US TV: "Power" button on the TV itself and the one on the remote do exactly the same thing: switch between "on" and "standby". The only way to get it off is to unplug the mains cord.
Who cares? The CRT in my TV is turned off (to the point that it takes about 10 seconds to fully come back on), so the component that takes 99.9% of the power isn't drawing a thing. The only thing required for standby is the IR receiver circuit. How much current can that possibly draw (at low voltages to boot) when idle?
but there are a lot of people who bring upon these conditions because of their own lazyness & over-indulgence. Fixing their hearts won't nessesarily make them want to improve other area's of their life which created the heart problem in the beginning.
You know, you can say that about a lot of diseases. You get lung cancer from smoking. You get liver problems from drinking. People primarily [1] get HIV through doing stupid things that are universally known to be stupid. But yet, we do what we can to treat them. Part of it is selfishness - I'd rather eradicate AIDS than see it mutate into an airborne version, for instance - but the biggest reason is because it's the right thing to do.
The men in my family tend to die of heart attacks, regardless of lifestyle. It's nice to know that if The Big One hits me while eating an oats-and-tofu breakfast after a morning jog, that the doctors might have another tool to keep me from dying. Screw you and your cost analysis.
[1] Yeah, children, transfusions, yada yada yada. Look up "primarily" before attempting to "educate" me.
My new approach to dealing with voucher opponents: replace the word "school" with "hospital", "teacher" with "doctor", and "student" with "patient". Make them explain why limiting your choice of which doctors and clinics to use when you're sick is good for society. Make them explain why allowing you to visit the doctor of your choice is racist and discriminates against the poor.
I can't let this common misconception stand. Most farmers are more advanced in their use of computers than the average person.
That's nice. I live in rural Nebraska and know more farmers than usual, and I can saw with confidence that most of the aren't interested in cutting-edge computer technology unless it directly relates to their work.
That doesn't mean they're stupid, or backward, but that the ones I know are more interested in getting the job done than the minor details. If Netscape 4 lets them do what they need, then so be it.
Also note that I'm not saying all farmers, but in my personal experience, that seems to be true for most of them.
Don't take this personally, but very few people other than bloggers regard the whole concept seriously. Its great that you've found a hobby that you truly enjoy - that's an important part of a well-rounded life - but you can't expect the rest of the world to get as excited about it.
In other words, your ideas are not intriguing to us and we do not wish to subscribe to your RSS feed. Nothing personal.
Depending on whether you have BLOCKSIZE set in the environment, the tests are 4/2GB and 1/0.5GB.
By the way, good catch. That's actually 4GB and.5GB, respectively.
Oh, and another thing: re-run the test on both machines to see what effect filesystem caching has on those times. As per another post of mine above, OS X (10.4.4 here) doesn't seem to cache anything at all, while FreeBSD tries to do its best to run entirely from RAM.
Any chance you could re-run that test using UFS as your file system on the Mac instead of HFS+?
It's unlikely. The iMac's my wife's desktop machine, and I'm pretty sure she'd shoot me for repartitioning it to run an experiment for Slashdot.
pOTOH, we're probably getting an external Firewire drive in the near future, and I'll definitely be running the test there. That'll be too late for this thread, though.
Man would I love a disk that can read 2GB (not to mention 4) in 1.6 seconds, where can you get those?
Beats me. Since "--file=/dev/null" is a microoptimization that doesn't actually read the source file, just its stat() information, I've never actually seen a disk that fast.
Where'd you get the gtar? I'm using gnutar from the base system (2005-03-21, 186024 bytes), since one of the features of 10.4 is that tar and other related programs support resource forks.
Depends on if you consider x% of the interweb population to be valuable to your business.
Someone who is using Netscape 4 (as an example) is either 1) too broke to afford a machine that can run newer versions, or 2) technophobic, or 3) determined to make the world bend to their will. How much money do you want to spend herding any of those three to your website, assuming you're in a high-tech business?
