Page 3 of 4 FirstFirst 1234 LastLast
Results 33 to 48 of 57

Thread: Do you get an 'XP rating' applied when you o/c?

  1. #33
    Resident abit mourner BUFF's Avatar
    Join Date
    Jul 2003
    Location
    Sunny Glasgow
    Posts
    8,067
    Thanks
    7
    Thanked
    181 times in 171 posts

  2. #34
    Admin (Ret'd)
    Join Date
    Jul 2003
    Posts
    18,481
    Thanks
    1,016
    Thanked
    3,208 times in 2,281 posts
    Originally posted by Austin
    Please show us any figures that show how much faster the AthlonXP is compared to the original Athlon. Then show how the AthlonXP fairs against the P4's out at that time (400FSB P4 to early AthlonXP and then 533FSB P4 to the rest). Which would you say is closer?
    From AMD's independently audited benchmark figures :-

    Code:
          Test                               Athlon    Athlon    Athlon      P4
                                              1400     XP1600    XP2000    2.0GHz A
    
    
    Business Winstone 2001                    100       100.6     111.0      82.4
    Sysmark 2001 Off. Prod.                   100       102.2     113.0     103.7
    
    Office Productivity                       100       101.4     112.0      93.0
    
    Content Creation Winstone 2001            100       103.2     112.1      95.4
    Content Creation Winstone 2002            100       111.4     122.3     112.9
    Sysmark 2001 Internet Content Creation    100       125.6     142.2     137.2
    
    Digital Media                             100       113.4     125.5     115.2
    
    
    3D WinBench 2000 D3D software             100       101.4     105.5     103.3
    3D WinBench 2000 hardware T&L             100       102.4     108.3     100.3      
    3DMark 2001 D3D software                  100       106.4     117.0     108.2       
    3DMark 2001 hardware T&L                  100       103.1     108.9     102.7      
    AquaMark (1024x768x32)                    100       101.5     107.4     102.0
    Dronez Generic (1024x768x32 generic)      100       109.5     114.8     116.9
    Evolva Benchmark (1024x768x32)            100       101.5     107.0     100.8
    Expendables Benchmark (1024x768x32)       100       105.0     115.8      91.6
    Half Life Smokin' (1024x768x32)           100       103.5     119.3      97.8
    MDK2 (1024x768x32)                        100       111.2     125.2     108.3
    QuakeIII Demo 2 (640x480x16)              100       106.1     115.7     115.8
    Return to Castle Wolf 3D (1024x768x32)    100       105.2     117.0     105.1
    Serious Sam  (1024x768x32)                100       104.8     117.5      87.5
    Serious Sam - 2nd Encounter (1024x768x32) 100       104.1     116.0      94.0
    Unreal Tournament (1024x768x32)           100       103.0     110.5      91.5
    
    3D Gaming                                 100       104.6     113.7     101.7
    
    
    Desktop Overall                           100       106.5     117.1     103.3
    Note : These figures are all normalised to the Thunderbird 1400 figures - hence the row of 100's. The raw data is available too.

    I don't know if you understand the normalisation process, but essentially, it uses a reference point to adjust raw data to give comparative figures. In this case, the reference point is the Athlon 1400MHz Thunderbird. Simplistically put, take the ACTUAL result for the Athlon 1400, and work out what factor you need to apply to get a figure of 100. So if the Athlon 1400 scored 120, you need to apply a factor of 0.83333 to get that to100. Therefore apply the same factor to the other scores to normalise them to the Athlon 1400 as a base of 100. For the statistically pedantic, yes, I know I've grossly oversimplified here, and yes, I did stats at University. I'm trying to illustrate the intent of normalisation rather than the actual technique. As I said, AMD do also provide the raw results data. I'm a firm believer in the saying attributed to Disraeli "lies, damned lies and statistics". I therefore do not propose to get into a debate as to whether AMD could have picked statistical techniques that suit their case or not - this is why I said in my earlier post that this sort of stuff ought to be done by an independent and trusted third-party. I'm also someone who, when being told by my bank some years ago, that interested added to a savings account was based on the "average" balance, put the manager right on the spot by asking which "average" he meant?

    He didn't know what I was talking about. So I said that mean, median and mode were all averages, and would be likely to give different results.

    "Oh, he said", "I mean the mean" and grinned, satisfied with his smart reply. "Okay," says I, "which mean?".

    "Huh?" he said, grin fading.

    "Well," says I, " you could mean the geometric mean, or maybe the harmonic mean, for instance". By now I was grinning and he was sweating.

    Anyway, back on topic.

    As requested, the XP1600 runs at 1400MHz, and the above figures suggest a significant performance increase over a Thunderbird Athlon at the same clock speed.

    Austin, you asked for some figures. From the above, does it appear that the XP2000 and P4-2000 were being compared to each other as of equal performance or, as AMD pointed out at the time, does an XP1600 give a P4-2000 a damn good run for it's money, and an XP2000 show it a clean pair of heels in virtually every regard?

