Originally Posted by sdlvx
I don't want to be that guy, but read the last few pages of the thread. You may learn something about how Blizzard compiled code is crippled on AMD hardware and it's not representative of its full performance.
As far as I am concerned, the only benchmarks that are a true and even test of raw CPU performance are ones where the tester compiles it themselves so they're aware of of the compiler optimizations done and that the source is available and the tester is capable of understanding it.
People are learning that if one product is running x87 code when it doesn't even have the x87 part of the CPU and the other is using SSE natively in an application, that a benchmark of that application is meangingless.
I could *easily* go into Linux, compile Blender on an Intel to not even use SSE, let alone AVX, SSE versions > 1, etc and then bench it against Piledriver with AVX, SSE, FMA, etc.
Does that seem fair to you? Because that's basically what Anandtech Bench is. Although SYSMARK is an exception, because Intel basically owns BAPCO.
So really, is it much of a surprise that 3570k is winning "80%" of the time like you claim?
If you want to prove that 3570k is superior to FX 8350, you need to prove that all of those benchmarks are using fair instruction sets on both platforms. Otherwise all you're doing is proving that a specific piece of software is faster on a specific piece of hardware. You're basically testing a small sample size that you have no idea how it works, and then saying that every piece of software in the entire world will behave the same.
What's the point when most applications I use favor Intel? This is not about Intel vs AMD. This is about choosing the best deal for me.
The reasons why AMD guys don't trust Anandtech is that they've been knowingly using benchmarks that are compiled or programmed in a way that more efficiently uses Intel's CPUs resources than AMD's.
Many review sites benched 8350. The verdict is mostly the same, i5 is superior.
I'm sure you can find another 20 sites that say literally the same thing.
I don't think you "lol look at the graphs!!! THIS ONE IS BETTER BETTER GRAPHS LOL!!!" understand that I could compile benchmarks my own way and make FX 8350 humiliate 3570k. Most benchmarks don't have anything to do with raw performance and more to do with the compiler. Hell, I'd bet I could "deoptimize" a 3930k to lose to an FX 8320 if I had to.
Okay, point taken, but what application does this have in everyday use? For people like me, my mom, my sister, my grandma, my classmates, people who play league of legends, my co-workers who only use office software, etc? i5-2570k offers me much more than FX 8350, and the same can be said for many, if not most people.
I showed you before the extreme example with CRAY and I've given you my anecdote about how the same margin of improvement applies to Blender as well. Then I linked you to FX 8350 and 3770k Blender benchmark and let you extrapolate. What was I showing you? That software optimizations can make something twice as fast. When you sit here and call something absolutely superior because it's 10% faster in some benchmarks, you look like you're completely inept and you have no idea what's actually being tested. Only that you go "ohh me get big bar graph bigger better SYSMARK and Starcraft 2 are like every other piece of software created and compiled with the only compiler in the world!!!!!!"
You're absolutely right, but most of the programs I run favor Intel. I'm willing to be that most people only run programs that favor Intel, which is why I favor i5 2570K over an FX 8350. I don't understand what you're trying to explain to me. I know that Intel compilers favor Intel, and that programs can be written in other compilers, in other languages, in other ways to favor AMD over Intel. I never denied that, but for me and most other people who will build a computer, an i5 3570K is the better deal.