*PROOF* that a 3930k bottlenecks SLI Titans.

[Jumper118]

this is in on of the most cpu intensive games in a cpu intensive environment. 

[martinrox1568]

You picked THE MOST INTENSIVE POSSIBLE WORKLOAD FOR THE CPU. A 64 player bf3 map. Try the same with single player metro last light and come back to us.

Edited November 17, 2013 by airdeano
Language

[Cacao]

You picked THE MOST INTENSIVE POSSIBLE WORKLOAD FOR THE CPU. A 64 player bf3 map. Try the same with single player metro last light and come back to us.

Ummmm, you mad or something?

[martinrox1568]

Ummmm, you mad or something?

Nah. Just forgot that I'm not in the YouTube comments anymore and have to be more respectable

[airdeano]

this might be weak, but here goes..

 

are you sure it is CPU related? meaning, networking through PCIe lanes develops

buffering lags. i believe your network is filling the pipeline as fast as it can and the

CPU threads (not cores) are pulling the dragging updates into a buffer, until the

"block" of information is assembled, then the CPU core develops and sends

the completed info through to destination.

 

something i did learn was that the instruction of networking through CPU takes a

big challenge in the Intel camp. the AMD instruction seems to be more streamlined

to processing on the fly rather than "building in the buffer". and that's why i believe

TEK Syndicate (Logan) found that the Intel CPUs didn't fair well against the AMD

under a networking scenario or gaming stream.

 

below is a tool for game devs, app builders, and system engineers.. but it goes to

what i am alluding to in CPU performance.

 

Intel® Performance Bottleneck Analyzer

 

The Intel® Performance Bottleneck Analyzer (Intel® PBA) framework seeks to relate knowledge of static assembly and latest performance monitoring techniques to automatically find and prioritize architectural bottlenecks for the Intel® Core™ Processor Family and the Intel® Atom™ processor. The bottlenecks which cannot be explained are prioritized and tagged for further analysis. To accomplish this goal, the tool utilizes performance monitoring data to recreate the hottest paths of instruction execution through a binary. The recreated paths of execution are then passed through an analysis which combines searches for well known code generation issues with knowledge of hundreds of the performance monitoring events. The toolset is designed, written, and maintained by performance engineers who work in the field to resolve performance bottlenecks every day. All features enabled in the toolset have already been used to study and identify performance opportunities on software.

 

told you is was weak..