Initial AMD Hammer will have a PR rating of 3400+. I'll take two.

[MadRat]

Rainsford-

I was showing that to compare to the Pentum4 with an Athlon you need to either leave a cushion in your numbers or use non-standard bus speeds.

Pentium4/400s should be compared to Athlons at 90-100fsb, although more accurately a 178fsb fits the comparison
Pentium4/533s should be compared to Athlons at 133fsb
Pentium4/667s should be compared to Athlons at 150-166fsb

You suggest that AMD should multiply the Pentium4 speed times .76 to get the speed at which their Athlon core compares. I refute that AMDs own formula points to .9 as the more accurate factor. The HAMMER family will need to be 20% more efficient per clock cycle in order to justify a factor of .76 in comparison to the Pentium4.

None of my numbers reflect comparison of true performance or IPC between the Pentium4 and Athlon families. Rather I am using AMDs statistical history to show the innaccuracy of thier own ratings. Don't get me wrong, I like AMD. I just do not like the logic behind their PR scheme.

 

[Adul]

i said it once and i will say it again. What baout intels new laptop cpu? It has a higher IPC and a lower clock thena P4. How will intel market this?

 

[Rainsford]

Originally posted by: MadRat
Rainsford-

I was showing that to compare to the Pentum4 with an Athlon you need to either leave a cushion in your numbers or use non-standard bus speeds.

Pentium4/400s should be compared to Athlons at 90-100fsb, although more accurately a 178fsb fits the comparison
Pentium4/533s should be compared to Athlons at 133fsb
Pentium4/667s should be compared to Athlons at 150-166fsb

You suggest that AMD should multiply the Pentium4 speed times .76 to get the speed at which their Athlon core compares. I refute that AMDs own formula points to .9 as the more accurate factor. The HAMMER family will need to be 20% more efficient per clock cycle in order to justify a factor of .76 in comparison to the Pentium4.

None of my numbers reflect comparison of true performance or IPC between the Pentium4 and Athlon families. Rather I am using AMDs statistical history to show the innaccuracy of thier own ratings. Don't get me wrong, I like AMD. I just do not like the logic behind their PR scheme.

Ok, I see what you're saying. I agree that the RP rating is a little odd, but I don't think AMD will just blindly use the current rating if it starts to be innacurate. And since I don't trust RP ratings or clock speeds, I don't care what they choose to call it. I for one am glad they have a RP rating to allow them to compete in the eyes of consumers with the P4.

 

[Woodchuck2000]

I still can't believe that people bothered so sit down any type pages of math to try and disprove marketing.
As I see it, the facts are as follows:

1)The P4 is able to reach much higher clock speeds than the AthlonXP.
2)The Athlon XP performs better, clock for clock, than the P4.
3)The average consumer buys on MHz.
4)The PR rating makes the XP sound like it's performance-equivalent P4.
5)The man-on-the-street therefore is more likely to but an XP.
6)And therefore there is more competition.
7)Marketing is not an exact science and so cannot be disproved.
8)It's far from ideal, but AMD have been conservative and fair in their ratings.
9)It doesn't affect anyone on this forum in any way since you're all sufficiently intelligent to buy based on performance rather than marketing. so why do you care?

As far as the Hammer goes, I think there were some rough benchmarks floating around that said the alpha silicon was performing on a par with a P4 1.6
Assuming performance and clockspeed scale linearly, getting a 3400+ would need maybe 1.8GHz, which is fairly plausible. That said, I'm waiting for the benchmarks. It'll be interesting to see if AMD keep the PR system fair or if they try to capitalise on the "64-bit" factor.

 

[sandorski]

Originally posted by: Diable

Originally posted by: sandorski

Anyway, it always shocks me how some are so vehement about the alleged "deception" of AMD's PR rating. Were you equally vehement when Intel kept the MHZ rating from PIII to P4? Was/is a P4 1400mhz actually faster than a PIII at 1000mhz?

A 2.53GHz P4 actually runs at 2.53GHz(19x133mhz) and a 1.2GHz P3 its actually runs at 1.2GHz(9X133mhz), so its indeed proper for Intel to keep the same mhz rating for the P4's that they used with the P3's, plus math hasn't changed .

