There’s a bunch of reviews about the 133Mhz PIVs out there; I’m sure you can find them.
You’ll see the XP getting consistently beaten; sometimes by a little, sometimes by a lot.
That in and of itself is unimportant. It’s New vs. Old.
What is important is whether or not you’ll get the same general results when the contest becomes New vs. New.
The question is not “Does a new PIV beat an old Athlon XP?” but “Can it beat a new Throughbred?”
Bargain My Butt
It seemed to be almost a requirement in these articles to say something like:
“Well, yes, the PIV does better at everything, but it’s so much more expensive than the Athlon XP.”
Well, that’s true if you don’t overclock.
If you do, it’s a whole ‘nother story, and one that’s changing. The LSNs are doing better than they did when first introduced.
If you scan our CPU database for the newest CPUs (hint: these will generally be those with codes starting with “320”), you’ll find that people are hitting 2.4GHz-2.5GHz with little or no voltage increase, usually with air, sometimes even with the stock Intel cooler. Pretty consistently. Those who haven’t either have other bottlenecks, or just didn’t try.
In short, just as good (actually, probably a little better due to a higher FSB) as those “expensive processors.”
So we’re not talking about $500 or $600 processors being the competition, we’re talking about a $130 processor.
The only AMD cost benefit right now (and it’s a very real one for many) is that Intel’s $130 will also need a new mobo. This effectively doubles the replacement cost to someone currently using an AMD platform and expecting to pop in a TBred one of these days.
That seems to be a cost-effective strategy, but is it a good performance strategy? Is it realistic to expect a TBred, which is simply a shrunk-down XP to catch up to that little overclocked wonder? Or will it take Barton or even Clawhammer to do so?
Let’s look at some benchmarks and try to project how well a Throughbred will do.
I did this by seeing benchmarking done at a few places, looking at how the XP scaled, and roughly projecting what it would probably take a TBred to match the performance of a 2.4-2.5GHz overclocked PIV, if it can be done at all.
(Anal compulsives/number junkies, please go away, this is for more highly evolved members of our audience. Yes, I know you don’t understand. 🙂 )
SiSoft Sandra CPUBench The Athlon holds up very well in this bench; it takes a 2.8GHz PIV to beat an XP overclocked to 1.9GHz. This bench pretty much measures raw CPU power and isn’t affected by other components; the TBred should have no problem remaining competitive here. The same can be
said for Sandra’s Multimedia Bench.
SiSoft Sandra Memory At any given memory speed, a PIV board is going to win, simply because there’s more available memory bandwidth for the memory. AMD overclockers can compensate for this somewhat by running high FSBs, but not completely. That’s not going to change with TBred, Barton, or maybe even the desktop Hammer. This is the most glaring weakness in AMD’s armor, but it doesn’t seem to do much harm at the moment.
PCMark 2002 This new CPU benchmark doesn’t seem to like the XP quite as much as Sandra, and it will take about 2GHz (actual) for a TBred to catch up with a 2.4-2.5GHz PIV (see link above).
SysMark2002 This benchmark gave Northwoods a big boost over the initial Willamettes (about 10% in the Office Productivity part, and 15% in the Internet Content Creation part). See here (but note that the 2.0 and 2.0A scores are reversed). Since the only significant difference between the two is the amount of cache; it’s pretty safe to say that AMD will trail in this category pretty badly until it releases Barton with a 512K cache.
3DMark 2001 SE The Athlon still does alright here, and TBreds should be fine.
Business Winstone 2001 The XP only trails by a bit; a TBred should have no problem staying competitive.
Content Creation Winstone 2002 The news is not so good here. Based on the scaling of the XP shown here, it looks like it will take XP2700 or 2800 (or 2.1-2.2GHz) speed for the TBred to catch up with 2.5GHz PIV speed.
Media Encoding It will probably take the equivalent of 2GHz, give or take a few dozen MHz, for the TBred to catch up, but it should remain at least competitive in Lame MP3 decoding.
DivX encoding may or may not be a different matter. This place came up with a relatively small gap (though that still would require about 2.1GHz for a TBred to close), while that place showed an unclosable 50% gap.
SSE-2 Heavy Applications The PIV will wipe the floor with the XP for such (Adobe After Effects 5.5, Lightwave) applications. AMD fans will have to wait for Hammer for redress.
Not all 3D rendering programs are like that, though, and for those that aren’t (3D Studio Max, Maya, Cinema 4D), a TBred should be able to catch up. However, this used to be an area where the Athlon used to be able to wipe the floor with the PIV. That is no longer the case.
Games If it’s Quake, or based on the Quake engine (like Jedi Knight 2), the PIV will win, but even Rip van Winkle knows that.
Comanche 4 likes 512K cache, AMDers will need to await for Barton to even the playing field here.
Anything else, the TBred should be able to at least match.
Conclusions: The Athlon no longer rules. In the overclocking world, the PIV is now an even competitor, maybe even a bit better at this point with the 1.6As doing so well. While the TBred can be expected to catch up in most areas, the most that can reasonably be expected is an overall draw.
However, in boxing, it’s usually not enough to edge out the champ on points. You have to knock him out, and the PIV doesn’t do that.
The majority of hobbyists settled on AMD platforms need a knockout to switch, much as many of them needed to see a knockout of the PIII by the Athlon to switch from Intel in the first place.
Outside of heavily SSE2 optimized programs (where choosing the PIV is a no-brainer), the AMD platform will likely trail a bit throughout the year, but the gap won’t become big enough to compel a change. If your strategy is to pop in a TBred (or even better, a Barton), it’s a reasonable strategy. If you don’t want to leave the AMD platform because of the additional cost of a mobo, or just because you love AMD/hate Intel, that decision won’t cost you much.
If this were going to be PIV vs. Athlon for the next three years, the Athlon would look decidedly shakier and would eventually get at least knocked down by SSE2 and/or better bandwidth at some point. However, Athlon just basically has to hang on for about a year; it can do that.
You won’t have to change in 2002. Unless you want to.
What the PIV now represents is a fully satisfactory alternative to the overclocker. It’s not a must-have; it’s a can-have.
You now have a choice where you didn’t before. That’s the real change.