Source
Source
When mantle released I was impressed at how good it was. Now you see Nvidia's regular drivers getting close to mantle performance but this works in all games and not just the hand full of mantle supported titles.Originally Posted by sumitlian
Man, AMD DX drivers are literally horrible than Nvidia.
Not sure if the "You" you're implying in your sentence is toward me or its for everybody.Originally Posted by Derp
When mantle released I was impressed at how good it was. Now you see Nvidia's regular drivers getting close to mantle performance but this works in all games and not just the hand full of mantle supported titles.
Now when you look at results like this you end up thinking "wow AMD's DX drivers are terrible" instead of "wow Mantle is amazing". I don't think this was AMD's intention.
Huh? Slipping behind? its a 3 FPS diff, with the AMD chip clocked way higher......I would not call that slipping. Keep in mind also how old is SB now? And how much did both cost at launch? Yea, not something id be proud of......Originally Posted by SoloCamo
Interesting to see the 2500k start slipping behind slightly after all this time even compared to the higher clocked AMD's... Nothing a 4ghz+ OC wouldn't fix but still.. At least I can say my 9590 is quicker than the 2500k (at stock) for once!Speaking of which, it's interesting how the 8350 performs better than the 2500k with the NV drivers but not mantle nor DX.![tongue.gif]( "tongue.gif")
edit: and LOL at the 8150 getting trumped pretty badly even by a 6300.. regardless of PD being a refresh it's gains over BD were quite noteworthy
Certianly an odd thing indeed considering this is exactly the other way around when we look at AMD openCL vs Nvidia CUDA where CUDA is extremely CPU bound openCL does barely touch the CPU.Originally Posted by sumitlian
With Nvidia GPU, 4770k and 6300 are almost equal.
But with AMD GPU, 4770 is 43% faster than 6300. Man, AMD DX drivers are literally horrible than Nvidia.
And looks like mantle is doing fine what it was made for, 80% more fps with Mantle (FX 4100).
Agree with you totally, I thought the same, but here its DirectX API, not OpenCL or Cuda.Originally Posted by maarten12100
Certianly an odd thing indeed considering this is exactly the other way around when we look at AMD openCL vs Nvidia CUDA where CUDA is extremely CPU bound openCL does barely touch the CPU.
Thanks for providing real world experience. I was very skeptical of this too, but by looking this now It seems wonder driver has become even more mature, and if its with Frostbite 3, then it is supposed to perform same with upcoming FB3 games. And whats the point of having a low level API and performing almost same to nvidia's dx optimization specially when you(AMD) initially called DX an overhead.
That sounds fundamentally wrong. Do you have a link for that?Originally Posted by maarten12100
Certianly an odd thing indeed considering this is exactly the other way around when we look at AMD openCL vs Nvidia CUDA where CUDA is extremely CPU bound openCL does barely touch the CPU.
I don't think Nvidia does this good in most cases actually their wonder drivers didn't work any wonders for me. Low level tweaks could also be in place it would be no new thing to leave out some detail for both of the companies.
Quote:Originally Posted by TFL Replica
That sounds fundamentally wrong. Do you have a link for that?
I think the folding editor knows his stuff. well I hopeOriginally Posted by anubis1127
I tend do agree with @dman811
. Picking up used AMD GCN 1.0 cards is the way to go right now for PPD/$$$. Even 7850s / R7 265s are a pretty nice option due to only needing one 6-pin power cable, and they do around 75-85k PPD per card. Get 4, or 5 of those in a rig, and you're looking at upwards of 350k PPD.
Plus the nice thing about AMD cards is the OpenCL implementation on them doesn't take much CPU at all, where on the NV side you pretty much need to dedicate 1 fast CPU thread per GPU, by fast I mean 3.5Ghz sandybridge or newer.
Capable?Originally Posted by chemicalfan
So, where are the Intel benches? Haswell's iGPU is capable now (compared with Sandy & Ivy), and considering PvZ is a pretty casual game, it's entirely possible that the target audience wouldn't have a dedicated GPU.
They are wrong to begin with I mean this game is not cpu bound and is a mainstream title by EA there has been a huge countdown on origin for a reason (yes I saw it while starting BF3 lol)Originally Posted by LazarusIV
You guys can bag on Mantle all you want but the way I look at it is this:
If AMD had never come out with Mantle, nVidia would never have optimized their DX11 drivers to compete with it and M$ would never have talked about DX12. Ever. I believe AMD mainly used Mantle to shake up the industry and thought if they could establish another solid graphics API along the way then they would. I for one am very thankful they did SOMETHING to cause some ripples where no one has rocked the boat for YEARS.
Oh, I don't see a countdown on Origin, but then again I'm not on the client. I'm not at home with my main PC, I've only got my Linux laptop right now. One week from today I'll be home though, then it's on like Donkey Kong!Originally Posted by maarten12100
They are wrong to begin with I mean this game is not cpu bound and is a mainstream title by EA there has been a huge countdown on origin for a reason (yes I saw it while starting BF3 lol)
I see no reason to doubt the Folding Editor's word, but do note that he's only referring to two specific F@H implementations. It is just logically unsound to assume that this info can be applied to OpenCL/CUDA as a whole.Originally Posted by maarten12100
I think the folding editor knows his stuff. well I hope![redface.gif]( "redface.gif")
The 181 min fps on the various CPUs using NV GPUs seems more like some sort of a GPU limitation than anything to do with cpu overhead. If the game was still somewhat CPU bound for the minimums you'd see higher end CPUs getting higher minimums than lower end ones.Originally Posted by geoxile
Hmm, Mantle is really good for bringing up the minimum FPS.
Also, Nvidia has a pretty sizable advantage, 20%+, for the CPU overhead in DX11. That said, it doesn't seem like something that is impossible to overcome. Surely, AMD could do it if they put more resources into DX driver development
I think you're looking at a different chart from everyone else.Originally Posted by anujsetia
So Basically in this game:
AMD R9 295x2 with Mantle at 4K: 91 Min, 102 Avg
Nvidia 780ti SLI at 4K: 74 Min, 83 Avg
R9 295x2 has higher minimum FPS than 780ti SLI Average FPS![biggrin.gif]( "biggrin.gif")
AMD R9 290x with Mantle at 4K: 49 Min, 54 Avg
Nvidia 780ti at 4K: 36 Min, 44 Avg
R9 290x has higher minimum FPS than 780ti Average FPS![biggrin.gif]( "biggrin.gif")
3rd chart shows DirectX Performance of both AMD & Nvidia Setups.Originally Posted by Kane2207
I think you're looking at a different chart from everyone else.
The 3rd chart shows 4K at 83 for both camps?
Let me make this clear, I DO NOT CARE about the AMD chip being clocked higher (I even indicated the performance can be fixed with a small oc on the 2500K), I DO NOT CARE about how old SB is now and quite frankly I DO NOT CARE how much they cost when they came out. I'm talking about the here and now. Where did I indicate that anyone is proud, let alone myself as well? All of these things will take this thread down the amd vs intel path which I have no interest in debating at this point.Originally Posted by Dimaggio1103
Huh? Slipping behind? its a 3 FPS diff, with the AMD chip clocked way higher......I would not call that slipping. Keep in mind also how old is SB now? And how much did both cost at launch? Yea, not something id be proud of......
