It's not all that bad. Think of the plus side: free porn. It is not the strongest of the species that survive, nor the most intelligent, but the one most responsive to change.
I think its not the strongest speciest nor most intelligent the one that BREEDS most!
Get a OLD VAX, lock the front cover. Keep your laptop inside of it! No one is going to steal that one. Uhh. If you dorm room is in florida don't turn it on. If your dorm is in canada open the window and keep it on through the winter.
Uhh. Think again.
There is more to power than the CPU power. The solution I would prefer would be QUAD opteron. You could get full tower case, put a single MB in it, and run it. 2.5Ghz opteron is over double as fast as those poor transmeta chips, now having 4 of them in a single system image with verylow memory latency will make it even faster relatively since the TMsolution would use the Gb ethernet for connecting eachothers memory!. The 2nd solution is replacement for the deskside case for which it doesn't matter if its bigger than the TM solution since your not going to use that area in anything else anyway, so there could be full rack as well. Power comsumption of 2nd solution is bigger but then again its MUCH cheaper than shelling off 100k$ for the 96 processor TM solution! Getting myrinet and Dual opterons in a rack would definitely beat the TM solution for similar price.
Get quad opteron. It should get about better performance for same price as the other but without need for clustering, for the small system. For bigger system, you could use myrinet and dual opterons. Oh what the heck. It costs 600-800$ per processor to build a rack of Athlon64 based cluster with Gb ethernet. So this effineon based cluster would be beaten with system costing less than their solution. So it is beaten from two different solutions. A) Getting single image opteron system if communicationlatencies are important. B) Getting cluster of AMD64 if price/performance was important. ONLY thing they bring is density of A for system type of B while costing more than A.
Tribes: Veangeance ->new game, new servers the problem you mentioned gone. So besides getting the new gamers to the game they can move the crowds from old game to the new game simply by suplying them with annoying newbies.
apt-get update apt-get dist-upgrade 12 years from now...[6 years to implmentation and assume they use it atleast 6 years after that...] Its Another point is the movie and entertainment. Streaming video from home computer to home computer at 3000x2000 resolution. Video conferences so that there will be 16 of said streams needed, [30" flatpanels are going to be cheap.] Consider the idea that there are 4 family members sharing the connection too.
IPV6 is doing something on latency, the thing their design is going to do something on latency too. [By reducing conversion between optical network and other network.]
For the backbone bandwith. If you can wave multiplex 200 different frequencies and run at 50Ghz operation and bundle 1000 fibers to a single cable you are lacking some bandwith so they need some improvement on backbone to get em the bandwith, but still they could just assume for first generation that not everyone is going to use entire bandwith at same time, so you could just over subsribe it a lot, and just draw the 10Gb lines for last mile as those lines are NOT going to be upgraded. And upgrade the backbone as soon as there is technology todo that.
The backbone bandwith requirement will be huge which makes me wonder the powercomsumption of said network equipment!
There will be some die shrinks before 2010 so they may put single chip router, for it. With address table on same die/package as the routing logic. Hopefully mostly optical networking, etc... The cooling of said single chip router maybe liquid, so they need extremely reliable liquid cooling available.
Here's your armour lets have a snow fight. hehehehehehe cesium muahhahhaha, boom HEHEHHE boom. Snow+cesium boom, water and cesium boom ice[T>-195'c] and cesium boom. No radio action hehehe just chemistry muahhahhah.
Adolf Hitler at first USED religion when he wasn't in power as a aid towards his goal, then PERSECUTED IT! Thats correct lutheran priests in nazi germany had some persecution going. As for Wermach belts that historical, not something that Hitler put there.
the famous Lutheran priest Martin Niemöller from Nazi Germany and his laconic remarks: "First, they took the communists," said Niemöller, "and I did not protest, since I am not a communist. Then, they abducted the Jews, but I did not say anything, for I am not a Jew. Afterwards, they arrested the catholics, but why should I bother, I am not a catholic? And when they came to get hold of me, there was, strangely enough, nobody left to protest."
