traditional etch/deposition system works by leaving putting on a layer (in this case 3 atoms thick) then etching off the stuff you don't want.
What I can't see is how one can lay down anything 3 atoms thick (or wide) reliably (in the sense of real-world mass manufacture, not one of a time in-the-lab productions) using scaled versions of existing Fab tachnologies and without some nano-assembler type technology. Worst case you'll get 3 atoms somewhere in the middle of the wafer and maybe 5 or 0 at the edges....
This sort of tech will come one day - but I beleive it's going have to be by revolution, not evolution....
the real potential here is not for some glass tube replacement - but for very different sorts of things - cheapo projection systems where your current screen just becomes a white board on the wall, or a translucent screen with back projection - or even better - the Snowcrash sort of system where it tracks your eyeballs and paints low intensity RGB lasers onto your retina
Ahem ... look inside your Airport ....
on
Open Networking
·
· Score: 3
there's the same WaveLan PCMCIA card that's in my Linux laptop.... 802.11 is something that Apple adopted, not something they invented.... like lots of other coold stuff (like CD Drives, bitmapped displays etc etc) Apple is a great early adopter but just because you first saw it from them don't assume that they thought up the idea.... (oh yeah I love my Apple Airport - makes a great/cheap 802.11 RF router for my Linux network)
I do something similar to this at home - but use and Apple Airport plugged into my local ether - this works great with the linux drivers but requires a Mac somewhere on the same lan to configure it - I have it set up to only accept connections with fixed MAC addresses and to require that those machines know the actual name of the net.
The only downside is that it's set up for NAT and now I can no long remote print because stupid LPR bitches that I'm no longer using a priviliged socket.... anyone got any ideas for a quick fix?
Next step of course will be to find a way to get the local Cafe wired.....
worries about those parking building where you drive in circles the same way going in as you do coming out... and that that will cause the earth to spin down..... we have a special place in hell for you..... running in circles making up for all the people driving in parking buildings....
well let's see - everywhere else in the world already uses the same standard and the US has 2-3 of it's own which one will you pick:-)
seriously though - I went to a funeral in NZ recently, my Uncle was there he had his GSM phone on international roam - his works at my Mom's house in the NZ Alps, at his house on an island off of Brisbane and where he's currently consulting near London... the same number rings worldwide - except where I live in the US (but it does work in Baltimore)
as soon as someone starts deploying a GSM network here I'm buying one
I found myself walking with my kids in an 'interesting' part of town recently when my 7yr old asked me "Daddy... who is that lady talking to..." and I honestly couldn't tell whether she had one of those hands-free phones..... or her own personal demons
It's either going to get a lot harder for us to tell who's crazy..... or an awfull lot easier for the crazier amoung us to fit in.... just put on some OK clothes and go downtown to the financial distruct and commune with the voices
well - I get your point - but I also beleive there's a class of genuinely non-obvious, innovative ideas for which patents are appropriate (not the bulk of the crap being patented at the moment) - IMHO this is one.
I suggest the following test to anyone considering patenting something - "would you feel proud explaining your idea to Mr Edison? or embarassed?"
yup they have a cool patent on their writeback buffer - basicly it stalls to clean points where exceptions are resolved - that way they don't have to worry about having 'clean' exceptions - just toss the memory/register changes and drop back to interpretting the code instruction by instruction
Is it
fair to say that the Rambus system carries a big latency penalty
compared to prevalent memory technologies, but it is a model in which
increases in bandwidth can be designed in without incurring further
costs in latency (ie. we have bandwidth scaleability)?
It's not that big - the thing to realize is that because the outgoing address bus is multiplexed (at a ~GHz rate) you're wasting a clock or so compared with other drams to send enough of the address to start a RAS cycle in the DRAM (this used to be more like 4 clocks on older versions of Rambus - the current rdrams are basicly V3 of their protocol). Because the data comes back multiplexed (16-bits at a time) the latency to first data is about the same for a traditional wider type of dram, but the time to last data (important if you then need to send the data over a channel that can't be flow controlled - or to a CPU who's cache system can't use partial cache lines) is longer.
