off.
^_^ Due to Godwin's law, we're unable to even discuss the possibility of committing small scale genocide against her bloodline.
As far as a consumer is concerned, usage is a factor in the value of an item. I get many more hours of entertainment out of a DVD then I EVER have out of a CD, after all, how long can you listen to the best 3 to 5 songs on a CD 'fore you're done.
At least with a DVD, I get ~2 hours of enjoyment everytime I open it up and watch it, or flip through the specials, or read a translator's notes and see if I catch any new meaning behind the movie from the notes.
Even when buying a game, I consider hours of gameplay and replayability. There are some games where you just want to beat them once and that's that, there are others that you can replay over and over and the enjoyment doesn't quite go away. Sure it diminishes over time, but I've always gone back to my favorites in 6 months to a year to feed the addiction, as light as it is.