Intel vs AMD

the 3570K (OCed) tops the 8350 (OCed) at all but possibly video rendering as far as power. It also OC's better and cooler, handles high freq DRAM multitudes better, and higher amounts of DRAM the same, generally runs cooler overall when OCed. (Actually I wouldn't recommend doing it as a tech level thing, but you can actually go back to Sandy Bridge and get a 2500K while will pretty much edge out the 8350 in most place also)

In Benchmarks Intel in real world there is no difference, I suggest you go with which offers more for cheap!

Real world is where it's at, BMs simply give a lowend baseline of what something does with a single task, You can go back to a 2500K (Sandy Bridge 2 Gen CPU that stands right with the much newer 8350), You get what you pay for

Well, you know, it's not like that old bootleg commercial: "More stripes - more Adidas".

The thing is, with Intel you get 4 cores, with AMD 4 modules /with 2 crippled cores each/.
Also, Intel is way better at any single-threaded and lightly threaded apps.
AMD shines then it's heavily loaded.
Intel is more efficient, by the rule of thumb AMD needs at least 30% more on the clock to fight on equal ground.
AMD is cheap as ****, while Intel keeps its prices pretty much the same for many years.
Intel runs cooler and it is easier to cool.
AMD overclocks easier as long you have adequate cooling.

Also - the FX8350 was set as a competition to the i5. They are somewhat equal, the only difference is whether you want to sacrifice everything for better times at encoding video and fooling around with heavy-workload apps with the FX...

Even then, other than video, for the most part I've found i5's perform better with heavy workloads (to a degree I think it is that it handles more DRAM and higher freqs better) and multi tasking

I am not sure where the "crippled cores" analogy is coming from...but an AMD core is significantly more useful than an Intel HTT process. HTT is basically a software "simulating" a core with a register stack.

AMD's cores are an integer core, they can functionally perform all the same tasks as a normal core when it comes to integer calculations without any down time.

They do share a front end, but the Cache is large enough to cover that.

Additionally...each "module" does share a flex FPU, but the FPU is capable of multi threading itself...so it can multitask 2 FPU ops per cycle.

So, while the innards are a completely different setup in the AMD, you sacrifice a slight bit of FPU performance to gain 2x the integer performance.

With hardware moving more toward HSA, and offloading physics to GPUs these days...FPUs in cores are less utilized anyway.

AMD's lag in single threaded performance has nothing to do with "core strength"...in any way. It has everything to do with integer pipelines, schedulers and decoders (the shared front end). That's why Steamroller's mantra is "feed the cores faster". If AMD can execute Steamroller true to the designs, then that should all but close the gap in single threaded performance while maintaining a substantial lead in multithreaded applications. Steamroller will be doubling up on decoders/schedulers and doubling the size of the FPUs in each module, while decreasing the length of the pipelines.

In Intel speak...this would be a little of both a "tick", and a "tock" as well. As the architecture is being heavily refined and restructured, with a die shrink, but it's not a complete redesign incorporating a complete change of design (even though intel is still running virtually the same architecture since P4 with minor adjustments in each generation...they claim "new architectures" every tick).

I cannot believe you actually paid that much money for your memory. Current gen Intel chips do not really benefits from high speed memory, AMD Piledriver show more gains however.

I build high end systems, quite a few of them, and I work with them all the time - you can read all the reviews and BM you want...sorry, but it doesn't relate to real world performance gains that are possible...BMs on nothing more than an entry level figure - they do one thing, period, run the software that was programmed in. Reviews can be and often are worse, especially if you really READ them, you often find things said that make no sense and are utterly wrong - i.e. I seen a these 'Expert' writers often say 'The DRAM DIMMS didn't load the proper timings under XMP from the SPD' as if they thing the sticks them selves go in and program the computer...wrong! what's on the SPD is nothing more than info. It's up to the BIOS to take that info and implement it as close as it can in the system - now then that can be real hard to do when say you have a set of stick that conform to XMP and need a tRFC of 278 or of 314 and the BIOS itself has a limit of 255 in it - yet they advertise they can run these things - Memory updates make up the bulk of most all BIOS updates and are the ones you don't hear about. Another good one is people reviewing DRAM (and this is my favorite but there's plenty similar) a reviewer (who claims to be a memory expert) did a review on a 16GB set of DRAM and claimed to have run it under XMP, and talked about low performance....Hmm, looking at the review, that 16GB 'set' was actually half of a 32GB set , the label was shown in one of the pictures, he tried it on an old BIOS that didn't support the config (and obviously could do it manually), so he dropped to 16GB and enabled XMP which can work and will run, the problem with performance - the 8 sticks set (8x4GB) is the basis for the XMP timings, and require a tRFC about 100 higher than would a 4x4GB set...so he was running the 16GB at a tRFC 100 higher than it should be...so of course performance lagged...there's tons of these out there.

What I like to do is set up identical systems, or use a single system and simply have people sit and do stheir thing on say a system running 1600 and again on one running 2133 or better, (and not just email or a browsing session), I want them to multi-task, open numerous apps, use large data sets, have a virus check in the background, etc. You have them do this blind, meaning not knowing what DRAM they are running - about 85% always pick/want the rig with the faster DRAM and comment that it's much faster, what did you do.

As far as costs, that's another thing that I look at both for me and for clients IS IT IN THE BUDGET? DO YOU NEED IT? for me yes, I am often using 20+ GB of DRAM between apps, a RAMDisk, VMs, photos, video, GIS, etc at a single session

So no offense please don't tell me that 'current gen intel chips' don't show benefits, the do, as did the third gen, IB, as did the 2nd gen SB and even back to socket 775. And further AMD Piledriver is at the same level of DullDozer the 8150, in general limited to 1 DIMM PER CHANNEL AT 1866, their own words, that I finally got them to admit and have kept the emails as proof. What truly stands out in DRAM use gains is AMD, yes, but definately not the Piledriver, it's their APUs that love fast DRAM for their GPUs