    For basic comparison, AMD used DDR-based P4 test systems, so as to gauge the relative performance of the CPU's rather than memory sub-systems. They do however, provide similar data against P4's using RDRAM, and the specifications of all systems used is, as I said, published.

    So, take a look at the Overall Desktop figures, as highlighted (by me) in red, and tell me again how an XP2000 is rated as equivalent to a P4-2000 by AMD, bearing in mind that the above is a subset of the figures AMD base their claims on, and that these figures have been independently audited.

    I'm no AMD expert, but the figures presented here are a very small part of the figures I went through with AMD's Marketing Manager.

    Far be it from me to argue AMD's case for them, but I think they put up a credible argument for their claim, and they most emphatically do claim, that the PR rating is based on a comparative to Thunderbird Athlons and not P4's.

    So, let's see your figures supporting your claim that the PR rating DOES relate to P4's not Thunderbirds

  3. #35
    Admin (Ret'd)
    Join Date
    Jul 2003
    Posts
    18,481
    Thanks
    1,016
    Thanked
    3,208 times in 2,281 posts
    Thanks for that. It's a much more recent and updated version of the White paper I was looking for.

  4. #36
    Banned
    Join Date
    Jul 2003
    Location
    House without a red door in Birmingham
    Posts
    1,595
    Thanks
    0
    Thanked
    0 times in 0 posts
    I'm sure I'll get the time soon. Until then you have to consider the following points.

    * What did you think AMD were going to say? LOL!
    * I 'only' did an A-Level in Maths but I certainly covered normalisation.
    * I'm sure you understand that statistics can show practically anything you want depending how you derive them and how you present them. Apple have the fastest CPU, really?
    * Using a P4 with DDR was better than using one with SDR (LOL) but DDR at the time was PC2100 which killed P4 perf, at that time they were also using 400FSB and 256k L2.
    * The changes in the AMD XP/PR rating when we hit TbredB, 512k L2, increased FSB and esp the release of the XP3200+ (in my opinion) all tried to keep up with the ever changing P4 and stay reltive to it. IMHO AMD knew people were comparing the XP rating to the P4.
    * The figures you show demonstrate what I already knew, except AMD threw a few choice bms in there to inflate the XP scores a little. The AthlonXP1600+ 1.4ghz was roughly 5% faster than the Athlon 1.4ghz but the XP rating reflects more than a 14% increase. The XP1500+ was often SLOWER than the Athlon 1.4ghz.
    * If you come away from AMD's own 'research' and simply compare the AMD and Intel CPUs along side each other it is very clear the perf is VERY relative to the P4 to the point of being as identical as two architecturally different peices of hardware are ever likely to get.

    I stand by my guns as my research is what I've done independently from many INDEPENDENT sources from a neutral's point of view. I didn't and don't care whether the XP rating is derived from the P4 or the std Athlon but I certainly know which is closer. I'd say watch this space but we are kind of off topic now, I'll start another thread ASAP when I check that PDF and gather my evidence a little. BTW I'd just like to point out that there's nothing wrong with disagreeing nor diffs of opinion.

    PS. BTW crime is down and the NHS has never been more efficient. Well according to the governments figures. The media's figures are often very different while the independent individual's personal experiences are soemwhere inbetween and probably closet to reality for good reason.

  5. #37
    Admin (Ret'd)
    Join Date
    Jul 2003
    Posts
    18,481
    Thanks
    1,016
    Thanked
    3,208 times in 2,281 posts
    OK, so you wanted figures, but now dismiss them as proving anything with statistics??
    Originally posted by Austin
    I'm sure I'll get the time soon. Until then you have to consider the following points.

    * What did you think AMD were going to say? LOL!
    AMD, in that document referenced, provided P4 figures too, and the stats make it clear to me that they are basing their ratings on exactly what they said they were basing it on.

    Originally posted by Austin
    * I'm sure you understand that statistics can show practically anything you want depending how you derive them and how you present them. Apple have the fastest CPU, really?
    Indeed you can - it is a point I made above. But none the less, AMD provided their ratings, they explained REPEATEDLY what they were basing them on, they published the statistical basis for those figures, they had them externally audited and they provided P4 comparisons as well. What more could they have done?

    Originally posted by Austin
    * Using a P4 with DDR was better than using one with SDR (LOL) but DDR at the time was PC2100 which killed P4 perf, at that time they were also using 400FSB and 256k L2.
    As I said, they did comparisons against RDRAM system AS WELL, and those figures are in that report too. BUT, their figures were, as they said, to show the basis for their PR ratings which are based on their own systems, not P4's, and therefore RDRAM is not directly relevant as AMD do not do RDRAM systems. In the figures they give, they seek to provide a comparison between Athlon and P4 performance, so keeping as many other factors as possible the same is the ONLY sensible way to try to compare raw CPU performance. AMD did not, however, try to hide that P4's gain from better memory performance, and included test results for RDRAM as well. The benchmarks they did were about their comparative CPU performance, not about overall real-word system performance. If they had been, then the testing regime would have been different.

