Visual studio 2013 (exe release) runs slower compare to VS2010

Started by
4 comments, last by xlope01 9 years, 4 months ago

Hello, I convert a project that was in VS 2010 to VS2013, everything went smoothly, but the performance (for the release exe, with same exact code, and optimization settings) drop almost in half compared to VS2010. Also the cpu usage was reduced to half compared to VS2010. Any ideas of what is going on?

In detail:

Both have the same code, SDK, scene, optimization settings.

VS2010 release exe (launched outside (and inside) Visual Studio yields same results)

105fps, cpu usage 8-9% (single thread, 100%(usage) / 12(threads) = 8.3% full using a single thread, makes perfect sense, since the application is single thread for the time being)

VS2013 release exe (launched outside (and inside) Visual Studio yields same results)

65fps, cpu usage 4-5% (it only using around half the power of a single thread)

Advertisement

There are tons of options, plus there are code generation differences between the compilers. Could be lots of things.

For anything performance related, your profiler is your best friend.

If your code is using less than one full core, that means you're giving up CPU time to other processes.

Typical reasons why this would happen:

- Your process is voluntarily waiting for something (Sleep, GPU, disk I/O, network I/O, etc)
- Another process on the system is preempting you (this almost never happens to a noticeable degree on a system with that many cores...unless all of your other cores are in use, or you and another process have restricted your thread affinity to only one core)


Personally, I would investigate any uses of Sleep first.

Is this comparison fair? By that I mean same CPU, same GPU and same drivers?

If the old benchmarks are from an old run and now you're comparing it to the new run; there's a high chance your drivers changed in between. Or you changed the GPU.


Hello, I convert a project that was in VS 2010 to VS2013, everything went smoothly, but the performance (for the release exe, with same exact code, and optimization settings) drop almost in half compared to VS2010.


Is this comparison fair? By that I mean same CPU, same GPU and same drivers?

for a fair test, only the compiler/linker/ide can change, and you must replicate compiler and linker settings as best as possible. no hardware or drivers on the PC used for speed tests can change.

is this the case? if so, the difference is of interest, but probably due to some difference in compiler or linker settings etc that you didn't notice yet.

if not, then comparisons are meaningless. apples and oranges.

Norm Barrows

Rockland Software Productions

"Building PC games since 1989"

rocklandsoftware.net

PLAY CAVEMAN NOW!

http://rocklandsoftware.net/beta.php

It turns out that I forgot to disable CF when I was creating the profile in CCC...

Anyway, now it is 6% faster compared to VS2010 release .exe

This topic is closed to new replies.

Advertisement