Your testing is definitely and obviously flawed and skewed, which is why there was so much negativity in your last thread, which this is a copy and paste of. You can't show all these numbers and charts with who knows how many different combinations of hardware, settings, and software, and expect to be taken as a serious, objective test.
We're on Anandtech, so let's look at
Anand's CPU Bench showing that stock vs stock, 2600K beats 8150 at almost every single test run. The ones where FX beat out 2600K, we're literally by a hair.
About x264 benchmark,
polzyp said:
In this benchmark, the single core performance of an overclocked AMD FX 8150 CPU @ 4.8 Ghz is better than a 3.7 Ghz (tubro) i5 2500k, but worse than a 4.0 Ghz i5 2500k.
So, you're saying a single 4.8GHz core is faster than a single 3.7GHz core? Astounding. Notice how a slight overclock to 4GHz easily gives the edge back to the lowly 2500K.
FX 3dmark11 performance is meh. My
2500k@4.8 and 6950@1000/1525 is beaten by nearly double, in graphics, which is expected with 6990 v 6950. But that's GPU, and we're focusing on CPU, so lets look at that. In the Physics test, my lowly 2500K wins by a couple frames.
A 4C/8T 2600K destroys an 8C/8T FX at 3dmark11 physics at similar ~4.8GHz clocks.
Now, the all important (/sarcasm) performance of 7zip:
polzyp said:
Intel's not ready for this one...
polzyp said:
7-zip really shows Bulldozer's strength.
polzyp said:
7-Zip Benchmarks!! FX is back!
I don't know about everyone else, but these quotes sound like they came straight off FX marketing slides. This is just my own opinion, but I can see why the last thread got derailed.
Anyway,
polzyp said:
Even a stock FX 8150 @ 3.6 Ghz manages to beat the 2600k @ 3.4 Ghz (both with Turbo enabled). Again this is just further proof that when all threads are used AMD shines. This is notable given the tremendous price difference. Good Work AMD!
Tremendous price difference between 2600K and FX 8150?
$269 FX vs
$329 2600K is by no means tremendous. And if you live near a Microcenter, 2600K is $279.
So, stock vs stock, FX 8150 beats a 2600K, not bad:
But, since you're comparing a 4.8GHz FX 8150, let's make the comparison fair, shall we? Let's not bother comparing to 3960X like you did, because THAT is a tremendous price difference, and was probably only done to say "look, FX nearly matches that $999 processor, beat that!".
Anyway, here's a benchmark (
link) of a 2600K at 4.9GHz, winning by ~3% at compression, losing by ~2% at decompression:
Some Bulldozer strength... it matches an overclocked 2600K.
For Unigine Heaven 2.5,
polzyp said:
We can see here that Nahelem bottlenecks heavily when compared to an Overclocked AMD FX 8150 @ 4.8 Ghz. Comparing at with a 6990 @ 830/1250 Mhz we notice a 24% increase in FPS, and when we overclock the 6990 we notice a 28% increase in FPS. This just comes to show that overclocking a 6990 with an i7 920 pushes it near its bottleneck.
This is another clear example of a skewed benchmark, and even more obvious. How do you compare a 3.6GHz Nehalem to a 4.8GHz Bulldozer, and act as if that's a fair comparison? This just goes to show that these "tests" of yours follow suit with Bulldozer marketing slides, nothing more.
If that Nehalem bottlenecked 6990 so much, why doesn't it bottleneck the 580 SLI? If Nehalem was a bottleneck, why is there a consistent rise in performance with higher performing GPU's? Or a rise in performance with GPU overclock? Pretty obvious this chart is testing GPU, not CPU:
Another chart that highlights GPU performance, showing spot on scaling with increasing GPU performance, which for some reason you are trying to pass off as CPU performance:
polzyp said:
We can see here that an overclocked FX 8150 @ 4.8 Ghz barely trails an OC'd 3960x @ 4.7 Ghz, but when the GPU is overclocked this difference is easily overcome.
Yes, obviously, because you are alleviating the GPU bottleneck by overclocking, increasing GPU performance, thereby increasing the score. GPU did that, not CPU. Overclock the 6990 for the i7 test and try again.
polzyp said:
It is also interesting to see that an overclocked 6990 easily beats an overclocked 7970, which is interesting given Heaven 2.5 is one of the benchmarks where the 7970 is supposed to shine most.
7970 has been shown to be able to match or slightly exceed dual GPU 6990 in some testing, like it does in your chart where it nearly matches the stock 6990 when 7970 is overclocked. But it's not at all surprising to see 6990 beating 7970 at anything, really, it is dual GPU vs single after all.
So, what we can extrapolate from these numbers, is that you have to overclock the hell out of an 8150 for it start to be somewhat competitive. On the other hand, for the FX to require almost a 1GHz overclock to be considered competitive is most definitely not a win in any shape, form, or fashion.
I guess if I really need to 7zip the hell outta some files, FX is my chip, because,
polzyp said:
7-Zip Benchmarks!! FX is back!
Well, alright.