The article is too negative and lacks detail
on
The History Of Pentium
·
· Score: 3, Insightful
The article lacks a lot of detail, especially about the Pentium I. It makes it look like the "addition of MMX" was to only enhancement of the Pentium I. Instead it went through at least two redesigns and shrinks. First from a BiCMOS based P60 and P66 to the later P75-P200 design. The "addition" of MMX brought many additional tweaks as a far improved branch prediction.
The article does also claim that the Pentium I FPU was sub par. This is not true, in fact the design gets the most out of a stack-based FPU without resorting to out-of-order exucution. The FPU of the much praised contender at that time, the 68060 was as much as three times slower due to lack of pipelining.
Some flaws in the Pentium I designs: Waste of resources for a dual read data cache, which is rarely utilized. Dog slow shift and integer multiplication as compared to motorolas offerings, but intel kept the strategy also in later CPUs.
You really think "the popular demo" looks like anything people would show at raves? I think you need to get out more...
Great for paranoid nuts, useless for real people
on
RF-Blocking Wallpaper
·
· Score: 3, Interesting
Ok, its nice. This wallpaper blocks a lot of RF radation. This means that you can not use WLAN, cellphones and terrestrial TV/Radio. Is this really what you want?
Seriously, if you'd want to have this in your living or work-room, you have issues. The only place I could this imagine at is a childrens room or maybe a cheesy nightclub.
Ok, this being slashdot and so on - I know lots of people like to make claims without checking any background first. Many people here claim the retraction of his PHD title is either insignificant or not justifified. In fact this is a pretty large science scam. I challenge you to actually read some background information:
Yep, Britain has a long and proud history of inventing things and then not capitalising on those inventions.
If you really want a good example for this, take Alan Turing. He was driven to suicide by british officials due to his homosexuality . And that despite his service for the country.
Transister Invented by Julius E. Lilienfeld, 1924. German emigrant to America. This is an undisputed fact, he even has patents on it
ICs Difficult to figure this out, but it is rather an evolutionary step from planar bipolar transistors. I think its fair to acknowledge the people who did the first working prototypes..
LCDs No, Merck in germany developed the first liquid crystals. I dont know who the first person was who used it for displays, but afair it was an american.
Why do they even try? As I take it they can hardly catch up with the NASA, taken all the envirowhiners in europe. Why do they even waste money on it then?
Good Article, seems to be quite accurate on details. And yes, I did actually read it.
I also made some attempts on vcs2600 programming some years ago. It could not be any more different from your daily C/php/... hacking. Think of microcontroller programming with even more demanding timing.
The machine has 128bytes (yes, bytes) of ram and 4-6kb of ROM. No video ram, everything is generated on the fly. The CPU does not support interrupts, all the timing is done by active waiting.
Not a lot of people in India speak French, German, Spanish, Italian, etc etc.
Dont worry, the new eastern european EU member have ample supply of cheap workers than are fluent in german, french and other languages. Spanish and Italian may not be required as these countries already have their own underpaid workers in the poorer parts of the country.
The application of Linux in India lowers cost of setting up a new software sweat shop. It also allows more people to be schooled in software development, that previously could not afford education. Implication => More US jobs can be outsourced, cheaper.
I am not sure any American software developer would like to do anything to help getting more IT jobs outsourced.
I'm pretty sure the V2 postdates 1938 by a half decade or so. WW-II didn't start until 1939, and V1s dropped on London some years after that.
The first successfull suborbital flight of an A-4 was in 1942 and you can be pretty sure that the flight control computer was already being developed for years before that. Well, it was just an example. I am pretty sure there are many similar developments in the 30ies.. "Analog computer" is a very broad definition.
Andy Bechtolsheim # Co-founder of Sun Microsystems
- One of the most impressing entrpreneurs of this time. Probably the only billionaire who still gets down and dirty..
Andy Grove # Co-founder and former president of Intel
- Not only the founder of the most successfull IC company, but he also did real research (He has some very interesting papers on crucial topics related IC stability)
Konrad Zuse # Inventor of the Z-1 through 3 machines, early program-controlled (using relays) computers
- Hands down, he build the first programmable computer. And does thus deserve credit.
