Compilers do not optimize for dual CPU configurations at all. Creating a second thread of execution is a programmers choice. A good compiler will then optimize the code for each thread independently to the best of it's ability. The guys at NASA were trying to do a quick and dirty CPU vs CPU comparison and bunged it up pretty good. These numbers are all pretty meaningless overall, a nice thought exercise, but for publishing, really poor test design and control--and that's being nice...
No, you misunderstood the paper. If your premise that vector ops are not significant, then why code them in the first place? The answer is they are not insignificant--on a PPC, but they are on a Pentium. Vectorized SSE2 code has a poor performance in this app because of the compiler, but the PPC version does not suffer this problem, a better paper would have shown those results as well--as the system is a sum of it's parts, not merely a single facet of one of those parts. Something the benchmark foul criers have been trying to say for awhile now, only it is convenient for them to fall silent when the results are to their liking despite the same issues being presented in reverse.
All pure speculation by a guy attempting to further his own ends--which appear to be outside the CS field. I just think he hadn't published anything lately that garnished enough PR. This kind of tripe gets his name out all over the place in a neutral field from his own. He can make claims and follow methodologies his peers would slam him for within the field, and he gains the appearance of NASA credibility just from the address under his name at the top of the paper. But it's just another un-refereed web page and hence relatively meaningless unless you agree with his write-up.
AAARRRGGGHHH! This paragraph is why bollocks like this is so BAD! There are no NASA Officials involved here, let alone any shred of official NASA organizational interest. This is a personal test by one guy who has access to the G5 because of who he works for, and an earlier Apple Press Opportunity the marketing folks thought they could take further advantage of.
The paper of the NASA guy is quite clear. There is two version of Jet3D, a vector one and scalar one. Both applications are doing the same job. Perhaps the scalar one is more precise, than the vector one, but the guy did not precise it.
Both the vector and scalar version of Jet 3D is able to take advantage of dual powermac. The guy give the result of a dual G5 in scalar : 498 Mflop and vector 5177 Mflop.
The guy from the nasa just emphasized with the single G5 thing, just because he wanted to compare the CPU, and not a dual versus a single.
If you look at the paper, you'll see that the dual smoke any P4 in scalar, and it annhilate any P4 while using the vector code.
If you look at the paper, you'll see that the dual smoke any P4 in scalar, and it annhilate any P4 while using the vector code.
I hate statements like "If you look at <blah blah>" because you are making the assumption that the another person came to a different conclusion than that of yours simply because he or she did not read the paper.
In addition, it is a fallacy to draw the conclusion you reached that a dual G5 will beat ANY P4 when the article only tested two types of P4 (that shipped about a year ago). In addition, the author wrote that "Clearly, the G5 would lag faster (up to 3.2GHz) P4 systems in Jet3D scalar floating point performance."
The author also (rightly) avoided hyperboles such as "smoke" and "annhilate" [sic].
OK, AirSluf, klinux and others, I've been pondering and conversing with some people -- discussing the G5 and benchmarks in general. Hopefully we can clear up a few things here.
First, Apple developed the G5 in direct response to what the high-end workstation customers wanted. It's that simple. The userbase needed 64bit and Apple gave it to them. There is a relatively large userbase and the software support is already in place and progressing nicely. We can't say that for the Wintelon crowd -- we really can't. Even though AMD-64 is around the corner (again), there are still a LOT of questions; questions that have been raised by myself and others right here on AI. All you have to do is search the message forum. Anyway, in a nutshell, AMD-64 looks rather peachy, but where is the full, trouble-free 64-bit Windows support? Is that going to be for the desktop-user crowd or the server lot? It's not clear at all. Similarly, where are the 64bit apps? Who's producing them? Who is planning on supporting them? How long will it take them to get their wares to market? and more importantly (as I've also noted in earlier threads) Where is the userbase for these things? There are many other questions that we've already been over time and again. The truth is that Apple is delivering. App developers are delivering, the OS is here *now* and working fine, and there is a nice userbase all ready to go. Just think about that.
Anyway, regarding the G5 evaluation and the vector performance... Keep in mind that this evaluation was done on completely un-optimized code for the G5. When the G5/Panther-specific optimization comes along, expect even better performance. Yes, it's that simple.
