AMD Mantle Update for Battlefield 4 delayed again - Pushed to February

Mantle will benfit CPUs more than GPUs. Its the CPU that sends info to the GPU and as such is the first bottleneck.
It's more to take the load off the CPU and get the GPU to do more, sort of like the way PhysX runs on the GPU if you've an Nvidia card, only far more extensive.

Speaking of which, funnily enough Nvidia were actually the ones to initiate this direction with things like PhysX and Cuda, although they did try to abuse it a lot.
 
Last edited:
Speaking of which, funnily enough Nvidia were actually the ones to initiate this direction with things like PhysX and Cuda, although they did try to abuse it a lot.

elaborate on that?
you mean in the sense of money milking?
 
elaborate on that?
you mean in the sense of money milking?

I think he means the fact that as with Mantle only being really available to AMD GPU's, CUDA and Physx only really work on Nvidia GPU's, Correct me if I misunderstood that though :)
 
It's more to take the load off the CPU and get the GPU to do more, sort of like the way PhysX runs on the GPU if you've an Nvidia card, only far more extensive.

Speaking of which, funnily enough Nvidia were actually the ones to initiate this direction with things like PhysX and Cuda, although they did try to abuse it a lot.

Nope not the case. Its more along the lines of utilizing CPU power to the max and have it process data far more efficiently. Its like using more with less(optimization). It doesn't take any load off the CPU rather than actually putting a bigger load on it but still maintain crazy performance(not just FPS).

When the CPU sends the data to the GPU its actually what the Dev wants the GPU to process. It takes this and turns it into whatever the dev desires and amazing things come out of this.

PhysX and Cuda are nothing like Mantle in any direction. PhysX never took off because it was IP and Cuda is great for rendering but OpenCL is starting to take over and it far favors AMD(like a 7870 beats a titan in OpenCL), yet its developed and backed by AMD,Nvidia, and about 5 other software companies.

Mantle was developed by AMD and can be used by any dev and any hardware. Not really comparable tbh.
 
Feel like I have to jump in here.

In layman's terms Mantle eliminates driver stalls by bypassing the driver thread.

Physx was a multithreaded physics SDK built upon CUDA.

CUDA is a GPGPU language for Nvidia CUDA-ready cards (GeForce series 8 or higher).

GPGPU languages are designed to accelerate any general-purpose application by maximising throughput - it's not restricted to graphics.
 
there won't be a 40% boost.
they said there was a 40% boost, they didn't say anything about averages or settings. so my guess is some lower end card had a framerate spike that boosted it 40% for a split second.
can't wait for the amount of bugs when nvidia feels like they need to make their own API, so developers have to deal with 3 different APIs.
i seriously hope mantle will be a failure or dx12 blows it out of the water, else you can get ready for even buggier games.

The test show a pretty decent improvement over DX11, but will this API really be that good? See it for yourself

Kaveri APU test: 14% faster
7970 test: 25% faster
R9 290x test: 58% faster
 
The test show a pretty decent improvement over DX11, but will this API really be that good? See it for yourself

Kaveri APU test: 14% faster
7970 test: 25% faster
R9 290x test: 58% faster

yeah, please don't mention the other parts that were used....
 
yeah, please don't mention the other parts that were used....
Its not hard to find out what they were.
Anyway, look at these (taken from the article linked by warchild)
  • CPU-limited scenario: 40.9% (1080p) and 40.1% (1600p) performance improvement under Ultra settings and 4xAA on the AMD A10-7700K with an AMD Radeon™ R9 290X.
  • GPU-limited scenario: 2.7% (1080p) and 1.4% (1600p) performance improvement under Ultra settings and FXAA on the Core i7-4960X with an AMD Radeon™ R7 260X
  • Average uplift for 1080p: 13.28% (Average of 290X and 260X data on the i7-4960X, A10-7700K, FX 8350 and i5-4670K)
  • Average uplift for 1600p: 11.35% (Average of 290X and 260X data on the i7-4960X, A10-7700K, FX 8350 and i5-4670K)

And also, this is what Dicehunter found
 
Last edited:
So Mantle benefits low-end CPUs more than high-end ones or GPUs for that matter. I wonder how far it'll extend. Could I pair my A6-6400k APU with a 290X and still get good framerates?

To me, it looks like it's a punch aimed far more at Intel than Nvidia.
 
Suddenly the high-end Kaveri boards like Gigabyte's G1.Sniper A88X start to make a lot more sense I guess.

On another note, I've got no need to feel bad about my recent purchase of a GTX 780 then... Yet anyway. ^_^
 
Nice! Shame i'm not at home to try star swarm out.

I find it odd that catalyst 14.1 is released after the games/benchmark is
 
Urgh, time to wait for the driver. I'll get some video made up showing the difference between mantle/non mantle performance when it is out.
 
Last edited:
Don't forget this a barebones demo and Mantle running on it is more like a worse case scenario with not fully finished software on both ends.
 
Don't forget this a barebones demo and Mantle running on it is more like a worse case scenario with not fully finished software on both ends.

i like that your name is red now. fits really well as you don't seem to be praising anything else
 
Back
Top