    Originally posted by Austin
    * The changes in the AMD XP/PR rating when we hit TbredB, 512k L2, increased FSB and esp the release of the XP3200+ (in my opinion) all tried to keep up with the ever changing P4 and stay reltive to it. IMHO AMD knew people were comparing the XP rating to the P4.
    AMD may well have known people were comparing P4's to Athlons, of whatever generation. After all, people are hardly likely to be comparing them to anything else.

    Bear in mind that Intel are gaining greatly from the use of clock speed as the final arbiter of system performance, and are in no great hurry to see that change - even if to do so would lead to better consumer understanding of the issues.

    Make no mistake - this is all about marketing, and AMD are fighting their corner. The PR rating is about marketing. It is about changing public perceptions. It is about combating the great unwashed's assumption that a P4 of a slightly faster clock speed must be better/faster than an Athlon of a slightly slower clock speed - because the great unwashed are all to likely to walk into the shop, pick the machine that has the fastest number and buy that because it must be the best. THAT is why AMD want this perception about clock speed changed, and why Intel don't.

    However, you didn't originally say what people were making of this - OUTSIDE AMD's control. You said
    Are you seriously trying to suggest that the 'XP rating' is remotely linked to the old Thunderbird Athlon?
    Well, as I said, yes, I am. That's your original claim, and that's what I gave the figures you asked for to refute.

    I did not say that people have not interpreted it otherwise. But AMD stated, again and again and again, what the PR rating was based on, yet some people simply refuse to accept what AMD used as the basis for their rating. Beyond telling you over and again what they have done and why they have done it, and providing the detailed figures (as per the extract above) to support it, exactly what were AMD supposed to have done to convince you that their rating was what they said it was.

    Originally posted by Austin
    * The figures you show demonstrate what I already knew, except AMD threw a few choice bms in there to inflate the XP scores a little. The AthlonXP1600+ 1.4ghz was roughly 5% faster than the Athlon 1.4ghz but the XP rating reflects more than a 14% increase. The XP1500+ was often SLOWER than the Athlon 1.4ghz.
    Then you are completely misunderstanding what the PR rating is all about. It is NOT, emphatically NOT saying that an XP1600 is any specific percentage faster than an Athlon 1400, or any other chip.

    That is the WHOLE POINT of ratings. You CANNOT compare relative performance by comparing clock speeds. Too many other factors affect performance. AMD want clock speed relegated to, at best, no more than ONE factor in assessing system performance - just like an engines cc is no great indicating of overall performance, as anyone driving a 4.2 litre Jag could tell you if they tried to out-accelerate an F1 car. There are many other factors affecting performance there too.

    Read that white paper. Their point is that it is not clock speed that matters, but work done. And work done per period of time is work per clock cycle X clock cycles per period. Doubling a CPU clock speed does NOT result in double the real world (or even benchmark) performance, so why should the ratio between PR rating and Thunderbird clock speed reflect the ratio of clock speeds? Indeed, if it did, it would destroy the argument that a PR rating was needed in the first place.

    The PR rating attempts to show what clock speed the Thunderbird would need to be running at to give the same overall performance. That is NOT the same as saying that increasing Thunderbird performance by x% results in x% performance gain.

    The underlying chip architecture has changed between T’Bird and Palomino, so direct clock speed comparisons are bound to be misleading, just as clock speed comparisons between Athlon (whether T’Bird, Palomino, T'Bred, Barton or whatever) are.

    Originally posted by Austin
    * If you come away from AMD's own 'research' and simply compare the AMD and Intel CPUs along side each other it is very clear the perf is VERY relative to the P4 to the point of being as identical as two architecturally different peices of hardware are ever likely to get.
    I've done my own research too, and quite possibly I did some of the testing for the research your talking about, including work for both PC Magazine and ZDNet on this issue. I'm not basing my comments just on reading about on the web, but on having actually done comparison testing at the time - on machines supplied by AMD and Intel for the purpose. AMD, for instance, provided me with an XP1500, XP1600, XP1700 and XP1800 processor, shortly after launch, specifically to do the comparison of real world performance.

    Originally posted by Austin
    * If you come away from AMD's own 'research' and simply compare the AMD and Intel CPUs along side each other it is very clear the perf is VERY relative to the P4 to the point of being as identical as two architecturally different peices of hardware are ever likely to get.
    But you cannot come away from AMD's own research. I'm not talking about what comparative P4 v Athlon performance is. THAT is another issue entirely. I'm talking about the point you challenged me on - which is what the PR rating was based on - and it is based on AMD "research". It's all very well cynically dismissing it just because AMD provided the data, but the fact remains, they did the tests and were audited on them, based their PR ratings on that and published both the basis for the ratings, and the test results.

