Thanks for the support guys, but there's no reason to flame someone because of his opinion.
jhalada expresses a valid concern, he is worried that the inclusion of SPEC benchmarks will show the Pentium 4 to be better than the Athlon. Before I get to my reasons for including it let's talk about a philosophy of mine that I believe in, this is the philosophy that I founded AnandTech on.
When you come to the site you are placing a great deal of trust in me. The way I picture it is like this, when you come to the site you're placing your money in my hands. Whether it be $100 for a cheap CPU upgrade, or $1000 for a nearly complete system upgrade, you are trusting me with your money. And even if only one person visited AnandTech, that's enough of your money to make me want to give you the most honest, unbiased recommendation I can put together based on the data I have. However a recommendation alone isn't enough, I also need to provide you all with enough data that you can formulate your own conclusions as well. I'm not here to tell you all what to buy, I'm here to educate, and with that education comes the ability to make a more well rounded decision. Now let's take a look at the big picture, AnandTech has close to 2 million individual readers now. Using the upgrade figures alone, you're looking at around $900,000,000.00 every month that all of my readers are entrusting me with (assuming $450 upgrade on average, I know the numbers don't always work out to that but for the sake of the argument let's assume that everyone is planning to upgrade this particular month). There isn't a sum of money on the face of this earth that I would take to jeopardize the hard earned money of each and every one of my readers. Understand that you all are the reason that I am here today, and I am forever thankful for the opportunity that you all have given me.
So there's reason one why I would never succumb to the pressure of a manufacturer assuming that one was present.
Secondly, you've got to give me some credit here. What you're assuming is that the minute someone flashes a green bill in my face I'll immediately drop to my knees and become their whore. I have just a *tad* more integrity than that. You're just going to have to take my word on this one since there's no real way I can prove it to you or anyone else other than by giving me a chance and letting my actions speak for themselves. The net is very self-regulating, and I've always believed that those that aren't reporting the truth or aren't making a conscious effort to provide honest information will eventually feel the effects of the self-regulating internet.
In close to four years that I've been running AnandTech I've been accused of being Intel biased, AMD biased, 3dfx biased, NVIDIA biased, Matrox biased, ATI biased and any other sort of bias that's even theoretically possible. Something tells me that you can't be both Intel and AMD biased at the same time, nor can you be biased towards all manufacturers, yet I've always received at least some email claiming one thing or another. This assumes yet another thing that is completely untrue, that I have something to gain by favoring one manufacturer over another. I don't own any stock in any of the companies I review products for, I don't hold positions on their boards, and I don't get a monthly bonus if I write one thing or another (I'll get to the issue of advertising in a second). I report the truth as I see it, and as I mentioned before, I provide all of the readers with the very same information I use to come to my conclusion in the reviews. It works like this, I use factors A, B and C to determine whether or not I recommend product Z. I provide explanations of those three factors in the review as well as whether or not I recommend product Z. Now if I just put together a quick thing on how product Z is awesome without even presenting those three factors then I can see where there would be a problem, but that's not how I do things.
Now we get to the issue of advertising. We've all got to make money somehow, so here's where we've got Anand, right? Wrong. I used to sell ads for AnandTech through about the end of 1998, and I quickly realized that I really hated selling ads. I wasn't here to do that, and I never really enjoyed it. I hated following up deadbeat advertisers, I hated negotiating with these guys, I hated everything about it. Ever hear of the separation of Church and State? Well, that's how advertising on AnandTech works. AnandTech, Inc. handles editorial content, and nothing more. Instead, I have a small sales force in California (with some reps in Taiwan) and they handle all of the ads. I don't know when a banner is going up until it actually appears on the site or until someone let's me know that the company has just signed on. I don't take advertising requests, I forward them all to my sales staff. And any editorial requests they receive are forwarded back to me. Don't get me wrong, often times manufacturers attempt to pull advertising based on things I have written, but those threats never make it to this side of the coast. I make it *very* clear that the only thing I deal with is editorial content and my sales staff makes it *very* clear that they hold no influence over what I write. Sure, I lose advertisers this way but I'm not here to please advertisers, I'm here because you all have placed your trust in me and your hard earned money in that trust as well. Would you turn down a $30,000 ad contract because it jeopardized that very philosophy?
