Yes and no: building codes are for structure etc. Sometimes they mingle mechanicals in as well. And yes, they do vary quite a bit, and there's at least a few model codes out there. But when it comes to electrical codes, there's really just one in the U.S.: the NEC. Local adoptions may modify it slightly, but it's usually minor. Smaller localities may simply adopt a completely local, usually short, code, to save some money. That's all.
Yeah, because buying a replacement is oh-so-hard. Yes, you can replace even on unibody Apple laptops with glued-in battery, all it takes is a bit of ingenuity and right tools.
Don't be stupid and talk about stuff you don't know anything about.
1. The normal GFCIs that are required by NEC are not for personal protection! They are designed for protection of property only (fires due to electrically leaky motors, etc). A standard 30mA GFCI will gladly let you die. You need a 5mA GFCI for personal protection, and you won't find that in Home Depot. A GFCI is not a circuit breaker, it won't "keep you safe in the event of [...]overdraw".
2. A distribution system designed to code is already sufficiently protecting the wiring from overload. The only reason to use larger wire is to keep voltage drops in check on longer runs. Otherwise, oversized wire is completely pointless, it doesn't protect you from anything.
3. I don't know what the heck you're doing, but wires come insulated, and you're only supposed really to remove insulation, not add to it. The only time I'd even add insulation is when it's part of listed splicing means or used for repair. I have rewired a bunch of stuff in my house, had it pass inspection, and I have not added insulation anywhere.
4. Code comes first. Yeah, common sense all right, advice from someone who is demonstrably clueless. Yay!
Huh? What does 300.18 got to do with this situation? Raceway is complete, but then you modify it and it is again complete before you pull new wire through it.
I do agree about the 24" spacing not being an issue in single-family residential construction.
The wiring up to the wall outlet is protected from overloads by the circuit breaker on the panel. The wiring up to the outlet on the extension cord/power strip is protected either by a local breaker in the strip (often they are integrated with the on/off switch), or is sized to be protected by the panel breaker. So, what's your problem again?
In normal circumstances, it's not really possible to overload an outlet strip.
Nope, not even close. Continuous-rated contacts are way bigger than those would be. They only need to carry the current for a minute, and at that point they can reach the maximum operating temperature. I could stuff one 50A Molex Mini-Fit Sr contact in a USB Type A shell, and that would be continuous duty.
Never mind that the last time I was there, one could get a chip card that served as an ATM card, a payment card, a transportation discount pass with your picture on it, and I'm sure I'm forgetting a couple other things.
Again, nobody advocates pushing 380A to the device. You push 10x less to the device, and an on-board power converter steps-down the voltage. That's like power distribution engineering 101. Since the "charging equipment" on your typical motherboards and GPU cards routinely supplies in excess of 100A, I think it's a non-issue. Just because legacy technology is huge doesn't mean modern techniques can't pack it in a very small volume.
Sigh. This is 8th grade physics stuff. Power = Voltage * Current (DC values). If the cells are at 3.5V, what you do is supply the thing with an off-the-shelf telecom supply at 48V, and even with losses you only need to push 30A or so if the cells need 380A charging current. Since the plug only needs to handle that current for a short time, it can be much smaller than already-small 30A-capable contacts. It'd easily fit in the envelope of a full-size USB type A connector.
I don't know where you got your numbers from, but the energy density this supercap has is on par with batteries: 20Wh/kg. Now look at the size of the caps she has. Those are samples that weigh grams. Whatever fits in a cellphone will weigh probably on the order of 100g, and will store on the order of 1 Wh. I don't see kilowatts for charging, never mind that you absolutely don't have to charge in 20s. A one minute charge cuts the power by a factor of 5, and anyone sane will go with what's economical and makes business sense, not with what was put on a headline somewhere.
Never mind that it doesn't matter much what the energy rate (power) is, what matters is the stored energy. 1 Watt-hour is 0.86 kg*Kelvin for water. Assuming whatever the phone and battery are made out of have the heat capacity comparable to water, releasing 1 W-h will heat things up by 10 degrees C.
