Back in the days when VRAM was used, it was because the bandwidth consumed just to feed the DACs was a significant proportion of the total available memory bandwidth. Then, having an output only port made a lot of sense.
Since those times, video memory bandwidth has increased enormously but DAC requirements have not. A 32 bit 3k x 2k LCD at 60Hz will only consume 1.4GB/s. Even maxing out a 400MHz DAC, you top out at 1.6GB/s with 32 bit colour. This is well under 10% of a modern graphics memory subsystems total bandwidth.
The only way I can think that there would be a return to dual port memory like VRAM of old would be with asymmetric ports. A fast one for the main I/O and a slow port (only say 2GB/s) for the DACs. If this 'slow' port was highly pin efficient, say a 2 bit wide packetised serial interface like rambus, this might be worth the die space/pin count. Even so, these pins might be better off employed making the main memory interface 10% faster or more efficient.
What killed Tomorrows World for me was when they went from Live broadcasts with the occasional reel of video to being a completely pre-recorded show.
Sure, being live meant that some of the demonstrations were not perfect, but it meant that you saw the technology warts and all, not a marketing piece. I remember the first ISDN based video conferencing demo, before decent compression, and it looked terrible. You couldn't have given those scratchy little black and white screened video phones away.
Going pre-recorded meant that they lost the spontaneity and it meant that they couldn't be as up to date. All they gained was a little slickness in the production, which didn't actually help make the programme any more watchable and if anything made it feel less cutting-edge.
Why not go the whole hog and have the Kernel in microcode on the processor? Worked well enough for the Transputer. Even modern processors have a hard time competing with the context switch time of a 30MHz T800.
What we are starting to see in the embedded market though is more use of embedding the processing core in an FPGA. If you only need 386/486 performance you can easily slot in an 8, 16 or even 32bit RISC core onto your FPGA, it might only take up a quarter of a decent sized one, add the your other logic to the rest of the FPGA. The only difference might be that you will be using a larger FPGA than you would have before (more gates, more interface pins for the cpu's memory and i/o busses). You don't then have to worry about your processor being obsoleted, only your FPGA (which you would have had to worry about anyway).
Mark..........
One solution to outgoing SMTP port blocks.
on
Broadband Crackdown
·
· Score: 1
For those who are interested, there are some providers who have SMTP AUTHenticated mail relays for their customers which work on ports other than port 25.
My own (Gradwell.com, with their email forwarding account) email host has their SMTP server listening on port 225 as well as 25.
I am sure that should port 225 be blocked, they would be happy to set up listeners on other ports too. Who knows, if the problem with such port blocking gets bad enough, there is no reason you coulsn't build a server whos every port connected you to an SMTP server. *8')
Take care,
I agree, work is just a means to an end. Even if you enjoy it, there are plenty of other things to enjoy.
Personally, I work a contracted 30 hour week. My company went through difficulties a while back and couldn't afford the raise they had been promising. Rather than lose me altogether they agreed to my request to move to a 4 day week.
They are happy that they didn't have to lose me, I am happy that i now get a three day weekend. I still earn more than enough to live on and have lots more free time.
Most of the people who I speak to about this are envious, in some respects it is just like being a student again, but now I have money.
The thing that I don't understand is they we are supposed to be talking about uncompressed HDTV resolution video. Now by my calculations 1280*1080 pixels at 30fps is only 42MB/sec if you use can encode with 1 byte per pixel! Thus even their 'uncompressed' video must be compressed a little.
This leads me to think that it might be compressed more than they are saying on the actual tape, but uncompressed by the drive on playback. Even so 42MB/sec is one hell of a bandwidth requirement.
Mark..........
Re:anyone remember getting flexi vinyl w/comp mags
on
Floppy CDs And DVDs?
·
· Score: 1
I remember them.
