Literature? No. If you don't want to take my word for it I can live with that. I know what I know and no one else will ever know all that I know. You can either agree with me or be wrong. (that's sarcasm in case someone takes it seriously)
I'm a system builder in Australia. I have 14 systems on the floor currently with various configurations of CPU and RAM. We're constantly testing, testing, and testing some more. Results are posted as we determine importance and as time permits.
We have direct suppliers for AMD and Intel, both of which only supply what the companies are currently producing. We wouldn't have it any other way as supporting a CPU that is no longer being produced is a problem for system builders in that we have to wait for old components to be shipped internationally (sucks when you advertise on-site 48 hour warranties). We're actually much further ahead than retail stores in the USA in that regard because I know before a company stops producing a model that its going to happen from these suppiers, who are in direct contact with the mother company on a daily basis. For AMD/Intel, its all about making a profit to continue research. Neither company is making any profit from selling 2-5 year old technology. Neither company really wants to cater to the end user who wants the cheapest CPU they can overclock to match the latest technology.
The 805D and 3600 X2 haven't been on our suppliers' lists for months for this reason. AU suppliers don't carry big inventories like the USA retail stores because they have to pay import tax and GST, which is also partly why our prices are inflated in comparison. Buying more CPUs than they can sell in a month is a losing proposition as they can't get their investment back when prices drop due to new technology.
AMD is currently still producing a few Semprons: 3000, 3200, 3400 and 3400. They are producing only the 3200 and 3500 in single core A-64s. Then they are producing the 3800, 4200, 5000, 5200, 5400, 5600 and 6000 in dual core in A-64. The older/slower models are being phased out because their new nomenclature is on the horizon and the low end will become the A-64 dual cores. That can be found on Google easily. It remains to be see if Sempron will continue or not, but my guess is no because they've already whittled down the line considerably.
Intel is the same. Only producing 4 Celerons, 4 Pentium Ds (925, 935, 950 and 960), and then it jumps to C2D and quads. The Penryn chips are on the horizon and the Nehalem is right behind it. Again, it remains to be seen if the Celeron and Pentium D will continue or not, but again my guess is no and that the C2D will become their low end before long.
Sorry, but I don't agree with that article nor your results at all. Let's take one example. Installing Windows XP Pro, from the time I click network settings button until the enter username window (I love this test as its one of the most relevant to speed of any that I've seen and is never used by anyone on reviews):
805D @ 4.0 GHz = 9 mins 40-55 seconds (done 5 times).
3800 X2 San Diego @ 3.15GHz = 13 mins 10-40 seconds (done 5 times). E6400 @ 3.2 GHz = 8 mins 40 seconds to 9 mins 5 seconds (done 5 times).
That's real time. That's what most people care about. That's what counts.
While clock for clock is all fine and dandy, memory comes into the equation for overall performance because of the difference between on-chip memory controllers and separate memory contollers on the motherboard. The higher the FSB goes, the better the memory performs on an Intel system too. Not so with an AMD. Thus the current Intel systems always perform better overall with higher OCs than the AMD system does. A person can argue that this is more than CPU, but actually it's not as the CPU determined the overall outcome.
I have a E6400 on my main system overclocked to 3.2GHz with Crucial DDR2-800 @ 4-3-3-8. The wife has the 805D overclocked to 4.0GHz with Corsair Value Select DDR2-533 @ 4-3-3-8. Real time for decoding, encoding, and file copying are within the ballpark (20-40 seconds) on both systems, even though mine cost heaps more. There's no way one system walks on the other for most applications, and both will walk on anything less than a 4800 X2 for most applications. There is a difference, but its only a seconds and hardly noticeable when the entire procedure takes 15 mins.
Sure, my system will beat hers handlily in FPS games and most AMD systems will also give it a heckofa run for its money in that situation, but then no one in our house plays FPS games so we couldn't care less.
The 805D is still a bargain and a great performer for the price, any way you slice it. And
no AMD chipset on the market to date will multi-task at the same level as an Intel 3.0 single core with hyperthreading, let alone a Pentuim D or C2D. Again, gamers don't care about this, but then most of the world is not gamers and does care.
Benchmarks are great if you understand what information they actually provide. My experience includes synthetics benchmarks and real time and does not reflect your claims. In fact, my results are totally the opposite of your claims. You've seen my real time results, now here's my benchmark results, which you can use to prove that a 3600 X2 is faster.. or not. Pay particular attention to the memory scores, which blow away any of AMD's:
http://oc.ultramaxcc.com.au/
The new Intels are giving up this advantage for memory controllers on the chipsets, like AMD has done for years. It remains to be seen if this is an overall better system or not, but there can be no doubt that there are advantages and disadvantages of on-chip memory controllers. The bottleneck changes from one point to another, but will always exist. Significant advances in CPU speed are necessary just to break even with the current effect on memory with separate memory controllers.
As for power consumption, when AMD made a faster CPU their marketing emphasis was on speed. Now that Intel is currently the leader of speed, AMD has focused marketing on power consumption, forcing Intel to address it as well. Bottom line is that I've seen no difference in my electric bill whether we run 65nm or 130nm CPUs, and unlike most here we run many computers 24/7. While your math is all good, the end results don't support it. I saw a considerable difference when we added a plasma TV though. Me thinks some people just buy into the marketing ploys when they see what they want to see and disregard the rest.
You keep posting links to TG. I take every TG review with a grain of salt as it's obvious to some of us that TG is in business to slant whatever company they're reviewing at the time toward looking good, which is easily done. Unfortunately, TG's editors have little to no technical knowledge which makes them useless in their roles. TG has been criticised for almost every article for the past 2 years for good reason. Do your own tests and post those results. I'll gladly do the same with whatever hardware we have in house at the time.