    Originally posted by Austin
    I stand by my guns as my research is what I've done independently from many INDEPENDENT sources from a neutral's point of view. I didn't and don't care whether the XP rating is derived from the P4 or the std Athlon but I certainly know which is closer.
    I'm not discussing which is the closer though - I'm talking about what it's based on. That was the challenge you threw down, and the claim that you made about it being based on the P4 comparison. For instance :-

    Originally posted by Austin
    ..... Anyway the XP rating is supposedly derived from some calculation based on the old Thunderbird Athlon but to anyone with basic knowledge of CPU perf it's abundently clear that AMD wished to avoid lawsuit type complications ... it's very apparent it is designed to give the Athlon a rating equivilent to P4.
    Originally posted by Austin
    The XP rating in reality does indicate the AthlonXP's perf relative to P4 although to avoid any legal strifes AMD claim it's all to do with the Thunderbird Athlons LOL!
    Originally posted by Austin
    Are you seriously trying to suggest that the 'XP rating' is remotely linked to the old Thunderbird Athlon? The fact that the XP rating ties in almost exactly with the relative P4 at the time is just a coincidence? If you truly think that I'd have to say you're in a very small minority (who can always be right of course).

    Please show us any figures that show how much faster the AthlonXP is compared to the original Athlon. Then show how the AthlonXP fairs against the P4's out at that time (400FSB P4 to early AthlonXP and then 533FSB P4 to the rest). Which would you say is closer?
    I'm not going to debate relative performance with you. That isn't my point. My point, which I think I've pretty much exhausted the mileage in, was explicitly was the PR rating was based on.

    You wanted figures comparing XP's to the old Athlons - I gave you some. You wanted figures comparing P4's to Athlon XP - I gave you some of them too. Remember, this is about what the PR rating is based on. AMD did the basing, and I merely quoted the figures AMD used to derive the rating.

    Originally posted by Austin
    PS. BTW crime is down and the NHS has never been more efficient. Well according to the governments figures. The media's figures are often very different while the independent individual's personal experiences are somewhere in-between and probably closet to reality for good reason.
    If you want me to defend government statistics, you have REALLY come to the wrong place. Most governments (and this one is much worse than most) could fall in a vat or truth serum and come up spluttering, and denying having fallen in the first place. I'm personally convinced that this government (and most others) wouldn't know the truth if it bit them in the behind and the day they say the Sun will rise in the East is the day I start looking West for it in the mornings.

    Originally posted by Austin
    BTW I'd just like to point out that there's nothing wrong with disagreeing nor diffs of opinion.
    Absolutely agree. I don't agree with your assertion about what AMD based the PR rating on, but am more than happy to see you continue to argue your case. I've had occasions in the past where I've debated a point (often political) and the arguments against seem to dry up somehow. It's as if some people don't likely arguing with a mod, or even less, an admin. Balls to that. If you don't agree with what I say, go right ahead and disagree. Ask DaBeeenster - we've had enough arguments, both here and elsewhere. We rarely agree, but both respect each other for arguing their point of view. Remember, Question Time is all about debate and that is my forum. A debate would be downright boring if it consisted of one person saying "I think ........." and everybody else saying "Yeah, I agree".

  6. #38
    Banned
    Join Date
    Jul 2003
    Location
    House without a red door in Birmingham
    Posts
    1,595
    Thanks
    0
    Thanked
    0 times in 0 posts
    Originally posted by Austin
    I'd say watch this space but we are kind of off topic now, I'll start another thread ASAP when I check that PDF and gather my evidence a little.
    Okay I think we can both agree that AMD did base their XP rating on the std Athlon and produced figures to support that, I've never denied that. However I think it is very clear that the XP number is way way closer to a P4 of that mhz than the std Athlon which AMD claim, or is that not what AMD claim? If the AthlonXP tends to be 5%ish faster than the std Athlon how do they justify deriving a number that indicates 14% perf gain over the std Athlon? Surely they're saying the XP1600+ part of the AthlonXP @ 1.4ghz would relate to a std Athlon @ 1.6ghz (or 1600mhz) which it clearly doesn't just like the AthlonXP1500+ was often slower than the Athlon 1.4ghz.

    However the AthlonXP at any speed rating VERY closely resembles the perf of a (non-800FSB) P4 at the same mhz (ie XP1600+ = P4 1.6ghz, XP2000+ = P4 2.0ghz). I've always admitted that AMD claim the AthlonXP's perf rating is derived from the std Athlon but I just think the reality is different, they must have had it in mind to give it equivilence to the P4 and it seems they adjusted it when needed to keep that in line.

    The different AthlonXP cores other than Barton didn't boost perf per clock. All the Tbred cores really did was enable the AthlonXP to clock faster not run faster clock for clock, hence no perf movement was necessary. TbredA introduced 0.13mu and an extra layer (total of 8) which enabled them to run cooler but not as (surely) intended much faster. The TbredB added one extra layer (9) but again all this did was enable the chip to run cooler and clock faster NOT gain perf per clock (so not any faster than the std Athlon). Yet the introduction of TbredB (XP2400+) is where AMD chose to reallign the XP rating which just happened to coincide with Intel's bump in speed with the 533FSB 512k P4B revision. Only the Barton enabled higher perf per clock courtesy of double the L2 cache, by this time Intel were ramping up the ghz, implimenting PC3200 (then Dual Channel PC2100) and soon moving to 800FSB ... for some reason the XP rating got inflated from both Barton AND 400FSB (as in the perf gain was less than the rating indicated compared to the 256k AthlonXP or Barton of lower FSB). Intel's top chip was 3.2ghz and the XP3200+ (400FSB but 2.1ghz at first IIRC) just happened to compete with it in the layman's eyes. Ah well.

