Does performance of SLI and Crossfire depends on number of CPU cores

[Vineet Reddy]

Does performance of having SLI or Crossfire depend on number of CPU cores. I have heard from many sites that transferring data between multiple GPU's at same time is hard for CPU's with less number of cores which negates the performance improvement achieved by SLI. They say having more cores do give better performance with SLI and Crossfire as transfer of data can be handled better if distributed among different cores than stressing on limited cores.

 

[cdawall]

No.

 

[cadaveca]

No.

Right, and things like Mantle prove AMD is striving to make CPU bottlenecks even less of a potential problem.

 

[BarbaricSoul]

^what cdawall said

How it was explained to you is not correct at all. As long as the CPU is strong enough for the game in question, reguardless of core count, it will not hamper SLI or crossfire performance.

 

[cdawall]

Right, and things like Mantle prove AMD is striving to make CPU bottlenecks even less of a potential problem.

Reviews even show it.

 

[GhostRyder]

Does performance of having SLI or Crossfire depend on number of CPU cores. I have heard from many sites that transferring data between multiple GPU's at same time is hard for CPU's with less number of cores which negates the performance improvement achieved by SLI. They say having more cores do give better performance with SLI and Crossfire as transfer of data can be handled better if distributed among different cores than stressing on limited cores.

No, even games that are heavily threaded don't benefit much from more cores beyond 4. You might see some slight increasing or decreases depending on the number of cores, but were talking within the margin of error and even a small % difference.

 

[...PACMAN...]

@cdawall Which site is that from? A 5960x getting beaten by an FX 8370.....that doesn't seem right?

 

[cdawall]

@cdawall Which site is that from? A 5960x getting beaten by an FX 8370.....that doesn't seem right?

http://www.technologyx.com/featured/amd-vs-intel-our-8-core-cpu-gaming-performance-showdown/

 

[erocker]

All of those benchmarks are CPU dependent games. Also, there's something wrong with the reviewers setup(s).. dunno why he's getting such low frames in some of the tests. That being said, the speed/power of the actual cores and overall architecture of the CPU has more to do with SLI/CrossFire performance than the amount of cores.

 

[cdawall]

All of those benchmarks are CPU dependent games. That being said, the speed/power of the actual cores and overall architecture of the CPU has more to do with SLI/CrossFire performance than the amount of cores.

Honest question did you state that backwards or say the FX was better than the 5960x?

 

[Bill_Bright]

No, even games that are heavily threaded don't benefit much from more cores beyond 4.

But that suggests if you have 2 cores or a single core processor, that performance differences compared to a quad is negligible. And that may or may not be the case.

So for everyone above who said "no". I respectfully disagree and say the answer is, "it depends". It depends on where you are starting from. That is, for example, specifically which dual core and which quad core - and which game too.

So for Vineet Reddy, since you did not specify in your opening post the specific CPUs, how many cores, their speeds and which games, so the answer is, "Maybe, maybe not. It depends". If comparing 1 or 2 cores to a quad core CPU, then yes, it could make a difference. If comparing a quad to a 6 or 8 core, probably not.

To further add to the muddiness of this, it depends on the game too. And not just the game, but the age of the game (many newer games take advantage of more cores).

And finally, (if not most importantly), it depends on the speeds of the processors in question.

So bottom line is the OP's question is too vague and generalized for a yes or no answer. So instead, the correct answer is, "it depends".

 

[erocker]

Honest question did you state that backwards or say the FX was better than the 5960x?

Nope. I didn't state any CPU over the other.

 

[cdawall]

But that suggests if you have 2 cores or a single core processor, that performance differences compared to a quad is negligible. And that may or may not be the case.

So for everyone above who said "no". I respectfully disagree and say the answer is, "it depends". It depends on where you are starting from. That is, for example, specifically which dual core and which quad core - and which game too.

So for Vineet Reddy, since you did not specify in your opening post the specific CPUs, how many cores, their speeds and which games, so the answer is, "Maybe, maybe not. It depends". If comparing 1 or 2 cores to a quad core CPU, then yes, it could make a difference. If comparing a quad to a 6 or 8 core, probably not.