John Von Neumann # Designer of EDVAC and IAS computers
- I'd rather credit him for inventing the concept of modern computers.
Ken Thompson # Co-developer (with Dennis Ritchie) of UNIX operating system for Bell Labs # Co-led (with Dennis Ritchie) team that developed the C programming language
- Naturally.. sorry Linus, they were first!
Antivote:
Philo T. Farnsworth # Inventor of modern television
-Statement is not true, this is an urban legend. I also do not see how this is related to computers?
George Philbrick # Inventor of the first fully electronic analog computer in 1938
-Doubt that too, there were many at the same time. For example the V2 flight control computer..
The transistor count on the video cards does not count the ram
How do you know? In fact, modern GPUs require a large amount of small scattered memory blocks. Texture caches, FIFOs for fragment/pixels/texels when they are not in sync, caches for vertex shader and pixel shader programs etc etc..
More recent GPUs are notorious for their incredibly long latencies. Long latencies imply that a lot of data has to be stored in chip..
Everything you say - theoretically ok. But thats not how it is done in reality. There is almost never any reason to process something in sequential frames because of design trade offs. Specifically in this application.
I have seen how it is done in reality and what is still up and coming.. YMMV but I stand with my point:)
Sorry, the quality of your arguments does not improve by insulting me.
Regarding your latency idea: The streamrate of pixels,lines and frames is well known. Are you trying to tell me, that you can get a delay of 4*20ms just by pipelining on a per line or per pixel basis? How many pipeline stages would that need? Where is the crossover between lots of slow sequential steps and fewer faster ones?
My point was, that the there is no way to drive the system cost down by increasing the latency by multiple frames. In fact that is an algorithmic necessity and not an attempt to cut cost, as already pointed out in numerous examples by others (3:2 pulldown, Mpeg2-decoding, motion interpolation)
And thats a sure fact, take it or leave it. I know the other side of these marketing descriptions, because I actually work with and know many people in that area.
The point of the manufacturers margin is void since all the big players in that biz have their own fabs. This means that yield is everything.
What the processor actually uses is something they don't make public knowledge
Nah, better do your homework:
http://www.semiconductors.philips.com/products/n ex peria/home/applications/tv/index.html
If you have a one chip soluation the cost analysis is quite simple: You have to settle down for one technology (usually dictated by whats available) and the die size. Any on chip framebuffer increases the die size by a two digit percentile and does thus drive yield down and cost up.
Any additional frame of latency will require an additional frame buffer. Believe me, in current technology there is no sweet spot which asks for increase the number of framebuffers over the bare minimum.
As I wrote - framebuffers are expensive, you do not want to increase the number over the bare minimum.
P.s.: We are talking about signal processing in a cost efficient architecture here, this is different from render-to-texture algorithms in overpowered 3D accelerators.
Let's say you have a pipeline of frames you're processing. There may be multiple frames in flight at different stages of processing. The longer it takes for for a frame to enter in one state and leave in it's final state doesn't nescicarily have any impact on the number of frames you can process
Yes theoretically you are right. However any processing step on a frame is still sequential. It does algorithmically not make sense to pipeline across frames. There are still several levels of hierarchy inbetween. (pixels, lines, fields.. ) If you think you have to pipeline entire frames, your algorithm is probably bad. As I wrote - framebuffers are expensive, you do not want to increase the number over the bare minimum.
Both original article and slashdot posting are quite inaccurate..
The delay is not caused by cheap processors, that is a myth. Just think about it, even delaying the video by 1 second will not reduce the required processing power...
In fact the delay is a technical neccessity for some of the algorithms employed in modern television. For example motion interpolation for 100Hz TV requires the knowledge of at least one frame in advance.
Also the "delay" in TFTs, as mentioned, has nothing in common with the delay due to video preprocessing....
The only remedy for this problem is to have an option to turn all the preprocessing off for video games and have an artificial audio delay, so it matches the video. Nothing that is out of bounds for an average TV...
Slashdot Mirror
The article lacks a lot of detail, especially about the Pentium I. It makes it look like the "addition of MMX" was to only enhancement of the Pentium I. Instead it went through at least two redesigns and shrinks. First from a BiCMOS based P60 and P66 to the later P75-P200 design. The "addition" of MMX brought many additional tweaks as a far improved branch prediction.
