AMD vs NV (290 vs 780): CPU bottleneck BF4

[beginner99]

So I've read about dx command lists (multithreading) and that only nv has this feature. My question now is how much better nv cards work with older cpus (I have a i7-870, lynnfield), especially in BF4 multiplayer.

I've googled but did not find any benchmarks especially with older CPUs. I do not care about single-player. Can anyone point me to such a link or has personal experience with cards from either vendor on an older cpu?

I'm thinking about a new gpu. for AMD speaks mantle and cheaper, for NV gsync (this upcoming asus 1440p screen sounds interesting) and maybe as said above better MT support.

 

[Gloomy]

Nvidia has better performance when CPU-limited. As for how that compares with Mantle, we don't know yet. If you can't wait for Mantle then Nvidia is the best choice.

 

[krumme]

The consoles run on 6 1.67ghz weak jaguar cores. Mantle will hands down beat nv solution for low cpu usage in bf4 and by a long stretch for weaker cores. Mantle is excactly what you want for your cpu. You can put my answer in your sig if thats not going to be the case. Going nv setup with your cpu now is stupid

 

[VulgarDisplay]

Does being CPU limited effect max fps or minimum FPS?

It seems that generally nvidia has proportionally lower minimum fps compared to their max fps if that means anything.

I'm not saying that nvidia's minimums are always lower than AMD's, but that they drop much farther in comparison to their max's than the latest AMD cards in the reviews I just looked at.

 

[3DVagabond]

Does being CPU limited effect max fps or minimum FPS?

It seems that generally nvidia has proportionally lower minimum fps compared to their max fps if that means anything.

I'm not saying that nvidia's minimums are always lower than AMD's, but that they drop much farther in comparison to their max's than the latest AMD cards in the reviews I just looked at.

That's more likely to do with memory bandwidth than CPU bottleneck. AMD cards generally have wider a memory buss.

 

[VulgarDisplay]

That's more likely to do with memory bandwidth than CPU bottleneck. AMD cards generally have wider a memory buss.

So the better CPU utilization allows for higher max fps then?

 

[Leadbox]

It seems that generally nvidia has proportionally lower minimum fps compared to their max fps if that means anything.

I'm not saying that nvidia's minimums are always lower than AMD's, but that they drop much farther in comparison to their max's than the latest AMD cards in the reviews I just looked at.

Thats boost 2.0 at work right there, it flies and throws up huge maximums (usually when there's nothing going on) and it tanks a good bit when you put on a bit of load. That could explain [H]'s subjective assertion that BF4 was smoother on the 290 than the 780 because the radeons run in a much tighter band for their minimums and maximums

 

[Carfax83]

So I've read about dx command lists (multithreading) and that only nv has this feature. My question now is how much better nv cards work with older cpus (I have a i7-870, lynnfield), especially in BF4 multiplayer.

I've googled but did not find any benchmarks especially with older CPUs. I do not care about single-player. Can anyone point me to such a link or has personal experience with cards from either vendor on an older cpu?

I'm thinking about a new gpu. for AMD speaks mantle and cheaper, for NV gsync (this upcoming asus 1440p screen sounds interesting) and maybe as said above better MT support.

If you're just going to be playing BF4, may as well go AMD, as Mantle will help you more than anything when it's finally released. Mantle will work best on lower spec CPUs, by removing lots of overhead..

BF4 doesn't even support DX11 multithreading anyway. A game has to be developed to use that particular feature, and insofar, few developers have done so; most likely due to AMD not supporting it.

NVidia's advantage lays primarily in how they optimize their actual driver code for multicore processors. Their optimizations are much more aggressive than AMD's, leading to less CPU bottlenecks with CPUs that can use four threads and more.

 

[Gloomy]

Thats boost 2.0 at work right there, it flies and throws up huge maximums (usually when there's nothing going on) and it tanks a good bit when you put on a bit of load. That could explain [H]'s subjective assertion that BF4 was smoother on the 290 than the 780 because the radeons run in a much tighter band for their minimums and maximums

That's a cool feature. Basically like single core boost for Intel chips? Neat.

 

[Carfax83]

Does being CPU limited effect max fps or minimum FPS?