"It would be misleading to give the impression that the persecution of Protestants and Catholics by the Nazi State tore the German people asunder or even greatly aroused the vast majority of them. It did not. A people who had so lightly given up their political and cultural and economic freedoms were not, except for a relatively few, going to die or even risk imprisonment to preserve freedom of worship."
I wouldn't call that an act of devouted christian! Nor I would call a Germany a christian nation if people just ignored the the arresting of large portion of its priests of majority churches and so on. I'd call Nazi Germany about as christian nation as US is, slighly less. I would call the Nazi Germany mostly atheists and christians by name but without acting on it. I think the thinking Hitler as christian may come from one certain political tool nazis took.
The Nazi-controlled German Christians Faith Movement dominated mainstream churches by mixing dogma with politics and proclaimed: "In the person of the Fuhrer we behold the One sent from God who places Germany in the presence of the Lord of History."
They simply spoke their own politics with few selected words from bible and called that a christian religion. Those who where devoted christians opposed and where send to laborcamps. But as always, most of the church members of a church where just inherited the membership, and took formal religion but didn't stand behind the christian faith nor no stand against Hitler.
There are plenty of others so but this is just an example Devoted Christian in Nazi Regime
Bonhoeffer, Dietrich b. February 4, 1906 d. April 9, 1945 Memorial only. Lutheran priest and theologian, active anti-Nazi. Arrested 1943 and died in Flossenburg concentration camp.
IBM had dualcore before POWER4. It was in G6. [Not PPC G6 but S390/G6] And that was long before Power4. And I wouldn't be surpriced if some one did it earlier.
The last century's biggest massacres was ANTI-religion force.[Polpot, Hitler, Stalin, Lenin...] So I think you might have some point to it. Face it people cannot control they badness in overall. There is no better or great found inside of human minds. Face it. People are selfish and bad, and never are going to grow up from that. The set of irrational beliefs attacking each other the worst of last century definitely is atheism. If you look backwards from time, its always been that way.
If its religion or any other means people are manipulated to believe something and then go and slaughter others, top persons usually where usually more interested in power and money than any goal that their God has given, and the religion for slaughter usually where twisted. Christianity definitely was extremely twisted for crusades from what bible tells.
Face it no matter what believe system person is under (Atheism is considered as one) it won't change it for good. Some times christianity may have changed drugaddicts when they converted from atheism, but on other direction there hasn't been any other improvement than stop claiming something is sin. Think, start drinking [No benefit] , having fun in women[Not applicable on slashdot] , stops helping others, and just takes care of himself. On other way-> the worst thing is that at first few years the radical new christian shouts around, of goodness of their path and are ignored mostly. And at some point shut ups most of the time. Face it atheism as a religion is no better than christianity. And any view system if twisted and washed enough can be moved to a mass murdering machine. Now call me cynical, I simply believe that in next 100 years, there will be a decade when billions will be slaughtered, and its done by atheists getting rid of religion. And still they fail in it, because christianity is such a strange thing that more you threat and slaughter them, more they convert others. But let em have idle decades and just influence from around world will start creeping in, and slowly converting people away.
Do you mean this one I think he uses some fossil fuel... Whose fan is in his hand, and he will throughly purge his floor, and will gather the wheat into his garner; but the chaff he will burn with fire unquenchable [Luke3:17]
Burn Baby BURN. In satanist russia YOU rule in hell.
Get a condom that is strong enough, and put hamster in it. And there's your hamster powered vibrator, when, one hamster stops giving you the vibes change hamster.