In other words: yes, it's as harmless as any current battery technology. Get over it.
My other gripe: couldn't that girl publish this stuff like everyone else out there? All we have is stupid news articles and a single-page PDF summary. Weren't they supposed to have write-ups submitted with the project? WTF is all this stuff? To me it's not $5k-worthy science if it's secret.
It is the basis of property law - the idea that raw land becomes valuable property only after some form of human labour is involved
I don't know what planet in what solar system you're talking about here, because certainly it's not the basis of the property law here on Earth. Unimproved land prices depend on zoning and desirability for given use, sometimes they are also influenced by speculators.
Whoa, whoa, not so fast. I'm talking about small freshwater post-glacial inland lakes. The ones I like are usually between an acre and 50 or so. No whales, no real bottom "features" (it's just silt deposited from whatever stream happens to run through the lake), I don't see much net use in those places either. I'm not saying I'm using dynamite, just that if I was there and was forced to feed a big group and it was legal to use, I'd do so.
I understand that some people fish for the heck of it, but when I'm bothered enough to do it, it's because I want some fresh fish to eat. I'd use dynamite a heartbeat if it were legal and I had a big group to feed.
Division operations for line drawing? Ever heard of Bresenham's algorithm? It can be further tweaked based on the organization of the frame buffer.
There are many state-of-the-art designs that are memory constrained. Parallax's Propeller has 512 32-bit words for high-speed cog memory. Their new Propeller II adds a third memory space (besides cog and hub memories), but the basic 512 word limit remains unchanged. XMOS XS1 architecture has 64 kbytes of RAM shared between code and data for a multi-core CPU tile. Each tile has at least 4 cores running at 125Mips, from that shared memory. All are quite capable chips, each in its own way (Prop II only exists as an FPGA testbed at the moment).
Recently on eBay I saw a bunch (like 500 or so) of HP Xeon servers with 72GB of DDR3 RAM each, each for under $500 shipped buy-it-now, with some sort of a volume discount, too. Since that's likely registered server RAM that IIRC costs more than non-registered consumer stuff, it seems like a steal. I'm not sure what sort of memory bandwidth requirements do modern compute nodes have, perhaps DDR3 isn't enough, but still - RAM is cheap, it may be cheaper than pushing data via a custom interface to a fast hard drive array.
Now that is some very informative stuff. Thank you. I'll avoid 2910s, then, just in case. There's plenty of selection and all I need is a couple 48 port L2 units with PoE.
Keep it in memory. It's not that expensive. If you decouple the snapshot memory (DDR3 or somesuch) from the main number-crunching CPU, and just keep that memory refreshed and error-corrected, you should be fine. You don't need disks.
For good checkpoint/restore, you probably need a custom node design that would accommodate it efficiently, but I can't see why it'd have a 70% overhead. Doing a copy of your memory contents to another memory that has same bandwidth and capacity, and then lazily moving that off-node without the main CPU being involved is no biggie. You probably could implement the memory bridge and the recover CPU on a simple FPGA. The main, fast CPU is crunching numbers, then stops, the FPGA takes a memory copy, the main CPU resumes, then the FPGA and the service CPU can keep the data locally on the node until the node fails and the snapshot contents are to be transferred to a replacement node. If one is doing linear algebra on a node only using linpack/lapack style stuff, then probably a GPU or an FPGA/ASIC would be more efficient than any off-the-shelf CPU anyway.
Slashdot Mirror
Yes and no: building codes are for structure etc. Sometimes they mingle mechanicals in as well. And yes, they do vary quite a bit, and there's at least a few model codes out there. But when it comes to electrical codes, there's really just one in the U.S.: the NEC. Local adoptions may modify it slightly, but it's usually minor. Smaller localities may simply adopt a completely local, usually short, code, to save some money. That's all.