Now, if you sell tractor parts, then you have a legitimate point. If you're selling music downloads or something else new, hip, and low margin, then forget it - support costs will eat any profit you might have made, and that's ignoring the opportunity costs.
Is HFS+ journaling turned on? Do you have journaling enabled on the FreeBSD machine? Comparable system buses? Hard drive performance differences: rotaional speed, hard drive caching, transfer rates? What about file locality on each system?
Let me retract my answer; while I think it was accurate, it didn't really address the main issue.
From what I can tell, even though the machine has 358MB of free RAM at the moment, it doesn't seem to do any caching at all:
$ time gnutar 2>/dev/null cf/dev/null. gnutar cf/dev/null . 2>/dev/null 2.34s user 9.72s system 42% cpu 28.541 total $ time gnutar 2>/dev/null cf/dev/null. gnutar cf/dev/null . 2>/dev/null 2.36s user 10.06s system 37% cpu 32.970 total $ time gnutar 2>/dev/null cf/dev/null. gnutar cf/dev/null . 2>/dev/null 2.31s user 9.68s system 42% cpu 28.516 total
Why isn't it using some of the memory we paid for to cache the directory layout? The example above isn't referring to a terraserver with millions of files. I'd imagine that all of the drive accesses used to complete that operation should fit easily in the available RAM, and yet it's like it's not even trying.
In comparison, here's another FreeBSD machine at home traversing the ports collection. Notice that the second and later runs are much faster than the first:
$ find/usr/ports | wc 102424 102424 4194165 $ time gtar 2>/dev/null cf/dev/null/usr/ports gtar cf/dev/null/usr/ports 2>/dev/null 1.05s user 4.00s system 24% cpu 20.911 total $ time gtar 2>/dev/null cf/dev/null/usr/ports gtar cf/dev/null/usr/ports 2>/dev/null 1.01s user 3.74s system 76% cpu 6.227 total $ time gtar 2>/dev/null cf/dev/null/usr/ports gtar cf/dev/null/usr/ports 2>/dev/null 0.95s user 3.79s system 79% cpu 5.979 total
Even during the first pass, it was much faster (by a wider margin than you'd expect given the similar hardware specs - you'll have to take my word on this one). After that, though, I doubt it ever pulled a single byte from the drive.
Now, feel free to tell me that I missed something stupidly obvious in the Mac's settings like enabling the caches, as long as you also tell me how to fix it.:-) I would be thrilled to discover that it was my own ignorance that made this system's hard drive so slow. I really don't think that's the case, though.
Do you have journaling enabled on the FreeBSD machine?
Yes. But if all else were exactly equal, that would be an indictment of OS X's design, wouldn't it?
Comparable system buses?
No. The iMac is much faster than the laptop in pretty much ever respect.
Hard drive performance differences: rotaional speed, hard drive caching, transfer rates?
The laptop has a slow drive with almost no cache on a UDMA-33 bus.
What about file locality on each system?
I have no idea. All systems have been significantly used over the course of several years. Again, if UFS handles "aging" much better than HFS+, then that doesn't look good for the latter.
The same commands on the FreeBSD source code tree on my laptop, which has a K6-3 CPU and a 10GB IDE hard drive:
$ find . | wc 40965 40965 1440726 $ du | tail -n1 453178. $ time gtar 2>/dev/null cf/dev/null. gtar cf/dev/null . 2>/dev/null 1.69s user 4.28s system 98% cpu 6.061 total
So, it has 20% more files to scan, but still finishes in less than one fourth the time.
I mention this for the benefit of people who think I'm comparing an entry-level iMac to some hyper-powered 16-way Unix box with SCSI RAID. By all rights, the Mac should spank this laptop, but that's not how it usually works out.
Other than that, the iMac is pretty snappy and extremely usable. I just wish I could find the "[_]make my drive not crawl" system setting.
If Intel's latest and greatest dual-core is only 10-15% faster than the single-core G5, he was spot-on about performance claims before the architecture change.