Your commit about the P3 beating a P4 is so dated that it make you look stupid. When the 1.4GHz P4 was released no software on the market was SSE2 optimized so yes a 1GHz P3 could beat a 1.4GHz P4. But now that most new cpu intensive apps(3dStudio, Photoshop etc.)are SSE2 optimized that same P4 will spank that P3.

Yes, Intel is correct, in that a P4 running at X mhz actually is running at X mhz, but as the benchmarks have shown time and time again, P4 mhz |= PIII mhz or Athlon mhz.

Any links showing a P4 1.4ghz beating a PIII 1ghz? 2 apps doesn't equate as "most".

 

[ElFenix]

Originally posted by: Wingznut PEZ
FUD... Probably the single worst acronym on the net...

Fear... Nope, I'm certainly not in fear of anything I've said.
Uncertainty... I'm certain of everything I said above.
Doubt... Nope, I have no doubt in what I said.


So, what does FUD have to do with anything???

And what did I say that wasn't the absolute truth?

you're trying to discredit amd's model rating system by mentioning that arthur andersen audited the benchmarks. as we all know arthur andersen audited enron's books, and every knows what happened to them. you put into your own post the possibility that arther andersen and AMD are pulling a fast one like enron did ("I'm not saying that there's necessarily any inconsistancies going on... But who's going to trust Arthur Andersen, about anything, any more?"), but the actual benchmarks from sites like the one you're posting on serve to show the current validity of amd's naming convention. i don't need to trust arthur andersen, i have anand.

 

[MadRat]

Don't get me wrong, I think AMD needs to use the PR system.

But I also don't think anything backed by Arthur-Andersen is credible. If they are going to stand behind the numbering system then it needs to scale linearly down to zero. If they want to scale back their ratings to compensate for the inneficiencies of the front-side bus at higher speeds then that would be an excellent opportunity for an AMD ad slogan to say, "We at AMD are above playing the simple MHz game, lets talk true performance!" If AMD feels that the magic factor is .76 then they need to express this, otherwise they need to stick to the .9 factor used by the original Palomino core and not inch it to absurdity with each new speed grade!! The truth only hurts the soft-skinned ninnies who cannot accept reality; AMD has made a good living out of avoiding lies and the present PR system tears at this established reputation. I've already demonstrated that they can compare the Pentium4 to the Athlon family more fairly by scaling the front-side bus to match the various Pentium4 speed grades. Not only can they scale to match the Pentium4, they can do it on the same terms of Intel by using 100MHz, 133MHz, and 166MHz front-side buses.

AMD needs to think really hard about rectifying the logic of their numbering system.

 

[bret]

oh wow that is fast...i leave for 8 months and look how much technology grows...i need to catch up quick nice post.

 

[ElFenix]

Originally posted by: MadRat
Don't get me wrong, I think AMD needs to use the PR system.

But I also don't think anything backed by Arthur-Andersen is credible. If they are going to stand behind the numbering system then it needs to scale linearly down to zero. If they want to scale back their ratings to compensate for the inneficiencies of the front-side bus at higher speeds then that would be an excellent opportunity for an AMD ad slogan to say, "We at AMD are above playing the simple MHz game, lets talk true performance!" If AMD feels that the magic factor is .76 then they need to express this, otherwise they need to stick to the .9 factor used by the original Palomino core and not inch it to absurdity with each new speed grade!! The truth only hurts the soft-skinned ninnies who cannot accept reality; AMD has made a good living out of avoiding lies and the present PR system tears at this established reputation. I've already demonstrated that they can compare the Pentium4 to the Athlon family more fairly by scaling the front-side bus to match the various Pentium4 speed grades. Not only can they scale to match the Pentium4, they can do it on the same terms of Intel by using 100MHz, 133MHz, and 166MHz front-side buses.

AMD needs to think really hard about rectifying the logic of their numbering system.

and yet the numbering system still stands as a pretty good indicator of performance vis-a-vis intel's number system (what, adding 10% MHz doesn't increase processing power 10%?!? intel should change their scheme!

) regardless of what company audited amd's numbers.

 

[SexyK]

For the record, when people are working on all this math, almost every rumor that i've heard (besides the one mentioned in this thread) has the 3400+ Hammer running at a true 2.0GHz not 2.6GHz. 2.6GHz would be the speed of an Athlon XP (palamino) 3400+. I assume the Hammer will perform clock for clock better than the Pally, so it's safe to assume there will be a shift in the PR system. Anyway, aside from all that, I'll say that I really don't care what AMD does, it's their business and they've done fine the last couple years. For me personally, the whole system seems to defy logic (as Wingz and other have said) so I personally choose not to like it, and to hold AMD in a lower light because of it. I don't care what you guys want to do, all I can say is how it strikes me, and I'm sure others.

