5 replies to this topic

[Flyby]

Hi all,
Flyby here. First let me say that among you here I am totally out of my league. I'm just some dumb sumbidge PC gamer who is trying to sort out a PC upgrade path. My gaming interest is pretty narrow. I enjoy PC flight sims, especially combat flight sims. In my world such sims were/are known to be CPU-intensive. Most such sims are older, and only make use of one cpu core. But newer combat flight sims are out or near release that make use of multiple cores. I'm just wanting to be ready with an appropriate processor architecture, as I ponder a new gaming system.

So my question is this; will the i7 series have longer legs for future gaming than the i5? As some incentive, I've got designs on an i7-920 at Micro Center for $200.00 USD. Hard to pass up, but if it's overkill for future gaming engines, I'd like to know.
Perhaps someone here who knows something about gaming engines has an opinion about whether or not the i7 will be more relevant in the near future than the i5 series. Right now it all seems to be about pushing the Ghz to about 3.6 as a sweet spot. But again, that's flight sims. I know there are other pc-gaming genre out there now that make use of multiple cores, but I like what I like. There will be more combat flight sims using multiple cores coming.

thanks for taking a moment to consider my (end-user of your efforts) plight. I await your considered replies.
Flyby out

[alphadog]

(Post was moved to the Lounge since it wasn't a coding/math/physics question.)

As for answering your question, my general approach in buying my custom workstations is to buy near, but not in, the cutting-edge, but not in it, and upgrade more often, rather than try to "futureproof" my purchase. Prices go up exponentially when you start going above the mainstream, and it's usually not worth it because technologies with lots of sizzle come in at about a 1-year cycle.

IOW, I'd rather spend $400-500 on a box now and upgrade again in 9-12 months, versus spending three times that hoping it lasts 2+ years. In terms of overall spend, it breaks-even, but you always have a pretty decent rig...

[Vilem Otte]

Hello,
as for our team working on game engine (and along with that cooperating currently on 3 games, though none of them is flight sim) it is important to have as much cpu cores (threads) as possible, though frequency is NOT so important.

This is due to how we use CPUs. Our engine has been designed as very parallelised application - you have thread(s) for physics, thread(s) for artifical intelligence, thread(s) for audio computation (yeah we really can use ray tracing for computing physics of sound), thread(s) for ray tracing in hybrid rendering solution, etc. ... so it is important to have multi core system (the best quad-core like i7, or even hex-core system) with average frequency (2.4 Ghz is enough imnsho) ... of course you can turn off most effects (it can affect gameplay!) and run our system just on single core Pentium 4 @ 2.4 Ghz.

About memory, RAM memory for CPU isn't really a problem right now (most data fits into it if data streaming from HDD to memory is good written), the bottleneck is graphics card. The same graphics data we have in memory for CPU (for ray tracing) we need on GPU, that means loading them here and streaming them. Though GPU memory is today much more limitted (especially as we use deferred shading on GPU to have faster and better blending of G-buffers generated with ray tracing and those generated by OpenGL).

Anyway to the CPUs - for now and next several years I'm sure to buy an i5 it will be enough for most games ... i7 is still way too expensive (but I think that 2.66 Ghz version is not so expensive - I think I saw it even for $284 thats "just" EUR 199) ... it's all about how much do you want to invest in your new PC?

[rouncer]

To get benefit out of threads you need to have them doing equal work.

I reckon the future is dx11, sporting easy to do displacement mapping, i cant wait - ill have my demo up in here straight away. Probably something that looks like starwars.

[Vilem Otte]

Quote

To get benefit out of threads you need to have them doing equal work.

Of course you need to divide work correctly between them (still thinking about some good adaptive or progressive solution on this) to use them as best as possible. F.e. ID tech 5 has really bad threading work from what I've heard.

Quote

I reckon the future is dx11, sporting easy to do displacement mapping, i cant wait - ill have my demo up in here straight away. Probably something that looks like starwars.

Ah... I presume you mean tessellating the object and then performing displacement mapping. Under OpenGL this was possible even before one or two years, Radeon HD 2900xt that supported it (extension GL_AMDX_vertex_shader_tessellator and later GL_AMD_vertex_shader_tessellator). I think it was possible to use this even in D3D9 using the same tessellator. And as soon as NVidia next generation will have HW tessellators for Direct3D 11 I think this extension will promote to OpenGL core.
Although tessellating models was never in D3D core (nor OpenGL core), so you can support it on few cards, but you also need to not-do-it on cards where extension is not present.

[rouncer]

I faced it, getting real displacements with dx10 stuff is too much trouble than its worth, i cant wait to get dx11 and finally do it properly.

dx9 can do it, amazingly enough, with full sillhuette, but its still more of a backyard job.