[filler]Now I have to add some filler at the bottom if I want nice formatting. Too few characters per line? Why bother checking this? I think the moderators know pretty soon whether or not a post a legit. People like myself that just want to make a nicely formatted post have to put all kinds of junk into our posts as filler.[/filler]
Hey, sorry but I wrote this and want to have my name on it. Ignore my AC post please.
Contrary to what most posters here are advising, maybe we should set up a group, like a division of Groklaw for example, that has as much leaked closed-liscence code as possible.
The purpose of this closed-liscence division would be to run independant comparisons of new OSS contributions against a library of leaked closed-liscence code to ensure nothing gets slipped by the project managers and poisons the project source.
I was initially going to suggest that the project manager do this comparison, but that would be too risky for the project (closed-source legal teams might have a go at it). Instead using a trusted OSS community party to do the checking saves us the hassle of each project manager having to download all the latest leaked closed-source. The "source-notary" would have a central repository of leaked material, which would not be redistributed by them, only made available to the original authors and for use to run comparisons on new OSS project code submissions and therefore avoid having a company pay a developer to salt the OSS project with leaked code.
I think this is a pretty mature way of handling this and should satisfy all parties.
The way it works is that you set your acceptable bounds as one of the constraints that you feed to the solver. So if you may have an altitude constraint that is very narrow, while other constraints are looser.
Starting intervals:
x = [-250', 4000']
y = [300', 700']
z = [1', 3']
Where z is the altitude with respect to the landing field and x and y are the locations.
The solver will immediately throw away any answers that don't fit in those intervals and concentrate on finding the best solutions within those bounds.
Yes, I agree that it will be a long time before we see any of this stuff running a real-time flight control system. A more reasonable line of questioning might be "Was any of the modelling software used by Rutan's group based on Interval Arithmetic?"
For the record, while real-time systems may be a while off, there are several companies offering good commercial modeling software that are built on these concepts. The reason they aren't widely known is that they are used mostly by heavy industries like mining and oil companies.
Again (see other post in thread), I offer this link to Dr. van Emden's intro to the subject here at. I direct your attention to the "How" section that lists a few commercial products.
Say you have an optimization problem that gives you too many solutions to evaluate with conventional numerical methods. Take for example an oil refinery that wishes to find the very best efficient and safe way of refining oil. There are many variables, equations, and nasty loop back cycles in this process. Traditional numerical analysis using floating point arithmetic will be struggling with errors carried over from each stage of the process, and will usually only find solutions near by to where you told it to start.
With interval arithmetic (and constraint processing) you will know that your answers are always correct. Any uncertainty will simply be reflected in a wider interval to contain the Real that is being approximated, the worst case being [-inf, inf].
Dr. Maarten van Emden over at the University of Victoria has a nice introduction page to the subject here at:. That is one of his current areas of research.
Here is a little quote also taken from Dr. van Emden.
"While this may seem to be of only theoretical interest, it is not. If one wants to solve systems of nonlinear inequalities or perform global optimization with side conditions of that form, interval constraints are more effective than conventional numerical methods."Quote source:
What I'd like to point out about the method they are using to transport the heat from the hotspots to the sinks is that Zalman is using heatpipes, which eventually lose the fluid inside them and stop working.
Before I go and dish out $1G for this case, I would want to be sure that there is a refill valve in the system similar to refridgerators, or that there is an assurance that there will be a lifetime supply of replacement heatpipe units for this machine. I mean my lifetime.
The product PDF states lifetime rated at 50k hours. This is about 5 years time, which isn't as long as you'd expect given that there are no moving parts other than gas/liquid.
So, definitely still a winner in my book, but with a shorter lifetime than expected.
I did that to another card long time ago. I found it really helped a lot.
I found that when I touched the top of the card where the chip was mounted underneath, the card was quite hot.
So since the amount of heat needed to be dissipated from the card is constant regardless of where it's dissipated from, I mounted a heat sink on the back of the card too. Double the heat sink size by having half below, half on top. The contact area on the back/top of the card is smooth and is lightly covered in paste. If you look at the picture, you'll see that the sinks are mounted lengthwise opposite to the card (does that make sense?). I did that otherwise the sink on the back/top would come into contact with some of the solder points. This way, the only thing it touches is the back of the card.