I'd still take that VT100 :)
Yeah, because buying a replacement is oh-so-hard. Yes, you can replace even on unibody Apple laptops with glued-in battery, all it takes is a bit of ingenuity and right tools.
Don't be stupid and talk about stuff you don't know anything about.
1. The normal GFCIs that are required by NEC are not for personal protection! They are designed for protection of property only (fires due to electrically leaky motors, etc). A standard 30mA GFCI will gladly let you die. You need a 5mA GFCI for personal protection, and you won't find that in Home Depot. A GFCI is not a circuit breaker, it won't "keep you safe in the event of [...]overdraw".
2. A distribution system designed to code is already sufficiently protecting the wiring from overload. The only reason to use larger wire is to keep voltage drops in check on longer runs. Otherwise, oversized wire is completely pointless, it doesn't protect you from anything.
3. I don't know what the heck you're doing, but wires come insulated, and you're only supposed really to remove insulation, not add to it. The only time I'd even add insulation is when it's part of listed splicing means or used for repair. I have rewired a bunch of stuff in my house, had it pass inspection, and I have not added insulation anywhere.
4. Code comes first. Yeah, common sense all right, advice from someone who is demonstrably clueless. Yay!
Running Romex (NM-B) within conduit is fine, I don't know where you got the idea that it's not. The conduit needs to be properly sized, though.
Huh? What does 300.18 got to do with this situation? Raceway is complete, but then you modify it and it is again complete before you pull new wire through it.
I do agree about the 24" spacing not being an issue in single-family residential construction.
The wiring up to the wall outlet is protected from overloads by the circuit breaker on the panel. The wiring up to the outlet on the extension cord/power strip is protected either by a local breaker in the strip (often they are integrated with the on/off switch), or is sized to be protected by the panel breaker. So, what's your problem again?
In normal circumstances, it's not really possible to overload an outlet strip.
Except that at least one major market - the U.S. - is closed to such things. The NEC, IIRC, doesn't let you install any receptacles in the countertop.
Nope, not even close. Continuous-rated contacts are way bigger than those would be. They only need to carry the current for a minute, and at that point they can reach the maximum operating temperature. I could stuff one 50A Molex Mini-Fit Sr contact in a USB Type A shell, and that would be continuous duty.
Never mind that the last time I was there, one could get a chip card that served as an ATM card, a payment card, a transportation discount pass with your picture on it, and I'm sure I'm forgetting a couple other things.
Unfortunately, the is no article besides a one-page summary. It sucks. That's not science, that's bullshit to me.
Again, nobody advocates pushing 380A to the device. You push 10x less to the device, and an on-board power converter steps-down the voltage. That's like power distribution engineering 101. Since the "charging equipment" on your typical motherboards and GPU cards routinely supplies in excess of 100A, I think it's a non-issue. Just because legacy technology is huge doesn't mean modern techniques can't pack it in a very small volume.
Sigh. This is 8th grade physics stuff. Power = Voltage * Current (DC values). If the cells are at 3.5V, what you do is supply the thing with an off-the-shelf telecom supply at 48V, and even with losses you only need to push 30A or so if the cells need 380A charging current. Since the plug only needs to handle that current for a short time, it can be much smaller than already-small 30A-capable contacts. It'd easily fit in the envelope of a full-size USB type A connector.
A charger designed to supply that sort of a current, which would be built into the phone. It's called engineering. You must have never heard of it.
No, that's what happens when you do it at a university lab, not at home while being clueless to boot. That's all.
I don't know where you got your numbers from, but the energy density this supercap has is on par with batteries: 20Wh/kg. Now look at the size of the caps she has. Those are samples that weigh grams. Whatever fits in a cellphone will weigh probably on the order of 100g, and will store on the order of 1 Wh. I don't see kilowatts for charging, never mind that you absolutely don't have to charge in 20s. A one minute charge cuts the power by a factor of 5, and anyone sane will go with what's economical and makes business sense, not with what was put on a headline somewhere.