The article does also claim that the Pentium I FPU was sub par. This is not true, in fact the design gets the most out of a stack-based FPU without resorting to out-of-order exucution. The FPU of the much praised contender at that time, the 68060 was as much as three times slower due to lack of pipelining.
Some flaws in the Pentium I designs: Waste of resources for a dual read data cache, which is rarely utilized. Dog slow shift and integer multiplication as compared to motorolas offerings, but intel kept the strategy also in later CPUs.
You really think "the popular demo" looks like anything people would show at raves? I think you need to get out more...
Ok, its nice. This wallpaper blocks a lot of RF radation. This means that you can not use WLAN, cellphones and terrestrial TV/Radio. Is this really what you want?
Seriously, if you'd want to have this in your living or work-room, you have issues. The only place I could this imagine at is a childrens room or maybe a cheesy nightclub.
Nanoweb is an HTTP server written in PHP, ..
Yes, but not TCP/IP stack.
Yes, it is new. The difference is that it is not only a web server, but also an entire TCP/IP stack!
Ok, this being slashdot and so on - I know lots of people like to make claims without checking any background first. Many people here claim the retraction of his PHD title is either insignificant or not justifified. In fact this is a pretty large science scam. I challenge you to actually read some background information:
Comprehensive background information
Yep, Britain has a long and proud history of inventing things and then not capitalising on those inventions.
If you really want a good example for this, take Alan Turing. He was driven to suicide by british officials due to his homosexuality . And that despite his service for the country.
Transister
Invented by Julius E. Lilienfeld, 1924. German emigrant to America. This is an undisputed fact, he even has patents on it
ICs
Difficult to figure this out, but it is rather an evolutionary step from planar bipolar transistors. I think its fair to acknowledge the people who did the first working prototypes..
LCDs
No, Merck in germany developed the first liquid crystals. I dont know who the first person was who used it for displays, but afair it was an american.
Why do they even try? As I take it they can hardly catch up with the NASA, taken all the envirowhiners in europe. Why do they even waste money on it then?
Good Article, seems to be quite accurate on details. And yes, I did actually read it.
I also made some attempts on vcs2600 programming some years ago. It could not be any more different from your daily C/php/... hacking. Think of microcontroller programming with even more demanding timing.
The machine has 128bytes (yes, bytes) of ram and 4-6kb of ROM. No video ram, everything is generated on the fly. The CPU does not support interrupts, all the timing is done by active waiting.
Not a lot of people in India speak French, German, Spanish, Italian, etc etc.
Dont worry, the new eastern european EU member have ample supply of cheap workers than are fluent in german, french and other languages. Spanish and Italian may not be required as these countries already have their own underpaid workers in the poorer parts of the country.
The application of Linux in India lowers cost of setting up a new software sweat shop. It also allows more people to be schooled in software development, that previously could not afford education. Implication => More US jobs can be outsourced, cheaper.
I am not sure any American software developer would like to do anything to help getting more IT jobs outsourced.
btw:
Impressing
I'm pretty sure the V2 postdates 1938 by a half decade or so. WW-II didn't start until 1939, and V1s dropped on London some years after that.
The first successfull suborbital flight of an A-4 was in 1942 and you can be pretty sure that the flight control computer was already being developed for years before that. Well, it was just an example. I am pretty sure there are many similar developments in the 30ies.. "Analog computer" is a very broad definition.
The CRT computer monitor is a close-enough relative of the TV to count as "yes, that's Farnsworth too".
c at hoderaytube.htm
The CRT was invented by Ferdinand Braun in 1897.
http://inventors.about.com/library/inventors/bl
Well spoken..
The list is quite arbitrary, they should categorize their hall of fame.
Andy Bechtolsheim
# Co-founder of Sun Microsystems
- One of the most impressing entrpreneurs of this time. Probably the only billionaire who still gets down and dirty..
Andy Grove
# Co-founder and former president of Intel
- Not only the founder of the most successfull IC company, but he also did real research (He has some very interesting papers on crucial topics related IC stability)
Konrad Zuse
# Inventor of the Z-1 through 3 machines, early program-controlled (using relays) computers
- Hands down, he build the first programmable computer. And does thus deserve credit.