To further add to the muddiness of this, it depends on the game too. And not just the game, but the age of the game (many newer games take advantage of more cores).

And finally, (if not most importantly), it depends on the speeds of the processors in question.

So bottom line is the OP's question is too vague and generalized for a yes or no answer. So instead, the correct answer is, "it depends".

The Pentium G3580 is a perfect example of how it really doesn't the architecture will make as much difference as number of CPU threads. The age old intel limitation of 16/4 for slot breakdown is a prime example, FSB before that.

Nope. I didn't state any CPU over the other.

I was under the impression those were GPU bound not CPU bound. Hence why CPU change made next to no difference.

 

[newtekie1]

In the summer months I run my 4790K almost exclusively as a dual-core. I have yet to come across a game that this makes a noticeable difference on with my 970s in SLI.

 

[neatfeatguy]

Like others have said, it doesn't depend on how many cores, but the power of the CPU in question. If you want an idea (though the CPU and GPUs I had are kind of dated by today's standards) of what to expect from going from an older CPU (Phenom II x4 940) to a newer CPU (i5-4670k), you can check my post here.

It'll show you the improvement I gained on a couple of games running 570s in SLI after going from an OC'ed PII x4 to a stock i5-4670k. I went from 4 cores to 4 cores. The only difference is the sheer power that the i5 has over the PII.

 

[GhostRyder]

But that suggests if you have 2 cores or a single core processor, that performance differences compared to a quad is negligible. And that may or may not be the case.

So for everyone above who said "no". I respectfully disagree and say the answer is, "it depends". It depends on where you are starting from. That is, for example, specifically which dual core and which quad core - and which game too.

So for Vineet Reddy, since you did not specify in your opening post the specific CPUs, how many cores, their speeds and which games, so the answer is, "Maybe, maybe not. It depends". If comparing 1 or 2 cores to a quad core CPU, then yes, it could make a difference. If comparing a quad to a 6 or 8 core, probably not.

To further add to the muddiness of this, it depends on the game too. And not just the game, but the age of the game (many newer games take advantage of more cores).

And finally, (if not most importantly), it depends on the speeds of the processors in question.

So bottom line is the OP's question is too vague and generalized for a yes or no answer. So instead, the correct answer is, "it depends".

That's why I stated "Beyond 4" which is the ideal amount of cores for a gamer IMHO. 2 cores can run games fairly well as well even in an SLI situation (Mostly involving i3's and the G3258 overclocked). But my point was that 4 was ideal and anymore really makes no difference, I did not insist that a single or dual core is just as good as a quad in gaming.

In the summer months I run my 4790K almost exclusively as a dual-core. I have yet to come across a game that this makes a noticeable difference on with my 970s in SLI.

With or without HT? Just curious...

 

[Bill_Bright]

The Pentium G3580 is a perfect example of how it really doesn't the architecture will make as much difference as number of CPU threads. The age old intel limitation of 16/4 for slot breakdown is a prime example, FSB before that.

That's why I stated "Beyond 4" which

Those examples illustrate my point. Because Vineet Reddy (the OP) did not specify the processors or the game, the answer must be unspecific too.

If you cherry pick specific CPU examples to compare, you can illustrate how a specific quad will blow the socks off a specific dual on a particular game.

Likewise, you can cherry pick a specific dual that will blow the socks off a specific quad in a different specific game.

And, you can cherry pick a game that does not care at all.

We need specifics to give a specific answer.

 

[cdawall]

Those examples illustrate my point. Because Vineet Reddy (the OP) did not specify the processors or the game, the answer must be unspecific too.

If you cherry pick specific CPU examples to compare, you can illustrate how a specific quad will blow the socks off a specific dual on a particular game.

Likewise, you can cherry pick a specific dual that will blow the socks off a specific quad in that specific game.

And, you can cherry pick a game that does not care at all.

We need specifics to give a specific answer.

Most of us were a part of his "server gaming rig" thread and are basing our posts off of that.

 

[newtekie1]

With or without HT? Just curious...

With HT, but I was running without HT until FarCry 4...