Do high end user use AMD instead of Intel?

[MiddleOfTheRoad]

So then you saying AMD is misleading consumers by showcasing their Cinebench performance.

I'm saying that Cinebench is misleading all consumers.

 

[MiddleOfTheRoad]

Which is exactly what the benchmarks do.

Not ones written for specific hardware. Cinebench is written with an Intel compiler.

Definition of Benchmark:
a neutral standard or point of reference against which things may be compared or assessed.

Can't make it any more simple. Key of the definition is neutral -- you can't be neutral if you are running optimizations for Intel hardware. Null and voided Cinebench as an accurate benchmark.

 

[Phynaz]

Not ones written for specific hardware. Cinebench is written with an Intel compiler.

Definition of Benchmark:
a neutral standard or point of reference against which things may be compared or assessed.

Can't make it any more simple. Key of the definition is neutral -- you can't be neutral if you are running optimizations for Intel hardware. Null and voided Cinebench.

Can't help but notice you put the word neutral in there. Can you link the definition that has that in there, because all the ones I see say nothing about that. You realize that posting information know to be false is against the forum rules, right?

BTW, you realize that the Intel compilers produce the fastest code for an AMD CPU, right?

 

[Phynaz]

I'm saying that Cinebench is misleading all consumers.

Then why would AMD use a misleading benchmark if they are not trying to do the same?

 

[MiddleOfTheRoad]

Then why would AMD use a misleading benchmark if they are not trying to do the same?

Has AMD marketing done anything competent within the past 10 years?
The jury is still out.....

 

[MiddleOfTheRoad]

Can't help but notice you put the word neutral in there. Can you link the definition that has that in there, because all the ones I see say nothing about that. You realize that posting information know to be false is against the forum rules, right?

I pulled it off the PCMark brochure I had sitting on the desk (handed out at an IDF). Directly above the line:
"We create PCMark in cooperation with the world's leading technology companies using an open process that guarantees fair and neutral benchmark results."

I'm not exactly sure PCMark can throw around the word neutral, either.... Considering the PCMark2005 scandal deliberately distorting scores for Via and AMD Cpu's: http://www.osnews.com/story/22683/I...quot_Cripple_AMD_quot_Function_from_Compiler_

 

[TheELF]

Not ones written for specific hardware. Cinebench is written with an Intel compiler.

As is 99% of software a user will use.
Benchmarks are supposed to show you what to expect in real life,and in real life everything will be written on intel compilers.

 

[2is]

In scientific integer computing -- it does win by a mile.
Twice the threads = twice the workload.

Even processing on slower AMD cores, an i5 can't keep up with the workload of an FX octocore (4 projects vs 8 projects simultaneously).
But there are limits, no FX can touch Haswell i7's -- which is why I now run 2 i7 4790K's on WCG.

But i7 > FX Octocore > i5 on integer computing by a pretty substantial margin

Please link me some benchmarks of actual programs that people use showing FX beating a haswell i5 by a mile.

So we have now went from AMD fans posting individual cherry picked charts out of complete reviews to posting theories that seem to have no real-world end user implications. So again, lets see all those applications running "miles" better on an FX.

 

[MiddleOfTheRoad]

As is 99% of software a user will use.
Benchmarks are supposed to show you what to expect in real life,and in real life everything will be written on intel compilers.

Windows market share is not 99%.

 

[TheELF]

I'm not exactly sure PCMark can throw around the word neutral, either.... Considering the PCMark2005 scandal deliberately distorting scores for Via and AMD Cpu's: http://www.osnews.com/story/22683/I...quot_Cripple_AMD_quot_Function_from_Compiler_

Well what do you expect?
Intel produces intel cpus and knows them inside out ,they have no idea how amd implemented any extra command sets,so of course with anything else than intel they use the most compatible commands,why? Because it's not some random public domain compiler but the industry standard, they have to make sure that factories don't explode and satellites don't crash back to earth.

 

[TheELF]

Windows market share is not 99%.

