Originally Posted by sdlvx
Happy birthday Gertrude
Yeah, they definitely are. You can tell by what Intel is doing.
First, Intel is focusing on GPU. Why? Because they know HSA/OpenCL is going to be a thing, and if they're shipping HD4000 quality GPUs as their high end, they're not going to get a piece of the pie. Ask yourself how many regular joes who would buy an Intel laptop care about iGPU performance? Most of them don't even know what an iGPU is, and the ones that do know and want to do something besides Facebook and Farmville with their GPUs get something from AMD or Nvidia.
Second, look at how Intel is framing their inability to increase performance and tweak their ancient P6 derived architecture. They won't admit it, they won't go, "well we can't get more than 7% IPC increase out of our chips now", but they will go "performance doesn't matter anymore!" The rumors and that supposed die shot of Steamroller leaves pretty much everyone that has an extremely good grasp on CPU architectures (much beyond my own) expecting at least a 30% increase in performance at the same clocks. Intel knows this. In fact, they know it so well, that my Google Reader for Intel this morning was filled with "Intel replacing Pentiums and Celerons with new Atom cores." And do you know why? Because they know their P6 derived architectures are finished and they're going to try and scale Atom up and turn it into a big core. It's all Intel has left, Itanium failed, Netburst failed, and original Atom failed but they're working hard to make it into something because it's the only CPU where Intel can extract more performance out of it.
Intel has been stuck in a rut for a long time. Nehalem was their last real significant IPC increase, SB was mainly just a clock bump. If you find the time to compare i7 975 to 2600k, you will find they end up pretty darn close, specially considering 975 has lower clocks.
It is the giant, white elephant in the room that no one wants to talk about, the fact that Intel, has, for the last 5 years, only been able to deliver <10% IPC increases. That, and the fact that Intel cheats at measuring their process node, and their 22nm is a lot closer to 26nm. http://www.electronicsweekly.com/mannerisms/manufacturing/the-honest-process-guy-2013-05/
There's a few more sources, but you get the gist.I think that once we start seeing games get optimized for AMD, and if we see more things like this HWBOT Prime benchmark that is designed with fair compilers which optimize evenly for all CPUs, that people are going to realize that Intel isn't that strong, they only know how to make themselves feel strong.
We've seen massive amounts of scrutiny over frame times and frame rates last year, and now I believe that if people are vocal enough about it, we can see people realize that CPU benchmarks that don't use a fair compiler are breaking CPU reviews.
To be able to recompile LAME and see a 60% increase in performance at the same clocks, and to use the slower version of LAME as a representation of a CPU's total performance is just flat out deceptive. If you want to play that game I might as well put a first time, 16 year old driver behind the wheel of a supercar, watch him crash and have a horrible lap time, and then declare the car is slow and doesn't handle. It's essentially the same thing.
To be totally honest with you, if Skyrim was compiled the same way LAME official is, you'd see a 60% performance increase on the AMD, which would actually put it above the 3770k. I really, really hope you guys seeing this kind of performance by running an open sourced, fair benchmark on top of an open sourced, fair runtime as a brutal blow to how CPU reviews are currently performed, when reviewers don't consider compilation options or instructions used on each CPU and program and instead declare benchmarks as an accurate representation of CPU performance without actually knowing anything about the benchmark.