For instance, one difference is in the pipelines and it *might* show in the test since the code used was designed for G4. The pipelines grew longer with G5 so given that, the G5 will likely burn a lot of cycles (i.e., no work getting done) -- loops need to be unrolled to the length or the G5's pipeline to achieve the same per-clock performance as the G4. It's a wonderful thing that a processor like the G5 can scale with frequency; especially given the new PPC architecture of the G5. The G5 scales nicely, as does the AltiVec unit with respect to performance. If you (choose) to recall, this degree of scaling DID NOT occur with the transition from Pentium3 to Pentium4. Lower clocked P3's were actuallt faster than P4's. And the simple truth is that the Pentium's vector performance can't hold a candle to AltiVec. The test didn't measure the G4 vs. G5 bus. It was more or less CPU specific. The bus in the G5 is simply MONSTROUS. Any results will likely show unbelievable wins for the G5.
When the G5 is released it will lay to rest any skepticism regarding real-world performance. Synthetic benchmarks won't mean as much as they used to. I think a lot of Wintelon folk will be surprised. They're already howling at the top of their lungs. Will they be in denial? Who really cares?
It's likely that there will be many obscure, sneaky, underhanded tests being cooked by PC/Wintelon-centric sites and individuals attempting to counter any performance advantage that the G5 will show. I guess we'll have to live with it. there is no trickery, no magic spells, no illusions. The performance is already evident with un-optimized code. It's up to the developers to make it *really* shine. Furthermore, there is a pattern that many people reading these threads should zero-in on. Why does everything always seem to come down to comparisons of scalar performance? Many people choose to ignore the glaring answer. The fact is that the performance crown for SIMD/vector was handed to the G4 long ago. there is simply nothing in the Wintelon world that can compete with AltiVec -- not Intel, not AMD; no one. Yes, it's that good and the G5 just extends that lead even further. Regarding scalar, people should keep in mind that the PPC 970 was derived directly from the Power4. People with a clue already know that the Power4 is world-renowned for it's scalar performance. What makes people think it's going to be any less effective with 970/G5? What specific part about the 970 isn't spectacular compared to the G4 or even the Intel line for that matter? Can anyone think of something? Doubt it.
Ed, I think you have already made up your mind so why are you debating this?
My position has always been: unless one has their hands on a production G5, one cannot draw any real conclusions until three months away from now. Therefore, you do not find me writing a protracted post trying to convince you that P4 is better than G5 or vice versa. G5 may be better than the discovery of fire but that is not the point. The point is that a fallacy presented in an argument can be disproved right now.
Case in point, saying that the paper showed dual G5 will beat any P4 is incorrect because the paper does not say that at all.
Another example: you second paragraph is filled with fallacy of burden of proof (or just plain lack of evidence!). (See http://www.nizkor.org/features/falla...-of-proof.html) You wrote that "the userbase needed 64bit and Apple gave it to them. There is a relatively large userbase and the software support is already in place and progressing nicely. <snip> The truth is that Apple is delivering. App developers are delivering, the OS is here *now* and working fine, and there is a nice userbase all ready to go."
What is Apple delivering? I do not see a shipping 64 bit CPU. I do not see a shipping 64 bit operating system. It is not here *now* and none of us know that it is working fine. In fact, by the fact that it is not shipping and that it is in beta/developer release, the correct logical conclusion would be that Panther is not working fine yet and bugs are still being resolved.
Also, where is that Apple user base needing 64 bit? How many people/corporation/university is that? What software support is in place? Can you be more specific? In fact, the same questions you posed for the x86 world applies to Apple as well but you chose not to address them.
In addition, you are making a 'poisoning of the well' argument in your last paragraph when the debate shifts to real-world benchmarking performed by the general computing public when G5 ships. (See http://seercom.com/bluto/skepticism/....poiswell.html) By labelling the opposite as 'sneaky' and 'underhanded' through ad hominem attacks, you are arguing that any claim that group makes is false before they make that argument or before ascertaining whether their claim is true or not.