Now this absolute latency is important for a system where every transaction is received by an idle memory controller - but if you have a transaction arriving at the dram while the drams are already busy and you can start multiple transactions in parallel mean latency may actually be lower (most modern drams can do one row sense while doing a data transfer - rdrams tend to have a finer granularity allowing more concurrency as well as a protocol that allows things like speculative row senses and out of order data transfers).
Mind you the down side of designing a memory controller that supports this amount of concurrency is that it's HARD - really HARD and hard to test - normally you just manage a queue of pending transactions in order - managing an out-of-order memory queue is difficult - it actually probably makes sense to integrate all this stuff closely with a CPU - into the out-of-order part of the CPU's memory controller.
Is this the
strength that many designers have seen in Rambus?
I think it depends on what you use it for - the systems I've used it for have been more bandwidth sensitive (and pin sensitive) - basicly they sort of fit a sweet spot as VRAMs stopped being usefull - but cost has always been an issue meaning that using them was always an iffy choice - doing chip designs usually means betting on supporting chips like DRAMS being available at a particular price a year in the future - choices are hard.
Caveat - I've done designs with the previous 2 generations of rdrams, not the latest 'direct' ones which IMHO (from reading the data sheets, not experience) are simpler to use.
they are talking about 'bandwidth-per-pin' - us chip designers can pump bandwidth by using more pins (modulo packaging, noise and power issues), and by doing interleave (the graphics accelerators I was designing back then were, from memory running 2-clock interleaved controllers at 50MHz so data was flowing at 25MHz [96-bits wide plus the VRAM fill hardware support gave us 1.5Gb/sec fill rates - not too shabby even by modern standards:-]).
However another thing that may not be obvious - today's 133MHz DRAMs being used in PCs are top-of-the-line - back in 1989 the fastest DRAMs were only being used in high-end servers because of the price premium.
(some background on why Rambus is good/bad in general) I've done designs with many of these technologies (traditional async ras/cas, sdram, rambus, not DDR) over the years - the older rambus designs were certainly harder to implement with (they used more of a network protocol paradigm) but not by much. The main thing about rambus is that at some level it trades off latency for bandwidth - there are some places where this is actually a good thing - display controllers for example.
Rambus also is a win in places where lots of concurrent transactions are available - the finer grained banking allows parallel row senses - reducing average latency, even speculative row senses for CPUs doing speculative instructions.
I beleive this is the main reason Intel went for rambus - they are building CPUs that are highly parallel at the low level - and can issue many overlapping memory requests at once - but they screwed up - this would have been great if they were hooking the rambus channels directly to their CPUs - but instead they are making them over the slot1 bus which forces complete serialization losing any possible advantage - AMD's slot A would have been a better choice but these buses still do a very basic serialization that's going to make obtaining almost any concurrency at the RAM channel level difficult (which is why IMHO rambus on Intel hardware sucks).
It was a decision made because they found integer math was faster this way. Digital Alphas came to the same conclusion.
There is a valid argument that this is true for a system where the memory accesses are smaller than an integer size (for example a 32-bit 286 memory to register add with a 16-bit bus). But for an Alpha - a RISC machine that doesn't have combined memory/ALU ops, and has a bus at least as large as it's integer size, and aligned values - this isn't an issue.
I think the thing people are talking about is the use of a bit order for 1-bit displays that doesn't match the instructions in the machine. The x86s are little-endian and their bit operations do all the 'right' things for am 8-bit byte where bit 0 of byte 0 is the left most pixel, bit 1 of byte 0, the next,.... bit 7 of byte 0 the next, bit 0 of byte 1 the next... etc etc
But for whatever reason IBM chose a display controller for the first PC (a 6845 I thik) that was designed for a big-endian CPU family - there bit 7 of byte 0 is the left most on the screen, bit 6 byte 0 next,.... bit 0 byte 0 next, bit 7 byte 1 next,.... - which wouldn't be that big a deal except that it's really hard to write a good blitter (the core of a 2d graphics library) if you're CPU's instruction set doesn't support shift instructions that work this way
Anyway - as pointed out below it was IBM who made the mistake of mixing endianess - not Intel who have a consistant architecture
Ummm.... not really - it was just an 8088 with a little extra support circuitry - no wierder than an 8088 - I used one to build a cheapo Mac card that paid for the down payment on my house - I can't complain:-)
According to this, the DEC PDP-11 and VAX lines were little-endian too - I'm not old enough to remember these machines (apart from FORTRAN torture at college on a VAX), but I know enough that they form a huge chunk of the history of the Internet, and Unix.