I would, and I have on many occaisions and I will continue to do so. Don't believe me? Disguise yourself as an advertiser and just see how far you'll get with a request to control editorial.
What next, ah yes, stipulations for review samples. I wake up in the morning, go to my morning classes, and when I get back I see packages at the office. I open them up, and what do I see? CPUs, video cards, motherboards, what have you, but no where in the package are a list of stipulations that I have to follow, benchmarks I have to run, or anything like that. The first example that was brought up was the Naturally Speaking benchmark. Let's look back at that review (Pentium III 500) which was published on February 22, 1999. On one page of the 13 page review I included three benchmarks provided by Intel, all three used heavy SSE optimizations and I clearly stated that in the review. If anything, the benchmarks illustrated what SSE *can* do, and we do know that SIMD instructions do play a key part in the performance of today's and future CPUs. Take a look at my Desktop CPU Comparison published in September 1999, read the second section on the first page to read my take on the problem with benchmarks:
Desktop CPU Comparison - September 1999
So this brings us to the final point, why on earth would I include SPEC numbers in a review and what do they have to do with anything at all? The Pentium 4 will do wonderfully in some of the SPEC tests, no doubt about it, but is that my reason for including the benchmarks? No. I give both manufacturers an equal chance of succeeding here, I use the most optimized compilers for AMD and the most optimized compilers for Intel. I use the config files that AMD uses as well as the config files that Intel uses. The AMD 760 platform is currently the fasted x86 platform in the SPEC CPU2000 benchmarks, you could just as easily accuse me of following AMD's benchmarking guidelines by including SPEC in the DDR review. But if you've paid any attention to everything I've written above, you may begin to realize that this isn't how I function. Let's look at what SPEC shows us.
jhalada asked the following:
<<
BTW, how often do you run a weather prediction algorithm on your computer? Do you simulate Northbridge Earthquake? Do you use your computer for face recognition? How about simulation of "large systems of molecules"? >>
And the answer is obvious, you don't. But how often do you run Linpack? How often do you play Expendable? The point of the benchmark isn't to benchmark a specific task that you do, instead it's supposed to point out architectural advantages and disadvantages. Remember that these algorithms and tests were the things that were run on supercomputers not too long ago and the performance under these tests can help predict the performance of a particular CPU/architecture in the future. How long do you think it will take before face recognition is something that is important to desktop users? Do you think you'll be typing in a password to login to your OS indefinitely? Why would you want to run a weather prediction algorithm on your computer? You wouldn't, but analysing the performance of your system when dealing with such a large amount of data can help to predict how a particular CPU/architecture can perform in other situations that aren't as extreme. The point of these benchmarks is to help analyse performance, I don't want you to come and read the review without paying attention to what the benchmarks are saying, I want you to gain a greater understanding from them. If a platform can handle something with a large data set, then what does that say about its FSB, memory bus and cache? If it can't where a competing platform can, what does that say?
I hope I've covered all bases there.
jhalada, like I said, you express a valid concern, and you're entitled to an explanation. I've made mine, I hope this can remain as an example to everyone of exactly how passionately I feel about the integrity behind AnandTech. This post isn't directed at
jhalada alone, but anyone that has any questions about the honesty and integrity behind AnandTech.
Why is it that me and the rest of the AnandTech team kill ourselves on a daily basis to bring you the most thorough information we can? Why is it that every leading member of our editorial staff is currently juggling school and putting in up to 80 hour work weeks to keep cranking out the reviews? If it were about making the most money with the least amount of work, I guarrantee you that things would be different. There wouldn't be any all nighters, no reviews being posted at 6AM, and definitely no reviews with 100s of benchmarks. Would I run 12 hour SPEC tests just to make Intel happy? Would I buy a $560 copy of Compaq Visual Fortran so I could give AMD a fair chance just to make the Pentium 4 look better? You've got to be kidding
Take care,
Anand