You're right, of course. However, the long-term question is what relative performance will be like five years from now. Do you think the PPC will see as many speed boosts between now and then as the Intel chips? Steve apparently didn't. In fact, he basically bet the farm that Intel would widen that narrow margin.
From that perspective, he timed this move perfectly. A few years ago, everyone would have screamed bloody murder at migrating to a slower chip (regardless of whether it actually was slower - I'm staying out of that debate). A few years from now, everyone would have screamed bloody murder as their systems fell further behind the curve. Instead, he made the migration just at the time when performance was about equal, so the worst we can say is that there's no huge benefit from switching at this point in time.
I'm not a Mac fanboy by any stretch, but I'll give them credit for pulling this one off about as well as could be hoped.
For those rare tasks that are written to be multithreaded it'll be ~1.8x as fast (thread overhead, bus contention, etc.)
Pardon my ignorance, but why would multi-threading be more important than multi-processing on an SMP system (which is what this basically is)?
If nothing else, it'd be nice to have a whole core to run my application while the OS uses the other core for system stuff, so even single-threaded apps should benefit nicely.
...with the exception that I/O Kit and the HFS+ filesystem seem to think a hard drive is a floppy and do their best to set its performance to that level.
I fully understand that my wife's iMac isn't an Xserve, but holy cow, the drive is slow. I'm not exaggerating when I say that the estimate stage of an Amanda backup - that is, basically running "tar --file/dev/null" - takes over an hour to complete on 20GB of content.
For a (not very) quick comparison, here's how long that process takes to run on my home directory on my FreeBSD desktop:
$ find . | wc 52788 60297 3122487 $ du | tail -n1 4270686. $ time gtar 2>/dev/null cf/dev/null. gtar cf/dev/null . 2>/dev/null 0.42s user 1.04s system 89% cpu 1.635 total
On the Mac, though, we see:
$ find . | wc 34346 35311 1997441 $ du | tail -n1 1026640. $ time gnutar 2>/dev/null cf/dev/null. gnutar cf/dev/null . 2>/dev/null 2.27s user 9.41s system 41% cpu 27.874 total
Even though my home directory in the Mac has 35% fewer files and directories to glance at, the tar run takes 17 times longer.
Now, I don't want to be that "a file copy takes 20 minutes!" guy, but this thing really is incredibly slow at certain operations. Just because parts of OS X have a Unix heritage doesn't mean that the whole package has Unix-like performance.
Buy a Mac because you like the OS and applications. We did. If you buy one because you think it's going to dominate all available benchmarks, though, then you're going to be sadly disappointed.
No kidding! Result #5 was "Roseann Barr's Loose Meat Sandwich.". I almost sprained an eyeball in my haste to avoid looking at that the description of that link.
Flamebait or troll or not, I accuse you of flagrant exaggeration, if not outright lying.
Nuts - you beat me to it. IQ tests are usually normalized to a standard deviation of either 15 or 16 points, so his "190 IQ" would put him between 5.6 and 6 standard deviations above normal. A quick Google only returned z-to-P calculators that work up to z=5, and I'm too lazy to look too hard, but even z=5 would put him in the top 0.999999399 of the population - making him smarter than all but about 3600 people.
Hey, someone has to be at the top, but I doubt it's some guy on Slashdot whining about how no one respects him.
What part of our infrastructure are we adding more copper to?
Aluminum can carry power just fine as long as you use the proper interconnects; the lack thereof is why aluminum home wiring got a bad reputation.
Computers don't use a whole lot of copper and are probably moving toward more exotic technologies anyway (carbon nanotubes, optical switches, etc.).
Networks are beginning to migrate over to fiber or RF.
People are increasingly switching from landlines to cell phones.
I'm sure Bad Things would happen if we ran out of copper, but is there anything we're currently using it for that couldn't be converted to use something else?
Try telling that to a woman that's been dying to pee all day. There's a method to their madness...
