"Byte-manipulation isn't one of Alpha's strong points..."
Sorry... I should have said "memory copying" since I use memcpy(). As for Byte-manipulation not being a strong point of Alpha, well, isn't the point of a general-purpose CPU/memory subsystem benchmark to uncover these kind of things?;)
"If you want a real idea of memory bandwidth... use...Stream..."
Nope, I wasn't trying to assess memory bandwidth. By "memory access speed" I was referring to memory latency, a critical variable to system performance.
"...benchmark using the applications and data that are important to you."
heheh. Yeah, sure. The reason synthetic benchmarks and "application conglomeration" benchmarks exist is largely because of: 1) It's frequently damned hard to benchmark systems using the real application, complete with users, data, and all that. 2) Some computers are used in a very general purpose way. I do all manner of work on my little linux boxes, for example, and that pales in comparison to the wide-ranging use of some of the larger systems where I work.
"...it's irresponsible to post benchmark results without making the benchmark available...".
Dang. You've got me there. I wonder if we can get the Imark benchmarks from Intel, btw, to see how they've been biasing them toward the newer instructions as they introduce extended instruction sets...;)
Anyway, my page contains a link to the code for the 1 megabyte version of the benchmark in question.
Same benchmark as above but using a 64megabyte array. BTW, there is also a fair amount of string copying in the benchmark. With the 64MB benchmark, memory access speed starts to dominate the results more. The Celeron 366 won this round, which doesn't say a lot for the memory architecture of the EV6 (192.233.54.85). The Celeron, for those not familiar with it, has a full core speed 128K level 2 cache on the chip and accesses memory via a 66-MHz memory bus. (some people overclock the chip and use faster memory -- but not me;)
Celeron366: 17.3 seconds CPU EV6@500MHz: 19.3 seconds CPU PentiumPro: 27.0 seconds CPU (200MHz, 256Kcache)
I ran a homemade benchmark on the RH EV6 and on my Pentium Pro 200 and my Celeron 366.
The benchmark allocates a 1 MB array (256K x 32-bit unsigned int (no, I didn't use "long" on the alpha;) ) and then uses a random number generator to select and increment the cells in a pattern that biases it heavily toward the front, creating a curve of memory activity much like many applications. Note that there is no floating point code in the benchmark. The results are below along with my celeron 366 and pentium pro 200 (256K cache):;)
ev6@500MHz: 7.23seconds cpu celeron366: 8.49seconds cpu pentiumpro: 13.67seconds cpu
The ev6 beat my humble 366c but I'm afraid my friend's 500 MHz Celeron beats it soundly. I don't remember the numbers he told me yesterday, but I remember it beating my 366 by 25% or so.
You get taxed indirectly when you buy something made in the U.S.A. because the corporate tax load is passed on (obviously) to the end consumer.
Inflating the money supply (inflation) is essentially a tax on the money supply. Your share of the "inflation tax" is proportional to the share of the money supply you hold.
Your pay a tax on real property you own. If you rent, of course, the tax is passed on to you, as in the example above. Does this mean we don't actually own real property but instead rent it from the government?
I'm sure we missed a few...
I remember a tax scheme in New York back in the BBS days where you would be taxed based on the "value" of shareware you downloaded(!) It was a sales tax on bits, essentially, with the value being purely arbitrary. Scary. It got flamed so badly it went away in record time.
Don't forget the third option: buy the generic CD from cheapbytes or equivalent. Fast, painless, and *cheap*. At about $7US, including shipping, it's probably cheaper than the download for most folks.
www.linuxone.com doesn't actually point to transmeta -- however, the comments in the source of the page credit Transmeta for the "this page is not here yet" page design!;)
Re:Linus ... architect of the net... ?
on
Weaving The Web
·
· Score: 1
Maybe not an architect of the net or the web, but he's the architect of the platform that supports my little piece of the web -- and a huge number of others. I wonder just how many Internet nodes are running his little operating system... In any event, Linus has definitely made the net bigger and better.
Since, to be patentable, this "process" must be nonobvious to one "skilled in the art", we don't need prior art to render it invalid.