The Vax was religiously little endian, the pdpd series a little more mixed. The internet on the other hand (well TCP/IP protocols anyway) is mostly big-endian.
Errr... and by the entirety of civilisation too.
Ummm - not quite - what we call 'big endian' is reall a mixed endian system - a true big endian system would count bytes down from the end of memory to the beginning (yes I know this is identical to a 'little endian' system thru a simple transform - but read on...) - so think about where our numbers come from.... arabic.... which is written the other way from our language... what this means is that our numbering system was originally designed to be written LSB first (from the point of view of an arabic writer) - this makes sense if you're a medievil trader - what you'e mostly doing is addition and subtraction - and those operations are performed LSB to MSB - so writing them that way makes sense.
Arabic and other left-to-right writing systems that include arabic numbers as we normally write them are true big endian writing systems, while english and most other european writing systems are mixed endian systems - which is why 'big-endian' systems that number memory from left to right, but multi-byte integer data right-to-left seem natural to us - it's an articfact of the way our culture writes its numbers withing its test
I beleive our 'modern' mixed endian way of including the right-to-left arabic numbering system in left-to-right text is a result of it being adopted that way by spanish monks looting the moorish libraries after they were driven south.... if they had been a bit smarter about what they did and had reversed the order of the digits so that it worked the same way as they did in arabic we would all consider 'little endian' numbering systems as natural and 'big-endian' as wierd
or for that matter the x86 architecture in toto - they had lots of great role models (vax/pdp11 even the 360) at the time and Motorolla/National and just about everyone else took up architectures with cleaner, easier to program memory models
Now, how many people have actually held onto their houses since 1978? Knowing government (remember Prop 39 this year), they'd have taxed us to the point where taxes are more than the mortgage.
More than you might think given that the states and counties have passed a number of laws allowing people to carry their exemptions over when buying/selling homes.
The ridiculousness of this law was driven home to me when I discovered that my boss was paying 50% less property tax than me on his million-dollar mansion with pool in the hills (same city)....
I predict that prop 13 will die when the number of people getting the cheapo locked in low rates decreases (due to them dieing off) below the number of homeowners who don't....
I used to design graphics accelerators for Macs. Color fidelity/matching/whatever was a very big issue - marketting was very hot on it and we worked hard to do 'the right thing'.
The problem is that you can't satisfy everyone - even though you might be able to get it close to 'right' for one person everyone's eyes are different - the numbers of rods and cones vary widely enough that the way we perceive colors from phosphors and reflected from paper is different enough from person to person that you can't get it right, just close (for example at one extreme is the 10% of the male population who are red-green color blind).
Another example of this is the way that ambient light plays in our perceptions - colors can look totally different in the morning than in the evening in the same room because the color composition of the ambient light changes - people in publishing who are serious about this sort of stuff have 'white rooms' with known lighting and no outside windows to look at stuff in.
In my experience this area is enough of a sinkhole that you can/will get lots of competing schemes for color matching with lots of area for arguing - IMHO any color calibration system that doesn't calibrate for the individual user's eyes is worthless - but at the same any system that does so is so subjective that it can't be reliably measured.
Oh yeah - and look very closely at any system that performs liner math (multiplication, matrix ops etc) on gama corrected (logrithmic) pixels [hint almost any HDTV system that does picture scaling does this]
For example, we pay a property tax that is related to what our house MAY bring IF we sell it.