You might say, oh yeah, prove that <insert chip/system> here is better than G5 then? However, as I have said before, that is not what I am here to do. It has been a while since my logic/critical thinking class but if are not going to change your mind regardless, than why should I bother?
[[[Ed, I think you have already made up your mind so why are you debating this?]]]
Um... is that a speculative fallacy you just committed? Yep. Add to that the
"Fallacy of interrogative presupposition" ie., loaded questions or statements.
[[[My position has always been: unless one has their hands on a production G5, one cannot draw any real conclusions until three months away from now.]]]
This is completely false. Accurate predictions can be made given that results can be extrapolated from prior knowledge about something. Now that the chip has been released and Apple and IBM have released much technical material about the CPU and the system, a LOT can be accurately predicted. You are ignoring the facts and the conclusions that can be accurately drawn from them -- even if an actual, physical system didn't exist (which many obviously do). This is an example of
"Slothful induction": the conclusion of a strong inductive argument is denied despite the evidence to the contrary. Similarly, it is "A priorism" in that you are attempting to deduce facts from abstractions and principles (your logic) rather than inducing from facts (we know quite a bit about the G5/970). In other words, we know how the Power4 performs and we know how AltiVec performs. We know how the G4 performs. The tech papers are out for the 970 as well as the G5 system...etcetera. Therefore, we can make accurate predictions. For example (and to quote Mr. Spock ;-) "If I drop a wrench on a planet with a positive gravity field, I need not see it fall, nor hear it hit the ground, to know that it has, in fact, fallen. "- Spock
[[[Therefore, you do not find me writing a protracted post trying to convince you that P4 is better than G5 or vice versa.]]]
No, but it seems that you are attempting to tarnish an otherwise outstanding system. To what end? Why split hairs?
[[[G5 may be better than the discovery of fire but that is not the point. The point is that a fallacy presented in an argument can be disproved right now.]]]
Denial.
[[[Case in point, saying that the paper showed dual G5 will beat any P4 is incorrect because the paper does not say that at all.]]]
And was I arguing that specifically? Is that a lie? Assumption on your part? It did show that a single G5 can fair quite well with unoptimized code against a P4.
[[[What is Apple delivering? ... I do not see a shipping 64 bit CPU. I do not see a shipping 64 bit operating system. It is not here *now* and none of us know that it is working fine.]]]
The exact thing that AMD was supposed to have out MONTHS ago? Look, the G5 isn't vapor. Not the system, not the processor, not the OS. It's here people have used it, people have tested it. Apple is already taking orders. It's up for sale on their website. So, to quote Mr. Spock once again... "since we are here, your statement is not only illogical but also unworthy of refutation."
[[[Also, where is that Apple user base needing 64 bit? How many people/corporation/university is that? What software support is in place? Can you be more specific? ]]]
This is a clear example of "Multiple questions or assertions" (plurium interrogationum). You are asking a complex question or a series of questions, or stating a complex assertion or multiple assertions, while only allowing for a single/simple response, and then assuming the oppositions inability to adequately respond is indication that their position is wrong and therefore false.
[[[In addition, you are making a 'poisoning of the well' argument in your last paragraph when the debate shifts to real-world benchmarking performed by the general computing public when G5 ships. (See http://seercom.com/bluto/skepticism...f.poiswell.html) By labelling the opposite as 'sneaky' and 'underhanded' through ad hominem attacks, you are arguing that any claim that group makes is false before they make that argument or before ascertaining whether their claim is true or not.]]]
I think you're reaching, but if you say so... ;-) However we could take SPEC as an example here. It's a benchmark that Windows-using people LOVE to tout when it's in their favor, even thought SPEC means noting to anyone unless all you do is sit around and run SPEC benchmarks all day. It's synthetic and can be used to oversimplify results. An as with any synthetic benchmark it can be infinitely abused. And the results can mean taking an exception and using it to paint broad generalities. For example, if a 3GHz Pentium CPU scores exceptionally well on SPEC it does NOT imply/translate into "all P4-based *systems* must therefore also be representative of that result" That would be an accident, but I've witnessed people make this mistake.
[[[ However, as I have said before, that is not what I am here to do.]]]
Then why are you here? Good intentions?
