Of game engines, CPU cores and threading...

[BFG10K]

It shouldn't be that much work in reality. the Os is supposed to handle threading task. Look at apple for instance. the Os handles the threading not the software. So it has to run through 2 layers of BS to get to where it needs to go. Total BS and a complete fialure on microsofts ends.

Nonsense; extracting ILP from general-purpose code is a very tough problem and there’s no silver bullet that solves it. An OS can’t magically speed up code that executes serially.

Likewise, simply spinning multiple threads at the program level doesn’t guarantee any kind of speedup on multi-core if the threads have dependencies on each other.

 

[BFG10K]

Try playing ARMA2...it will gobble up CPU cores like no other game.

That game runs piss-poor no matter what CPU is thrown at it.

Sure, there are a handful of games that show meaningful performance gains at reasonable settings, but multi-core proponents focus far too much on them given they’re a drop in the bucket. They also focus on low detail settings that no 4 GHz i7 owner with a high-end GPU setup is ever going to be using.

For the vast majority of games, a reasonably fast dual-core is good enough to saturate the graphics system if you’re running at reasonable detail levels. That’s the fundamental point here.

Add in that it looks like 50% af the games tested are console ports...

I don't follow your reasoning here. Console ports already have some multi-core optimizations because the 360 and PS3 are multi-core systems, so if anything that should help to some degree.

PC exclusives (e.g. Crysis, Stalker) are far more taxing on the GPU than console ports, so they show even less benefit from the CPU than console ports do.

 

[BFG10K]

They use only one graphics card at a high resolution, hence they could easily end up gpu bound and not cpu bound in the test. Pretty much makes the article and the results somewhat pointless.

It’s not pointless at all. It proves yet again that somebody gaming at realistic settings for their target hardware is almost always going to find they're GPU bottlenecked, while their CPU makes little to no difference.

I’ve seen this thing time and time again, both in real-world gaming scenarios and also with benchmarking. The fact is, the importance of the CPU is vastly overstated on the internet.

It’s more than reasonable to expect a 5870/GTX470 owner to be gaming at 1080p or higher. That’s the whole point of purchasing high-end graphics cards. I have a GTX470 and I generally don’t game at anything less than 1920x1200 with 2xAA.

If anything, the article was biased towards the CPU because they didn’t use AA, which again is something that can be reasonably expected to be enabled by owners of such cards.

Test should of been down with tri-fire or tri-sli and at a lower resolution and lower gfx settings.

Uh, what? Do you seriously think tri-GPU owners run at lower settings than a single GPU owner? What then is the point of such systems?

Tell me, do you expect somebody that drops down $1K for an Intel hex-core and another $1K for tri-GPU is the kind of person that will be playing games at 1280x1024 with no AA?

Seriously, if you want to run at such prehistoric settings then buy yourself a cheap passively cooled HTPC GPU.

 

[thedosbox]

Using fallcies...you must be kidding...didn't you just accuse me of the same?!
+ performance = it's using the added core for better performance...not matter how you try and spin it.

Perhaps you ought to read the article before commenting. From the very first paragraph:

PC gaming is often a central concern in the purchase of a new PC, but knowing how many CPU cores to buy - whether it's worth the upsell from two to three or four, or even six, is a key issue.

There's no debate about whether additional cores improve performance, just whether that improvement is worthwhile.

Did you not read the -cpuCount=x part? :thumbsdown:

It's obvious you didn't read the linked post

Pot, Kettle, black, eh?

I call BS when I see it. But it's a fair point that I'm not a moderator, so I'll refrain from continuing to call you on it.

 

[pmurgs]

It’s not pointless at all. It proves yet again that somebody gaming at realistic settings for their target hardware is almost always going to find they're GPU bottlenecked, while their CPU makes little to no difference.

I’ve seen this thing time and time again, both in real-world gaming scenarios and also with benchmarking. The fact is, the importance of the CPU is vastly overstated on the internet.

It’s more than reasonable to expect a 5870/GTX470 owner to be gaming at 1080p or higher. That’s the whole point of purchasing high-end graphics cards. I have a GTX470 and I generally don’t game at anything less than 1920x1200 with 2xAA.

If anything, the article was biased towards the CPU because they didn’t use AA, which again is something that can be reasonably expected to be enabled by owners of such cards.


Uh, what? Do you seriously think tri-GPU owners run at lower settings than a single GPU owner? What then is the point of such systems?

Tell me, do you expect somebody that drops down $1K for an Intel hex-core and another $1K for tri-GPU is the kind of person that will be playing games at 1280x1024 with no AA?

Seriously, if you want to run at such prehistoric settings then buy yourself a cheap passively cooled HTPC GPU.

They whole point I was making, if you take everything I said in context and don't cherry pick what to dispute and then present strawman arguments and trying to read things into what I said that I didn't say, is that the article is trying to see if more cores make a difference. And if so, how much of a difference.

If thats the point, then you want to make sure the test machine will not be gpu bound. Most likely people buying a 980 cpu will have multiple gfx cards.
And as the example I posted, of how my pc was cpu bound with a Q9400 @ 2.66Ghz in BC2, it proves my point that with only a 5850 clocked to 5870 performance, you can still be cpu bound, depending on the game.

I don't know how you managed to become a super moderator with attitude you've just shown to other members of this board.