Actually I think that basing property taxes on current valuation is the norm throughout the western world, and probably everywehere else too. Otherwise little old ladies who've lived in their houses for 60 years will inflate away their tax basis and end up paying far less than their share (Mind you here in California the LOLs banded together and passed an initiative that froze their tax basis at an early '80s level meaning that the rest of us end up having toshoulder much more than our share of the burden)
"Microsoft is going to be releasing a 'subscription version' of Office 10. This version will actually stop allowing a user to create new documents after the subscription period ends.
So day 1 you make an empty document of each type and archive it..... from then on you just duplicate empty documents on the desktop rather than using office to make them for you.... or better yet - download those warez empty documents from the net..... can you just see M$ going to court trying to ban the giving away of empty documents....:-) "but your Honor - they're a device designed to 'subvert an access mechanism'" - "in rebutal - 'we made them with Windows - it's time it was banned'"
if your screen's phosphors are long persistance (TV for example) they limit how fast you screen can refresh anyway (it's a tricky balance - too fast and dumb VGA screens flicker, too slow and fast refresh rate screens blur...)
how fast is the display subsystem - large screen refreshing above about 250MHz (say 85Hz) is pretty close to impossible due to the imense amount of data needed to be sucked for the refresh process - this takes bandwidth away from poly rendering - above 250MHz is also a big problem for the analog portion of the subsystem (DACs, cables etc - the results can be smeared pixels)
Slashdot Mirror
What I can't see is how one can lay down anything 3 atoms thick (or wide) reliably (in the sense of real-world mass manufacture, not one of a time in-the-lab productions) using scaled versions of existing Fab tachnologies and without some nano-assembler type technology. Worst case you'll get 3 atoms somewhere in the middle of the wafer and maybe 5 or 0 at the edges ....
This sort of tech will come one day - but I beleive it's going have to be by revolution, not evolution ....
the real potential here is not for some glass tube replacement - but for very different sorts of things - cheapo projection systems where your current screen just becomes a white board on the wall, or a translucent screen with back projection - or even better - the Snowcrash sort of system where it tracks your eyeballs and paints low intensity RGB lasers onto your retina
there's the same WaveLan PCMCIA card that's in my Linux laptop .... 802.11 is something that Apple adopted, not something they invented .... like lots of other coold stuff (like CD Drives, bitmapped displays etc etc) Apple is a great early adopter but just because you first saw it from them don't assume that they thought up the idea .... (oh yeah I love my Apple Airport - makes a great/cheap 802.11 RF router for my Linux network)
The only downside is that it's set up for NAT and now I can no long remote print because stupid LPR bitches that I'm no longer using a priviliged socket .... anyone got any ideas for a quick fix?
Next step of course will be to find a way to get the local Cafe wired .....
worries about those parking building where you drive in circles the same way going in as you do coming out ... and that that will cause the earth to spin down ..... we have a special place in hell for you ..... running in circles making up for all the people driving in parking buildings ....
seriously though - I went to a funeral in NZ recently, my Uncle was there he had his GSM phone on international roam - his works at my Mom's house in the NZ Alps, at his house on an island off of Brisbane and where he's currently consulting near London ... the same number rings worldwide - except where I live in the US (but it does work in Baltimore)
as soon as someone starts deploying a GSM network here I'm buying one
It's either going to get a lot harder for us to tell who's crazy ..... or an awfull lot easier for the crazier amoung us to fit in .... just put on some OK clothes and go downtown to the financial distruct and commune with the voices
I suggest the following test to anyone considering patenting something - "would you feel proud explaining your idea to Mr Edison? or embarassed?"
yup they have a cool patent on their writeback buffer - basicly it stalls to clean points where exceptions are resolved - that way they don't have to worry about having 'clean' exceptions - just toss the memory/register changes and drop back to interpretting the code instruction by instruction
It's not that big - the thing to realize is that because the outgoing address bus is multiplexed (at a ~GHz rate) you're wasting a clock or so compared with other drams to send enough of the address to start a RAS cycle in the DRAM (this used to be more like 4 clocks on older versions of Rambus - the current rdrams are basicly V3 of their protocol). Because the data comes back multiplexed (16-bits at a time) the latency to first data is about the same for a traditional wider type of dram, but the time to last data (important if you then need to send the data over a channel that can't be flow controlled - or to a CPU who's cache system can't use partial cache lines) is longer.