Self-righteousness: confusing good intentions with actual good or truth.
[[[It has been a while since my logic/critical thinking class]]]
It would seem.
[[[but if are not going to change your mind regardless, than why should I bother?]]]
Loaded question. <sigh...>
You also commit the "Fallacy of opposition" since those who disagree with you must be wrong and not thinking straight or thinking logically. And then there is
"Nothing but objections": continually raising objections as a means of avoiding the issue.
Look I really don't want to split hairs. I'm sure if there was something hosed up with what I posted, Amorph or Programmer would have pointed it out and corrected me for my faulty reasoning. Why must this stuff be debated to the Nth degree of absurdity? And the splitting of hairs.... Picking nits. Jeez. So, let's keep the thread on topic and avoid all this senseless bickering.
I hate statements like "If you look at <blah blah>" because you are making the assumption that the another person came to a different conclusion than that of yours simply because he or she did not read the paper.
Actually, you're pointing people to the parts of the paper that your argument is based on, which is perfectly fine. Maybe they read it and weighted it differently than you did, or maybe their knowledge set made them qualify it in a way that yours didn't, or maybe they just missed it. Perfectly intelligent and capable people make mistakes and oversights. As long as they're pointed out the way Powerdoc did, i.e. "look here to see where my reasoning came from" rather than "you are clearly some sort of congenitally inferior muck dweller because anyone with two neurons to rub together can see that I'm right," no harm done.
Quote:
In addition, it is a fallacy to draw the conclusion you reached that a dual G5 will beat ANY P4 when the article only tested two types of P4 (that shipped about a year ago). In addition, the author wrote that "Clearly, the G5 would lag faster (up to 3.2GHz) P4 systems in Jet3D scalar floating point performance."
You can extrapolate pretty easily from those P4 results to the currently shipping ones, because the P4 hasn't changed significantly since the models he tested.
Also, the G5 lost in single-threaded scalar floating point compiled for another CPU, not with the version of Jet3D he'd actually run on the G5.
Quote:
The author also (rightly) avoided hyperboles such as "smoke" and "annhilate" [sic].
That makes him a rare bird indeed, but it has little bearing on the discussion.
Comments
Originally posted by AirSluf
Compilers do not optimize for dual CPU configurations at all. Creating a second thread of execution is a programmers choice. A good compiler will then optimize the code for each thread independently to the best of it's ability. The guys at NASA were trying to do a quick and dirty CPU vs CPU comparison and bunged it up pretty good. These numbers are all pretty meaningless overall, a nice thought exercise, but for publishing, really poor test design and control--and that's being nice...
No, you misunderstood the paper. If your premise that vector ops are not significant, then why code them in the first place? The answer is they are not insignificant--on a PPC, but they are on a Pentium. Vectorized SSE2 code has a poor performance in this app because of the compiler, but the PPC version does not suffer this problem, a better paper would have shown those results as well--as the system is a sum of it's parts, not merely a single facet of one of those parts. Something the benchmark foul criers have been trying to say for awhile now, only it is convenient for them to fall silent when the results are to their liking despite the same issues being presented in reverse.
All pure speculation by a guy attempting to further his own ends--which appear to be outside the CS field. I just think he hadn't published anything lately that garnished enough PR. This kind of tripe gets his name out all over the place in a neutral field from his own. He can make claims and follow methodologies his peers would slam him for within the field, and he gains the appearance of NASA credibility just from the address under his name at the top of the paper. But it's just another un-refereed web page and hence relatively meaningless unless you agree with his write-up.
AAARRRGGGHHH! This paragraph is why bollocks like this is so BAD! There are no NASA Officials involved here, let alone any shred of official NASA organizational interest. This is a personal test by one guy who has access to the G5 because of who he works for, and an earlier Apple Press Opportunity the marketing folks thought they could take further advantage of.
The paper of the NASA guy is quite clear. There is two version of Jet3D, a vector one and scalar one. Both applications are doing the same job. Perhaps the scalar one is more precise, than the vector one, but the guy did not precise it.
Both the vector and scalar version of Jet 3D is able to take advantage of dual powermac. The guy give the result of a dual G5 in scalar : 498 Mflop and vector 5177 Mflop.