 

[Dribble]

It’s not pointless at all. It proves yet again that somebody gaming at realistic settings for their target hardware is almost always going to find they're GPU bottlenecked, while their CPU makes little to no difference.

I’ve seen this thing time and time again, both in real-world gaming scenarios and also with benchmarking. The fact is, the importance of the CPU is vastly overstated on the internet.

There's a big difference between:
a) the game uses > 2 cores.
b) you can get by with a dual core if it's fast enough.

There are two types of people who are reading that article that really care:
a) people with existing cpu's (dual or quad) who want to know how badly their current cpu is limiting them. Is it's worth them buying that new 5870 or do they need a new cpu to make it worth it? Perhaps they could upgrade their dual core 2 cpu to a quad, would that make the difference? In which case only benching with the fastest cpu on the planet isn't much help. e.g. I have an old Q6600 @ 3.2, if I buy a flash new graphics card will I see the benefit?

b) people looking to buy a new machine. Remembering people generally upgrade their cpu's much less often then their graphics cards is it worth them going for a quad, or will a dual cpu be enough? Obviously if it looks like quads are used then that makes it certain you want a quad, if not for today then in a little while when their new i? cpu isn't fast enough as a dual core any longer. An article that hides that information by using a cpu that's so fast it's gpu bottlenecked as a dual core isn't helpful.

 

[zebrax2]

They whole point I was making, if you take everything I said in context and don't cherry pick what to dispute and then present strawman arguments and trying to read things into what I said that I didn't say, is that the article is trying to see if more cores make a difference. And if so, how much of a difference.

If thats the point, then you want to make sure the test machine will not be gpu bound. Most likely people buying a 980 cpu will have multiple gfx cards.
And as the example I posted, of how my pc was cpu bound with a Q9400 @ 2.66Ghz in BC2, it proves my point that with only a 5850 clocked to 5870 performance, you can still be cpu bound, depending on the game.

I don't know how you managed to become a super moderator with attitude you've just shown to other members of this board.

Your argument is also flawed because it does not show that your were core bound. It merely shows that you were CPU bound which could be from having a low cpu speed.

 

[ModestGamer]

Nonsense; extracting ILP from general-purpose code is a very tough problem and there’s no silver bullet that solves it. An OS can’t magically speed up code that executes serially.

Likewise, simply spinning multiple threads at the program level doesn’t guarantee any kind of speedup on multi-core if the threads have dependencies on each other.

Before you flap those gums again. I am saying the OS should be handling threads not the game.

 

[pmurgs]

Your argument is also flawed because it does not show that your were core bound. It merely shows that you were CPU bound which could be from having a low cpu speed.

Yes, your right. I could of been core bound. I personally suspect BC2 needs at least 3 cores to run well and a high clock speed to really shine, because I'm guessing there is only one thread (thus on one core) thats ultimately feeding the graphics card.

In this specfic case, the games 30 odd threads still use 100% of my processing power, at 2.66Ghz and 3.9Ghz.

If the review on how many cores you need, had taken away the chance of being gpu limited, then it would of been easier for me to see at what point, more cores do not help the game. If they had given us different runs for at a low/medium/high Ghz core speed, that might of helped understand better whether a game is core or Ghz limited, or what kind of combination of both.

 

[zebrax2]

Was turbo left ON on the review?

 

[edplayer]

At the moment Core i5 750 is just so much better at running two threads compared to Clarkdale.

Can you explain this more? I don't understand

 

[cbn]

Can you explain this more? I don't understand

http://www.anandtech.com/show/2901/1

In the Anandtech review 2.66 Ghz Core i5 750 is getting higher FPS than 3.3 Ghz Core i5 661 in classic dual threaded games. I know Turbo modes are also involved, but that doesn't explain everything.

 

[Schmide]

Before you flap those gums again. I am saying the OS should be handling threads not the game.

That doesn't make sense? The only thing the OS does is context switches and partitioning. It is up to the program to control how one thread works with another. If you don't manage your data dependencies, you'll gum up your flapping.

 

[BFG10K]

They whole point I was making, if you take everything I said in context and don't cherry pick what to dispute and then present strawman arguments and trying to read things into what I said that I didn't say, is that the article is trying to see if more cores make a difference. And if so, how much of a difference.

Okay, so we benchmark a tri-GPU system at 1280x1024 with no AA and show some differences with cores. Then by running at realistic settings (like the article), such differences are nullified. Tell me then, how are the 1280x1024 results relevant in any way?

Most likely people buying a 980 cpu will have multiple gfx cards.

Right, and such people will also be running at 2560x1600 with 4xAA or 8xAA. They won’t be running at 1280x1024 with no AA can so they show six cores getting 250 FPS while four cores “only” get 200 FPS. You don’t drop down $2K-$3K on a system to do that.

And as the example I posted, of how my pc was cpu bound with a Q9400 @ 2.66Ghz in BC2, it proves my point that with only a 5850 clocked to 5870 performance, you can still be cpu bound, depending on the game.

That’s lovely, but how does that disprove the article, or my statements about using appropriate detail levels for a GPU system?

I don't know how you managed to become a super moderator with attitude you've just shown to other members of this board.

Why the personal insults and the mod-callout? Are you angry or something?

 

[BFG10K]

Before you flap those gums again. I am saying the OS should be handling threads not the game.

The OS already manages threads as far as distributing them across available CPU resources goes. But an OS can’t make a game multi-threaded, nor can it remove inter-thread dependencies from said game.