Results 1 to 13 of 13
Thread: XP PR Madness
10-12-01, 10:23 AM #1
XP PR Madness
This site is a big proponent of AMD's new advertising gimmick, the PR rating. Aside from any subjective opinions on whether or not the methodology has any validity or not, here's a simple analysis of the assignment of PR vs. XP clock rate which easily proves it to be unreliable.
Accepting at face value AMD's assertion that a 1333 MHz XP is equivalent to PR1500+ we come up with a ratio of XP MHz vs. PR of:
So, using this factor, an XP processors running at their respective MHz should have PRs of:
1400*1.125=1575 not PR1600+
1467*1.125=1650 not PR1700+
1533*1.125=1725 not PR1800+
So, how can AMD claim the inflated PR ratings they advertise? And don't tell me it's more complicated than that because the only difference between the various XP processors is the clock rate which is a linear parameter. And note that, with each succeeding PR designation, the exaggeration of the PR becomes greater and greater. With the next logical release, an XP running at 1600 MHz, the PR rating will be a full 100 MHz higher than it should be - PR1900+ vs. PR1800+.
This is crap, pure and simple, and has no place in this industry. Unfortunately, since AMD has already launched this advertising gimmick, there's no way they can take it back.
10-12-01, 10:35 AM #2
I do agree the PR rating is a crap idea but there is one reason why AMD are doing this.
THE MASSES ARE STUPID
AMD should just release a good ad Campaign to try and educate the dumb.
But no matter what AMD do at the moment I will still be an AMD fan.
10-12-01, 10:52 AM #3
- Join Date
- Sep 2001
- Boulder, CO
its not linear.
you just proved that. what they are saying is that their chips run roughly 200mhz faster thatn p4's.
1.2 = 1400+
1.4 = 1600+
1.6 = 1800+
for the last time people, this is not that fu**ing hard!
its REALLY simple. amd chips are really faster than p4's. not by a little, by a lot. by about 200mhz (a little more actually). so instead of having the general population think amd is just slow, amd has to take action to make people understand that its not just numbers. if you dont like the way amd is doing it, go out and buy an intel, and get a <sarcasm> real </sarcasm> 2.0ghz chip.
honstely, you are smart enough to know that the 1800+ is really a 1.6ghz. will you feel cheated if you get it home, and find out its really a 1.6ghz. my guess is no. and if you do feel cheated, even though its clearly marked as a 1.6ghz, well, i feel sorry for you.
i really just cant understand why everyone is having such a hard time with all this.
10-12-01, 10:57 AM #4
A simply way to comvert: -
PR = (MHz-333)x1.5
MHz = (PR/1.5)+333
When converting you may get an answer like 1600.5 but just round to the nearest 100 for PR and remember that the MHz will end in a 00 or a 33 or a 67.
1600 PR = 1399.9999999999 ie 1400
1700 PR = 1466.3333333333 ie 1467
1400 MHz = 1600.5 ie 1600
1467 MHz = 1701 ie 1700
Hope this will help end the confusion.
10-12-01, 12:06 PM #5
This is getting weird. What I did was use the 1333 MHz XP, rated a PR1500 to establish a clock ratio between the XP and the P4 or PR clock rate. That comes out to 1500/1333=1.125, meaning that:
each XP clock cycle is equal to 1.125 P4 clock cycles
I'm willing to accept this factor as the intrinsic performance advantage inherent in the XP architecture. This ratio cannot change, however, when the base clock of the XP processor changes. The rate of change is linear. I'm not disputing that the XP is, clock of clock, is a better performer than the P4. That is a given with all the evidence from the various benchmarks. All I'm saying is that this PR rating doesn't follow this relationship of 1 XP clock being equal to 1.125 P4 clocks can't change just for the sake of a higher PR rating. So, there's obviously some fudging going around with these ratings and they can't be accepted as being valid.
10-12-01, 02:02 PM #6
But why do you base your ratio on the 1500?? Say you based it on the 1800, then 1800/1533 = 1.174. Then take that and you'll see that 1333*1.174 = 1564.9, but the 1333 Mhz chip has a PR of 1500, which is modest, it should be called the model 1565?? Right?? But this is the exact opposite of what you're saying, so obviously it isnt linear. Besides, I read on their website that the numbers are based on a ton of benchmarks on each chip, they didnt do some tests with on chip and come up with a ratio, as you suggested. Also, many people don't understand that PR isnt relative to the PIV, but instead relative to current Athlon's performance.