The guy from the nasa just emphasized with the single G5 thing, just because he wanted to compare the CPU, and not a dual versus a single.
If you look at the paper, you'll see that the dual smoke any P4 in scalar, and it annhilate any P4 while using the vector code.
Originally posted by Powerdoc
If you look at the paper, you'll see that the dual smoke any P4 in scalar, and it annhilate any P4 while using the vector code.
I hate statements like "If you look at <blah blah>" because you are making the assumption that the another person came to a different conclusion than that of yours simply because he or she did not read the paper.
In addition, it is a fallacy to draw the conclusion you reached that a dual G5 will beat ANY P4 when the article only tested two types of P4 (that shipped about a year ago). In addition, the author wrote that "Clearly, the G5 would lag faster (up to 3.2GHz) P4 systems in Jet3D scalar floating point performance."
The author also (rightly) avoided hyperboles such as "smoke" and "annhilate" [sic].
First, Apple developed the G5 in direct response to what the high-end workstation customers wanted. It's that simple. The userbase needed 64bit and Apple gave it to them. There is a relatively large userbase and the software support is already in place and progressing nicely. We can't say that for the Wintelon crowd -- we really can't. Even though AMD-64 is around the corner (again), there are still a LOT of questions; questions that have been raised by myself and others right here on AI. All you have to do is search the message forum. Anyway, in a nutshell, AMD-64 looks rather peachy, but where is the full, trouble-free 64-bit Windows support? Is that going to be for the desktop-user crowd or the server lot? It's not clear at all. Similarly, where are the 64bit apps? Who's producing them? Who is planning on supporting them? How long will it take them to get their wares to market? and more importantly (as I've also noted in earlier threads) Where is the userbase for these things? There are many other questions that we've already been over time and again. The truth is that Apple is delivering. App developers are delivering, the OS is here *now* and working fine, and there is a nice userbase all ready to go. Just think about that.
Anyway, regarding the G5 evaluation and the vector performance... Keep in mind that this evaluation was done on completely un-optimized code for the G5. When the G5/Panther-specific optimization comes along, expect even better performance. Yes, it's that simple.
For instance, one difference is in the pipelines and it *might* show in the test since the code used was designed for G4. The pipelines grew longer with G5 so given that, the G5 will likely burn a lot of cycles (i.e., no work getting done) -- loops need to be unrolled to the length or the G5's pipeline to achieve the same per-clock performance as the G4. It's a wonderful thing that a processor like the G5 can scale with frequency; especially given the new PPC architecture of the G5. The G5 scales nicely, as does the AltiVec unit with respect to performance. If you (choose) to recall, this degree of scaling DID NOT occur with the transition from Pentium3 to Pentium4. Lower clocked P3's were actuallt faster than P4's. And the simple truth is that the Pentium's vector performance can't hold a candle to AltiVec. The test didn't measure the G4 vs. G5 bus. It was more or less CPU specific. The bus in the G5 is simply MONSTROUS. Any results will likely show unbelievable wins for the G5.
When the G5 is released it will lay to rest any skepticism regarding real-world performance. Synthetic benchmarks won't mean as much as they used to. I think a lot of Wintelon folk will be surprised. They're already howling at the top of their lungs. Will they be in denial? Who really cares?
It's likely that there will be many obscure, sneaky, underhanded tests being cooked by PC/Wintelon-centric sites and individuals attempting to counter any performance advantage that the G5 will show. I guess we'll have to live with it. there is no trickery, no magic spells, no illusions. The performance is already evident with un-optimized code. It's up to the developers to make it *really* shine. Furthermore, there is a pattern that many people reading these threads should zero-in on. Why does everything always seem to come down to comparisons of scalar performance? Many people choose to ignore the glaring answer. The fact is that the performance crown for SIMD/vector was handed to the G4 long ago. there is simply nothing in the Wintelon world that can compete with AltiVec -- not Intel, not AMD; no one. Yes, it's that good and the G5 just extends that lead even further. Regarding scalar, people should keep in mind that the PPC 970 was derived directly from the Power4. People with a clue already know that the Power4 is world-renowned for it's scalar performance. What makes people think it's going to be any less effective with 970/G5? What specific part about the 970 isn't spectacular compared to the G4 or even the Intel line for that matter? Can anyone think of something? Doubt it.