Part of the problem for me i guess is that back in the day AMD used a PR system, but then once it looked like they were going to beat Intel to 1GHz, all of a sudden the pure MHz meant something again, but oops, then when Intel's new arcitecture comes out, can't compete in clock speed again, so it's time to bring back the PR system. Their waffling on the issue only discredits them more in my book.

Again, you can think whatever you want, and I really won't mind, but you probably won't be able to convince me all of a sudden that the PR system is completely legit, so don't bother screaming at me about competition and fairness. I like AMD, I like them being around, and I think they make a good product... However, I personally prefer Intel chips and I don't like the PR system, I don't think that's to terrible of me.

Kramer

 

[Rand]

It makes no sense that the delta between PR and actual speed should get larger, with zero core improvements.

I'll agree with you on that one, the Model Rating methodology is clearly flawed. But while flawed, it's still for the moast part quite accurate overall.

Like SexyK stated earlier, if you follow this PR model, then you come to the conclusion that a Willamette 1300mhz is just as fast as a 1300mhz Tbird. How is that not completely flawed???

It is of course, but I don't see how that has anything to do with this. The ThunderBird core was not PR rated.
It's no more flawed then comparing a Willamette 1.3GHz to a Tualatin PIII 1.3GHz.

The other reason is benchmarks. Who's to say what benchmarks are used? (Well, in this case, obviously AMD is the one who says.)

Take 3D rendering, for example. If you use Maya, then the AMD 1800mhz's PR seems to fit. But if you use Adobe or Lightwave, then the 1800mhz is actually slower than a 1600mhz Willamette!

In audio... Use LAME, and the AMD competes. Use MusicMatch or MP3 Maker, and it falls short of it's PR.

Those are only two examples. But I can find an example like that for pretty much every benchmark.

For ANY processor there will always be benchmarks they are unusually good at. Hell, I can pick out a benchmark in which the K6-2 can perform quite well compared to Willamette 1.5GHz.
It's not even difficult to find benchmarks in which the VIA C3 performs extremely well and can beat out any P6 core processor.
I can pick out benchmarks in which the Willamette 1.3GHz walks away from an AthlonXP 2200+. I can also pick out benchmarks in which the ThunderBird 1.2GHz can outperform a Northwood 2.53GHz.

If you look hard enough you can find a benchmark to "prove" almost anything you want.
Outlying benchmars that are an extreme rarity are not the focus of a Model Rating scheme aimed at the general public though, we need to find an 'average' performance on most benchmarks and base it off of that.
There will always be applications that under/overperform on any given performance scale.

The day that AMD's PR rating runs into the brick wall, is the day they release the Athlon 1933mhz. This will be called a "2400+", but it will not live up to a P4-2.4b in the majority of the latest benchmarks.

At the current pace I expect it to become unrealistic at about the time of a hypothetical 2500+, but that's beside the point. What matters now is that it's accurate now, or at least the overwhelming consensus views it as accurate.
When it becomes inaccurate then I'll proclaim against AMD, though I would strongly hope the modify the PR metric prior to then.

Again, you can think whatever you want, and I really won't mind, but you probably won't be able to convince me all of a sudden that the PR system is completely legit, so don't bother screaming at me about competition and fairness.

It isnt complately legitimate, it clear that the differential between a PR rated AthlonXP and a P4 should not get higher at increasing MHz/PR #'s.
The PR rating is far from ideal, and I fully agree that AMD should have integrated a better system.
I do think there NEEDS to be some PR scale however, as MHz is clearly flawed itself. It's obvious to entirely different microarchitectures are not going to scale and perform identically at any given clockspeed.
AMD's PR scale is not the ideal scale, but as long as it's realistic I consider it "good enough"... or at least no worse then judging on pure MHz.

Thankfully the PR scale was initially set so extremely conservatively against the Willamete so as to ensure they wouldnt have to modify it too quickly for the Northwood. Otherwise it would have been completely inaccurate long ago.