Intel C++ Compiler, also known as icc or icl, is a group of C and C++ compilers from Intel available for OS X, Linux, Windows and Intel-based Android devices.

http://en.wikipedia.org/wiki/Intel_C++_Compiler

 

[MiddleOfTheRoad]

http://en.wikipedia.org/wiki/Intel_C%2B%2B_Compiler

And again -- you're acting like it's the only Compiler in existence. There are many, many others.... Microsoft even makes one. If I'd had to guess, I'd say ICC is used for maybe 30% of software (being generous).

And the reason why ICC sucks for benchmarks -- from the exact same page you just linked (and probably didn't read):

"Intel was caught in a case of suspected "benchmarksmanship", when it was shown that the object code produced by the Intel compiler for the AnTuTu Mobile Benchmark omitted portions of the benchmark in order to show increased performance compared to ARM platforms."

The numbers are pretty much cooked for anything on Intel hardware -- so benchmarks written on ICC are a total scam.
If you want an accurate hardware benchmark -- run something purely synthetic.

 

[MiddleOfTheRoad]

Well what do you expect?
Intel produces intel cpus and knows them inside out ,they have no idea how amd implemented any extra command sets,so of course with anything else than intel they use the most compatible commands,why? Because it's not some random public domain compiler but the industry standard, they have to make sure that factories don't explode and satellites don't crash back to earth.

Try again. Re-Read The Posted Article for great quotes like:

"Ars found out that by changing the CPUID of a VIA Nano processor to AuthenticAMD you could increase performance in PCMark 2005's memory subsystem test by 10% - changing it to GenuineIntel yields a 47.4% performance improvement!"

They cooked the compiler to be biased to specific CPUID's. Change the name to Intel -- and suddenly, the exact same hardware is miraculously 47% faster. This is not an isolated incident. These marketing games are exactly why purely synthentic benchmarks are the only ones I accept. It is exactly why both AMD and NVIDIA pulled out of the benchmark consortium. There is a not a lot of common ground between those two companies....... But we found at least one.

 

[ShintaiDK]

Dear AMD,
Using Cinebench to show the performance of your chips is complete BS.

Its pretty bad attempt as excuse when even the company they try to defend says they are wrong.

 

[MiddleOfTheRoad]

Its pretty bad attempt as excuse when even the company they try to defend says they are wrong.

Shintai, if Intel stated the world was flat -- you'd believe them. You do realize everyone already knows what you are going to say before you say it.
Intel should add you to the payroll (if they haven't already). Carry on.

Personal attacks are not allowed here.
Markfw90

 

[2is]

Shintai, if Intel stated the world was flat -- you'd believe them. You do realize everyone already knows what you are going to say before you say it.
Intel should add you to the payroll (if they haven't already). Carry on.

What I find funny is you entered this thread with a thinly veiled disguise to sound objective, but as soon as you were called out on your extremely optimizing outlook on AMD performance you abandoned that approach and have made as many excuses and tin foil hat conspiracy theories as anyone, all in a span of less than 24 hours.

I'm still waiting for those benchmarks btw. I'm sure you have them buried in your favorites somewhere right? Oh don't forget, benchmarks of actual applications. A Sandra integer benchmark isn't very useful unless that performance can be applied to something people actually run.

 

[TheELF]

Try again. Re-Read The Posted Article for great quotes like:

"Ars found out that by changing the CPUID of a VIA Nano processor to AuthenticAMD you could increase performance in PCMark 2005's memory subsystem test by 10% - changing it to GenuineIntel yields a 47.4% performance improvement!"

They cooked the compiler to be biased to specific CPUID's. Change the name to Intel -- and suddenly, the exact same hardware is miraculously 47% faster. This is not an isolated incident. These marketing games are exactly why purely synthentic benchmarks are the only ones I accept. It is exactly why both AMD and NVIDIA pulled out of the benchmark consortium. There is a not a lot of common ground between those two companies....... But we found at least one.

1st off ... an 2008 article?Seriously!?
Did you read my post you quoted?
Did anyone check if with the better numbers you have worse stability?
Intel puts its name on its compiler and has to be able to guarantee 100% proper execution,how are they going to guarantee this on someone else's cpu?


And even if everything you say is correct,why in the world would anybody buy a cpu that is being buried instead of a cpu that is always going to run with top speed?
Everything you say is a huge selling point for intel cpus.