--
Ed
--
Ed
My position has always been: unless one has their hands on a production G5, one cannot draw any real conclusions until three months away from now. Therefore, you do not find me writing a protracted post trying to convince you that P4 is better than G5 or vice versa. G5 may be better than the discovery of fire but that is not the point. The point is that a fallacy presented in an argument can be disproved right now.
Case in point, saying that the paper showed dual G5 will beat any P4 is incorrect because the paper does not say that at all.
Another example: you second paragraph is filled with fallacy of burden of proof (or just plain lack of evidence!). (See http://www.nizkor.org/features/falla...-of-proof.html) You wrote that "the userbase needed 64bit and Apple gave it to them. There is a relatively large userbase and the software support is already in place and progressing nicely. <snip> The truth is that Apple is delivering. App developers are delivering, the OS is here *now* and working fine, and there is a nice userbase all ready to go."
What is Apple delivering? I do not see a shipping 64 bit CPU. I do not see a shipping 64 bit operating system. It is not here *now* and none of us know that it is working fine. In fact, by the fact that it is not shipping and that it is in beta/developer release, the correct logical conclusion would be that Panther is not working fine yet and bugs are still being resolved.
Also, where is that Apple user base needing 64 bit? How many people/corporation/university is that? What software support is in place? Can you be more specific? In fact, the same questions you posed for the x86 world applies to Apple as well but you chose not to address them.
In addition, you are making a 'poisoning of the well' argument in your last paragraph when the debate shifts to real-world benchmarking performed by the general computing public when G5 ships. (See http://seercom.com/bluto/skepticism/....poiswell.html) By labelling the opposite as 'sneaky' and 'underhanded' through ad hominem attacks, you are arguing that any claim that group makes is false before they make that argument or before ascertaining whether their claim is true or not.
You might say, oh yeah, prove that <insert chip/system> here is better than G5 then? However, as I have said before, that is not what I am here to do. It has been a while since my logic/critical thinking class but if are not going to change your mind regardless, than why should I bother?
Um... is that a speculative fallacy you just committed? Yep. Add to that the
"Fallacy of interrogative presupposition" ie., loaded questions or statements.
[[[My position has always been: unless one has their hands on a production G5, one cannot draw any real conclusions until three months away from now.]]]
This is completely false. Accurate predictions can be made given that results can be extrapolated from prior knowledge about something. Now that the chip has been released and Apple and IBM have released much technical material about the CPU and the system, a LOT can be accurately predicted. You are ignoring the facts and the conclusions that can be accurately drawn from them -- even if an actual, physical system didn't exist (which many obviously do). This is an example of
"Slothful induction": the conclusion of a strong inductive argument is denied despite the evidence to the contrary. Similarly, it is "A priorism" in that you are attempting to deduce facts from abstractions and principles (your logic) rather than inducing from facts (we know quite a bit about the G5/970). In other words, we know how the Power4 performs and we know how AltiVec performs. We know how the G4 performs. The tech papers are out for the 970 as well as the G5 system...etcetera. Therefore, we can make accurate predictions. For example (and to quote Mr. Spock ;-) "If I drop a wrench on a planet with a positive gravity field, I need not see it fall, nor hear it hit the ground, to know that it has, in fact, fallen. "- Spock
[[[Therefore, you do not find me writing a protracted post trying to convince you that P4 is better than G5 or vice versa.]]]
No, but it seems that you are attempting to tarnish an otherwise outstanding system. To what end? Why split hairs?
[[[G5 may be better than the discovery of fire but that is not the point. The point is that a fallacy presented in an argument can be disproved right now.]]]
Denial.
[[[Case in point, saying that the paper showed dual G5 will beat any P4 is incorrect because the paper does not say that at all.]]]
And was I arguing that specifically? Is that a lie? Assumption on your part? It did show that a single G5 can fair quite well with unoptimized code against a P4.