    As for using DDR on the P4 to isolate the CPU is complete pants IMHO. Did they try dropping the P4 down to 200FSB (4x50mhz) because the AMD used 200FSB (2x100mhz)? The simple fact is that most P4 mobos before PC2700 were RDram and for very good reason ... the P4 was (and still is) always FAR more reliant on high FSB and memory bandwidth. I'll check that PDF (as soon as I get chance). If your point is that AMD say the XP rating is derived from the std Athlon and gave evidence to support it I fully agree. If you're saying they really based it on that then that is where I do disagree.

    PS. Watch the length of you posts, the mods may ban you!

  7. #39
    Admin (Ret'd)
    Join Date
    Jul 2003
    Posts
    18,481
    Thanks
    1,016
    Thanked
    3,208 times in 2,281 posts
    Originally posted by Austin
    Okay I think we can both agree that AMD did base their XP rating on the std Athlon and produced figures to support that, I've never denied that.
    .... yet you said
    Originally posted by Austin
    Are you seriously trying to suggest that the 'XP rating' is remotely linked to the old Thunderbird Athlon? The fact that the XP rating ties in almost exactly with the relative P4 at the time is just a coincidence? If you truly think that I'd have to say you're in a very small minority (who can always be right of course).
    and the other similar quotes I referred to in the last post.

    That's precisely what kicked this discussion off in the first place - the statement that the rating had nothing to do with performance relative to the old T'Bird Athlons.

    Originally posted by Austin
    However I think it is very clear that the XP number is way way closer to a P4 of that mhz than the std Athlon which AMD claim, or is that not what AMD claim?
    AMD assert that the PR rating is (or at least, WAS) designed to indicate the Mhz at which a T'Bird would have to be running to give the same performance in their benchmark suites as the XP chip bearing that rating. In other words, a T'Bird Athlon would have to be running at 1800MHz to give the same performance as an Athlon XP1800.

    Originally posted by Austin
    If the AthlonXP tends to be 5%ish faster than the std Athlon how do they justify deriving a number that indicates 14% perf gain over the std Athlon? Surely they're saying the XP1600+ part of the AthlonXP @ 1.4ghz would relate to a std Athlon @ 1.6ghz (or 1600mhz) which it clearly doesn't just like the AthlonXP1500+ was often slower than the Athlon 1.4ghz.
    Because MHz is NOT a fair indicator of system performance. That's the whole point of the PR rating.

    I presume you're getting that 14% figure from the fact that XP1600 @ 1400MHz implies that 1600 is 14% higher than 1400? i.e. 1400 x 1.14 is 1600 (ish).

    But that presumes that performance increases in a linear fashion, and atthe same rate, as clock speed. But it doesn't. We all know it doesn't. Is an Athlon XP3200 exactly twice as fast (in a range of benchmarks) as XP1600? Would a P4 chip of double the clock speed give double the performance of the one half the clock speed?

    Performance does not increase in a 1:1 relationship with clcok speed. Would that it did!!

    The PR Rating indicates the MEGAHERTZ at which a T'Bird would haveto be running to give the same performance as the XP carrying that rating, but that does NOT mean that increasing clock speed by 14% (from 1400 to 1600) results in a 14% increase in performance. In fact, the law of diminishing marginal returns applies, doesn't it?

    Clock speed is FAR from the only factor affecting CPU performance. Other factors would include main memory speed and design, bus speed, the type of calculations you ask the chip to do, on-board cache size, cache type, the effectiveness of the cache algorithm, and so on.

    I wouldn't mind betting that if you simply kept increasing clock speed in increments, with no other changes to chip internals or supporting hardware, that every time you increased the clock speed a bit more, the actual relative performance gain you achieved would be less than last time. Probably, the time would come when the other factors would be providing such a bottleneck that increasing clock speed ALONE provided no performance gain at all.

    Therefore, consider that the PR rating is supposed to be providing a performance comparison with an equivalent T'Bird machine, and it is NOT saying that it directly relates to a given PERFORMANCE increase in percentage terms - because that increase is not linear with clock speed. This is what I meant when I said that I didn't think you understood what the PR rating was actually all about.

    If you increased a T'Bird 1400 clock speed by 14%, do you gain 14% increase in those benchmark results. I'd bet you don't. Does 6% sound like a more realistic increase in benchmark results.

    THAT is what the PR rating says.

    1) Take an XP chip running at a given clock speed and benchmark it.

    2) Work out what clock speed a T'Bird would have to be running at to achieve that same benchmark result, and stick THAT figure on the XP as a PR Rating.

