Is Core Parking still an issue with WoW and Windows 7?

[Skelly]

I know back in Wrath, WoW would only use two cores even if you had more, and it would take a reg edit to get the third to help out. Just wondering if anyone in the community has any new info on the situation, as I know WoW and SC2 can both use up to 3 cores.

[Fuzzykins]

I know back in Wrath, WoW would only use two cores even if you had more, and it would take a reg edit to get the third to help out. Just wondering if anyone in the community has any new info on the situation, as I know WoW and SC2 can both use up to 3 cores.

They fixed that in like 3.3...

[discombobulate]

The well done, biblical megathread.
http://www.mmo-champion.com/threads/...tweaks...Guide

[vesseblah]

I know back in Wrath, WoW would only use two cores even if you had more, and it would take a reg edit to get the third to help out. Just wondering if anyone in the community has any new info on the situation, as I know WoW and SC2 can both use up to 3 cores.

WoW has been able to use up to about 70 cores since TBC beta when the game was multithreaded but it uses only 3 cores efficiently, anything above that is wasted. People with Intel's Hyper Threading processors had some performance issues that required changing WoW's config.wtf (not registry) but that problem should be fixed somewhere during WLK as well.

[Cilraaz]

The well done, biblical megathread.
http://www.mmo-champion.com/threads/...tweaks...Guide

You mean the guide that's no longer useful since it refers to the now deprecated processAffinityMask CVar? WoW will use any cores available to it, so the processAffinityMask CVar isn't needed to force it to see all available cores. I did some testing in this thread that showed there was no viable gain from messing with processAffinityMask.

[Plasmon]

Well firstly, you are debunking something that was never claimed in that guide thread. The author (strunker) knows that WoW by default recognizes all cores and virtual cores, he was never claiming that the affinity mask is needed to force WoW to see the other cores. His point was that WoW only heavily uses 3 threads so it's better to force it to physical cores 2, 3, and 4, and to skip core 1 so that all of the operating system and background CPU usage (which mostly uses core 1), does not have to compete with WoW for CPU utilization. He was also advocating that it's better to ensure WoW's main 3 threads are spread across physical cores only rather than utilizing hyperthreading and risk loading both threads of one core while leaving another physical core at a reduced load. (although wow doesn't seem to do that according to Wries testing)

Both your test and Wries test in that thread are not thorough enough to come to the strong conclusion that you claim. (which was that disabling core 1 doesn't help) All you should really say is that under your specific test you didn't see any significant FPS difference, but that your results are inconclusive due to major the lack of rigor. I don't blame you for the lack of rigor, I definitely don't care enough to do any kind of thorough testing about this topic, but my point is just that your conclusion is way too strong compared to the depth of your testing.


Here are some of the issues that I see with your test:

1. Browser open with 21 tabs, ok but weren't they all pretty much sitting idle and therefore not utilizing the CPU much? That's much more of a RAM stressor than a CPU stressor.

2. It looks like you didn't test with anything open that puts a moderate load on the CPU, which means the first core was probably nearly idle. I realize that most people won't be putting much extra load on their CPU while playing a game, but some probably do, and also sometimes things like virus scanners or software updaters decide it's a good time to operate which can cause FPS drops due to the competing CPU load. If that stuff is on it's own core and WoW is on the other three cores then maybe WoW wouldn't take as much of a performance hit during those moments and other times where additional programs are running.

3. You didn't have hyperthreading then so you couldn't test it, which was one of the main points in that guide thread. (Testing affinity mask 84 vs 252 vs 255.) Wries has a quad core with HT but didn't test 84 or 252, which are the most important ones to test! (Both disable core 1 and it's virtual thread. 84 is physical cores 2, 3, 4 with no HT. 252 is cores 2, 3, 4, with HT. 255 is all 8 threads.) Additionally he did his tests by using process manager instead of the config.wtf code modification, which may or may not be important, but that guide thread advocates the config modification. The only big conclusion that I think can be drawn from Wries' tests is that test #5 demonstrated that WoW doesn't use more than one thread per physical core at all, but it sometimes switches from the physical thread to the virtual thread of the same core.