Hey they shouldn't be designing for average but for something more than average. Besides there should be left some room for growth as we are talking about future;) Besides they should dedicate some weight for grocery... I'm not sure that what the TOP 1% of americans weight, but I do have some image in my mind.
http://www.plagueangel.net/grotto/id11.html
Hey I didn't mean the fuel anyway. Think, any flying car that is designed to carry big american family and their grocery will definitely be able to carry atleast half a ton of explosives in the cabinet.
http://news.bbc.co.uk/1/hi/world/americas/2151754. stm
Hey pig's flying is an a engineering problem. So you could learn the trade and start some serious genetical engineering. After a some point you could make a flying pig, I'm pretty sure that in few decades there exists a technology for such breakthrough to be realized, so you should dedicate your life in learning genetical engineering as you seem to have motivation for getting it done. Imagine that you could give a pet flying pig to any female who says she will date when pigs fly. I'm pretty sure that if you would farm couple of thousand of them the slashdot crowd would make you RICH, as soon as you would put it ebay.
Flying cars capable of carrying 500Kg of explosives [Assume a flying car designed to carry 2 USsians], in reprogrammed autopilot could do the trick. No it wouldn't have power of huge aeroplane in destroying entire buildings, but just a small problems like destroying a local FBI office. Now that wouldn't be something that alqaeda would do but consider it possible that some disturbed american individuals could do to get back on goverment on some occasion... If someone wan'ts to do serious damage then it would be simply fill-a big trailer with explosives, drive to location sure 20 tons of explosives could do some serious damage when it explodes below a skyscraper. Or better drive two trailers there. That would be more of alqaeda thing not the small amount of explosives a flying car could carry.
Id elaborate exacly, I got a bug that everything appeared and disappeared mystically all the time, and all the locations of the things on the screen jumped around randomly when scrolling. I went through ENTIRE swing documentation of EVERY single class in my application and ALL their superclasses and wasn't able to find single reason WHY it happened the way it did. I went through all the tutorials I could find on swings. Finally I decided to do random code transformations and suddenly after I changed the order of declaring JLayeredPane and JScrollPane That correct I was hoping of putting LayeredPane inside Scrollpane but it just didn't work as simply as documentation said it was supposed to work. To put things worse, the documentation was so huge that looking for it took weeks. No the worst problem isn't speed, the worst problem is undocumented features of HUGE system that cause bugs that are impossibble to hunt down based on available information and only randomaction was able fix that. Its sheer size of documentation to read for finding IF there was feature in documentation at SOME level where I inherited it where the cause of pain would be found. So in overlla ITS PANE IN THE ASS.
One small point. The complexity simply reduces your chances of creating something truly spectacular, by forcing you to design around those complexities. Intel probably is 2 generations ahead of supposed EV8 introduction in a two years. And we see what kind of single threaded performance intel is going to bring desktop by then, since a full die shrink to 0.65u EV8 would of been 100mm (based on papers) Oh wait, you don't do such huge complex complex microarchitectures because you keep bumping on limits of manageable complexity and validation tasks for even something much smaller. The complexity the designers face always creeps in end product. Either for lack of features that could be desined, or lack of something else. You can trow lots of engineers to the problem, but simply increasing team size won't bring more novel microarchitecture, but makes you capable of working around big layouts, and low level design points. But overall high level design is still going to be something that single thread couldn't even come close to what certain other architecture could bring...
Well... [So I'd better stop treating you just another, apologist x86 software writer who is in love its own tools.] The extra design effort on X86 and work arounds in everyplace, do cost something, if you have simpler ISA you don't need cludges to work around it so much. And every single cludge will take power, some cludges will either add latencies, or increase cycle time. Think where you need to put the extra transistors! Especially how do you deal with 8 and 16 and 32 bit operations, with subword dependencies and hunting them. There are dozens of things and dozens of places where things get complicated. There is not a single place where x86 crufts Anyway perhaps we just look things from completely different perspective. The simpler ISA makes easier to do great things. I'm personally a EE guy who has enough software background to know what I can throw away from MPU, and what kind of software interface is needed, to be real general purpose CPU, And I've been spoiled by that freedom, by trowing out things that any sane assembler coder would object, but I don't design for assembler coders anyway, I'm designing ISA, and compiler, and first implementation by myself, there is certain optimizations available, when I look at everycorner how I could make it do more in parallel put less hw doing things, and everything, that affects in the performance. This freedom is... Hmm tainting my opinions towards anything that limits options for optimization. I'll balance the ease of programming vs performance equation just like old seymour, only difference is that these days things can be abstracted in software so that others won't have to deal with it. Well my opinions towards X86 is mainly tainted by fact that everything looks so cludgy...