    Bear in mind this is marketing, and aimed at non-PC literate users, in the main. Users who were used to the T'Bird generation. What AMD are saying is "forget MHz, the XP rating tells you the relative performance of this XP machine compared to that old T'Bird design. Buying a XP1600 is like buying a T'Bird at 1600MHz". They are NOT claiming it's 14% faster.

    Originally posted by Austin
    I've always admitted that AMD claim the AthlonXP's perf rating is derived from the std Athlon but I just think the reality is different, they must have had it in mind to give it equivilence to the P4 and it seems they adjusted it when needed to keep that in line.
    But thats the point. They don't just claim it. They explained the basis for the calculation and published the figures - in some detail.

    As for the changes later in the product range, I'm not even going to start arguing what that meant or how it was done, as I haven't followed it closely enough. You may well be right, I don't know. What I DO know is that I said early on in this debate that the PR rating is far from perfect, and even AMD will admit that the original scheme does not scale well, especially when getting far away from the original T'Bird processors and when changing technologies.

    AMD were trying to get an international independent standard set up to assess relative performance but it seems to have led nowhere. I do not officially know why. I have been told unofficially (and I'm NOT attributing this to AMD, as it did NOT come from them) that a large part of the reason is that for such an independent scheme to work, it has to have the active cooperation of all major chip-makers, and that Intel are not really very keen on the idea. They, I was told, are quite happy with customers assuming double the clock speed equals doublethe performance, as it works in their favour.

    I would ASSUME that any realignment in the PR rating is to do with these problems in scalabilty, but as I haven't researched it, and nor have I had any briefings from AMD on it, it is JUST an assumption.

    As for the length of the posts, I figure if I can't argue you into agreement, I can at least try to bore you to death.

  8. #40
    Banned
    Join Date
    Jul 2003
    Location
    House without a red door in Birmingham
    Posts
    1,595
    Thanks
    0
    Thanked
    0 times in 0 posts
    I simply said (and always have) that AMD's official line was that the XP rating was based solely on the Tbird Athlon and it's obvious IMHO that it was to give it a rating relative to the P4 it was competing with.

    So for AMD's theory of what the XP rating means is true then a std Athlon would have to run at 1500mhz to give the perf of an XP1500+ which is easy to see it's total carp! Even at 1.4ghz the std Athlon often beat the XP1500+! O/c'ing the std Athlon to 1.5ghz gave perf well above the XP1500+ and easily competed with the XP1700+! The 14% speed increase I talked about was based on the "XP1600+ indicated that it gives the perf of a std Athlon at 1600mhz" ... still I can't see how AMD claim the slight enhancements of the XP core could derive an 1400mhz chip with a 1600mhz rating because the optimisations would have to make the CPU much faster than 5% per clock. Again when you compare the XP1600+ to P4 1.6ghz or XP2000+ to P4 2.0ghz the XP rating ties in almost precisely (again rem archys are very diff) ... obv that's with RDRAM which is the only fair comparison IMHO. Bottom line, a Tbird Athlon at 1600mhz totally wipes the floor with the AthlonXP1600+ ... and saying the XP1500+ is remotely faster than the Tbird Athlon 1.4ghz is also a joke. Unless you pick bms very precisely it all falls apart ... until you compare the AthlonXP to the relative P4 that is.

    So to be breif (LOL) I agree (and always have) that AMD say the XP rating is derived fromt he std Athlon but you don't have to look very far to know that's total rubbish and that it is designed pure and simple to give a speed rating relative to the P4.

  9. #41
    Admin (Ret'd)
    Join Date
    Jul 2003
    Posts
    18,481
    Thanks
    1,016
    Thanked
    3,208 times in 2,281 posts
    Originally posted by Austin
    .... So for AMD's theory of what the XP rating means is true then a std Athlon would have to run at 1500mhz to give the perf of an XP1500+ which is easy to see it's total carp! Even at 1.4ghz the std Athlon often beat the XP1500+!
    NO, it didn't. Look at the figures in the AMD document you were given the link to.

    How many times do I have to say this. It's based on AMD's benchmark regime!!!

    The figures are there in black and white. Yes, sure, on individual tests in some circumstances a 1.4 T'Bird will outperform an XP1600, but then in some tests, a 1.4 T'Bird will outperform a P4 2.4 running RDRAM and, for that matter, in some tests, a P4 2.4 running DDR will outperform a P4 2.4 running RDRAM. But some tests don't matter a damn when evaluating the PR rating. The ONLY tests that matter are those in the PR Rating.

    It's not about what level of performance comparisons you get in tests, or I get in tests, or Ziff Davis Labs get in tests, it's about the specific suite of benchmarks that AMD use in developing the rating, and everything else is irrelevant.

    AMD designed a test. They picked a series of benchmarks, performed a testing process, repeating each test at least three times, and more if the coefficient of variation was too large, until they reached a set of figures. They published the benchmarks used, the raw data and the process they used, and they had it audited.