John Von Neumann
# Designer of EDVAC and IAS computers
- I'd rather credit him for inventing the concept of modern computers.
Ken Thompson
# Co-developer (with Dennis Ritchie) of UNIX operating system for Bell Labs
# Co-led (with Dennis Ritchie) team that developed the C programming language
- Naturally.. sorry Linus, they were first!
Antivote:
Philo T. Farnsworth
# Inventor of modern television
-Statement is not true, this is an urban legend. I also do not see how this is related to computers?
George Philbrick
# Inventor of the first fully electronic analog computer in 1938
-Doubt that too, there were many at the same time. For example the V2 flight control computer..
The transistor count on the video cards does not count the ram
How do you know? In fact, modern GPUs require a large amount of small scattered memory blocks. Texture caches, FIFOs for fragment/pixels/texels when they are not in sync, caches for vertex shader and pixel shader programs etc etc..
More recent GPUs are notorious for their incredibly long latencies. Long latencies imply that a lot of data has to be stored in chip..
ok.. now we are at the start of the thread again.
:)
Everything you say - theoretically ok. But thats not how it is done in reality. There is almost never any reason to process something in sequential frames because of design trade offs. Specifically in this application.
I have seen how it is done in reality and what is still up and coming.. YMMV but I stand with my point
Have a nice day..
Sorry, the quality of your arguments does not improve by insulting me.
Regarding your latency idea: The streamrate of pixels,lines and frames is well known. Are you trying to tell me, that you can get a delay of 4*20ms just by pipelining on a per line or per pixel basis? How many pipeline stages would that need? Where is the crossover between lots of slow sequential steps and fewer faster ones?
My point was, that the there is no way to drive the system cost down by increasing the latency by multiple frames. In fact that is an algorithmic necessity and not an attempt to cut cost, as already pointed out in numerous examples by others (3:2 pulldown, Mpeg2-decoding, motion interpolation)
And thats a sure fact, take it or leave it. I know the other side of these marketing descriptions, because I actually work with and know many people in that area.
The point of the manufacturers margin is void since all the big players in that biz have their own fabs. This means that yield is everything.
What the processor actually uses is something they don't make public knowledge
n ex peria/home/applications/tv/index.html
Nah, better do your homework:
http://www.semiconductors.philips.com/products/
If you have a one chip soluation the cost analysis is quite simple: You have to settle down for one technology (usually dictated by whats available) and the die size. Any on chip framebuffer increases the die size by a two digit percentile and does thus drive yield down and cost up.
Any additional frame of latency will require an additional frame buffer. Believe me, in current technology there is no sweet spot which asks for increase the number of framebuffers over the bare minimum.
I believe you have no idea of the algorithms..
As I wrote - framebuffers are expensive, you do not want to increase the number over the bare minimum.
P.s.: We are talking about signal processing in a cost efficient architecture here, this is different from render-to-texture algorithms in overpowered 3D accelerators.
Let's say you have a pipeline of frames you're processing. There may be multiple frames in flight at different stages of processing. The longer it takes for for a frame to enter in one state and leave in it's final state doesn't nescicarily have any impact on the number of frames you can process
Yes theoretically you are right. However any processing step on a frame is still sequential. It does algorithmically not make sense to pipeline across frames. There are still several levels of hierarchy inbetween. (pixels, lines, fields.. ) If you think you have to pipeline entire frames, your algorithm is probably bad. As I wrote - framebuffers are expensive, you do not want to increase the number over the bare minimum.
Both original article and slashdot posting are quite inaccurate..
The delay is not caused by cheap processors, that is a myth. Just think about it, even delaying the video by 1 second will not reduce the required processing power...
In fact the delay is a technical neccessity for some of the algorithms employed in modern television. For example motion interpolation for 100Hz TV requires the knowledge of at least one frame in advance.
Also the "delay" in TFTs, as mentioned, has nothing in common with the delay due to video preprocessing....
The only remedy for this problem is to have an option to turn all the preprocessing off for video games and have an artificial audio delay, so it matches the video. Nothing that is out of bounds for an average TV...