A) Fetch,Decoding is where the most power comsumption lays, because its the most complex part most instructions go through. And you say that X86 decoding is NOT more complex than risc? An you say Flags computation per instruction takes no power either? A hint, every transistor STAGE change costs power. B) I was speaking of high performance risc, not the one that gets less than half performance compared to alpha. And yes the X86 legacy prevents relaxed memory ordering. C) Extending ISA beyond extended ISA... D1) Its main purpose is keep X86 decoders away from main pipeline. D2) They are low power item, only thing that those loop caches are not used, trace cache has MORE than just a cache. The fetch part takes some power too. Cache doesn't consume huge amounts of power since most of it is not active, and its designed avoid leakage. But if you have small cache that is all the time active you consume power because of activating it all the time. E) Programmer visible registers, Flags dependencies, lack of triadic operations make more stops in the pipeline. Or the heroics. F) Trace cache simply hurts instruction hit rate. And there is more to trace cache than simple cache logic. Well those transistors, need to take care of ALL corner cases near critical path. Besides validation problem it hurts but not too much, but its still part of the crud... All riscs flags are handled differently, than X86 and hmm. OKay I'll admit. Alpha is in my ideas the RISC, and most others are simply something between risc and cisc, power for instance is only partially risc with all its Enchantchements, sparc has its complex register file that holds its down, ARM has its own problems with shift unit trick it made in early days, and while MIPS is cleaner it still has its cruft from 80's. Those riscy things have bits that turn off the flags for an instruction so those flags don't hurt IPC only clock speed. FLAGS PROBLEM IS REORDERING AND DEPENDENCY CHECK! It increases the number of imput operands for instruction by 50% so that much more reordering dependency checking increases.. G) Yes it is. AMD is heavily competing intel in its own territory so AMD has to go with full custom design too. Its design methology and process that counts in clock speed. None of the surviving riscs use full custom design methology. There is plenty of difference from putting ASIC designers and ASIC design methology with slight tweaks for MPU developement like power does, than having engineers putting rectangles and tweaking for clock speed. Something to think why G5 has 2 cycle ALU latency at its clock speed, its mostly because of design methology used. AMD sells 20-30M processors per year, and thats order of magnitude bigger volume than ANY high performance risc. That gives them ability to tweak their process specifically for high performance MPU manufacturing, have multiple revisions on already made processor. [For comparable risc they don't made such revision.] And speed binning is more agressive definitely.
RISC has less LOGIC transistors, for given execution width. The clock speed is more or less function of design methology, not just amount of engineers. EV5 was designed by engineering team whose size was 10% from Ppro team... Simplicity of ISA gave it advantage. Last but not least, the high performance risc volume is so small that it won't warrant as big fixed costs improvement by design team. AMD even as a smaller x86 player has order of magnitude bigger volume than biggest highperformance risc volume. [I won't count embedded uses for power as their volume as they are not using same parts as low end.] That volume improves amount of bin splits, and thats something that gives immediately 200 to 400mhz clock speed advantage for AMD over PPC. And for SPARC, well that is limited by ONE legacy design decision made in 80's. Alpha was the purest form of RISC and everything that survived in risc arena was not
Slashdot Mirror
NoNoNoNo You got it wrong, the revolution won't be North vs South. It will westcoast Vs eastcoast!
>Now.. a mutation enabling you to work, talk on the phone, and have sex at the same time, that would be an evolutionary advantage!
So the hookers who do both physically present customers, and call girl duties at same time are going to inherit the earth?
If man would be evolved from monkey why we weren't left as a bonobos?