[[[What is Apple delivering? ... I do not see a shipping 64 bit CPU. I do not see a shipping 64 bit operating system. It is not here *now* and none of us know that it is working fine.]]]
The exact thing that AMD was supposed to have out MONTHS ago? Look, the G5 isn't vapor. Not the system, not the processor, not the OS. It's here people have used it, people have tested it. Apple is already taking orders. It's up for sale on their website. So, to quote Mr. Spock once again... "since we are here, your statement is not only illogical but also unworthy of refutation."
[[[Also, where is that Apple user base needing 64 bit? How many people/corporation/university is that? What software support is in place? Can you be more specific? ]]]
This is a clear example of "Multiple questions or assertions" (plurium interrogationum). You are asking a complex question or a series of questions, or stating a complex assertion or multiple assertions, while only allowing for a single/simple response, and then assuming the oppositions inability to adequately respond is indication that their position is wrong and therefore false.
[[[In addition, you are making a 'poisoning of the well' argument in your last paragraph when the debate shifts to real-world benchmarking performed by the general computing public when G5 ships. (See http://seercom.com/bluto/skepticism...f.poiswell.html) By labelling the opposite as 'sneaky' and 'underhanded' through ad hominem attacks, you are arguing that any claim that group makes is false before they make that argument or before ascertaining whether their claim is true or not.]]]
I think you're reaching, but if you say so... ;-) However we could take SPEC as an example here. It's a benchmark that Windows-using people LOVE to tout when it's in their favor, even thought SPEC means noting to anyone unless all you do is sit around and run SPEC benchmarks all day. It's synthetic and can be used to oversimplify results. An as with any synthetic benchmark it can be infinitely abused. And the results can mean taking an exception and using it to paint broad generalities. For example, if a 3GHz Pentium CPU scores exceptionally well on SPEC it does NOT imply/translate into "all P4-based *systems* must therefore also be representative of that result" That would be an accident, but I've witnessed people make this mistake.
[[[ However, as I have said before, that is not what I am here to do.]]]
Then why are you here? Good intentions?
Self-righteousness: confusing good intentions with actual good or truth.
[[[It has been a while since my logic/critical thinking class]]]
It would seem.
[[[but if are not going to change your mind regardless, than why should I bother?]]]
Loaded question. <sigh...>
You also commit the "Fallacy of opposition" since those who disagree with you must be wrong and not thinking straight or thinking logically. And then there is
"Nothing but objections": continually raising objections as a means of avoiding the issue.
Look I really don't want to split hairs. I'm sure if there was something hosed up with what I posted, Amorph or Programmer would have pointed it out and corrected me for my faulty reasoning. Why must this stuff be debated to the Nth degree of absurdity? And the splitting of hairs.... Picking nits. Jeez. So, let's keep the thread on topic and avoid all this senseless bickering.
--
Ed
Originally posted by klinux
I hate statements like "If you look at <blah blah>" because you are making the assumption that the another person came to a different conclusion than that of yours simply because he or she did not read the paper.
Actually, you're pointing people to the parts of the paper that your argument is based on, which is perfectly fine. Maybe they read it and weighted it differently than you did, or maybe their knowledge set made them qualify it in a way that yours didn't, or maybe they just missed it. Perfectly intelligent and capable people make mistakes and oversights. As long as they're pointed out the way Powerdoc did, i.e. "look here to see where my reasoning came from" rather than "you are clearly some sort of congenitally inferior muck dweller because anyone with two neurons to rub together can see that I'm right," no harm done.
In addition, it is a fallacy to draw the conclusion you reached that a dual G5 will beat ANY P4 when the article only tested two types of P4 (that shipped about a year ago). In addition, the author wrote that "Clearly, the G5 would lag faster (up to 3.2GHz) P4 systems in Jet3D scalar floating point performance."
You can extrapolate pretty easily from those P4 results to the currently shipping ones, because the P4 hasn't changed significantly since the models he tested.
Also, the G5 lost in single-threaded scalar floating point compiled for another CPU, not with the version of Jet3D he'd actually run on the G5.
The author also (rightly) avoided hyperboles such as "smoke" and "annhilate" [sic].
That makes him a rare bird indeed, but it has little bearing on the discussion.