4. You only tested in one specific game situation which was standing in Orgrimmar for 60 seconds. A more thorough test would involve multiple scenarios like a flight path test, a raiding test, a battleground/Tol Barad test, and repetition of those, etc.

5. Your test is statistically inaccurate. Your first one was with affinity mask set to 15, and your second one was with no affinity mask modification, which by WoW default is 15 for quad core with no HT. Those two tests should have produced identical FPS results under the same conditions because there is no difference in settings. However, your results were 55 and 58 FPS which proves that the testing conditions were not identical. You would need to do a series of test repetitions to come to an averaged value for each, and after doing that you'd probably get nearly equal results for those two scenarios. The margin of error demonstrated here suggests that without repetition you would be unable to accurately detect FPS differences with any of your other scenarios unless it's a large difference, like a lot more than 5%. The benefit of disabling core 1 might only be in that 5% range, so your test wouldn't detect it.

6. You only did tests with a good computer, and your CPU was highly overclocked. The impact of affinity masks should be more prominent at stock speeds or with slower CPUs because the CPU utilization from background and other non-WoW threads would be significantly higher.



Your testing was a good start, I just don't think you should be so certain about your conclusion or about using your findings to completely discount the guide thread written by "Strunker" as being useless. He might still be right. It's also your word against Strunker's... in his own testing he found that the highest FPS was achieved with affinity mask at 84 (using physical cores 2, 3, 4) compared to 255 (all 8 threads available). His test is probably also lacking in rigor, so I don't know what the real truth is, but I don't see a reason to value his testing any less than yours. I also do get the reasoning behind leaving the most used core for windows and other programs while keeping WoW on the other three, at least until it becomes a bit more multi-thread friendly than it is currently.

[Deleted]

Fixed.... cant count on my hands how long ago.

[Cilraaz]

Well firstly, you are debunking something that was never claimed in that guide thread.

The thread I linked was originally to debunk a separate "guide" that stated disabling core 1 doubled framerate. Yes, that thread claimed double framerate. The testing just seemed to fit this scenario, so I threw it in.

In response to your points:

1. It wasn't meant to be something to stress the CPU. I was simply stating the test conditions.
2. It was a good test condition, as it's the way most people play WoW. I don't know anyone who does anything that would even moderately stress the CPU while trying to get good WoW framerate.
3. As below, the reason his testing worked was a problem with the way WoW handled virtual cores/hyperthreading. That's long since been fixed.
4. I don't have all the time in the world to benchmark a game that can't really be benchmarked well. If you had read my thread, I stated that if the performance gains were going to be anything other than marginal percentage gains, it would have been noticeable in that test.
5. 5.4% margin of error based on the complete lack of control of other people. That's statistically acceptable. If disabling core 1 only gives you that 5%, then it's not even worth worrying about. Again, palpable results are what I was looking for. When I overclocked my i5 750, I gained a 40-80% framerate boost, depending on the scenario. That's a gain worth mentioning.
6. Perhaps it would do more on a bad computer. Unfortunately, I don't have one to test on. Considering the guide author was using a hyperthreaded quad core, I don't think he had bad computers in mind, though.

in his own testing he found that the highest FPS was achieved with affinity mask at 84 (using physical cores 2, 3, 4) compared to 255 (all 8 threads available).

Again, this was a problem with the way WoW handled hyperthreading that was resolved multiple patches ago. If he would have tested physical cores 1-4 vs physical cores 2-4, fps would have been near identical. Again, the guide's point is moot.

[Almighty1]

this thread is touching on some pts that i feel would help a great deal to reduce my lag issues. Unfortunately my knowledge in this area is limited,but i am confident that with some decent instructions i can figure it out. I created a thread about my lag problems a few days ago. If anyone can check it out and shed some light on my situation please do.
also is this advisable to do on a laptop? I am mostly worried about overheating issues