...or as a webmaster/designer for a porn site.
It's not all that bad. Think of the plus side: free porn.
It is not the strongest of the species that survive, nor the most intelligent, but the one most responsive to change.
I think its not the strongest speciest nor most intelligent the one that BREEDS most!
No no, you got it totally wrong it takes bios upgrade and 2000$ , or perhaps just 1000$ for lower end models...
Get a OLD VAX, lock the front cover. Keep your laptop inside of it! No one is going to steal that one.
Uhh. If you dorm room is in florida don't turn it on. If your dorm is in canada open the window and keep it on through the winter.
Uhh. Think again.
There is more to power than the CPU power. The solution I would prefer would be QUAD opteron. You could get full tower case, put a single MB in it, and run it. 2.5Ghz opteron is over double as fast as those poor transmeta chips, now having 4 of them in a single system image with verylow memory latency will make it even faster relatively since the TMsolution would use the Gb ethernet for connecting eachothers memory!.
The 2nd solution is replacement for the deskside case for which it doesn't matter if its bigger than the TM solution since your not going to use that area in anything else anyway, so there could be full rack as well. Power comsumption of 2nd solution is bigger but then again its MUCH cheaper than shelling off 100k$ for the 96 processor TM solution! Getting myrinet and Dual opterons in a rack would definitely beat the TM solution for similar price.
Get quad opteron. It should get about better performance for same price as the other but without need for clustering, for the small system. For bigger system, you could use myrinet and dual opterons. Oh what the heck. It costs 600-800$ per processor to build a rack of Athlon64 based cluster with Gb ethernet. So this effineon based cluster would be beaten with system costing less than their solution. So it is beaten from two different solutions.
A) Getting single image opteron system if communicationlatencies are important.
B) Getting cluster of AMD64 if price/performance was important.
ONLY thing they bring is density of A for system type of B while costing more than A.
Tribes: Veangeance ->new game, new servers the problem you mentioned gone. So besides getting the new gamers to the game they can move the crowds from old game to the new game simply by suplying them with annoying newbies.
apt-get update ...[6 years to implmentation and assume they use it atleast 6 years after that...]
apt-get dist-upgrade
12 years from now
Its Another point is the movie and entertainment. Streaming video from home computer to home computer at 3000x2000 resolution. Video conferences so that there will be 16 of said streams needed, [30" flatpanels are going to be cheap.] Consider the idea that there are 4 family members sharing the connection too.
IPV6 is doing something on latency, the thing their design is going to do something on latency too. [By reducing conversion between optical network and other network.]
For the backbone bandwith. If you can wave multiplex 200 different frequencies and run at 50Ghz operation and bundle 1000 fibers to a single cable you are lacking some bandwith so they need some improvement on backbone to get em the bandwith, but still they could just assume for first generation that not everyone is going to use entire bandwith at same time, so you could just over subsribe it a lot, and just draw the 10Gb lines for last mile as those lines are NOT going to be upgraded. And upgrade the backbone as soon as there is technology todo that.
The backbone bandwith requirement will be huge which makes me wonder the powercomsumption of said network equipment!
There will be some die shrinks before 2010 so they may put single chip router, for it. With address table on same die/package as the routing logic. Hopefully mostly optical networking, etc...
The cooling of said single chip router maybe liquid, so they need extremely reliable liquid cooling available.
Here's your armour lets have a snow fight. hehehehehehe cesium muahhahhaha, boom HEHEHHE boom. Snow+cesium boom, water and cesium boom ice[T>-195'c] and cesium boom. No radio action hehehe just chemistry muahhahhah.
Adolf Hitler at first USED religion when he wasn't in power as a aid towards his goal, then PERSECUTED IT! Thats correct lutheran priests in nazi germany had some persecution going. As for Wermach belts that historical, not something that Hitler put there.
the famous Lutheran priest Martin Niemöller from Nazi Germany and his laconic remarks: "First, they took the communists," said Niemöller, "and I did not protest, since I am not a communist. Then, they abducted the Jews, but I did not say anything, for I am not a Jew. Afterwards, they arrested the catholics, but why should I bother, I am not a catholic? And when they came to get hold of me, there was, strangely enough, nobody left to protest."