I never figured out why chipset manufacturers mount the chips and sinks on the underside. Heat rises. It would make more sense from a thermal point of vue to mount everything on the other side of the card. Maybe there they don't want to get too close to the CPU on some motherboards who knows. I haven't bothered to look it up but there's probably some standard somewhere that says it must be so.
So back to your question, I don't care much for aesthetics or what the standard is. I just think it makes lots of sense to take heat away in the direction it wants to move.
Hey, thanks but I think you missed the point.
I'm not cooling the video memory, I'm cooling the GPU. I know that I can get heat sinks for cheap, but none of the stores in my town had them in stock and I used what I had kicking around the parts bin. The Zalman video cooler is way more expensive than what I did. Not only that but a second hand stock AMD heatsink is less that $15. Besides, it took me half and hour, and most of all, I had fun doing it.
I mean, look at it; how fucking funny is that?
The sound was driving me crazy one day so I got out the hacksaw.
Just take any old stock AMD or P4 heatsink and chop it in half. I didn't have proper heatsink fasteners on my card so drilled it out and zip tied it down. The bottom is still smooth and the paste was properly applied.
The only problem was getting the stock fan off as it was glued on. I put my card in a ziplock bag and chucked it in the freezer for half an hour. Then I used a screwdriver to pry off the fan assembly (with an old library card to protect the pcb).
Check it out (it's not a swiss watch but it gets the job done).
This spring, I attended a presentation by Dr. Andy Schloss. A musician who maps instruments and sounds to his three-dimensional electronic sensor that he invented in the 80s, he does quite a few live performances and has thought of many applications for his instrument outside of the music world.
More Details
Slashdot Mirror
Heh
You must mean this one.
I'm just a buffalo spammer,
In the heart of America...
Singing, woe yoe yoe, woe woe yoe yoe
Woe yoe yoe yo, yo yo woe yo woe yo yoe
Beautiful panda reference.
Here are the videos from the video gallery page.
Rocketman-LAN.wmv
trinidad-pov.wmv
egypt.wmv
MTV.wmv
brazil.wmv
bush.wmv
olympics.wmv
boystoys.wmv
rocketeer.wmv
extreme_machines.wmv
beach.wmv
knightrider.wmv
superbowl.wmv
media.wmv
stadium.wmv
Robocop.wmv
[filler]Now I have to add some filler at the bottom if I want nice formatting. Too few characters per line? Why bother checking this? I think the moderators know pretty soon whether or not a post a legit. People like myself that just want to make a nicely formatted post have to put all kinds of junk into our posts as filler.[/filler]
My favorite item is the telco powered cell-phone near the bottom of the page.
Look at the flash slideshow and see for yourself...
clicky
Hey, sorry but I wrote this and want to have my name on it. Ignore my AC post please. Contrary to what most posters here are advising, maybe we should set up a group, like a division of Groklaw for example, that has as much leaked closed-liscence code as possible.
The purpose of this closed-liscence division would be to run independant comparisons of new OSS contributions against a library of leaked closed-liscence code to ensure nothing gets slipped by the project managers and poisons the project source.
I was initially going to suggest that the project manager do this comparison, but that would be too risky for the project (closed-source legal teams might have a go at it). Instead using a trusted OSS community party to do the checking saves us the hassle of each project manager having to download all the latest leaked closed-source. The "source-notary" would have a central repository of leaked material, which would not be redistributed by them, only made available to the original authors and for use to run comparisons on new OSS project code submissions and therefore avoid having a company pay a developer to salt the OSS project with leaked code.
I think this is a pretty mature way of handling this and should satisfy all parties.
Not that much.
The way it works is that you set your acceptable bounds as one of the constraints that you feed to the solver. So if you may have an altitude constraint that is very narrow, while other constraints are looser.
Starting intervals:
x = [-250', 4000']
y = [300', 700']
z = [1', 3']
Where z is the altitude with respect to the landing field and x and y are the locations.
The solver will immediately throw away any answers that don't fit in those intervals and concentrate on finding the best solutions within those bounds.
Yes, I agree that it will be a long time before we see any of this stuff running a real-time flight control system. A more reasonable line of questioning might be "Was any of the modelling software used by Rutan's group based on Interval Arithmetic?"
For the record, while real-time systems may be a while off, there are several companies offering good commercial modeling software that are built on these concepts. The reason they aren't widely known is that they are used mostly by heavy industries like mining and oil companies.