I never had a problem with them because rather than copying them to tape first, I used to connected the computer directly to the stereo. Having said that the short playing time (even at 33rpm) didn't leave room for very complex programs. *8')
Ah, fond memories of the Sinclair ZX81 (Timex 1000 for those of you stateside).
This is an interesting device, both for the way it was implemented and the peculiarities this produced.
The hemispherical display area is apparently 10" (250mm) diameter. The rotating nature of the screen means that the the voxels are a swept segment of cylinder. They are approximately 0.3mm square in the direction of travel, but their length of travel is different depending on how close to the centre they are.
This means that there is more detail at the centre of the hemisphere than at the edges. In fact, at the edges, the voxels are stretched out to ~ 0.3mm * 0.3mm * 4mm!
Interesting, but hardly revolutionary. I would be more interested in the use of LEPS or TFT style technology laid down in layers, possibly using a photo-lithography or similar process. Spinning disks reminds me far too much of John Logie Baird's mechanical television.
The last of these is particularly interesting (if somewhat self contradictory, as it describes Univacs miniaturised planes which are only 4.5" square, and cost $6000 each for 1 kilobits of storage.
For another, there's no way I've ever heard of to actually "quad pump" a clock signal....
DDR doesn't quite work as you described. No current memory uses a single clock line, almost all have many clock lines. DDR for instance has matched pairs of clocks 180o out of phase, so when one set is rising the other is falling.
With Quad pumped memory you have 4 sets of clock signals, each 90o out of phase with the previous, thus each part of the cycle is controlled by one of the 4 sets of clock lines.
And no, Kingston's "quad-pumped" SDRAM isn't really quad-pumped either; it's just DDR which is cleverly interleaved to essentially make it twice as wide.
But DDR is just SDR that is cleverly interleaved to essentially make it twice as wide. *8')
Slashdot Mirror
"w2k is effected as well"
Something cannot be effected. Something can be affected, but to affect something you effect a change.
Back in the days when VRAM was used, it was because the bandwidth consumed just to feed the DACs was a significant proportion of the total available memory bandwidth. Then, having an output only port made a lot of sense.
Since those times, video memory bandwidth has increased enormously but DAC requirements have not. A 32 bit 3k x 2k LCD at 60Hz will only consume 1.4GB/s. Even maxing out a 400MHz DAC, you top out at 1.6GB/s with 32 bit colour. This is well under 10% of a modern graphics memory subsystems total bandwidth.
The only way I can think that there would be a return to dual port memory like VRAM of old would be with asymmetric ports. A fast one for the main I/O and a slow port (only say 2GB/s) for the DACs. If this 'slow' port was highly pin efficient, say a 2 bit wide packetised serial interface like rambus, this might be worth the die space/pin count. Even so, these pins might be better off employed making the main memory interface 10% faster or more efficient.
Take care,
Mark..........
What killed Tomorrows World for me was when they went from Live broadcasts with the occasional reel of video to being a completely pre-recorded show.
Sure, being live meant that some of the demonstrations were not perfect, but it meant that you saw the technology warts and all, not a marketing piece. I remember the first ISDN based video conferencing demo, before decent compression, and it looked terrible. You couldn't have given those scratchy little black and white screened video phones away.
Going pre-recorded meant that they lost the spontaneity and it meant that they couldn't be as up to date. All they gained was a little slickness in the production, which didn't actually help make the programme any more watchable and if anything made it feel less cutting-edge.
Why not go the whole hog and have the Kernel in microcode on the processor? Worked well enough for the Transputer. Even modern processors have a hard time competing with the context switch time of a 30MHz T800.
Take care,
Mark..........
What we are starting to see in the embedded market though is more use of embedding the processing core in an FPGA. If you only need 386/486 performance you can easily slot in an 8, 16 or even 32bit RISC core onto your FPGA, it might only take up a quarter of a decent sized one, add the your other logic to the rest of the FPGA. The only difference might be that you will be using a larger FPGA than you would have before (more gates, more interface pins for the cpu's memory and i/o busses). You don't then have to worry about your processor being obsoleted, only your FPGA (which you would have had to worry about anyway).