 

[Enigmoid]

@Middle of the Road

We already had this discussion about the 3770k vs. 8350 on linux but it was removed. I will repost.

 

[MiddleOfTheRoad]

@Middle of the Road

We already had this discussion about the 3770k vs. 8350 on linux but it was removed. I will repost.

Exactly -- So riddle me this: Why does an FX-8350 run head to head with an i7 3770k under Linux across the board..... But suddenly the exact same FX CPU barely can keep up with a lowly dual core i3 under Windows? Software bullshiting...... I mean "optimizations." 'Nuff said. Intel has already been caught pulling the same crap with ARM benchmarks.

 

[Enigmoid]

Exactly -- So riddle me this: Why does an FX-8350 run head to head with an i7 3770k under Linux..... But suddenly the exact same hardware barely can keep up with a lowly i3 under Windows? Software bullshiting...... I mean "optimizations." 'Nuff said.

The 3770k is 20% ahead. That is hardly neck and neck. 4 tests out of 25 the 3770k is > 50% ahead.

 

[2is]

Exactly -- So riddle me this: Why does an FX-8350 run head to head with an i7 3770k under Linux across the board..... But suddenly the exact same FX CPU barely can keep up with a lowly dual core i3 under Windows? Software bullshiting...... I mean "optimizations." 'Nuff said. Intel has already been caught pulling the same crap with ARM benchmarks.

The first thing I did after reading this, is question my own self. I've never looked up the meaning of "neck and neck" I only assumed I knew what it meant. I was fully willing to accept that I was interpreting that phrase incorrectly for the majority of my life, so I looked it up.

even in a race, competition, or comparison.
"we have six contestants who are neck and neck"


synonyms: level, equal, tied, side by side, close; informaleven-steven
"going into the third lap, Christian and Perry are neck and neck"

That's essentially what I had always thought it meant.

What does that phrase mean to you?

 

[Burpo]

Come on guys, don't confuse him with facts.. You spoil all the fun!

 

[MiddleOfTheRoad]

The 3770k is 20% ahead. That is hardly neck and neck. 4 tests out of 25 the 3770k is > 50% ahead.

That is disingenuous.

The stock clocked FX-8350 won 8 of the 25 benchmarks.... I'd call that reasonably neck and neck. At the least, competitive. If it didn't win any of the benchmarks and merely got close, I'd agree with you. Keep in mind -- these are chips that supposedly don't compete with i7 according to Intel marketing.... Yet it is actually faster in 8 categories.

The stock clocked FX-8350 was faster than the stock i7 3770k in:
1) NAS Parallel Benchmarks v3.3 - MG.B
2) NAS Parallel Benchmarks V3.3 SP.A
3) John The Ripper V1.6.9-jumbo-7 Blowfish
4) x264 v2012-10-03 H.264 Video Encoding
5) GraphicsMagick v1.3.16 Operation Sharpen
6) Time Linux Kernel Compilation v.3.1 Time To Compile
7) C-Ray v1.1 Total Time
8) Parallel Bzip2 Compression v1.1.6

Yet, somehow this same FX chip suddenly can't keep up with dual core i3's under Windows..... Some would call that "hinky." I call it software bullshiting.

 

[2is]

Soooooo now we have a situation where <33% = neck and neck and comparing overclocked to stock. This just keeps getting better and better. By better I mean laughably ridiculous.

 

[MiddleOfTheRoad]

Soooooo now we have a situation where <33% = neck and neck and comparing overclocked to stock. This just keeps getting better and better. By better I mean laughably ridiculous.

Negative -- the stock clocked FX-8350 was faster in 8 benchmarks versus the stock clocked i7 3770k. I excluded all the results of the overclocked FX in my previous post (if you choose to include the overclocked one -- then the FX actually wins 11 out of the 25 benchmarks versus the i7 3770k).

Again, my point isn't the benchmarks themselves -- but the huge disparity in performance between Windows and Linux for AMD chips. I am pretty confident the Intel compiler is artificially holding the hardware back under Windows. The vast majority of people wouldn't even think an FX-8350 can outrun an i7 3770k in a single benchmark, much less 8. But it does. How? Common belief on this forum is that it performs roughly like an i3.

So how exactly do the non-believers explain that disconnect?