I still wanna give Bulldozer a chance. AMD announced back in Dec. 29 that they're coming out with some new FX SKUs, namely the FX-4170 and -6200, and I'm obviously interested in the FX-6200. Doing the math based on PassMark, the FX-6200 at stock can theoretically deliver almost similar aggregate
performance as an i5-2500/2500K. Now, PassMark is a multi-threaded benchmark and is apparently not very popular here at TR or at any other prominent tech site, but the numbers give a ballpark figure of where each chip lies in the performance spectrum, when all cores are working, at least. You can also use the numbers to calculate the individual performance of each core of a particular CPU.http://www.cpubenchmark.net/high_end_cpus.html
Here are the numbers from PassMark (taken Feb. 16, 2012 ... they change everyday) :
Core i5-2500K : 6,743
FX-8150 : 8,250
FX-6100 : 5,593
FX-6200 : To be computed
Although this benchmark is multi-threaded and stresses all the cores it can find, we can easily compute the per-core performance of a CPU by simply dividing the score by the # of cores, so given the data above, a 2500K core gets 1,686/core, while a single FX-8150 core gets 1,031. These numbers (1,686 and 1,031) are in line with what TR measures with real-world, lightly-threaded apps, which show the 2500K to be around 60% faster than the FX-8150. Aggregate performance of the FX-8150 is 22% higher than the 2500K, which is also not too different from what TR sees in highly-threaded apps.
Doing the math,
1,686 / 1,031 * 3,600 = 5,887
.. we can estimate that the FX-8150 has to run at a whopping 64% faster (5.9GHz!!!) to reach the per-core performance of a 2500K! Of course, if each FX core matched each i5-2500K core's performance, the 5.9GHz 8-core FX's (model no.?) aggregate performance would be twice the 2500K's aggregate performance using twice as many cores. It won't cost $270 if it existed now, that's for sure.
I've also found that PassMark's numbers across a processor family (i.e. same die and enabled features) are pretty much in line with each other. You can easily come up with the numbers for a different SKU using another SKU if you use clock speeds as your other factors. Hence, we can compute/estimate how the upcoming FX-6200 will score in PM:
3,800 / 3,300 * 5,593 = 6,440
That's pretty close to the 2500K's score. Clock it just a bit higher and you get
4,000 / 3,800 * 6,440 = 6,779
This means you have to overclock an FX-6100 or an FX-6200 to 4.0GHz to reach (approximately) the aggregate performance of an i5-2500K. Each FX core @ 4.0GHz will score 1,130, which still falls way below an i5-2500K core's score of 1,686 ... but you have two more cores.
You can either grab a $160 FX-6100 right now and bring it up to 4.0GHz and convince yourself you got a great deal comparable to a $230 2500K, or you can wait for the FX-6200, clock it at 4.0GHz also, and possibly enjoy higher OC headroom should you wish to leave the 2500K (at stock) behind for just $15 more. It has a higher TDP, yes, but clocking the FX-6100 at 4.0GHz will undoubtedly suck around that much juice as well anyway, with the -6200 quite possibly being able to reach higher clocks. Also remember that you need a P67 or Z68 board to OC your K CPU with any real level of satisfaction, which also usually cost about $40 more than AM3+ boards, so don't think it's just the difference between the FX-6100's and 2500K's prices that you have to fork over to get OC goodness on that Intel system -- it's more like (230 - 160) + 40 = $110 more.
NEC V20 > AMD Am386DX-40 > AMD Am486DX2-66 > Intel Pentium-200 > Cyrix 6x86MX-PR233 > AMD K6-2/450 > AMD Athlon 800 > Intel Pentium 4 2.8C > AMD Athlon 64 X2 4800 > AMD Phenom II X3 720 > AMD FX-8350 > Zen?