If, however, this is deemed "nonobvious" then I'm gonna get patents on barcoding ( or otherwise marking with machine-readable identification;) livestock, real estate, public lamp posts, and little blue pills. I would list more but I don't want someone to beat me to it. Meanwhile, if anyone tries to patent one of the "processes" listed above, I can still cite this comment as prior art!;)
"You are 10 times more likely to die on the road on a ten minute trip to the airport than a 12 hour flight to Europe"
Hmmm... somehow I doubt this. Perhaps you meant to say "You are 10 times more likely to die on the road than on a commercial flight, assuming you drive and fly an average amount". Or, possibly, "You are 10 times more likely to die in 1000 miles of driving than in 1000 miles of flying on commercial flights".
Of course, if you really meant that a single 10-minute drive is 10 times as likely to kill you than a single flight across the Atlantic, I'd love to see some supporting data.;)
"Reagan was clueless and did nothing significant to end the "cold war"..."
I beg to differ. The military run-up was planned in advance as a way to break the cold war stalemate since it was not expected that the Soviet economy could sustain a build-up on anywhere near the scale that the US could. And, yes, Reagan did play a significant role in that.
Re:Intelligent (karma as an average of moderation)
on
Moderation Ideas
·
· Score: 1
Of course, the problem with deriving karma almost directly from an average of moderation is that it would discourage those with high karma from posting. For example, if I had high karma I might want to protect it under the system you suggest by only posting when I have a very solid comment to make. However, other comments that might rate a +2 I might not want to post since it would bring down my karma.
I actually like Rob's karma formula as it is now -- I'm just curious about how it is calculated.;)
I also live in Tallahassee and was in Hurricane Kate in November, 85. Kate was a minimal hurricane when it hit our beautifully forested city of Tallahassee. heheh. The 70-ish MPH winds of Kate knocked down so many trees that we, living well inside the city limits, had no electricity for days. Some people in the outlying neighborhoods and further had no power for 2 or 3 weeks!
But the thing I remember most vividly was the howling wind that went on for hours. A large tree, about 20 inches in diameter, fell behind our apartment -- and we didn't even hear it fall. We discovered it the next day lying about 10 feet from the house.
A minimal hurricane, even hitting where there is no possibility of flooding, is a pretty scary thing. Stuff flying through the air, trees and limbs falling, and the ever present possibility of a "tornado" or superswirl focuses your attention every time the howl reaches a higher pitch.
I'm alternately fascinated and terrified by hurricanes. We've had many much larger ones threaten our area since '85 but we haven't been hit again. But we choose to live down here and we take our chances. My friends and relatives living right on the Gulf made their own choices too.
As for those who think there is something wrong with a webcam sitting in the path of the hurricane, remember: you don't have to look. But, please, don't judge those of us who appreciate the view.;)
"...they fail to get the point across that you'd be forced to run AIX.
*shiver*"
Seconded. I worked with AIX for years until we replaced all of our IBM boxes with Sun's. I won't be going back either without a major compensation boost!;)
The Linux test score of 241 users appears to have made the top twenty on ideasinternational's 2-tier SAP benchmark list. (That appears to be the Linux/Siemens machine in 12th place.)
Though still a long way from the leader, a Sun 10000 rated at 1410 users, it is closing in on machines like the 12-CPU IBM AS/400 at 330 users.;)
Re:But is it really faster than a Starfire?
on
IBM takes aim at Sun
·
· Score: 2
Um, not exactly. The model second on the list is actually five different computers in a cluster running (at least) five different instances of Oracle. The Sun at the top of the list is a single computer running a single instance of Oracle. This list is the top non-cluster configurations and no RS/6000's are in the top ten.
I'll believe it when it shows up on the top of this list, for example. It's cheap to claim you have the fastest computer -- much more impressive to prove it in an open forum.
Does anyone know what they are basing their claims on?
Actually, in your scenario, the most active pages will be in RAM twice -- once in the ramdisk and once in the disk cache. Since the bulk of the serving will be out of the cache, the ramdisk will sit there mostly idle -- and a waste of 50 MB of RAM.
Unfortunately, with writable dynamic content, the ramdisk will have to be written to disk periodically, adding complexity, overhead, and, quite possibly, more disk IO than using a disk directly!
My server is a Celeron with 320MB RAM running Linux 2.0.36. I configured it with a 128 MB ramdisk and did a great deal of testing. Performance was significantly better, especially during peak loads, than running straight from the disk. Of course, I had a considerably more complicated set of scripts and still stood to lose some transactions if something bad happened.
