Originally Posted by Homeles
I don't think you've been paying much attention to the industry as of late. Intel's not competing with AMD anymore. The pressure is still definitely on Intel, just from another source — ARM.
Let's take a look at your list of complaints:
8 core processors?
That's a valid one. A competitive AMD would put pressure on Intel to release chips with higher core counts for consumers.
Higher stock clocks?
I don't think you've heard of the term "leakage."
Why does it need more cache? More cache would be very expensive. They doubled L1 and L2 bandwidth, what more do you want?
AVX2 is going to replace SSE, which is the most widely used instruction set AFAIK
. AVX2 is pretty huge.
x86 performance is stagnating as a result of simple physics, not because of a lack of competition.
I know ARM is now Intel's greatest threat. But seriously, most consumers won't be doing any serious gaming or content creation on a tablet or smartphone anytime soon.
I also realize that current leakage is a big problem. That's why Intel went to FinFET/tri-gate transistors. Higher clock speeds are attainable if Intel tried. IBM recently released a server-grade 5.5 GHz processor so it can be done.
Okay, there are definitely diminishing returns with cache.
Programs will need to be at least recompiled if not re-written entirely to fully utilize AVX2 instructions. Also, AVX2 will not be of any benefit to legacy applications. In time, AVX2 will make a tremendous impact but not right off the bat.
x86 is stagnating in part because Intel can afford to hold back on innovation, not just because of simple physics. Silicon transistors have not reached the fundamental limits of physics yet.
Originally Posted by Brutuz
How is it for the benefit of all consumers? Most won't notice any performance difference, but they will battery life difference.
Yes, Intel is catering to the needs of mainstream consumers who only care about shiny ultrabooks. But here at OCN, we are power users and enthusiasts. We demand the very best out of our hardware.
Desktop performance is potentially being held back by Intel's relentless focus on mobile performance. Take Ivy Bridge. Do true gamers care about the HD 4000? No. They get a real graphics card. How about the crappy thermal paste that Intel used? Mobile users won't notice, but overclockers certainly will. For all the tri-gate hype, Ivy Bridge only offers at best +10% IPC improvement over Sandy Bridge (which is largely negated because Sandy Bridge has greater overclocking headroom).
Haswell is clearly designed with ultrabooks in mind, not desktop computers. Desktop processor optimization shouldn't become an afterthought just because server processors and mobile processors are so profitable.
My fundamental argument is that Intel is settling for 'good enough' x86 performance, when in fact they can do better. If you were purchasing a 3930k or 3960x wouldn't you want all 8 cores fully functional, rather than a crippled hexacore chip? Yes, the TDP would go up, but if you're willing to pay for "extreme" performance you would also be willing to invest in a high-performance cooling solution.