So I can take 2 seconds to lift the seat (with the side of my shoe; see post above), but she can't take two seconds to put it down? Title IX says that they can put the seat down just as easily as we can lift it.
Unless we're talking about my wife, in which case all of the above (especially the part about feet) was all in jest, and I would never use the toilet without cleaning the whole thing with Lysol afterward for your butt's convenience. Honest.
But less so the undersides where women want men to put their fingers to raise and lower toilet seats!
You use your fingers to lift toilet seats? Good God, man - that's why shoes have cleats. If a woman wants me to lift the seat each time, then she can darn well sit on my footprints.
Next you'll be telling use that you don't flush public toilets by stepping on the lever, and we'll all laugh at you.
My initial reaction is that Xen is the way to go because it is REALLY running different Linux instances.
It depends on what you're trying to accomplish. If you want to host full-blown virtual machines, each with their own static resources, then Xen is great.
If you're only trying to isolate one specific application, though, and you're not sure how much memory, disk, or CPU to allocate to it, then the style of virtualization that OpenVZ (among others) uses may be exactly what you needed.
I've been using FreeBSD jails (similar idea) for a few years, and that kind of virtualization is a great way to isolate potential security exploits from the rest of the system. Want to try a binary-only application? Fire it up inside a jail. Need to run some PHP app with a sketchy track record? Host it on a jailed Apache server.
Many of us are starting to use jails where we would have used chroot a few years ago. I know I'm hopelessly geeky, but my desktop machine is running three full-fledged jail systems - that is, complete images that I can SSH into and play around inside of - as I write this. When you can do such things with almost zero runtime overhead, there's no great reason why not to.
Slashdot Mirror
Who cares? The CRT in my TV is turned off (to the point that it takes about 10 seconds to fully come back on), so the component that takes 99.9% of the power isn't drawing a thing. The only thing required for standby is the IR receiver circuit. How much current can that possibly draw (at low voltages to boot) when idle?
You know, you can say that about a lot of diseases. You get lung cancer from smoking. You get liver problems from drinking. People primarily [1] get HIV through doing stupid things that are universally known to be stupid. But yet, we do what we can to treat them. Part of it is selfishness - I'd rather eradicate AIDS than see it mutate into an airborne version, for instance - but the biggest reason is because it's the right thing to do.
The men in my family tend to die of heart attacks, regardless of lifestyle. It's nice to know that if The Big One hits me while eating an oats-and-tofu breakfast after a morning jog, that the doctors might have another tool to keep me from dying. Screw you and your cost analysis.
[1] Yeah, children, transfusions, yada yada yada. Look up "primarily" before attempting to "educate" me.
My new approach to dealing with voucher opponents: replace the word "school" with "hospital", "teacher" with "doctor", and "student" with "patient". Make them explain why limiting your choice of which doctors and clinics to use when you're sick is good for society. Make them explain why allowing you to visit the doctor of your choice is racist and discriminates against the poor.
That's nice. I live in rural Nebraska and know more farmers than usual, and I can saw with confidence that most of the aren't interested in cutting-edge computer technology unless it directly relates to their work.
That doesn't mean they're stupid, or backward, but that the ones I know are more interested in getting the job done than the minor details. If Netscape 4 lets them do what they need, then so be it.
Also note that I'm not saying all farmers, but in my personal experience, that seems to be true for most of them.
Number of occurrences of "blog": 6
Number of occurrences of "blogosphere": 1
Don't take this personally, but very few people other than bloggers regard the whole concept seriously. Its great that you've found a hobby that you truly enjoy - that's an important part of a well-rounded life - but you can't expect the rest of the world to get as excited about it.
In other words, your ideas are not intriguing to us and we do not wish to subscribe to your RSS feed. Nothing personal.
By the way, good catch. That's actually 4GB and .5GB, respectively.