As my next excercise, I tried to duplicate that performance without the ramdisk. By tuning the values in/proc/sys/vm/bdflush I was able to actually get the performance higher than when using the ramdisk. The reason for the improvement was that I now had more memory for caching the disk by not double-caching in the ramdisk and the cache as I had been doing.
The trick that worked for me was to increase the percentage of dirty buffers before forcing a flush to 80% and to increase the timeout for dirty buffers before flushing them to disk to 10 minutes. That does include some of the disadvantages of the ramdisk but my UPS is good for over 10 minutes so I don't worry much (the Internet connectiond drops when power is lost so my machine, while still up, goes idle). My startup/shutdown/backup scripts are much simpler as a result though.
I'm not worried that the Mozilla team will create another unmanageable mess of crashware. Large projects can be built solidly, if done right. I suspect that much of the reason Mozilla has taken so long was the need to rewrite so much existing code. There will be a payoff from all that work.
Bring on the IRC and messaging clients! Woohoo!!
I intend to convert to Mozilla as soon as the team tells me to.;)
This article is pretty low quality stuff, even as FUD goes. It's riddled with factual errors (no SMP support for Linux, confusing Java applet/application differences with general Java limitations).
It's backdrop of opinion is not generally accepted nor is it strongly argued. (Java has failed, Network Computers have failed, Linux hype has peaked). In this industry, success ranges from days or weeks (games), months or years (new hardware architectures, applications, websites), and years to decades (programming languages, operating systems, CS theories). Java and Linux appear to be still climbing the success curve, even a bit faster than most successful languages and OS's.
And, last, perhaps the biggest journalistic offense of all: there is no article here, just a sampling of opinions and quotes from industry sources. By soliciting quotes from 10 or 15 well-selected sources I can produce an article to make any point I want just by throwing out the ones that I can't weave into the point I'm trying to make.
Yesterday's paper USA Today cover story was about the new plan for grading SAT's on a curve. A student would get extra points for "striving" if that student came from a school where scores were lower or somesuch. (I couldn't find the article on their web site via their search. Anyone have a link?)
Slashdot Mirror
"Byte-manipulation isn't one of Alpha's strong points..."
;)
;)
Sorry... I should have said "memory copying" since I use memcpy(). As for Byte-manipulation not being a strong point of Alpha, well, isn't the point of a general-purpose CPU/memory subsystem benchmark to uncover these kind of things?
"If you want a real idea of memory bandwidth... use...Stream..."
Nope, I wasn't trying to assess memory bandwidth. By "memory access speed" I was referring to memory latency, a critical variable to system performance.
"...benchmark using the applications and data that are important to you."
heheh. Yeah, sure. The reason synthetic benchmarks and "application conglomeration" benchmarks exist is largely because of: 1) It's frequently damned hard to benchmark systems using the real application, complete with users, data, and all that. 2) Some computers are used in a very general purpose way. I do all manner of work on my little linux boxes, for example, and that pales in comparison to the wide-ranging use of some of the larger systems where I work.
"...it's irresponsible to post benchmark results without making the benchmark available...".
Dang. You've got me there. I wonder if we can get the Imark benchmarks from Intel, btw, to see how they've been biasing them toward the newer instructions as they introduce extended instruction sets...
Anyway, my page contains a link to the code for the 1 megabyte version of the benchmark in question.
Same benchmark as above but using a 64megabyte array. BTW, there is also a fair amount of string copying in the benchmark. With the 64MB benchmark, memory access speed starts to dominate the results more. The Celeron 366 won this round, which doesn't say a lot for the memory architecture of the EV6 (192.233.54.85). The Celeron, for those not familiar with it, has a full core speed 128K level 2 cache on the chip and accesses memory via a 66-MHz memory bus. (some people overclock the chip and use faster memory -- but not me ;)
Celeron366: 17.3 seconds CPU
EV6@500MHz: 19.3 seconds CPU
PentiumPro: 27.0 seconds CPU (200MHz, 256Kcache)
I ran a homemade benchmark on the RH EV6 and on my Pentium Pro 200 and my Celeron 366.
;) ) and then uses a random number generator to select and increment the cells in a pattern that biases it heavily toward the front, creating a curve of memory activity much like many applications. Note that there is no floating point code in the benchmark. The results are below along with my celeron 366 and pentium pro 200 (256K cache): ;)
The benchmark allocates a 1 MB array (256K x 32-bit unsigned int (no, I didn't use "long" on the alpha
ev6@500MHz: 7.23seconds cpu
celeron366: 8.49seconds cpu
pentiumpro: 13.67seconds cpu
The ev6 beat my humble 366c but I'm afraid my friend's 500 MHz Celeron beats it soundly. I don't remember the numbers he told me yesterday, but I remember it beating my 366 by 25% or so.