It seems that generally nvidia has proportionally lower minimum fps compared to their max fps if that means anything.

I'm not saying that nvidia's minimums are always lower than AMD's, but that they drop much farther in comparison to their max's than the latest AMD cards in the reviews I just looked at.

I've never noticed that at all. If anything, NVidia's GPUs perform exceptionally well, better than their AMD counterparts given that they don't have similar bandwidth for the most part. Kepler is a very efficient architecture, make no mistake.

What could cause that though, is less VRAM. The 2GB GTX 770 has lower minimums than the 4GB version in certain games that use lots of VRAM.

So can boost for that matter, but unless you run your machines in a very warm environment, I don't think it should make that much of an impact.

 

[3DVagabond]

So the better CPU utilization allows for higher max fps then?

No, I'm talking about why AMD might have higher minimums. As Leadbox said, that could be attributed to nVidia's boost. It sounds reasonable.

 

[desprado]

Go for GTX 780.
Enough said

 

[BrightCandle]

There does seem to be a small advantage to Nvidia cards with lower amounts of CPU performance, but its a relatively small advantage. We know that Nvidia cards show wider CPU usage than AMD and sometimes this brings a decent performance advantage, but not in all games. I don't think its a good reason to buy one card over another, its too small a difference to be worth choosing on the basis of, especially since with each driver release the situation changes and I haven't seen a test of this for over a year.

 

[VulgarDisplay]

I've never noticed that at all. If anything, NVidia's GPUs perform exceptionally well, better than their AMD counterparts given that they don't have similar bandwidth for the most part. Kepler is a very efficient architecture, make no mistake.

What could cause that though, is less VRAM. The 2GB GTX 770 has lower minimums than the 4GB version in certain games that use lots of VRAM.

So can boost for that matter, but unless you run your machines in a very warm environment, I don't think it should make that much of an impact.

I'm talking about the 780ti vs 290x though. 290x has similar and sometimes better minimums even though the 780ti has higher max fps.

 

[blastingcap]

There does seem to be a small advantage to Nvidia cards with lower amounts of CPU performance, but its a relatively small advantage. We know that Nvidia cards show wider CPU usage than AMD and sometimes this brings a decent performance advantage, but not in all games. I don't think its a good reason to buy one card over another, its too small a difference to be worth choosing on the basis of, especially since with each driver release the situation changes and I haven't seen a test of this for over a year.

This

 

[Carfax83]

There does seem to be a small advantage to Nvidia cards with lower amounts of CPU performance, but its a relatively small advantage. We know that Nvidia cards show wider CPU usage than AMD and sometimes this brings a decent performance advantage, but not in all games. I don't think its a good reason to buy one card over another, its too small a difference to be worth choosing on the basis of, especially since with each driver release the situation changes and I haven't seen a test of this for over a year.

According to this PClabs.pl review, NVidia drivers scale best on at least four threads, provided the game can use that many threads to begin with. So either a dual core with hyperthreading, or a straight quad core for good performance.

For optimal performance, quad or hex core processors with hyperthreading can help in certain games like Crysis 3 and Battlefield 4, as both of those games can use up to 8 threads.

 

[parvadomus]

I would get an AMD video card for BF4. The mantle API should make the difference specially improve the CPU performance. You can always wait some days and see how it performs ofc.

 

[OCGuy]

You would have to be crazy to rush to purchase a GPU specifically for Mantle when all signs point to an imminent release.

Then just check a review that reflects your CPU and monitor resolution and make an informed decision. :thumbsup:

 

[parvadomus]

According to this PClabs.pl review, NVidia drivers scale best on at least four threads, provided the game can use that many threads to begin with. So either a dual core with hyperthreading, or a straight quad core for good performance.

For optimal performance, quad or hex core processors with hyperthreading can help in certain games like Crysis 3 and Battlefield 4, as both of those games can use up to 8 threads.

I couldnt extract a proper conclusion from that review. You think NV drivers are threaded cause it performs better with more lower performance cores? Thats too bad, it should perform almost the same as the raw CPU perf is the same (context switches might penalize performance a bit but not THAT BAD).
I would want to see more reviews about this.

 

[Erenhardt]

Ok. I see how much and how often this pclab article is confusing people. I will try translate the conclusions and what it is all about.