I'm not a proponent of the PR scale because I believe it's the perfect measurement of performance, I'm a proponent of it because at present I believe it compares performance between differing architectures better then MHz does.
When that ceases to be true I will speak against the PR scale, but for now it's the best of a poor set of alternatives IMHO.

I like AMD, I like them being around, and I think they make a good product... However, I personally prefer Intel chips

In that I agree with you, I'm running an Intel processor now and my next upgrade will almost assuredly be a Northwood/533. My last processor was also Intel.
Should AMD become a better option for me, I won't hesitate to switch however... providing the potential platforms are decent. Poor platform support was the reason I sold my last AMD processor (a K6-2).

 

[MadRat]

I spent a good hour plotting out the Athlon XP's at various front-side buses and multipliers using the .81 factor. If the 2200+'s factor of .81 is the magic factor for future Athlons (non-Hammer) then it would have made more sense to use that factor in the original Athlon XPs. Instead of a 1.33GHz 1500+ we'd of had a 1.25GHz 1500+! AMD sold their equipment way short by using the .9 factor early on in the Palomino's development. Remember that AMD did release 1.0GHz/133fsb Palominoes as MP 1.0GHz chips, so plotting back to 1.0GHz is not unreasonable.

Lets use a factor of .81 to match 100fsb Athlon XPs against Pentium4/400 and Pentium4/533 processors:

1200+ = roughly 982MHz = 10 x 100fsb (1000) NO COMPARABLE INTEL PENTIUM 4 PROCESSOR
1266+ = roughly 1036MHz = 10.5 x 100fsb (1050) NO COMPARABLE INTEL PENTIUM 4 PROCESSOR
1350+ = roughly 1105MHz = 11 x 100fsb (1100) COMPARE TO P4-1.3/400
1400+ = roughly 1145MHz = 11.5 x 100fsb (1150) COMPARE TO P4-1.4/400
1466+ = roughly 1200MHz = 12 x 100fsb (1200) COMPARE TO P4-1.4/400
1533+ = roughly 1254MHz = 12.5 x 100fsb (1250) COMPARE TO P4-1.5/400
1600+ = roughly 1309MHz = 13 x 100fsb (1300) COMPARE TO P4-1.6/400 or P4-1.6A/400
1650+ = roughly 1350MHz = 13.5 x 100fsb (1350) COMPARE TO P4-1.6/400 or P4-1.6A/400
1733+ = roughly 1417MHz = 14 x 100fsb (1400) COMPARE TO P4-1.7/400 or P4-1.6A/400
1766+ = roughly 1445MHz = 14.5 x 100fsb (1450) COMPARE TO P4-1.7/400 or P4-1.6A/400
1833+ = roughly 1500MHz = 15 x 100fsb 1500) COMPARE TO P4-1.8/400 or P4-1.8A/400
1900+ = roughly 1554MHz = 15.5 x 100fsb (1550) COMPARE TO P4-1.9/400 or P4-1.8A/400
1966+ = roughly 1609MHz = 16 x 100fsb, (1600) COMPARE TO P4-1.9/400 or P4-1.8A/400
2000+ = roughly 1636MHz = 16.5 x 100fsb (1650) COMPARE TO P4-2.0/400
2066+ = roughly 1690MHz = 17 x 100fsb (1700) COMPARE TO P4-2.0/400
2133+ = roughly 1745MHz = 17.5 x 100fsb (1750) COMPARE TO P4-2.0/400
2200+ = roughly 1800MHz = 18 x 100fsb (1800) COMPARE TO P4-2.2/400
2266+ = roughly 1850MHz = 18.5 x 100fsb (1850) COMPARE TO P4-2.2/400
2333+ = roughly 1908MHz = 19 x 100fsb (1900) COMPARE TO P4-2.2/400 or P4-2.26/533
2366+ = roughly 1945MHz = 19.5 x 100fsb (1950) COMPARE TO P4-2.2/400 or P4-2.26/533
2450+ = roughly 2005MHz = 20 x 100fsb (2000) COMPARE TO P4-2.4/400 or P4-2.4A/533
2500+ = roughly 2045MHz = 20.5 x 100fsb (2050) COMPARE TO P4-2.4/400 or P4-2.4A/533

Lets use a factor of .81 to match 133fsb Athlon XPs against Pentium4/400, Pentium4/533, and Pentium4/667 processors:

1200+ = roughly 982MHz = 7.5 x 133fsb (1000) NO COMPARABLE INTEL PENTIUM 4 PROCESSOR
1300+ = roughly 1064MHz = 8 x 133fsb (1066) COMPARE TO P4-1.3/400
1366+ = roughly 1118MHz = 8.5 x 133fsb (1133) COMPARE TO P4-1.3/400
1466+ = roughly 1200MHz = 9 x 133fsb (1200) COMPARE TO P4-1.4/400
1550+ = roughly 1268MHz = 9.5 x 133fsb (1266) COMPARE TO P4-1.5/400
1633+ = roughly 1336MHz = 10 x 133fsb (1333) COMPARE TO P4-1.6/400 or P4-1.6A/400
1700+ = roughly 1390MHz = 10.5 x 133fsb (1400) COMPARE TO P4-1.7/400 or P4-1.6A/400
1800+ = roughly 1472MHz = 11 x 133fsb (1466) COMPARE TO P4-1.8/400 or P4-1.8A/400
1866+ = roughly 1526MHz = 11.5 x 133fsb (1533) COMPARE TO P4-1.8/400 or P4-1.8A/400
1966+ = roughly 1609MHz = 12 x 133fsb (1600) COMPARE TO P4-1.9/400 or P4-1.8A/400
2033+ = roughly 1663MHz = 12.5 x 133fsb (1666) COMPARE TO P4-2.0/400
2100+ = roughly 1718MHz = 13 x 133fsb (1733) COMPARE TO P4-2.0/400
2200+ = roughly 1800MHz = 13.5 x 133fsb (1800) COMPARE TO P4-2.2/400
2266+ = roughly 1854MHz = 14 x 133fsb (1866) COMPARE TO P4-2.2/400 or P4-2.26/533
2366+ = roughly 1936MHz = 14.5 x 133fsb (1933) COMPARE TO P4-2.2/400 or P4-2.26/533
2433+ = roughly 1991MHz = 15 x 133fsb (2000) COMPARE TO P4-2.4/400 or P4-2.4A/533
2500+ = roughly 2045MHz = 15.5 x 133fsb (2066) COMPARE TO P4-2.4/400 or P4-2.4A/533
2600+ = roughly 2127MHz = 16 x 133fsb (2133) COMPARE TO P4-2.53/533
2700+ = roughly 2209MHz = 16.5 x 133fsb (2200) COMPARE TO P4-2.66/533
2766+ = roughly 2263MHz = 17 x 133fsb (2266) COMPARE TO P4-2.66/533
2850+ = roughly 2332MHz = 17.5 x 133fsb (2333) COMPARE TO P4-2.8/533
2933+ = roughly 2400MHz = 18 x 133fsb (2400) COMPARE TO P4-2.93/533
3000+ = roughly 2454MHz = 18.5 x 133fsb (2466) COMPARE TO P4-2.93/533 or P4-3.0/667
3100+ = roughly 2536MHz = 19 x 133fsb (2533) COMPARE TO P4-3.0/667
3166+ = roughly 2590MHz = 19.5 x 133fsb (2600) COMPARE TO P4-3.0/667
3250+ = roughly 2659MHz = 20 x 133fsb (2666) COMPARE TO P4-3.0/667
3333+ = roughly 2727MHz = 20.5 x 133fsb (2733) COMPARE TO P4-3.33/667

Lets use a factor of .81 to match 166fsb Athlon XPs against Pentium4/400, Pentium4/533 and Pentium4/667 processors:

2033+ = roughly 1663MHz = 10 x 166fsb (1666) COMPARE TO P4-2.0/400
2133+ = roughly 1745MHz = 10.5 x 166fsb (1743) COMPARE TO P4-2.0/400
2233+ = roughly 1827MHz = 11 x 166fsb (1826) COMPARE TO P4-2.2/400
2333+ = roughly 1908MHz = 11.5 x 166fsb (1909) COMPARE TO P4-2.2/400 or P4-2.26/533
2433+ = roughly 1991MHz = 12 x 166fsb (1992) COMPARE TO P4-2.4/400 or P4-2.4A/533
2533+ = roughly 2072MHz = 12.5 x 166fsb (2075) COMPARE TO P4-2.53/533
2600+ = roughly 2127MHz = 13 x 166fsb (2133) COMPARE TO P4-2.53/533
2700+ = roughly 2209MHz = 13.5 x 166fsb (2200) COMPARE TO P4-2.66/533
2833+ = roughly 2317MHz = 14 x 166fsb (2324) COMPARE TO P4-2.8/533
2933+ = roughly 2400MHz = 14.5 x 166fsb (2407) COMPARE TO P4-2.93/533
3000+ = roughly 2454MHz = 15 x 166fsb (2490) COMPARE TO P4-2.93/533 or P4-3.0/667
3150+ = roughly 2577MHz = 15.5 x 166fsb (2573) COMPARE TO P4-2.93/533 or P4-3.0/667
3250+ = roughly 2659MHz = 16 x 166fsb (2656) COMPARE TO P4-2.93/533 or P4-3.0/667
3333+ = roughly 2727MHz = 16.5 x 166fsb (2739) COMPARE TO P4-3.33/667
3466+ = roughly 2835MHz = 17 x 166fsb (2822) COMPARE TO P4-3.33/667
3550+ = roughly 2905MHz = 17.5 x 166fsb (2905) COMPARE TO P4-3.5/667
3650+ = roughly 2986MHz = 18 x 166fsb (2988) COMPARE TO P4-3.5/667
3750+ = roughly 3068MHz = 18.5 x 166fsb (3072) COMPARE TO P4-3.66/667
3850+ = roughly 3150MHz = 19 x 166fsb (3154) COMPARE TO P4-3.83/667
3950+ = roughly 3232MHz = 19.5 x 166fsb (3237) COMPARE TO P4-3.83/667
4066+ = roughly 3327MHz = 20 x 166fsb (3320) COMPARE TO P4-4.0/667
4166+ = roughly 3409MHz = 20.5 x 166fsb (3400) COMPARE TO P4-4.0/667

Granted we have no idea how the Athlon XP core will stack up against Prescott's 667fsb, but the 166fsb is sufficient to compete against it at the low end. AMD could have produced Athlon XP's all the way down to 1200+ to 1350+ as competition with the early Pentium 4/1.3's. The core would have been running alot cooler in comparison using this .81 factor right off the bat rather than the .9 factor at the low end. The original XP 1500+ ran at 1.33GHz on the 133fsb but could have been released at a 100fsb and only 1250MHz, again at a much less power draw and heat output. Its ridiculous to think that using a fixed constant would produce an XP 1633+ out of that same 1.33GHz/133fsb Palomino!! AMD really dropped the ball on that one.

Notice on the chart that not only is AMD able to target a wide variety of speed grades to fit the yields of their processors, but most of their processors end up at speed grades that Pentium4's cannot compete on a fair level. Consider the Pentium 4/1.6A which ends up competing against processors all the way up to a 1766+ rating. AMD could have also tuned their processors to fit the needs of small retailers that agressively marketing to the customer base of Intel-aligned OEMs like Dell. Hate to say it, but AMD may have dropped the ball big time on this by being too conservative. :Q

Lets look at what models compare to Intel Pentium4 processors:

No Comparison List: 1200+/10x100, 1266+/10.5x100, 1200+/7.5x133

Williamette
P4-1.3/400 vs. 1350+/11x100, 1300+/8x133, 1366+/8.5x133
P4-1.4/400 vs. 1400+/11.5x100, 1466+/12x100, 1466+/9x133
P4-1.5/400 vs. 1533+/12.5x100, 1550+/9.5x133
P4-1.6/400 vs. 1600+/13x100, 1650+/13.5x100, 1633+/10x133
P4-1.7/400 vs. 1733+/14x100, 1766+/14.5x100, 1700+/10.5x133
P4-1.8/400 vs. 1833+/15x100, 1800+/11x133, 1866+/11.5x133
P4-1.9/400 vs. 1900+/15.5x100, 1966+/16x100, 1966+/12x133
P4-2.0/400 vs. 2000+/16.5x100, 2066+/17x100, 2133+/17.5x100, 2033+/12.5x133, 2100+/13x133, 2033+/x166, 2133+/x166