You missed a few of the more sneaky taxes:
...
You get taxed indirectly when you buy something made in the U.S.A. because the corporate tax load is passed on (obviously) to the end consumer.
Inflating the money supply (inflation) is essentially a tax on the money supply. Your share of the "inflation tax" is proportional to the share of the money supply you hold.
Your pay a tax on real property you own. If you rent, of course, the tax is passed on to you, as in the example above. Does this mean we don't actually own real property but instead rent it from the government?
I'm sure we missed a few
I remember a tax scheme in New York back in the BBS days where you would be taxed based on the "value" of shareware you downloaded(!) It was a sales tax on bits, essentially, with the value being purely arbitrary. Scary. It got flamed so badly it went away in record time.
Don't forget the third option: buy the generic CD from cheapbytes or equivalent. Fast, painless, and *cheap*. At about $7US, including shipping, it's probably cheaper than the download for most folks.
"We believe when people hear that phrase they think of AOL." -Jim Whitney, AOL
;)
Close, Jim. Now when people hear that phrase they will think "AOL sucks". Is that what you want?
...That come from Microsoft? Would those be illegal also? ;)
www.linuxone.com doesn't actually point to transmeta -- however, the comments in the source of the page credit Transmeta for the "this page is not here yet" page design! ;)
Maybe not an architect of the net or the web, but he's the architect of the platform that supports my little piece of the web -- and a huge number of others. I wonder just how many Internet nodes are running his little operating system... In any event, Linus has definitely made the net bigger and better.
Since, to be patentable, this "process" must be nonobvious to one "skilled in the art", we don't need prior art to render it invalid.
;) livestock, real estate, public lamp posts, and little blue pills. I would list more but I don't want someone to beat me to it. Meanwhile, if anyone tries to patent one of the "processes" listed above, I can still cite this comment as prior art! ;)
If, however, this is deemed "nonobvious" then I'm gonna get patents on barcoding ( or otherwise marking with machine-readable identification
"You are 10 times more likely to die on the road on a ten minute trip to the airport than a 12 hour flight to Europe"
;)
Hmmm... somehow I doubt this. Perhaps you meant to say "You are 10 times more likely to die on the road than on a commercial flight, assuming you drive and fly an average amount". Or, possibly, "You are 10 times more likely to die in 1000 miles of driving than in 1000 miles of flying on commercial flights".
Of course, if you really meant that a single 10-minute drive is 10 times as likely to kill you than a single flight across the Atlantic, I'd love to see some supporting data.
"Reagan was clueless and did nothing significant to end the "cold war"..."
I beg to differ. The military run-up was planned in advance as a way to break the cold war stalemate since it was not expected that the Soviet economy could sustain a build-up on anywhere near the scale that the US could. And, yes, Reagan did play a significant role in that.
Of course, the problem with deriving karma almost directly from an average of moderation is that it would discourage those with high karma from posting. For example, if I had high karma I might want to protect it under the system you suggest by only posting when I have a very solid comment to make. However, other comments that might rate a +2 I might not want to post since it would bring down my karma.
;)
I actually like Rob's karma formula as it is now -- I'm just curious about how it is calculated.
I also live in Tallahassee and was in Hurricane Kate in November, 85. Kate was a minimal hurricane when it hit our beautifully forested city of Tallahassee. heheh. The 70-ish MPH winds of Kate knocked down so many trees that we, living well inside the city limits, had no electricity for days. Some people in the outlying neighborhoods and further had no power for 2 or 3 weeks!
;)
But the thing I remember most vividly was the howling wind that went on for hours. A large tree, about 20 inches in diameter, fell behind our apartment -- and we didn't even hear it fall. We discovered it the next day lying about 10 feet from the house.
A minimal hurricane, even hitting where there is no possibility of flooding, is a pretty scary thing. Stuff flying through the air, trees and limbs falling, and the ever present possibility of a "tornado" or superswirl focuses your attention every time the howl reaches a higher pitch.
I'm alternately fascinated and terrified by hurricanes. We've had many much larger ones threaten our area since '85 but we haven't been hit again. But we choose to live down here and we take our chances. My friends and relatives living right on the Gulf made their own choices too.
