The, shall we say, officially-sponsored benchmark numbers for Conroe are extremely impressive. Even the relatively lousy numbers are very impressive.
When you get to the gaming numbers, and the 20% improvement turns into 30%, they go from impressive to, well, maybe not immediately believable.
We’ll talk more about those later, but Intel’s numbers, if they substantially hold up, are the equivalent of one of those movie guys with a two-foot long run pointing it at us and saying, “Freeze” after a gun fight.
And we’re not exactly sure if he has any bullets left.
For anyone planning to buy a fast computer anytime soon, especially a gamer, Intel’s numbers essentially make them freeze until these products come out. After all, if they’re right, why spend a bundle today when you can do 30% better within six months?
What Intel did was quite clever. Had they not brought AMD into the picture, the message coming from IDF the next six months would have been, “Don’t buy our products for the next four-six months.” This is not exactly a sales booster.
By giving AMD a “starring” role, the message turns into “Don’t buy anybody’s products for the next four-six months.” That at least spreads the misery around, and I suppose somebody in Santa Clara hopes that if Intel gets the sales flu as a result of this, AMD will get pneumonia.
This is brilliant “when you have lemons, make lemonade” marketing strategy.
Legit or Not?
How legitimate are these numbers? Well, something you ought to keep in mind in any debate is just what’s at stake for the speaker.
For instance, “don’t buy a computer for the next few months” is certainly not going to be welcome news at VoodooPC, and not surprisingly, the head of VoodooPC, not wanting his business to go down the drain, really, really doesn’t like this comparison and frankly grasps any straw he can find to discredit this.
It’s not that some of his points don’t have a little legitimacy, but “little” is just about it. An outdated BIOS might make a little difference, but not a 40% difference in a game, and trying to compare a non-standard Intel-made timedemo to a non-standard Tom’s Hardware Guide-made timedemo just obfuscates the issue.
On the other hand, the two sites that put up what were for the most part Intel-designed benchmarks can’t just throw up their hands and say, “Gee, it looked legit to us.”
Update: Anandtech has just issued a follow-up report on Conroe. An updated BIOS was used, which yielded little to no difference. Human (Anandtech, not Intel) error resulted in the game that showed a 40% gap being tested at two different resolutions. Once corrected, the gap shrunk to a 20% Conroe advantage.
What are the possibilities here?
1) The results are completely legit. Perhaps SSE128 and other Conroe enhancements are enough to provide the necessary gaming boost for Conroe. We won’t know for sure if that’s the case until the product comes out.
2) There’s some subtle, maybe deliberate, maybe not, differences/omissions which adversely affect the results, but not significantly In the past, Intel has not been above subtly biasing the results towards a desired goal. I remember one comparison they did a long time ago between two or their own products where the favored child got the fastest hard drive available, while the other thing got a pretty slow one.
However, you’d have to come up with a ton of these little biases to yield a 30% difference. DDR vs. DDR2 is likely to be another minor bias, again, a matter of a few percentage points.
3) There are other factors Intel is manipulating to its advantage If a game is CPU-limited, more raw CPU firepower alone could have a disproportionate affect on scores. Yes, it’s an advantage, and it’s “real,” but perhaps a faster X2 will see a disproportionate jump in fps in these games, too.
4) Enhancements or Cheats Realistically, the only big ticket item, the smoking gun, so to speak, would be software that affects video performance.
However, there are enhancements and there are cheats.
For instance, if Intel used a driver which enabled SSE128, that’s clearly an enhancement, not a cheat, since the video card companies will certainly include anything like that in future drivers, and the AMD processor couldn’t use such a feature anyway.
On the other hand, if Intel modified a driver which allowed the Intel chip to use the best means of performing various tasks, while blocking a perfectly capable AMD chip from doing the same, that’s clearly a cheat. Subtly lowering video quality for the Intel but not the AMD chip would also be a cheat.
In other words, if it helps Intel, doesn’t hurt AMD, and might reasonably be expected to be included in future drivers, it’s legitimate. If it helps Intel, hurts AMD, and/or could not reasonably be expected to be included in a future driver, it’s a cheat.
No one has yet even begun to check this possibility out.
Comparing Past To Future
Some are complaining that comparing a future processor to a current one is somehow unfair. This is nonsense, simply because no one is going to buy a computer based on this assessment.
What is Intel saying with these numbers? They’re saying to potential AMD buyers, “Don’t buy an AMD system now, our future one will be 20% better.” Provided that 20% figure is legit, that’s pretty good advice for most people.
At least a month before any Conroes show up for sale, AM2 processors and systems will appear. They’ll certainly be measured and benchmarked, and that will become the AMD standard, good, bad or indifferent.
A month or two after that, Conroes will show up, and if you have the slightest doubt about Intel’s numbers, there will certainly be plenty of people checking all the nooks and crannies then.
Then, and only then, will (rational) people make the buying decision and open the wallet.
There’s three possibilities next July:
1) The Conroe advantage pretty much holds up, and AM2 is initially little better than socket 939: If that’s the case, you either buy the Conroe, or wait for AMD to do better. You certainly haven’t been hurt by waiting; you’re better off for it.
2) The Conroe advantage doesn’t hold up because AM2 matches or beats it: If that’s the case, you bless Intel for making you wait, bless AMD for rising to the challenge, then bless AMD again with your money. Again, you’re better off having waited for AM2.
3) The Conroe advantage doesn’t hold up because Intel proves to be a pack of lying dogs on this: Only here might you be hurt by waiting, but rest assured, Intel will get hurt by it more than you, which is why it’s extremely doubtful Intel has done any more than fib a little.
So you see, the Intel announcement is going to hurt AMD (along with itself) no matter what ends up happening (along with AMD OEMs and AMD fanboys planning world conquest), but it only hurts you if Intel is lying big-time.
For now, and probably a while, we’re not going to be able to check the Conroe side of the equation.
However, there’s nothing Intel can do to prevent people from checking the AMD side of the equation. Individual, websites, OEMs building luxury AMD systems :), whoever has an X2 processor and an X1900XT Crossifrecan run at least the F.E.A.R. benchmark to see if the AMD numbers hold up.
Indeed, it would be interesting if somebody with the setup along with a freeze unit ran an X2 at say, 2.6, 2.8, 3.0 and 3.2GHz just to see if performance gains disproportionate if the X2 goes faster in some of the gaming tests. For instance, an X2 running at 3.2GHz really shouldn’t be able to match the Conroe if performance gains were somewhat proportionate to CPU speed. If testing shows that it does, then Intel has rather less of an advantage than the current numbers indicate.
Or you can put your wallet back in your pocket, take it easy for the next few months, and check out the numbers available in July.
Need something else to amuse yourself with the next few months? Keep an eye on what AMD has to say about all this. I know I will. 🙂