    Argue about whether the PR rating reflects real-word performance all you like, but all that boils down to is whether one set of benchmarks or another represents a better reflection of real-world tests. The PR rating uses the benchmarks it uses. You can argue that there are better benchmarks, but unless you get an independent testing system, agreed by AMD and Intel, such arguments will go round and round. The PR rating does not attempt to be independent. It is AMD's way of comparing THEIR newer chips to THEIR older chips, using a defined and published method.

    You talk about an overclocked Athlon T'Bird outperforming an XP1500. But overclocked how? Upping the FSB perhaps???? In which case, you are clocking the rest of the system too and are improving memory performance as well. That is not the same thing at all. And again, using WHAT tests? Have you run the same test regime on that o/c'd machine as AMD use in the PR rating?

    I'll say it again. I sat down with AMD and went through these figures in some detail, and I flat out don't agree with you. They are (or at least, were, when launched), in my view, based on exactly what AMD said they were based on.

    Look at the figures from their Benchmark results. The XP2000 scores 117.1. The P4-2000 DDR scores 103.3 (which is between an XP1500 and XP1600), and even a P4-2000 RDRAM scores 108.4 (with the XP1600 at 108.5 and the XP1800 at 112.0).

    The figures are there, and it's totally daft to suggest that the PR rating compares XPxxxx to P4 running at xxxxGHz. It simply doesn't. OTHER benchmarks might give other results, but AMD aren't using other benchmarks - they are using the ones they stated and published.

    Now, this is my last post on the subject. You wanted figures and got them - chapter and verse. You've made repeated remarks about AMD tying in to P4 performance, or wiping the floor with or getting wiped by this or that, but they are all vague statements. No cold, hard figures, no benchmark results, no independent auditing. All assertions and statements, no supporting evidence.

    If the figures I've quoted, and that reference White Paper aren't enough for you, then so be it. I don't propose to spend any more time going round in circles.

    If this post comes across as fed up, Austin, it's not aimed at you. It's due to the headache I've got, and have had for about a week now, and that I feel I'm batting my head against a brick wall here.

    I've made my case, and have said all I'm going to on the subject.

  10. #42
    Member
    Join Date
    Aug 2003
    Posts
    160
    Thanks
    0
    Thanked
    0 times in 0 posts
    does anyone have a list of the presets for cpus so i can work out what to use to get ceratin ratings, eg. 13x166 = xp2700 etc.

  11. #43
    Senior Member
    Join Date
    Jul 2003
    Location
    Sheffield
    Posts
    529
    Thanks
    1
    Thanked
    0 times in 0 posts
    Originally posted by Saracen
    How many times do I have to say this. It's based on AMD's benchmark regime!!!
    The benchmarks were set up to justify the decision of AMD to start using the rating system again, not the other way round.

    Everyone knows this.

    The purpose of the rating system is blindlingly obvious regardless of any official AMD statement. Their crystal clear intent is to more fairly rate AMD vs P4 processor speed to an IT uneducated public.

    Back to the start of the thread. I much prefer the bios not to show the meaningless rating factor.

    My old palomino 2100+ (1.7ghz) is a far less power chip than either my 1700+ or 1800+ tbred B's. The rating system is worthless on a discussion forum like this.

  12. #44
    Sublime HEXUS.net
    Join Date
    Jul 2003
    Location
    The Void.. Floating
    Posts
    11,819
    Thanks
    213
    Thanked
    233 times in 160 posts
    • Stoo's system
      • Motherboard:
      • Mac Pro
      • CPU:
      • 2*Xeon 5450 @ 2.8GHz, 12MB Cache
      • Memory:
      • 32GB 1600MHz FBDIMM
      • Storage:
      • ~ 2.5TB + 4TB external array
      • Graphics card(s):
      • ATI Radeon HD 4870
      • Case:
      • Mac Pro
      • Operating System:
      • OS X 10.7
      • Monitor(s):
      • 24" Samsung 244T Black
      • Internet:
      • Zen Max Pro
    I prefer to have the raw MHz rating myself, the PR thing at bootup is only good to those that either A) have no real computer knowledge, or B) want to show off..
    (\__/)
    (='.'=)
    (")_(")

  13. #45
    dgr
    dgr is offline
    Senior Member
    Join Date
    Jul 2003
    Posts
    621
    Thanks
    0
    Thanked
    0 times in 0 posts
    Originally posted by Stoo
    I prefer to have the raw MHz rating myself, the PR thing at bootup is only good to those that either A) have no real computer knowledge, or B) want to show off..
    show off that i spent loads of money on a 3200+?

    guess i'm not cool, saving money on buying a 2800 and OCing it to a 3200+ ?

    ah well

    dgr
    dothan 745 @ 2.4ghz | 2gb Corsair XMS (2-3-3-6) | dual raptors (raid0) | ATI 9700pro | CM201 | dual lg 1810

  14. #46
    In a place called VERTIGO CrapshoT's Avatar
    Join Date
    Jul 2003
    Location
    Pushing his fingers into his eyes
    Posts
    1,340
    Thanks
    1
    Thanked
    0 times in 0 posts
    i nearly always get i rating off my cpu in my asus a7n8x mobo at the minute ive got a xp2500 its runnin at 12.5x174 fsb and its recognised as a xp2800 in my bios when i had it running at 11x204fsb it was recognised as a xp3200 but if i ran it at 10.5x204 it was recognised as a xp3000 but its 3d mark 2001 that worriees me my cpu is recocnised as a 2.17 duron lol
    Farts are like children.....You always love you're Own.!!!
    Wise Men learn from Other people's mistake's...Fool's learn from their OWN.!!!

