i3-4170 or FX-6300?

Prysin · March 13, 2016

Just now, Majestic said:

http://anandtech.com/bench/product/1543?vs=697

Check out for example the Dolphin Emulator. That does show the same difference, and it's also reflected by the way AMD's are unable to run games like Metroid Prime, Starfox and Super Mario Sunshine but Intel's are.

I think the gains in games are just too complicated to expect lineair gains.

well emulators are a bit tricky to use as a bench. Mostly because it depends how its written and how the things you ARE emulating affects the emulator software and also the hardware.

Clock numbers taken from Intels own product pages

i5 4690k 3.5/3.9GHz -> 7.6 (Devils Canyon/Haswell) 4 cores, less L3 cache, Better TIM....

I5 6600k 3.5/3.9GHz -> 7.93 (Skylake).... WTF? HOW CAN SKYLAKE BE SLOWER THEN EQUALLY CLOCKED HASWELL??

i7 2600k 3.4/3.8GHz -> 11.57 (Sandy)

i7 3770k 3.5/3.9GHz -> 10.67 (Ivy)

i7 4770k 3.5/3.9GHz -> 7.63 (Haswell) 4 cores + HT, more L3 cache, worse TIM....

i7 4790k 4.0/4.4Ghz -> 6.80 (Devils Canyon/Haswell)

i7 5770c 3.3/3.7GHz -> 8.02 (Broadwell)

i7 6700k 4.0/4.2Ghz -> 6.48 (DDR3L) (Skylake)

i7 6700k 4.0/4.2GHz -> 6.47 (DDR4) (Skylake)

i7 5960x 3.0/3.5GHz -> 8.6 (Haswell. 8 phys cores + HT, 20MB L3)

I also dont think Dolphin scales well or at all above 4 threads though... Which makes it an "unfair" test for FX... look at the scores, whilst they ARE questionable in some cases like the insane jump from Ivy to Haswell, despite similar clock speed.... wasnt Haswell like 15% faster then Ivy? THAT is not 15%, that is sure...

Broadwell, aka 5775C has 5-7% higher IPC then Haswell. So why is it slower by a noticeable margin then the 4770k which has near similar clock speeds?

200MHz shouldnt make THAT huge impact when IPC is higher....

However look at clock speeds and how it impacts performance of said program on AMDs side.......

FX 8150 3.6/3.9GHz -> 15.07 (Bulldozer)

FX 6350 3.9/4.2GHz -> 13.92 (Piledriver) <- 6 cores

A10 7870k 3.7/4.1GHz -> 13.75 (Steamroller) <- NO L3 CACHE, lower clock speed, 4 cores....

FX 8350 4.0/4.2GHz -> 13.53 (Piledriver) <- 8 cores, more L3 cache, marginally higher base clock.

again, WTF? Steamroller is barely 10% faster then Piledriver at EQUAL clocks... here it is faster then Piledriver at LOWER clocks and fewer threads AND no L3 cache... none at all....

We both know FX has weaker cores. It is a fact. However if it cannot use all its cores, said weakness is amplified. This is probably what we are seeing here.. However this Dolphin benchmark is all over the place.

One should expect linear results in similar software, in correlation with known, proven performance differences. But this makes no sense. NONE.....