Oh, and another thing: re-run the test on both machines to see what effect filesystem caching has on those times. As per another post of mine above, OS X (10.4.4 here) doesn't seem to cache anything at all, while FreeBSD tries to do its best to run entirely from RAM.
It's unlikely. The iMac's my wife's desktop machine, and I'm pretty sure she'd shoot me for repartitioning it to run an experiment for Slashdot. pOTOH, we're probably getting an external Firewire drive in the near future, and I'll definitely be running the test there. That'll be too late for this thread, though.
Beats me. Since "--file=/dev/null" is a microoptimization that doesn't actually read the source file, just its stat() information, I've never actually seen a disk that fast.
Where'd you get the gtar? I'm using gnutar from the base system (2005-03-21, 186024 bytes), since one of the features of 10.4 is that tar and other related programs support resource forks.
Someone who is using Netscape 4 (as an example) is either 1) too broke to afford a machine that can run newer versions, or 2) technophobic, or 3) determined to make the world bend to their will. How much money do you want to spend herding any of those three to your website, assuming you're in a high-tech business?
Now, if you sell tractor parts, then you have a legitimate point. If you're selling music downloads or something else new, hip, and low margin, then forget it - support costs will eat any profit you might have made, and that's ignoring the opportunity costs.
Let me retract my answer; while I think it was accurate, it didn't really address the main issue.
From what I can tell, even though the machine has 358MB of free RAM at the moment, it doesn't seem to do any caching at all:
Why isn't it using some of the memory we paid for to cache the directory layout? The example above isn't referring to a terraserver with millions of files. I'd imagine that all of the drive accesses used to complete that operation should fit easily in the available RAM, and yet it's like it's not even trying.In comparison, here's another FreeBSD machine at home traversing the ports collection. Notice that the second and later runs are much faster than the first:
Even during the first pass, it was much faster (by a wider margin than you'd expect given the similar hardware specs - you'll have to take my word on this one). After that, though, I doubt it ever pulled a single byte from the drive.
Now, feel free to tell me that I missed something stupidly obvious in the Mac's settings like enabling the caches, as long as you also tell me how to fix it. :-) I would be thrilled to discover that it was my own ignorance that made this system's hard drive so slow. I really don't think that's the case, though.
Yes.
Do you have journaling enabled on the FreeBSD machine?
Yes. But if all else were exactly equal, that would be an indictment of OS X's design, wouldn't it?
Comparable system buses?
No. The iMac is much faster than the laptop in pretty much ever respect.
Hard drive performance differences: rotaional speed, hard drive caching, transfer rates?
The laptop has a slow drive with almost no cache on a UDMA-33 bus.
What about file locality on each system?
I have no idea. All systems have been significantly used over the course of several years. Again, if UFS handles "aging" much better than HFS+, then that doesn't look good for the latter.
That seems reasonable, but not too likely to happen. I only have the internal drive and my wife would skin me alive for trying.
I mention this for the benefit of people who think I'm comparing an entry-level iMac to some hyper-powered 16-way Unix box with SCSI RAID. By all rights, the Mac should spank this laptop, but that's not how it usually works out.
Other than that, the iMac is pretty snappy and extremely usable. I just wish I could find the "[_]make my drive not crawl" system setting.
You're right, of course. However, the long-term question is what relative performance will be like five years from now. Do you think the PPC will see as many speed boosts between now and then as the Intel chips? Steve apparently didn't. In fact, he basically bet the farm that Intel would widen that narrow margin.
From that perspective, he timed this move perfectly. A few years ago, everyone would have screamed bloody murder at migrating to a slower chip (regardless of whether it actually was slower - I'm staying out of that debate). A few years from now, everyone would have screamed bloody murder as their systems fell further behind the curve. Instead, he made the migration just at the time when performance was about equal, so the worst we can say is that there's no huge benefit from switching at this point in time.
I'm not a Mac fanboy by any stretch, but I'll give them credit for pulling this one off about as well as could be hoped.