Northwood/400s
P4-1.6A/400 vs. 1600+/13x100, 1650+/13.5x100, 1733+/14x100, 1766+/14.5x100, 1633+/10x133, 1700+/10.5x133
P4-1.8A/400 vs. 1833+/15x100, 1900+/15.5x100, 1966+/16x100, 1800+/11x133, 1866+/11.5x133, 1966+/12x133
P4-2.0A/400 vs. 2000+/16.5x100, 2066+/17x100, 2133+/17.5x100, 2033+/12.5x133, 2100+/13x133, 2033+/10x166, 2133+/10.5x166
P4-2.2/400 vs. 2200+/18x100, 2266+/18.5x100, 2333+/19x100, 2366+/19.5x100, 2200+/13.5x133, 2266+/14x133. 2366+/14.5x133, 2233+/11x166, 2333+/11.5x166
P4-2.4A/400 vs. 2450+/20x100, 2500+/20.5x100, 2433+/15x133, 2500+/15.5x133, 2433+/12x166

Northwood/533s
P4-2.26/533 vs. 2266+/18.5x100, 2333+/19x100, 2266+/14x133. 2366+/14.5x133, 2333+/11.5x166
P4-2.4B/533 vs. 2450+/20x100, 2500+/20.5x100, 2433+/15x133, 2500+/15.5x133, 2433+/12x166
P4-2.53/533 vs. 2600+/16x133, 2533+/12.5x166, 2600+/13x166
P4-2.66/533 vs. 2700+/16.5x133, 2766+/17x133, 2700+/13.5x166
P4-2.8/533 vs. 2850+/17.5x133, 2833+/14x166
P4-2.93/533 vs. 2933+/18x133, 3000+/18.5x133, 2933+/14.5x166, 3000+/15x166

Prescott/667s
P4-3.0/667 vs. 3000+/18.5x133, 3100+/19x133. 3166+/19.5x133, 3250+/20x133, 3000+/15x166, 3150+/15.5x166, 3250+/16x166
P4-3.33/667 vs. 3333+/20.5x133, 3333+16.5/x166, 3466+/17x166
P4-3.5/667 vs. 3550+/17.5x166, 3650+/18x166
P4-3.66/667 vs. 3750+/18.5x166
P4-3.83/667 vs. 3850+/19x166, 3950+/19.5x166
P4-4.0/667 vs. 4066+/20x166, 4166+/20.5x166

AMD's Thoroughbred, launched at 1800MHz/133fsb would now be comparable to the P4-2.2 as it is now, and they would of ruled the low end before that time. One drawback of using the .81 factor across the board is that AMD releases beyond 1800MHz/133fsb will be hard pressed for yields. If the rumoured P4-2.8/533 does come out this fall then AMD will still need to get their .13 process to scale to 2.33GHz at 133fsb just to compete on a level playing field, a whopping 533MHz jump in six months!

Edit: Fixed charts, added P4 comparisons to XPs.

 

[Piano Man]

My head is starting to hurt with all of these numbers.

 

[Rand]

Does anybody know of any processors that have ever been PR Rated LOWER then their physical clockspeed?

(Besides the AMD 5x86, I know of that)

 

[RanDum72]

Does anybody know of any processors that have ever been PR Rated LOWER then their physical clockspeed?

I think if you do that to Cyrix C3's, they would be lower.

 

[sandorski]

Originally posted by: Rand
Does anybody know of any processors that have ever been PR Rated LOWER then their physical clockspeed?

(Besides the AMD 5x86, I know of that)

Nope, although IMO(not that it matters ) the P4 should have been.

 

[MadRat]

Can anyone hear honestly say they'd of been disappointed with the Pentium4 if it was PR'd lower? If the 2.53GHz P4 was PR2200 then I'd of been perfectly happy with it knowing it was equivalent to a 4GHz P!!! when doing some types of encoding.

I totally went back through those numbers in my last post and found alot of little typographical erros... reads easier now, too. I went back and made a chart of P4's and equivalent XPs to make it easier to compare the two on a .81 factor.

I'm wlling to bet that in most cases these linear-factor based revisions will be very accurate in benchmarking. Too bad we don't have the time to benchmark every speed grade (5x-20.5) of the Palomino and Thouroughbred cores.

 

[Smithy18]

AMD/INTEL Geeze why do we keep getting into this. People sould know better than to start bashing company's hardware...
I think my companys better....
NO NO mine is...

I have an Idea lets just have one big chip company so that they can control all of the cpu market and charge us whatever they want for their products. Think $$ on Intel is bad, immagine what it would be if AMD wanst around to keep them competitive. You better hope that Hammer makes a big impression or stuff is going to start falling apart.
Just my 2 cents

Regards
Stephen

 

[Valinos]

Originally posted by: Smithy18
AMD/INTEL Geeze why do we keep getting into this. People sould know better than to start bashing company's hardware...
I think my companys better....
NO NO mine is...