Now this absolute latency is important for a system where every transaction is received by an idle memory controller - but if you have a transaction arriving at the dram while the drams are already busy and you can start multiple transactions in parallel mean latency may actually be lower (most modern drams can do one row sense while doing a data transfer - rdrams tend to have a finer granularity allowing more concurrency as well as a protocol that allows things like speculative row senses and out of order data transfers).
Mind you the down side of designing a memory controller that supports this amount of concurrency is that it's HARD - really HARD and hard to test - normally you just manage a queue of pending transactions in order - managing an out-of-order memory queue is difficult - it actually probably makes sense to integrate all this stuff closely with a CPU - into the out-of-order part of the CPU's memory controller.
Is this the strength that many designers have seen in Rambus?
I think it depends on what you use it for - the systems I've used it for have been more bandwidth sensitive (and pin sensitive) - basicly they sort of fit a sweet spot as VRAMs stopped being usefull - but cost has always been an issue meaning that using them was always an iffy choice - doing chip designs usually means betting on supporting chips like DRAMS being available at a particular price a year in the future - choices are hard.
Caveat - I've done designs with the previous 2 generations of rdrams, not the latest 'direct' ones which IMHO (from reading the data sheets, not experience) are simpler to use.
However another thing that may not be obvious - today's 133MHz DRAMs being used in PCs are top-of-the-line - back in 1989 the fastest DRAMs were only being used in high-end servers because of the price premium.
(some background on why Rambus is good/bad in general) I've done designs with many of these technologies (traditional async ras/cas, sdram, rambus, not DDR) over the years - the older rambus designs were certainly harder to implement with (they used more of a network protocol paradigm) but not by much. The main thing about rambus is that at some level it trades off latency for bandwidth - there are some places where this is actually a good thing - display controllers for example.
Rambus also is a win in places where lots of concurrent transactions are available - the finer grained banking allows parallel row senses - reducing average latency, even speculative row senses for CPUs doing speculative instructions. I beleive this is the main reason Intel went for rambus - they are building CPUs that are highly parallel at the low level - and can issue many overlapping memory requests at once - but they screwed up - this would have been great if they were hooking the rambus channels directly to their CPUs - but instead they are making them over the slot1 bus which forces complete serialization losing any possible advantage - AMD's slot A would have been a better choice but these buses still do a very basic serialization that's going to make obtaining almost any concurrency at the RAM channel level difficult (which is why IMHO rambus on Intel hardware sucks).
There is a valid argument that this is true for a system where the memory accesses are smaller than an integer size (for example a 32-bit 286 memory to register add with a 16-bit bus). But for an Alpha - a RISC machine that doesn't have combined memory/ALU ops, and has a bus at least as large as it's integer size, and aligned values - this isn't an issue.
But for whatever reason IBM chose a display controller for the first PC (a 6845 I thik) that was designed for a big-endian CPU family - there bit 7 of byte 0 is the left most on the screen, bit 6 byte 0 next, .... bit 0 byte 0 next, bit 7 byte 1 next, .... - which wouldn't be that big a deal except that it's really hard to write a good blitter (the core of a 2d graphics library) if you're CPU's instruction set doesn't support shift instructions that work this way
Anyway - as pointed out below it was IBM who made the mistake of mixing endianess - not Intel who have a consistant architecture
Ummm .... not really - it was just an 8088 with a little extra support circuitry - no wierder than an 8088 - I used one to build a cheapo Mac card that paid for the down payment on my house - I can't complain :-)
The Vax was religiously little endian, the pdpd series a little more mixed. The internet on the other hand (well TCP/IP protocols anyway) is mostly big-endian.