Mark..........
For those who are interested, there are some providers who have SMTP AUTHenticated mail relays for their customers which work on ports other than port 25.
My own (Gradwell.com, with their email forwarding account) email host has their SMTP server listening on port 225 as well as 25.
I am sure that should port 225 be blocked, they would be happy to set up listeners on other ports too. Who knows, if the problem with such port blocking gets bad enough, there is no reason you coulsn't build a server whos every port connected you to an SMTP server. *8')
Take care,
Mark..........
Personally, I work a contracted 30 hour week. My company went through difficulties a while back and couldn't afford the raise they had been promising. Rather than lose me altogether they agreed to my request to move to a 4 day week.
They are happy that they didn't have to lose me, I am happy that i now get a three day weekend. I still earn more than enough to live on and have lots more free time.
Most of the people who I speak to about this are envious, in some respects it is just like being a student again, but now I have money.
Mark..........
--
Sorry to be picky, but what was wrong with calling it The UK Sunday Times?
Take care,
Mark..........
--
The thing that I don't understand is they we are supposed to be talking about uncompressed HDTV resolution video. Now by my calculations 1280*1080 pixels at 30fps is only 42MB/sec if you use can encode with 1 byte per pixel! Thus even their 'uncompressed' video must be compressed a little.
This leads me to think that it might be compressed more than they are saying on the actual tape, but uncompressed by the drive on playback. Even so 42MB/sec is one hell of a bandwidth requirement.
Mark..........
I remember them.
I never had a problem with them because rather than copying them to tape first, I used to connected the computer directly to the stereo. Having said that the short playing time (even at 33rpm) didn't leave room for very complex programs. *8')
Ah, fond memories of the Sinclair ZX81 (Timex 1000 for those of you stateside).
Mark..........
This is an interesting device, both for the way it was implemented and the peculiarities this produced.
The hemispherical display area is apparently 10" (250mm) diameter. The rotating nature of the screen means that the the voxels are a swept segment of cylinder. They are approximately 0.3mm square in the direction of travel, but their length of travel is different depending on how close to the centre they are.
This means that there is more detail at the centre of the hemisphere than at the edges. In fact, at the edges, the voxels are stretched out to ~ 0.3mm * 0.3mm * 4mm!
Interesting, but hardly revolutionary. I would be more interested in the use of LEPS or TFT style technology laid down in layers, possibly using a photo-lithography or similar process. Spinning disks reminds me far too much of John Logie Baird's mechanical television.
Take care,
Mark..........
Personally I feel that MRAM is a beautifully ioronic idea.
Anyone have a good place to send the kids to show them what CORE really was?
You might want to point people at http://www.psych.usyd.edu.au/pdp-11/ core.html which has a nice, brief summary of the technology.
Other sites worth looking at are:
http://web.mit.edu/6.933/www/core.html
http://www.computer-mus eum.org/collections/ferrite_mem.html
http://www.science.uva.n l/faculteit/museum/CoreMemory.html
http://www.fortunecity.com/ marina/reach/435/coremem.htm
The last of these is particularly interesting (if somewhat self contradictory, as it describes Univacs miniaturised planes which are only 4.5" square, and cost $6000 each for 1 kilobits of storage.
Mark..........
DDR doesn't quite work as you described. No current memory uses a single clock line, almost all have many clock lines. DDR for instance has matched pairs of clocks 180o out of phase, so when one set is rising the other is falling.
With Quad pumped memory you have 4 sets of clock signals, each 90o out of phase with the previous, thus each part of the cycle is controlled by one of the 4 sets of clock lines.
And no, Kingston's "quad-pumped" SDRAM isn't really quad-pumped either; it's just DDR which is cleverly interleaved to essentially make it twice as wide.
But DDR is just SDR that is cleverly interleaved to essentially make it twice as wide. *8')
Mark..........