Again (see other post in thread), I offer this link to Dr. van Emden's intro to the subject here at. I direct your attention to the "How" section that lists a few commercial products.
Yes, it's a very useful field.
Say you have an optimization problem that gives you too many solutions to evaluate with conventional numerical methods. Take for example an oil refinery that wishes to find the very best efficient and safe way of refining oil. There are many variables, equations, and nasty loop back cycles in this process. Traditional numerical analysis using floating point arithmetic will be struggling with errors carried over from each stage of the process, and will usually only find solutions near by to where you told it to start.
With interval arithmetic (and constraint processing) you will know that your answers are always correct. Any uncertainty will simply be reflected in a wider interval to contain the Real that is being approximated, the worst case being [-inf, inf].
Dr. Maarten van Emden over at the University of Victoria has a nice introduction page to the subject here at:. That is one of his current areas of research.
Here is a little quote also taken from Dr. van Emden.
"While this may seem to be of only theoretical interest, it is not. If one wants to solve systems of nonlinear inequalities or perform global optimization with side conditions of that form, interval constraints are more effective than conventional numerical methods."Quote source:
He means systems like interval arithmetic whereby you represent a Real number as an interval bounded by two machine floating point numbers.
First off, I really like this case.
What I'd like to point out about the method they are using to transport the heat from the hotspots to the sinks is that Zalman is using heatpipes, which eventually lose the fluid inside them and stop working.
Before I go and dish out $1G for this case, I would want to be sure that there is a refill valve in the system similar to refridgerators, or that there is an assurance that there will be a lifetime supply of replacement heatpipe units for this machine. I mean my lifetime.
The product PDF states lifetime rated at 50k hours. This is about 5 years time, which isn't as long as you'd expect given that there are no moving parts other than gas/liquid.
So, definitely still a winner in my book, but with a shorter lifetime than expected.
I did that to another card long time ago. I found it really helped a lot.
I found that when I touched the top of the card where the chip was mounted underneath, the card was quite hot.
So since the amount of heat needed to be dissipated from the card is constant regardless of where it's dissipated from, I mounted a heat sink on the back of the card too. Double the heat sink size by having half below, half on top. The contact area on the back/top of the card is smooth and is lightly covered in paste. If you look at the picture, you'll see that the sinks are mounted lengthwise opposite to the card (does that make sense?). I did that otherwise the sink on the back/top would come into contact with some of the solder points. This way, the only thing it touches is the back of the card.
I never figured out why chipset manufacturers mount the chips and sinks on the underside. Heat rises. It would make more sense from a thermal point of vue to mount everything on the other side of the card. Maybe there they don't want to get too close to the CPU on some motherboards who knows. I haven't bothered to look it up but there's probably some standard somewhere that says it must be so.
So back to your question, I don't care much for aesthetics or what the standard is. I just think it makes lots of sense to take heat away in the direction it wants to move.
Hey, thanks but I think you missed the point. I'm not cooling the video memory, I'm cooling the GPU. I know that I can get heat sinks for cheap, but none of the stores in my town had them in stock and I used what I had kicking around the parts bin. The Zalman video cooler is way more expensive than what I did. Not only that but a second hand stock AMD heatsink is less that $15. Besides, it took me half and hour, and most of all, I had fun doing it. I mean, look at it; how fucking funny is that?
The sound was driving me crazy one day so I got out the hacksaw.
Just take any old stock AMD or P4 heatsink and chop it in half. I didn't have proper heatsink fasteners on my card so drilled it out and zip tied it down. The bottom is still smooth and the paste was properly applied.
The only problem was getting the stock fan off as it was glued on. I put my card in a ziplock bag
and chucked it in the freezer for half an hour. Then I used a screwdriver to pry off the fan assembly (with an old library card to protect the pcb).
Check it out (it's not a swiss watch but it gets the job done).
Pic 1: http://fullcircletraining.com/images/quiet1.jpg
Pic 2: http://fullcircletraining.com/images/quiet2.jpg
You can see I did the same thing to the northbridge on the motherboard.
happy modding.
j.
This spring, I attended a presentation by Dr. Andy Schloss. A musician who maps instruments and sounds to his three-dimensional electronic sensor that he invented in the 80s, he does quite a few live performances and has thought of many applications for his instrument outside of the music world. More Details