Pardon my ignorance, but why would multi-threading be more important than multi-processing on an SMP system (which is what this basically is)?
If nothing else, it'd be nice to have a whole core to run my application while the OS uses the other core for system stuff, so even single-threaded apps should benefit nicely.
...with the exception that I/O Kit and the HFS+ filesystem seem to think a hard drive is a floppy and do their best to set its performance to that level.
I fully understand that my wife's iMac isn't an Xserve, but holy cow, the drive is slow. I'm not exaggerating when I say that the estimate stage of an Amanda backup - that is, basically running "tar --file /dev/null" - takes over an hour to complete on 20GB of content.
For a (not very) quick comparison, here's how long that process takes to run on my home directory on my FreeBSD desktop:
On the Mac, though, we see:Even though my home directory in the Mac has 35% fewer files and directories to glance at, the tar run takes 17 times longer.
Now, I don't want to be that "a file copy takes 20 minutes!" guy, but this thing really is incredibly slow at certain operations. Just because parts of OS X have a Unix heritage doesn't mean that the whole package has Unix-like performance.
Buy a Mac because you like the OS and applications. We did. If you buy one because you think it's going to dominate all available benchmarks, though, then you're going to be sadly disappointed.
No kidding! Result #5 was "Roseann Barr's Loose Meat Sandwich.". I almost sprained an eyeball in my haste to avoid looking at that the description of that link.
Nuts - you beat me to it. IQ tests are usually normalized to a standard deviation of either 15 or 16 points, so his "190 IQ" would put him between 5.6 and 6 standard deviations above normal. A quick Google only returned z-to-P calculators that work up to z=5, and I'm too lazy to look too hard, but even z=5 would put him in the top 0.999999399 of the population - making him smarter than all but about 3600 people.
Hey, someone has to be at the top, but I doubt it's some guy on Slashdot whining about how no one respects him.
Richard a capella darn near did it for me, and I wasn't even particularly depressed when I hit "play".
Aluminum can carry power just fine as long as you use the proper interconnects; the lack thereof is why aluminum home wiring got a bad reputation.
Computers don't use a whole lot of copper and are probably moving toward more exotic technologies anyway (carbon nanotubes, optical switches, etc.).
Networks are beginning to migrate over to fiber or RF.
People are increasingly switching from landlines to cell phones.
I'm sure Bad Things would happen if we ran out of copper, but is there anything we're currently using it for that couldn't be converted to use something else?
So I can take 2 seconds to lift the seat (with the side of my shoe; see post above), but she can't take two seconds to put it down? Title IX says that they can put the seat down just as easily as we can lift it.
Unless we're talking about my wife, in which case all of the above (especially the part about feet) was all in jest, and I would never use the toilet without cleaning the whole thing with Lysol afterward for your butt's convenience. Honest.
You use your fingers to lift toilet seats? Good God, man - that's why shoes have cleats. If a woman wants me to lift the seat each time, then she can darn well sit on my footprints.
Next you'll be telling use that you don't flush public toilets by stepping on the lever, and we'll all laugh at you.
It depends on what you're trying to accomplish. If you want to host full-blown virtual machines, each with their own static resources, then Xen is great.
If you're only trying to isolate one specific application, though, and you're not sure how much memory, disk, or CPU to allocate to it, then the style of virtualization that OpenVZ (among others) uses may be exactly what you needed.
Yes you do - you just don't know it yet.
I've been using FreeBSD jails (similar idea) for a few years, and that kind of virtualization is a great way to isolate potential security exploits from the rest of the system. Want to try a binary-only application? Fire it up inside a jail. Need to run some PHP app with a sketchy track record? Host it on a jailed Apache server.
Many of us are starting to use jails where we would have used chroot a few years ago. I know I'm hopelessly geeky, but my desktop machine is running three full-fledged jail systems - that is, complete images that I can SSH into and play around inside of - as I write this. When you can do such things with almost zero runtime overhead, there's no great reason why not to.