Mulitple Core - x64 Roadmap

[leakyD]

Some discussions here:

http://www.battlefront.com/community/showthread.php?t=90609&highlight=multiple+cores

and here:

http://www.battlefront.com/community/showthread.php?t=92463&highlight=multiple+cores

Are there any options for CMx2 implementation, or is this a CMx3 thing.

Is it even viable development, and instead spend dev time focusing on playability/features/modeling?

Can CM get by for 5yrs (more?) with single core x86 support?

Thinking about it, we'd probably prefer more playability/features/modeling than faster load times/higher pixel counts.

But then, greater playability/features/modeling might necessitate multiple cores, x64 support.

[ClarkWGriswold]

A bit disappointing that multi-core support is so far off (especially considering I'm looking at buying an 8 core machine next). But it seems as though multi-core support is not really mainstream for games right now - especially from small developers like BFC. I imagine that will change over the next few years though, as multi-cores become much more common.

I'm curious, has anyone ever developed any sort of benchmark for CMSF/CMBN? If the CMx2 engine has a built-in way to measure frames per second (I think I saw something in the beta VAARs), a very complex replay file could make for a great way to compare different PC configurations.

[Ranger33]

I would say dual-core is the bare minimum these days for a new PC or laptop. I just glanced over on newegg.com, the only computers not running dual-core or greater are super cheap netbooks, and even most of them have dual-cores. I just upgraded to quad-core myself.

Will Bulge, Eastern Front, etc. all be single core, because by the time they come out a game without multicore support will definitely be behind the curve. Maybe with the funds from CMBN you guys can find some way to get it implemented, whatever that entails. I can't help but think that CPU power is the main bottleneck for CMx2 and without multicore support the engine can't reach its full potential.

[YankeeDog]

I can't help but think that CPU power is the main bottleneck for CMx2 and without multicore support the engine can't reach its full potential.

I don't think so. Based on Steve & Co.'s past comments, I think it's pretty clear that the main bottleneck for CMx2's development is coding time and debugging/testing. There's lots of features they'd like to add to the engine, which current machines would be quite capable of handling with just one core, that they just don't have enough time to code and test. Flamethrowers and fire, for example. Not big deal in terms of processing power or VRAM, but requires all sorts of new weapons, animations, terrain features (what burns and what doesn't), and AI behaviors.

To be sure, additional processing power might make some features easier to add, because to a degree you could throw raw processing power at them and not worry as much about how efficient your code was. At best, a minor advantage, tho.

Steve said a while back that the performance increase for utilizing 2 cores w/ CMx2 was minor, but that theoretically much more substantial gains could be made with 4+ cores. So I'd expect that they'll make multi-core support a higher priority once their target minimum system spec is for a quad core machine -- currently, minimum spec for the game still a Core 2 Duo.

[13]

Combat Mission: CryEngine 3

[Ranger33]

Well, obviously they know more about than I do, I was just thinking in terms of large scale battles, where there is a lot going on under the hood. Those battles could certainly use as many processor cycles as they can get.

Don't pay any attention to me though, I don't have anything to do this weekend so I'm just rambling

[YankeeDog]

Well, obviously they know more about than I do, I was just thinking in terms of large scale battles, where there is a lot going on under the hood. Those battles could certainly use as many processor cycles as they can get.

Certainly processor cycles is more important if you wan to fight very large battles. Based on my experiences with CMSF, with regard to map size and sheer number of models, it actually seems that system memory may more often be the bottleneck than processor speed. Where processor speed really seems to come in is when there's a lot of ordnance flying around -- calculating all those ballistic paths and penetration probabilities seems to be almost exclusively a processor thing.

But if you have a reasonably new rig, you can play reinforced battalion-sized battles, on a 4km x 4km map right now in CMSF. I expect CMBN will allow at least the same. Honestly, anything bigger than this is getting pretty unwieldy to play for me, especially in WWII ETO, where the max. practical direct fire weapons range is generally under 2km.

[Redwolf]

I rather have Charles put in real functionality for me, and I buy a PC that suits my needs. Going multi-core is a major timesink and going *effective* multicore an even bigger one. This isn't single-programmer material.

I couldn't care less about losers who don't understand this and buy PCs with many slow cores. There are very few desktop applications where that is a good strategy.

What might be more doable is offloading some computation to the graphics card since the actual GPU is probably underutilized in CMx2.