  15. #47
    Banned
    Join Date
    Jul 2003
    Location
    House without a red door in Birmingham
    Posts
    1,595
    Thanks
    0
    Thanked
    0 times in 0 posts
    Saracen if you read as carefully as I did you should have seen long ago (when you first started on your rampage) that I was going to supply supporting evidence and begin a new thread on the subject rather than going off on an unecessary tangent in this thread ... shame you thought better to keep posting what clearly turned into a single minded rant here. As for having a headache do rem there are those worse off than yourself. I won't go on about how my daughter has been clinging on to her life for the last 4 years kowing each day with her could be the last, that my wife is in agony and can barely walk most days, nor that she had skin cancer in her hairline which though removed may make a comeback, that I have had agonising back pain for the last month, that we're facing all the other financial and family strains which many people face too. Again I realise there are plenty of people way worse off than myself or my family, if I find time to help on here I do and if I don't then I don't begin a thread or fresh 'opinion exchange' I am unable to finish. I still endevour to find time to patiently help people and discuss things on here, shame it seems you don't considering your position. If you had waited for a new thread on this matter I would have supplied supporting evidence and made my points, the fact is I've been in hospital more than out of it for the last few months.

    Thanks for supplying evidence but I already knew AMD had attempted to show people the XP rating was all about the std Athlon and nothing to do with the P4. I don't refute that but I do think you're wrong if you think the XP rating is nothing to do with the P4 regardless of what or how AMD stated or attempted to brainwash top members of the tech community (looks like that one worked).

    There are other comments you make which indicate you haven't really bothered to read what I've stated. Firstly that you ignored my request to wait and begin a new thread or for me to gather my evidence and make my points. Secondly that I never said AMD outwardly claim the XP rating is to do with the P4, only that it was a hidden agenda and blatently obvious to those with open minds that the XP rating is ALL about the P4. Thirdly the std Athlons were easily unlocked (pencil only) so o/c'ing was not a matter of increased FSB, the Athlon 1.5ghz was way faster than the XP1500+ every time. Fourthly I compared the XP1500+ to the Athlon 1.4ghz (so no o/c'ing) and pointed out the XP1500+ was almost always slower so again show how awful the XP rating was if it was based on the std Athlon. Fifthly I said you could quote AMD's own stats as much as you like as anyone who's worked in or studied stats will realise they can easily show what you want esp if you are biased as AMD so obviously were. Even using AMD's stats I don't see how they arrive at the bloated XP rating, bloated if they are comparing it to a std Athlon, as perfect as possible if you look at P4 vs AthlonXP. Did I get all the relevant hw and run exhaustive tests? No, noone was paying me for the time and cost of that. Much of the respect I've gained on the net is because unlike many review sites I come from a completely independent and unbiased position and that (when given chance) I always back up my statements and comments with plenty of reason and evidence.

  16. #48
    Banned
    Join Date
    Jul 2003
    Location
    House without a red door in Birmingham
    Posts
    1,595
    Thanks
    0
    Thanked
    0 times in 0 posts
    dadofsam this isn't 100% up to date but should more than suffice.

    AthlonXP SktA (462) clock rates (all 133/266FSB unless otherwise stated):
    1600+ 1.40ghz 10.5x
    1700+ 1.47ghz 11x
    1800+ 1.53ghz 11.5x
    1900+ 1.60ghz 12x
    2000+ 1.67ghz 12.5x
    2100+ 1.73ghz 13x
    2200+ 1.80ghz 13.5x
    2400+ 2.00ghz 15x
    2600+ 2.13ghz 16x (also available in 333FSB & 12.5x @ 2.08ghz)
    2700+ 2.16ghz 13x (333FSB)
    2800+ 2.25ghz 13.5x (333FSB)

    BARTON ATHLONXP (all 166/333FSB "")
    AthlonXP2500+ @ 1.83ghz 11x
    AthlonXP2800+ @ 2.08ghz 12.5x (note there's a TbredB with exactly the same name)
    AthlonXP3000+ @ 2.16ghz 13x
    AthlonXP3200+ @ 2.20ghz 11x (200/400FSB, also some early reviews used 2.10ghz)

Page 3 of 4 FirstFirst 1234 LastLast

Thread Information

Users Browsing this Thread

There are currently 1 users browsing this thread. (0 members and 1 guests)

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •