The reason why I post this is for context. Under our economic context, what Intel did there IS unethical in a free and open capitalistic economy. Its why we have government. But generally I agree with you both to most exstent. I live in California. A few years back, the "government" (pregovernator) chose to give everyone a chance to be winners in HS sports. IF they wanted to, they could be in the playoffs, regaurdless of record. To me, thats what this article reminds me of. If you want everyone to be able to play, then YOU pay for it, not me. And dont try to tell me that just because this team or that team needs a chance by whatever logic, or fogged distortions, is just as good cause they play too, I aint buyin it. ONLY when its worth buying it, THEN Ill spend my money and takes my chancesAll that aside, I wouldn't say Intel was unethical (since I use to word to denote actually evil actions). Instead I'd say it simply tried to prevent competition, instead of allowing it. If you made a product and your competitor went to all the re-sellers and paid them kickbacks to sell either none or very small amounts of your product, with non-disclosure agreements, you'd think of them as....______?
what did intel do? what is the proof?
I think the "author" (I use that term loosely) of the "article" (again, I'm being generous) was trying to state that the C2D line of processors was built to excel at benchmarks instead of actual applications. Think of stuff like 3DMark... sure, we all want higher scores, but that's not an actual game any of us play. I think that's where he was going with that.
How do you make a CPU that runs 3DMark better than applications?
MarkG said:How do you make a CPU that runs 3DMark better than applications?
Easy, usually benchmarking applications run different tests in isolation and after compute a score based on individual test's score.
Now if you improve parts of CPU like for example integer multiplication using a high order of magnitude like 4 times faster but memory access stays the same. in the benchmarks like SpecInt you will see a great increase but the real life applications won't see that much of a difference because the system didn't get an overall improvement, and SpecInt is a specific benchmark.
(Well I'm not implying that this is the case for Core 2 as tests clearly confirm that's it's a architecture that's improved overall even for memory access despite using old fsb)
So the answer to your question, optimizing a CPU for benchmarks means not making overall optimizations to the architecture, but local ones which results in greater benchmarks scores but possibly lower real applications benefits.
I just explained to him what the expresion 'optimized for benckmarks' means or could mean, I din't sugest anything. Where did the money part came from?Yeah.. spend all that money just to make the CPU good for benchmarks. They'll optimize the bench, not the CPU.
Indeed. Embarrassing is that article.
What was good and what was the purpose?
what did intel do? what is the proof?