As for those who think there is something wrong with a webcam sitting in the path of the hurricane, remember: you don't have to look. But, please, don't judge those of us who appreciate the view.
"Social graces are the packet headers of everyday life"
;)
Well said! And, in terms that even us geeks can understand. I propose we use that slogan to adorn the entrance of Geek Charm School.
"...they fail to get the point across that you'd be forced to run AIX.
;)
*shiver*"
Seconded. I worked with AIX for years until we replaced all of our IBM boxes with Sun's. I won't be going back either without a major compensation boost!
The Linux test score of 241 users appears to have made the top twenty on ideasinternational's 2-tier SAP benchmark list. (That appears to be the Linux/Siemens machine in 12th place.)
;)
Though still a long way from the leader, a Sun 10000 rated at 1410 users, it is closing in on machines like the 12-CPU IBM AS/400 at 330 users.
Um, not exactly. The model second on the list is actually five different computers in a cluster running (at least) five different instances of Oracle. The Sun at the top of the list is a single computer running a single instance of Oracle. This list is the top non-cluster configurations and no RS/6000's are in the top ten.
I'll believe it when it shows up on the top of this list, for example. It's cheap to claim you have the fastest computer -- much more impressive to prove it in an open forum.
Does anyone know what they are basing their claims on?
Actually, in your scenario, the most active pages will be in RAM twice -- once in the ramdisk and once in the disk cache. Since the bulk of the serving will be out of the cache, the ramdisk will sit there mostly idle -- and a waste of 50 MB of RAM.
/proc/sys/vm/bdflush I was able to actually get the performance higher than when using the ramdisk. The reason for the improvement was that I now had more memory for caching the disk by not double-caching in the ramdisk and the cache as I had been doing.
Unfortunately, with writable dynamic content, the ramdisk will have to be written to disk periodically, adding complexity, overhead, and, quite possibly, more disk IO than using a disk directly!
My server is a Celeron with 320MB RAM running Linux 2.0.36. I configured it with a 128 MB ramdisk and did a great deal of testing. Performance was significantly better, especially during peak loads, than running straight from the disk. Of course, I had a considerably more complicated set of scripts and still stood to lose some transactions if something bad happened.
As my next excercise, I tried to duplicate that performance without the ramdisk. By tuning the values in
The trick that worked for me was to increase the percentage of dirty buffers before forcing a flush to 80% and to increase the timeout for dirty buffers before flushing them to disk to 10 minutes. That does include some of the disadvantages of the ramdisk but my UPS is good for over 10 minutes so I don't worry much (the Internet connectiond drops when power is lost so my machine, while still up, goes idle). My startup/shutdown/backup scripts are much simpler as a result though.
Well, I guess that renders moot the comments about the mac not being slashdotted, eh? ;)
I'm not worried that the Mozilla team will create another unmanageable mess of crashware. Large projects can be built solidly, if done right. I suspect that much of the reason Mozilla has taken so long was the need to rewrite so much existing code. There will be a payoff from all that work.
;)
Bring on the IRC and messaging clients! Woohoo!!
I intend to convert to Mozilla as soon as the team tells me to.
This article is pretty low quality stuff, even as FUD goes. It's riddled with factual errors (no SMP support for Linux, confusing Java applet/application differences with general Java limitations).
It's backdrop of opinion is not generally accepted nor is it strongly argued. (Java has failed, Network Computers have failed, Linux hype has peaked). In this industry, success ranges from days or weeks (games), months or years (new hardware architectures, applications, websites), and years to decades (programming languages, operating systems, CS theories). Java and Linux appear to be still climbing the success curve, even a bit faster than most successful languages and OS's.
And, last, perhaps the biggest journalistic offense of all: there is no article here, just a sampling of opinions and quotes from industry sources. By soliciting quotes from 10 or 15 well-selected sources I can produce an article to make any point I want just by throwing out the ones that I can't weave into the point I'm trying to make.
Sheesh. What a waste of bandwidth.
...Well, maybe not perfect, but it's pretty libertarian of 'em. ;)
Yesterday's paper USA Today cover story was about the new plan for grading SAT's on a curve. A student would get extra points for "striving" if that student came from a school where scores were lower or somesuch. (I couldn't find the article on their web site via their search. Anyone have a link?)