"Desktop processors based on the "Palomino" core will be marketed as the AMD Athlon™ XP processor. As a way of communicating the performance improvements of the new AMD Athlon™ XP processor relative to the performance of the currently available AMD Athlon™ processor, AMD has developed a model numbering convention."
10-12-01, 02:16 PM #7
"The man in black fled across the desert, and the gunslinger followed."
well unfotunaly the mass stupid is the majority
some times you have to dumb somtehing down so the masses will bite
i dont see why they dont just do a tv add or something they can beam into the rooms of every stupid person.
the only people that are going to go to some werid location at 6 in the am are geeks that know the diffrece anyway.My Prime95 stats Huzza
1004.793 Years total
Noli turbare circulos meos!
i'm moving to Theory, Everything works there
' I want to be the unwobbling pivot at the center of an ever-revolving universe; I want to be still.'
10-12-01, 02:26 PM #8
am i missing something here? dont the pr ratings say 1500+ (note the plus after the number). they aren't saying that a 1.33 is EQUAL to a 1500, they're saying it's a 1500 or greater. thus there is NO SET MHZ ratio.
and yes, if you took the time to actually read what AMD said you'd see (as was already pointed out) that they're comparing it to the Athlon Tbirds, not to a P4. IMO these numbers are quite conservative when compared to p4's. i think you're kidding yourself if you think that 1 palomino clock cycle is equal to 1.125 P4 clock cycles.
please have a clue what you're talking about before flaming something. it's one thing to ask a question and state what you think for discussion. quite another to flame something you haven't even (obviously) taken the time to find out about.
having said that, i dont personally like the rating system, however i do think it's a very good idea by AMD considering most everyone doesn't know about the big differences in performance between the two chips at a given mhz rating.
10-12-01, 02:59 PM #9
Yes their mojo PR is formulated against the regular Athlon, but the marketing intention is to use it against the P4. It should be obvious that:
1. They're not trying to compete against their own products, what would be the point in that?
2. They know the mass marketing public this is aimed at are not going to read their PR White Paper.
I'll say it again, we all agree that, clock for clock, the XP is more efficient. You know, it does more work per clock cycle. But contrary to what has been posted here, they maintain in the White Paper that there is a relationship that is clock speed oriented. Specifically they state:
Processor Performance = (Work per Clock Cycle) x (Clock Speed)
So you could think of my relatively simplistic "ratio of 1.125" as being analagous to the addtional work per clock cycle at any given clock speed. Therefore, similar to the formula above, this should be a constant which can be applied to derive the PR as clock speed increases. But it doesn't because AMD martketeers have chosen to arbitrarily make every increase a nice round 100 PR points. That clearly invalidates any pretense that the final PR value follows some rigorous methodology. The tests might, but the results are being "adjusted" for that nice 100 point differential.
Plus I find all of this techie elitism sickening. Stop referring to those less knowedgible in this relatively insignificant field of knowlege like they are some lower form of life. I know some very intelligent people who are clueless about CPUs, but know a lot more than I do about significantly more important issues. A lot of this reminds me of that Saturday Night Live skit with the snot-nosed creepy Tech Support guy who berates anyone who has a problem.
10-12-01, 03:14 PM #10
still need to take into account the plus after the rating, they are saying it's above 1500 equivelant... that could be anywhere from 1501 to 3000 and beyond. if that's the case they dont have to give an exact number, they give a conservative number (yes it's nice and rounded to the nearest 100, so what?) and add the plus to it indicating it's an equivelant of that or higher.
i think NOT rounding it the way they do would make it seem like an exact equivelant. but by rounding out a nice even number and adding a "+" to the end, they've left it open.
for one thing, it'd be basically impossible to give an EXACT mhz equivelant comparing a p4 vs an athlon. so in the equation cl speed x work per cycle... the answer is most likely gonna be substantially over your 1.125 or whatever number. their chosing to give a fair approximation makes perfect sense, along with the room for error with the "+" afterwards.
granted, i haven't personally benched the two vs each other, but from reports and articles i've read, the AMD pr #'s are conservative.
10-12-01, 03:18 PM #11
by using the "+" afterward, you need to include that into your clock equivelant.
that would make your 1.125+ per cycle when compared.
10-12-01, 03:25 PM #12
I thought that the 1500+ is a model # not a rating.Hello:sn:
10-12-01, 04:22 PM #13
- Join Date
- Sep 2001
- Boulder, CO
it IS a model, but then again, my samsung 900p monitor is, well, 19 inches, whereas the 700p is 17 inches...
my toshiba drive model is SD640 or something like that, being that it's a 6x dvd, and 40x cdrom, etc.
it IS the model #, but it can also be indicitive of its speed.