"It would be misleading to give the impression that the persecution of Protestants and Catholics by the Nazi State tore the German people asunder or even greatly aroused the vast majority of them. It did not. A people who had so lightly given up their political and cultural and economic freedoms were not, except for a relatively few, going to die or even risk imprisonment to preserve freedom of worship."
I wouldn't call that an act of devouted christian! Nor I would call a Germany a christian nation if people just ignored the the arresting of large portion of its priests of majority churches and so on. I'd call Nazi Germany about as christian nation as US is, slighly less. I would call the Nazi Germany mostly atheists and christians by name but without acting on it.
I think the thinking Hitler as christian may come from one certain political tool nazis took.
The Nazi-controlled German Christians Faith Movement dominated mainstream churches by mixing dogma with politics and proclaimed: "In the person of the Fuhrer we behold the One sent from God who places Germany in the presence of the Lord of History."
They simply spoke their own politics with few selected words from bible and called that a christian religion. Those who where devoted christians opposed and where send to laborcamps. But as always, most of the church members of a church where just inherited the membership, and took formal religion but didn't stand behind the christian faith nor no stand against Hitler.
There are plenty of others so but this is just an example Devoted Christian in Nazi Regime
Bonhoeffer, Dietrich b. February 4, 1906 d. April 9, 1945
Memorial only. Lutheran priest and theologian, active anti-Nazi. Arrested 1943 and died in Flossenburg concentration camp.
IBM had dualcore before POWER4. It was in G6. [Not PPC G6 but S390/G6] And that was long before Power4. And I wouldn't be surpriced if some one did it earlier.
The last century's biggest massacres was ANTI-religion force.[Polpot, Hitler, Stalin, Lenin...] So I think you might have some point to it. Face it people cannot control they badness in overall. There is no better or great found inside of human minds. Face it. People are selfish and bad, and never are going to grow up from that. The set of irrational beliefs attacking each other the worst of last century definitely is atheism. If you look backwards from time, its always been that way.
If its religion or any other means people are manipulated to believe something and then go and slaughter others, top persons usually where usually more interested in power and money than any goal that their God has given, and the religion for slaughter usually where twisted. Christianity definitely was extremely twisted for crusades from what bible tells.
Face it no matter what believe system person is under (Atheism is considered as one) it won't change it for good. Some times christianity may have changed drugaddicts when they converted from atheism, but on other direction there hasn't been any other improvement than stop claiming something is sin. Think, start drinking [No benefit] , having fun in women[Not applicable on slashdot] , stops helping others, and just takes care of himself. On other way-> the worst thing is that at first few years the radical new christian shouts around, of goodness of their path and are ignored mostly. And at some point shut ups most of the time.
Face it atheism as a religion is no better than christianity. And any view system if twisted and washed enough can be moved to a mass murdering machine. Now call me cynical, I simply believe that in next 100 years, there will be a decade when billions will be slaughtered, and its done by atheists getting rid of religion. And still they fail in it, because christianity is such a strange thing that more you threat and slaughter them, more they convert others. But let em have idle decades and just influence from around world will start creeping in, and slowly converting people away.
Do you mean this one I think he uses some fossil fuel...
Whose fan is in his hand, and he will throughly purge his floor, and will gather the wheat into his garner; but the chaff he will burn with fire unquenchable [Luke3:17]
Burn Baby BURN.
In satanist russia YOU rule in hell.
Get a condom that is strong enough, and put hamster in it. And there's your hamster powered vibrator, when, one hamster stops giving you the vibes change hamster.
Hey they shouldn't be designing for average but for something more than average. Besides there should be left some room for growth as we are talking about future ;) Besides they should dedicate some weight for grocery... I'm not sure that what the TOP 1% of americans weight, but I do have some image in my mind.
http://www.plagueangel.net/grotto/id11.html
Hey I didn't mean the fuel anyway. Think, any flying car that is designed to carry big american family and their grocery will definitely be able to carry atleast half a ton of explosives in the cabinet. http://news.bbc.co.uk/1/hi/world/americas/2151754. stm
Hey pig's flying is an a engineering problem. So you could learn the trade and start some serious genetical engineering. After a some point you could make a flying pig, I'm pretty sure that in few decades there exists a technology for such breakthrough to be realized, so you should dedicate your life in learning genetical engineering as you seem to have motivation for getting it done.
Imagine that you could give a pet flying pig to any female who says she will date when pigs fly.
I'm pretty sure that if you would farm couple of thousand of them the slashdot crowd would
make you RICH, as soon as you would put it ebay.
Flying cars capable of carrying 500Kg of explosives [Assume a flying car designed to carry 2 USsians], in reprogrammed autopilot could do the trick. No it wouldn't have power of huge aeroplane in destroying entire buildings, but just a small problems like destroying a local FBI office. Now that wouldn't be something that alqaeda would do but consider it possible that some disturbed american individuals could do to get back on goverment on some occasion... If someone wan'ts to do serious damage then it would be simply fill-a big trailer with explosives, drive to location sure 20 tons of explosives could do some serious damage when it explodes below a skyscraper. Or better drive two trailers there. That would be more of alqaeda thing not the small amount of explosives a flying car could carry.
Id elaborate exacly, I got a bug that everything appeared and disappeared mystically all the time, and all the locations of the things on the screen jumped around randomly when scrolling. I went through ENTIRE swing documentation of EVERY single class in my application and ALL their superclasses and wasn't able to find single reason WHY it happened the way it did. I went through all the tutorials I could find on swings. Finally I decided to do random code transformations and suddenly after I changed the order of declaring JLayeredPane and JScrollPane That correct I was hoping of putting LayeredPane inside Scrollpane but it just didn't work as simply as documentation said it was supposed to work. To put things worse, the documentation was so huge that looking for it took weeks. No the worst problem isn't speed, the worst problem is undocumented features of HUGE system that cause bugs that are impossibble to hunt down based on available information and only randomaction was able fix that. Its sheer size of documentation to read for finding IF there was feature in documentation at SOME level where I inherited it where the cause of pain would be found. So in overlla ITS PANE IN THE ASS.
Swing programming in Java is Pane in the ass.
One small point. The complexity simply reduces your chances of creating something truly spectacular, by forcing you to design around those complexities. Intel probably is 2 generations ahead of supposed EV8 introduction in a two years. And we see what kind of single threaded performance intel is going to bring desktop by then, since a full die shrink to 0.65u EV8 would of been 100mm (based on papers) Oh wait, you don't do such huge complex complex microarchitectures because you keep bumping on limits of manageable complexity and validation tasks for even something much smaller. The complexity the designers face always creeps in end product. Either for lack of features that could be desined, or lack of something else. You can trow lots of engineers to the problem, but simply increasing team size won't bring more novel microarchitecture, but makes you capable of working around big layouts, and low level design points. But overall high level design is still going to be something that single thread couldn't even come close to what certain other architecture could bring...
Well... [So I'd better stop treating you just another, apologist x86 software writer who is in love its own tools.]
The extra design effort on X86 and work arounds in everyplace, do cost something, if you have simpler ISA you don't need cludges to work around it so much. And every single cludge will take power, some cludges will either add latencies, or increase cycle time. Think where you need to put the extra transistors! Especially how do you deal with 8 and 16 and 32 bit operations, with subword dependencies and hunting them. There are dozens of things and dozens of places where things get complicated. There is not a single place where x86 crufts
Anyway perhaps we just look things from completely different perspective. The simpler ISA makes easier to do great things. I'm personally a EE guy who has enough software background to know what I can throw away from MPU, and what kind of software interface is needed, to be real general purpose CPU, And I've been spoiled by that freedom, by trowing out things that any sane assembler coder would object, but I don't design for assembler coders anyway, I'm designing ISA, and compiler, and first implementation by myself, there is certain optimizations available, when I look at everycorner how I could make it do more in parallel put less hw doing things, and everything, that affects in the performance. This freedom is... Hmm tainting my opinions towards anything that limits options for optimization. I'll balance the ease of programming vs performance equation just like old seymour, only difference is that these days things can be abstracted in software so that others won't have to deal with it.