First and for the most the question its trying to answer is what CPU should you get for gaming. Two core, or 4 core.

One of the points was to bust or confirm a myth that you shouldn't buy nv card and 2 core cpu, because it will be held bach by that CPU, opposite to amd card.

Why 2 fast cores and 4 slow cores. Two reasons:
1. To maintain the same CPU processing power
2. To put nv cards in relatively good light by implying that 2 fast core intel CPU can be changed for 4 core amd CPU within the same budged - you dont need to jump to 4 core intel cpu that costs a lot more than 2core. It is relevant here - in post communist eastern EU country.

After game benchmarks there is quick roundup:
Maintaining the same CPU power - in the sense of number_of_CPU_coresxCPU_Speed, even in games that can fully utilize 4 core CPUs, 2 faster (4,6GHz) cores never run slower than on 4 slow cores (2,3GHZ). But that is only truth for amd cards. Someone using nvidia card should avoid pairing it with 2core 2 thread CPU.

In the final conclusian there are 3 parts:
1. What card with 2core, 2 thread CPU:
AMD, pairing 2 core CPU with nvidia is bad idea.
2. What card with 4 core ,4 thread CPU:
Nvidia, because 4 core CPU will not hold it back
What card with 8 thread CPU:
Nvidia, because 8 core CPU will not hold it back

In BF4 single player 290X is faster than gtx780, but in multiplayer is slower. Conclusion: amd cards are more CPU bound in BF4 MP.

The conclusions are bit flawed. And everything is based on multiplayers scenario, which was pointed out numerous times as inconsistent and in no way representative.

 

[beginner99]

You would have to be crazy to rush to purchase a GPU specifically for Mantle when all signs point to an imminent release.

Then just check a review that reflects your CPU and monitor resolution and make an informed decision. :thumbsup:

Yeah that was my initial idea but mantle us delayed and for all we know it will be march till we see any benchmarks and it is actually released.

 

[OCGuy]

Yeah that was my initial idea but mantle us delayed and for all we know it will be march till we see any benchmarks and it is actually released.

If in fact the rumors are true that Q1 was put on the latest slide for a reason.....I would take Mantle out of the equation completely.

 

[Erenhardt]

If in fact the rumors are true that Q1 was put on the latest slide for a reason.....I would take Mantle out of the equation completely.

I'm in two minds about it. There is first mantle driver leaked:

Absolutely. Downloading this is most likely innocuous, but there is absolutely no possible benefit (even if you own a Toshiba, no games currently support either tech) and there is always a possible downside when messing with display drivers.

And Johan Andersson negated that bf mantle will be deleyed to feb.

On the other hand is up to 45% performance worth it? Just get a nv card from higher tier. If you are considering G-sync - clearly aimed at wealthy enthusiast, you can afford more expensive and faster card/s aswell.

 

[OCGuy]

I'm in two minds about it. There is first mantle driver leaked:

And Johan Andersson negated that bf mantle will be deleyed to feb.

On the other hand is up to 45% performance worth it? Just get a nv card from higher tier. If you are considering G-sync - clearly aimed at wealthy enthusiast, you can afford more expensive and faster card/s aswell.

Well we are only a few days away from either reading reviews or another delay, so why speculate, right?

 

[Carfax83]

I couldnt extract a proper conclusion from that review. You think NV drivers are threaded cause it performs better with more lower performance cores? Thats too bad, it should perform almost the same as the raw CPU perf is the same (context switches might penalize performance a bit but not THAT BAD).
I would want to see more reviews about this.

The proper conclusion imo is that NVidia drivers scales best on multicore processors that offer at least four threads, whether dual core with hyperthreading, or an actual quad core assuming the actual game can take advantage of those threads.. Processors with eight threads are just extra gravy, and only a select few games can use that processing power..

AMD drivers on the other hand seem to have difficulty going beyond two threads. They still scale, but not to the extent that NVidia drivers do.

The actual game matters a great deal as well. Some games like Rome Total War II or StarCraft 2 do not scale beyond two threads whatsoever. However, most games these days are trending towards at least quad core support, as that's where the technology is heading.

If developers cannot multithread their games, they will go out of business as the next gen consoles rely heavily on multithreading for performance.