Never mind that it doesn't matter much what the energy rate (power) is, what matters is the stored energy. 1 Watt-hour is 0.86 kg*Kelvin for water. Assuming whatever the phone and battery are made out of have the heat capacity comparable to water, releasing 1 W-h will heat things up by 10 degrees C.
In other words: yes, it's as harmless as any current battery technology. Get over it.
My other gripe: couldn't that girl publish this stuff like everyone else out there? All we have is stupid news articles and a single-page PDF summary. Weren't they supposed to have write-ups submitted with the project? WTF is all this stuff? To me it's not $5k-worthy science if it's secret.
- Said someone who never saw a fuse in his life, and never got a piece of wire glowing by shorting a Ni-Cd cell.
IOW: So how is it different from the batteries we have now, as far as your pseudo-concerns are concerned?!
It is the basis of property law - the idea that raw land becomes valuable property only after some form of human labour is involved
I don't know what planet in what solar system you're talking about here, because certainly it's not the basis of the property law here on Earth. Unimproved land prices depend on zoning and desirability for given use, sometimes they are also influenced by speculators.
Whoa, whoa, not so fast. I'm talking about small freshwater post-glacial inland lakes. The ones I like are usually between an acre and 50 or so. No whales, no real bottom "features" (it's just silt deposited from whatever stream happens to run through the lake), I don't see much net use in those places either. I'm not saying I'm using dynamite, just that if I was there and was forced to feed a big group and it was legal to use, I'd do so.
I understand that some people fish for the heck of it, but when I'm bothered enough to do it, it's because I want some fresh fish to eat. I'd use dynamite a heartbeat if it were legal and I had a big group to feed.
Division operations for line drawing? Ever heard of Bresenham's algorithm? It can be further tweaked based on the organization of the frame buffer.
There are many state-of-the-art designs that are memory constrained. Parallax's Propeller has 512 32-bit words for high-speed cog memory. Their new Propeller II adds a third memory space (besides cog and hub memories), but the basic 512 word limit remains unchanged. XMOS XS1 architecture has 64 kbytes of RAM shared between code and data for a multi-core CPU tile. Each tile has at least 4 cores running at 125Mips, from that shared memory. All are quite capable chips, each in its own way (Prop II only exists as an FPGA testbed at the moment).
Recently on eBay I saw a bunch (like 500 or so) of HP Xeon servers with 72GB of DDR3 RAM each, each for under $500 shipped buy-it-now, with some sort of a volume discount, too. Since that's likely registered server RAM that IIRC costs more than non-registered consumer stuff, it seems like a steal. I'm not sure what sort of memory bandwidth requirements do modern compute nodes have, perhaps DDR3 isn't enough, but still - RAM is cheap, it may be cheaper than pushing data via a custom interface to a fast hard drive array.
Now that is some very informative stuff. Thank you. I'll avoid 2910s, then, just in case. There's plenty of selection and all I need is a couple 48 port L2 units with PoE.
Keep it in memory. It's not that expensive. If you decouple the snapshot memory (DDR3 or somesuch) from the main number-crunching CPU, and just keep that memory refreshed and error-corrected, you should be fine. You don't need disks.
For good checkpoint/restore, you probably need a custom node design that would accommodate it efficiently, but I can't see why it'd have a 70% overhead. Doing a copy of your memory contents to another memory that has same bandwidth and capacity, and then lazily moving that off-node without the main CPU being involved is no biggie. You probably could implement the memory bridge and the recover CPU on a simple FPGA. The main, fast CPU is crunching numbers, then stops, the FPGA takes a memory copy, the main CPU resumes, then the FPGA and the service CPU can keep the data locally on the node until the node fails and the snapshot contents are to be transferred to a replacement node. If one is doing linear algebra on a node only using linpack/lapack style stuff, then probably a GPU or an FPGA/ASIC would be more efficient than any off-the-shelf CPU anyway.