Well my opinions towards X86 is mainly tainted by fact that everything looks so cludgy...
Slashdot, stuff that matters.
A) Fetch,Decoding is where the most power comsumption lays, because its the most complex part most instructions go through. And you say that X86 decoding is NOT more complex than risc? An you say Flags computation per instruction takes no power either? A hint, every transistor STAGE change costs power.
= RW T060503232439&p=3
B) I was speaking of high performance risc, not the one that gets less than half performance compared to alpha. And yes the X86 legacy prevents relaxed memory ordering.
C) Extending ISA beyond extended ISA...
D1) Its main purpose is keep X86 decoders away from main pipeline.
D2) They are low power item, only thing that those loop caches are not used, trace cache has MORE than just a cache. The fetch part takes some power too. Cache doesn't consume huge amounts of power since most of it is not active, and its designed avoid leakage. But if you have small cache that is all the time active you consume power because of activating it all the time.
E) Programmer visible registers, Flags dependencies, lack of triadic operations make more stops in the pipeline. Or the heroics.
F) Trace cache simply hurts instruction hit rate. And there is more to trace cache than simple cache logic. Well those transistors, need to take care of ALL corner cases near critical path. Besides validation problem it hurts but not too much, but its still part of the crud...
All riscs flags are handled differently, than X86 and hmm. OKay I'll admit. Alpha is in my ideas the RISC, and most others are simply something between risc and cisc, power for instance is only partially risc with all its Enchantchements, sparc has its complex register file that holds its down, ARM has its own problems with shift unit trick it made in early days, and while MIPS is cleaner it still has its cruft from 80's. Those riscy things have bits that turn off the flags for an instruction so those flags don't hurt IPC only clock speed. FLAGS PROBLEM IS REORDERING AND DEPENDENCY CHECK! It increases the number of imput operands for instruction by 50% so that much more reordering dependency checking increases..
G) Yes it is. AMD is heavily competing intel in its own territory so AMD has to go with full custom design too. Its design methology and process that counts in clock speed. None of the surviving riscs use full custom design methology. There is plenty of difference from putting ASIC designers and ASIC design methology with slight tweaks for MPU developement like power does, than having engineers putting rectangles and tweaking for clock speed. Something to think why G5 has 2 cycle ALU latency at its clock speed, its mostly because of design methology used. AMD sells 20-30M processors per year, and thats order of magnitude bigger volume than ANY high performance risc. That gives them ability to tweak their process specifically for high performance MPU manufacturing, have multiple revisions on already made processor. [For comparable risc they don't made such revision.] And speed binning is more agressive definitely.
RISC has less LOGIC transistors, for given execution width.
The clock speed is more or less function of design methology, not just amount of engineers.
EV5 was designed by engineering team whose size was 10% from Ppro team... Simplicity of ISA gave it advantage.
Last but not least, the high performance risc volume is so small that it won't warrant as big fixed costs improvement by design team. AMD even as a smaller x86 player has order of magnitude bigger volume than biggest highperformance risc volume. [I won't count embedded uses for power as their volume as they are not using same parts as low end.] That volume improves amount of bin splits, and thats something that gives immediately 200 to 400mhz clock speed advantage for AMD over PPC. And for SPARC, well that is limited by ONE legacy design decision made in 80's. Alpha was the purest form of RISC and everything that survived in risc arena was not
http://www.realworldtech.com/page.cfm?ArticleID