Ironheart, you power estimates go far beyond absurd, easily to the point of being -to use your favorite word- laughable.
Not only would one be required to run their computer 24/7 to see those numbers; they'd have to run their computer
at full load 24/7. Basically, our hypothetical person would have to run Prime95 every second of every day or every week of every month for a year.
To use bend-over-backward liberal estimates that are at least
slightly less "laughable", let's take your most extreme 80W difference and assume full load for five hours a day, which really is bending over backward for any gamer since games are only going to utilize half an octacore chip, at most (so I guess we can assume they're doing something other than gaming?), and since that's really quite a lot of play time to average for anyone who's more than like... 12 (and has a job/college to think about)
5 hours a day is 1825 hours for the year 1825h*3600s/h*80j/s= 525,600,000
Congrats, you've used a bit over 500 additional megajoules. One kilowatt hour is 3.6MJ so at 26 cent per, you're looking at a hair less $38/year, or basically the cost of buying a decent cup of coffee once a month.
For a gamer running only 2-4 cores, and even then, only modestly in a modern DX11 title, that's likely to be vastly less.
As for the "laughable" sites presented on performance, I'm afraid yours is by far the least substantial. There is basically no information given on methodology by you, nor did you actually cite a verifiable source (just a general website, in which these alleged tests could be
anywhere), getting results that don't agree with those of the bigger, more reputable review sites.
We can't tell anything from the numbers that the site you're "citing" gets, because we have no information on how each performed in each specific test, the margin by which any given chip was ahead relative to any other in any particular test (which in turn would tell us how much they're using properly threaded software in that batch of tests... nearly a full quarter of which are
file compression tests (why?)), or even what each program was doing.
Going back to sources where we actually
can see exactly what happened-
The Guru3D review I linked focused on the i7 2600k, so the i5 2500k was only included as additional backdrop in most of the tests, and yet, where both were included, the 8150 won in 9 of those tests, effectively tied in 2 (within about 3% at most), and lost in 3. I'm too tired to average it all out, but looking at the margins by which the 8150 tends to win vs the 2500k, the result is a foregone conclusion.
Tom's Hardware got mixed results, but still ended up concluding that Bulldozer is handily faster than the 2500k in heavily threaded software (as we've been saying all along), which CPU-intensive software generally and ever-increasingly is.
Anandtech gets results like those of Guru3D. The FX8150 beats the 2500k in almost every program, except when they
artificially cripple Cinebench to only use a single core (artificially since Cinebench is absolutely mutlithreaded). Note of course how that particular detail is exactly the kind of thing missing from the alleged "tests" from your "source".
The only sector of software in which the 2500k shows any kind of consistent edge is gaming, and since most games aren't CPU-intensive, that's generally limited to either tests that artificially bottleneck the game on the CPU by using absurdly low resolutions, in older DX9 titles, and/or generally cases where the CPU is the bottleneck because the FPS is already extraneously high.
If games generally
do become CPU intensive, there's no reason to think they won't follow the pattern of all other software, and take advantage of CPU speed primarily by using more cores, which would favor Bulldozer immensely, yet, somehow despite its importance, your post fails entirely to even mention
either the increasing role of multithreading in software, or the fact that, again, gaming performance rarely varies except in meaningless cases.
Of course, I'm not saying Bulldozer is unequivocally the better chip to buy. There are many cases where I would recommend Intel chips, and I have an Intel chip in one of two of my gaming machines (a Phenom II in the other). The right decision's contingent upon particular software focus and assumptions about the future, but unlike you, I haven't made it a point to try as hard as possible to come up with the most biased analysis (if can call it that) possible, leaving out any fact that doesn't further that point (you didn't even ADDRESS upgradeability), and bending over in every way possible to favor Intel with any assumptions you make, or avoid citing actual sources with links we can actually follow to scrutinize methodology, and I
certainly wouldn't pretend, as you do, that there is no case where a Bulldozer chip might make sense, given the broad selection of software in which it handily outperforms its Intel contemporaries.
Maybe that wouldn't irk me so much had you not also made it a point to call far more substantial posts and followups by others "laughable".
Cochise, on 27 May 2012 - 06:42 AM, said:
Nice stuff Vulp,
As an AMD fanboy, I hate to say it, but the Intel i5-2500k is probably the best bang for buck in gaming
right now and bests the new AMD FX-8150 in most areas. They are also priced almost exactly the same.
I say "right now" because you will probably get more life out of a 990FX motherboard for AMD and who knows what windows 8 will bring to the FX-8150 but throwing all the "what ifs" out the window, it would probably be the i5. I would say that the AMD is more "future proof" than the i5 if you do more than just gaming. If not, then no.
Yet another FX-8150 comparo review
http://www.bjorn3d.c...25&pageID=11066
It will be interesting to see what the next AMD Piledriver will bring because I think with this iteration AMD was just laying the architecture out for what the future "will be" because with the FX-8150 it has not quite arrived yet.
Because I am a fanboy, I will probably get the AMD and I know that I will be able to use the next generation CPU in my motherboard and not have to re-invest in all that again.
Food for thought.
Well, that review isn't really clearcut since the 8150 handily wins in Sandra, Truecript, and half the Cinebench tests. It basically ends up as a draw there.
Edited by Catamount, 27 May 2012 - 06:59 AM.