can someone with BD/ZBZ check SC2's core thread scheduling?

[soccerballtux]

Just wondering if it runs on threads 1 and 3 (diff modules) or 1 and 2 (same module).

Also, is there any way of profiling this to figure out what exact hardware advantage the Intels have over the AMDs? I still feel pretty strongly that there might be some unfair compiler optimizations (http://arstechnica.com/gadgets/2008/07/atom-nano-review/6/) going on

 

[BrightCandle]

Microsoft released a patch for the scheduling that treats the modules like ht. That is that odd cores are used in preference first. StarCraft will behave just like any other app on windows or will use the real cores first.

 

[KingFatty]

I believe the physics engine for SC2 is based on Havok, which has some link with Intel (does Intel own Havok?). Not sure if that means anything, but interesting in that it could be the same model that NVidia practices with PhysX and having it run so much better on NVidia cards; perhaps there is some Intel optimization for Havok (just speculation here)?
But, don't forget to overclock the CPU-NB, separately from the CPU. It really helps SC2. At least this was true for Phenom II processors, I don't know anything about the FX chips.

 

[inf64]

If it's similar to Nv's model than it's x87 dependent. But I doubt that's the case since this is intel and not Nvidia we are talking about.

 

[bunnyfubbles]

so I guess AMD is just trying to save face by trying not to look like whiners by not saying anything first themselves, and not doing it when BD first came out last year?

 

[inf64]

It's only in a few titles we see this perf. discrepancy. It would be fun to see if there was ICC compiler check in the compiled exe of the game/app :

 

[KingFatty]

Are there similar discrepancies btw Intel CPU performance and AMD CPU performance on other games that rely on Havok to do the physics?

Is SC2 the only game that is really affected by use of Intel vs AMD, and other games that use Havok show no difference in performance regardless of CPU?

 

[inf64]

SC2 is the most popular game that has this "behavior" but there may be some more titles that exhibit this. I know that Metro 2033 is the exact opposite,it scores a lot better on Vishera and even 8150 than it does on SB/IB. As to why nobody knows for sure,it's not even that well threaded title.

 

[beginner99]

SC2 heavily relies on single threaded performance and at that resolution it will be CPU bound. Hence the large difference since IB easily is 40-50% faster single-threaded than vishera. And that is exactly the difference we see. So this is pretty much what you you expect.

Note that sc2 is probably the only game that will get cpu limited even on the fastest cpu with tri sli or whatever. 4v4 and zergling spam...

 

[soccerballtux]

Microsoft released a patch for the scheduling that treats the modules like ht. That is that odd cores are used in preference first. StarCraft will behave just like any other app on windows or will use the real cores first.

SC2, built on the WoW engine platform, I believe will act like WoW does, which is to say WoW manually handles the thread scheduling with a bitmapped variable that controls which "cores" get scheduled to. It overrides any Windows scheduling for the WoW threads.

 

[soccerballtux]

It's only in a few titles we see this perf. discrepancy. It would be fun to see if there was ICC compiler check in the compiled exe of the game/app :

I wonder if we could get a Via Nano system and try the same thing that they tried with PCMark05, by writing GenuineIntel to those registers. We wouldn't be able to test how much faster an AMD processor would be, but at least we'd be able to see if WoW and SC2 are playing unfair.

 

[KingFatty]

Does WoW use a physics engine at all?

I don't see WoW listed as a game that uses Havok:
http://en.wikipedia.org/wiki/Havok_(software)

 

[Ventanni]

I'd say it's likely attributed to single thread performance since, like WoW, SC2 only utilizes two cores, but I do find it strange that both the 8350 and the 8150 perform exactly the same.