I have an Idea lets just have one big chip company so that they can control all of the cpu market and charge us whatever they want for their products. Think $$ on Intel is bad, immagine what it would be if AMD wanst around to keep them competitive. You better hope that Hammer makes a big impression or stuff is going to start falling apart.
Just my 2 cents

Regards
Stephen

Ahh, the good ole days of $1200 for flagship Pentiums $300 used to be reasonable...now I think $600 is insane and $100 is reasonable.

 

[Sunner]

1)The P4 is able to reach much higher clock speeds than the AthlonXP.
2)The Athlon XP performs better, clock for clock, than the P4.
3)The average consumer buys on MHz.
4)The PR rating makes the XP sound like it's performance-equivalent P4.
5)The man-on-the-street therefore is more likely to but an XP.
6)And therefore there is more competition.
7)Marketing is not an exact science and so cannot be disproved.
8)It's far from ideal, but AMD have been conservative and fair in their ratings.
9)It doesn't affect anyone on this forum in any way since you're all sufficiently intelligent to buy based on performance rather than marketing. so why do you care?

I agree 100% with this.

 

[Zugzwang152]

Agreed, but back on topic, I doubt Hammer can get a PR rating of 3400+ to begin with. Their test sample debuted at only 800 Mhz correct? Plus I think moving directly up to 2.6Ghz would hurt their sales. I want to see Hammer at 3000+, and closer in Mhz to the last Barton or T-bred that comes out. At that point, we'll be able to see benchies showing just how much better Hammer is than equivalently clocked t-breds, or a p4 near 3Ghz.

 

[MadRat]

The correllation of the Hsmmer at this point to the Pentium4 is non-sensical. That would be like comparing Banias to the Palomino, one being in pre-production, the other core long since into past its prime. There is probably some significance to the 800MHz speed of the samples, like perhaps it matched the internal clock speed of its memory transfers. It doesn't mean that the Hammer cannot scale. That was A0 silicon, not a production run. It was only released to potential clients of the production chips so that they could troubleshoot x86-64 software engineering. Besides that its not a consumer chip, its a professional workstation chip. The price will be beyond most consumers.

 

[Zugzwang152]

Is that to say the only offering for consumer-level products from AMD from now on will be a T-bred or Barton based Athlon? That is ridiculous. We've already gone over how AMD doesn't have the resources to maintain 3 lines of desktop CPU's. When Hammer comes out, Barton/T-bred will be relegated to the Duron name, correct? What then, will take Athlon's place? None other than Sledgehammer, which therefore must be priced accordingly (at LEAST within P4 ranges). Perhaps they might reduce features and cache, but they have to come out with something to challenge P4, or they're desktop cpu division is toasted.

 

[SexyK]

Originally posted by: Zugzwang152
Is that to say the only offering for consumer-level products from AMD from now on will be a T-bred or Barton based Athlon? That is ridiculous. We've already gone over how AMD doesn't have the resources to maintain 3 lines of desktop CPU's. When Hammer comes out, Barton/T-bred will be relegated to the Duron name, correct? What then, will take Athlon's place? None other than Sledgehammer, which therefore must be priced accordingly (at LEAST within P4 ranges). Perhaps they might reduce features and cache, but they have to come out with something to challenge P4, or they're desktop cpu division is toasted.

That's wrong, Sledgehammer (Opteron) will almost definiely be beyond the price range of 99% of the people on this forum. Clawhammer will take over the Athlon name, and most likely have a nearly identical pricing structure to the P4 (~$600 at the extreme high end on down). I think it's unrealistic for people to expect AMD to keep selling chips for $59-$200. If you want them to be around for more than the next year or so, and like their technology so much, then you should be willing to pay at a minumum $350-$400 for their newest chip. They are losing money selling so low as it is, we all know that. The whole purpose of selling low was to gain market share and a user base, now that they have a throng of followers, it's time to start making money on their chips, if they can't do that, well, they're in trouble.

Kramer

 

[MadRat]

I'm thinking that AMD can sell their chips at those low margins because of their ability to build so many per wafer.

The real money losers would likely be P4-Celerons, not Durons or Athlons.