Is 2600+ the right model number for the new Athlon XP?
Given suggestions from people on coming up with a standardized benchmark, I looked at the suite of tests from Tom's Hardware to see what one kind of average might look like. I looked at how three versions of the Athlon have performed, and how conservative or aggressive the model number is. I looked at the 1600+, the 2200+, and the 2600+ performed relative to the Pentium 4 in each benchmark, and adjusted the model number to the fastest Pentium 4 that it was able to outperform (some guessing was required). Since the 1600+ and 2200+ fall within the megahertz range of Northwoods on the slower 400MHz bus, for the sake of comparison, the 400MHz parts were used. However, I thought the 2600+ should be compared against the faster 533MHz front side bus Northwoods, since the Pentium 4 has pretty much transitioned to the faster bus by the 2600 range.
Quake III 640x480 "Demo001"
- Athlon XP 1600+ -> Same
- Athlon XP 2200+ -> 2000+
- Athlon XP 2600+ -> 2200+
Quake III 1024x768 "NV15Demo"
- Athlon XP 1600+ -> 1800+
- Athlon XP 2200+ -> Same
- Athlon XP 2600+ -> 2400+
3DMark 2001 SE
- Athlon XP 1600+ -> 1800+
- Athlon XP 2200+ -> Same
- Athlon XP 2600+ -> 2300+
Comanche 4
- Athlon XP 1600+ -> Same
- Athlon XP 2200+ -> 2100+
- Athlon XP 2600+ -> 2300+
Lame MP3 3.92
- Athlon XP 1600+ -> 1800+
- Athlon XP 2200+ -> Same
- Athlon XP 2600+ -> Same
MP3 Movie Maker Platinum
- Athlon XP 1600+ -> 1400+
- Athlon XP 2200+ -> 1600+
- Athlon XP 2600+ -> 2000+
Xmpeg 4.5 und Divx 5.02 Pro
- Athlon XP 1600+ -> Same
- Athlon XP 2200+ -> 2000+
- Athlon XP 2600+ -> 2300+
Pinnacle Studio 7
- Athlon XP 1600+ -> Same
- Athlon XP 2200+ -> 2000+
- Athlon XP 2600+ -> 2300+
SysMark 2002 Office
- Athlon XP 1600+ -> Same
- Athlon XP 2200+ -> 2000+
- Athlon XP 2600+ -> 2100+
SysMark 2002 Content Creation
- Athlon XP 1600+ -> 1400+
- Athlon XP 2200+ -> 1600+
- Athlon XP 2600+ -> 1800+
Winace 2.2
- Athlon XP 1600+ -> Same
- Athlon XP 2200+ -> Same
- Athlon XP 2600+ -> 2300+
Lightwave 7.5
- Athlon XP 1600+ -> 1200+
- Athlon XP 2200+ -> 1500+
- Athlon XP 2600+ -> 1800+
Cinema 4D XL 7.303
- Athlon XP 1600+ -> 1800+
- Athlon XP 2200+ -> 2500+
- Athlon XP 2600+ -> 2800+
3D Studio Max 4.26
- Athlon XP 1600+ -> 1500+
- Athlon XP 2200+ -> 2000+
- Athlon XP 2600+ -> 2300+
Of course, some guesswork had to be involved, but from averaging the results, I can make the following conclusions.
If the Athlon XP 1600+ were based on the way it performed with respect to the Pentium 4 Northwood with 400MHz FSB on the applications and benchmarks from the Tom's Hardware suite, then on average, the model number fits it exactly. It is neither conservative nor aggressive in this case.
If the Athlon XP 2200+ were based on the way it performed with respect to the Pentium 4 with 400MHz FSB on the applications and benchmarks from the Tom's Hardware suite, then on average, the model number is slightly aggressive, but 2200+ would work as long as you round to the nearest 100+.
However, If the Athlon XP 2600+ were based on the way it performed with respect to the Pentium 4 with 533MHz FSB on the applications and benchmarks from the Tom's Hardware suite, then on average, the model number is far too aggressive, and should be decreased to around 2300+.
Well, obviously, that doesn't seem right that a 1.8GHz Athlon XP performs close enough to be called a 2200+, but a 2.13GHz Athlon XP only performs enough to be called a 2300+. However, Intel has released newer, faster front side bus, and it's obvious that AMD hasn't figured this in to their ratings system, even though the performance story is very different when compared against Intel's faster parts.
Either that, or AMD has changed the model ratings from comparing with RDRAM based systems, to comparing with DDR. Other reviews seem to indicate the 2600+ having a slight edge over the 2.53GHz Pentium 4 when using DDR. But again, it depends on the applications being used. Tom's Hardware has only one set of benchmarks out of many.
Either way, since AMD's model numbers are as yet undefined, any argument against them is fairly moot. The only real argument is that in the current benchmarks, the 2600+ doesn't outperform the 2.53GHz Pentium 4 as well as it should. AMD's newer model number is definitely less conservative than it was before, and this is despite some compensation already used in the model number formula.
It will be interesting to see how the 2.8GHz Pentium 4 compares against AMD's upcoming Athlon XP 2800+ that they plan to release using their faster bus interface. Given the trend in scalability, it seems that AMD needs the faster front side bus to maintain any credibility, since a number of reviewers have already begun to notice AMD's model number situation. Industry experts such as Paul DeMone and Dean Kent have also noticed this trend.
My suggestion would be for AMD to define their basis for model numbers, so at least there is a credible source. Right now, it just looks arbitrary, which is going to be the main source of any negative PR on the issue.
Just my honest opinion.
wbmw |