Errr... and by the entirety of civilisation too.
Ummm - not quite - what we call 'big endian' is reall a mixed endian system - a true big endian system would count bytes down from the end of memory to the beginning (yes I know this is identical to a 'little endian' system thru a simple transform - but read on ...) - so think about where our numbers come from .... arabic .... which is written the other way from our language ... what this means is that our numbering system was originally designed to be written LSB first (from the point of view of an arabic writer) - this makes sense if you're a medievil trader - what you'e mostly doing is addition and subtraction - and those operations are performed LSB to MSB - so writing them that way makes sense.
Arabic and other left-to-right writing systems that include arabic numbers as we normally write them are true big endian writing systems, while english and most other european writing systems are mixed endian systems - which is why 'big-endian' systems that number memory from left to right, but multi-byte integer data right-to-left seem natural to us - it's an articfact of the way our culture writes its numbers withing its test
I beleive our 'modern' mixed endian way of including the right-to-left arabic numbering system in left-to-right text is a result of it being adopted that way by spanish monks looting the moorish libraries after they were driven south .... if they had been a bit smarter about what they did and had reversed the order of the digits so that it worked the same way as they did in arabic we would all consider 'little endian' numbering systems as natural and 'big-endian' as wierd
or for that matter the x86 architecture in toto - they had lots of great role models (vax/pdp11 even the 360) at the time and Motorolla/National and just about everyone else took up architectures with cleaner, easier to program memory models
More than you might think given that the states and counties have passed a number of laws allowing people to carry their exemptions over when buying/selling homes.
The ridiculousness of this law was driven home to me when I discovered that my boss was paying 50% less property tax than me on his million-dollar mansion with pool in the hills (same city) ....
I predict that prop 13 will die when the number of people getting the cheapo locked in low rates decreases (due to them dieing off) below the number of homeowners who don't ....
The problem is that you can't satisfy everyone - even though you might be able to get it close to 'right' for one person everyone's eyes are different - the numbers of rods and cones vary widely enough that the way we perceive colors from phosphors and reflected from paper is different enough from person to person that you can't get it right, just close (for example at one extreme is the 10% of the male population who are red-green color blind).
Another example of this is the way that ambient light plays in our perceptions - colors can look totally different in the morning than in the evening in the same room because the color composition of the ambient light changes - people in publishing who are serious about this sort of stuff have 'white rooms' with known lighting and no outside windows to look at stuff in.
In my experience this area is enough of a sinkhole that you can/will get lots of competing schemes for color matching with lots of area for arguing - IMHO any color calibration system that doesn't calibrate for the individual user's eyes is worthless - but at the same any system that does so is so subjective that it can't be reliably measured.
Oh yeah - and look very closely at any system that performs liner math (multiplication, matrix ops etc) on gama corrected (logrithmic) pixels [hint almost any HDTV system that does picture scaling does this]
Actually I think that basing property taxes on current valuation is the norm throughout the western world, and probably everywehere else too. Otherwise little old ladies who've lived in their houses for 60 years will inflate away their tax basis and end up paying far less than their share (Mind you here in California the LOLs banded together and passed an initiative that froze their tax basis at an early '80s level meaning that the rest of us end up having toshoulder much more than our share of the burden)
So day 1 you make an empty document of each type and archive it ..... from then on you just duplicate empty documents on the desktop rather than using office to make them for you .... or better yet - download those warez empty documents from the net ..... can you just see M$ going to court trying to ban the giving away of empty documents .... :-) "but your Honor - they're a device designed to 'subvert an access mechanism'" - "in rebutal - 'we made them with Windows - it's time it was banned'"
But, but, but .... I already missed Battlefield Earth .... does this mean I have to go see it now? (shudder) .... you're evil ...
giving each coke machine a phone number ..... and causing us all to change our area codes every so often ....
and the issue isn't about PLAYing minesweeper - but deciding whether a particular position is self-consistant - a whole different thing
pretty soon they'll define commercials as part of the broadcast TV access mechanism .... and ban your mute button .....