The AMD Mantle Thread

[bystander36]

100%+ increase...pics or it didn't happen.

Btw, that is not a 100% increase. He is saying the minimums are improved by 100%. The averages may be much closer to each other.

 

[Carfax83]

so for me no nvidia card on the planet would give me this type of performance, unless i built a new PC from the ground up around a new top of the line i7

I think it's possible that you received such a large performance increase with Mantle seeing as you are completely CPU bound.

But the main reason why you got such a large increase with Mantle, is because AMD's Direct3D drivers perform so poorly to begin with.

Mantle does work very effectively for lower performing CPUs though admittedly..

 

[cytg111]

But the main reason why you got such a large increase with Mantle, is because AMD's Direct3D drivers perform so poorly to begin with.

Is this one of those things where if you repeat it enough times it becomes reality?

Not a living soul on this board knows to what level or what degree AMD's drivers are optimized in whatever regard. Just because Nvidia can pull a rabbit out of their drivers doesnt mean AMD can. It could be. It could also be not.
So, everyone here, who is a GPU driver developer AND has working experience on AMD's stack, raise your hands. Thats what I thought.
Anyway it may be, the why and why not speculation of AMD's Direct3D drivers has zero concern for this thread.
Just saying.

 

[Venomous]

Is this one of those things where if you repeat it enough times it becomes reality?

Not a living soul on this board knows to what level or what degree AMD's drivers are optimized in whatever regard. Just because Nvidia can pull a rabbit out of their drivers doesnt mean AMD can. It could be. It could also be not.
So, everyone here, who is a GPU driver developer AND has working experience on AMD's stack, raise your hands. Thats what I thought.
Anyway it may be, the why and why not speculation of AMD's Direct3D drivers has zero concern for this thread.
Just saying.

Last time I remember seeing AMD pulling some increased power from drivers were back when the 5xxx series came out. I think it was around the 5th month after release they extracted a good bit of performance.

However this also caused some issues for crossfire and eyefinity users as a side effect .

 

[Carfax83]

Is this one of those things where if you repeat it enough times it becomes reality?

Not a living soul on this board knows to what level or what degree AMD's drivers are optimized in whatever regard. Just because Nvidia can pull a rabbit out of their drivers doesnt mean AMD can. It could be. It could also be not.
So, everyone here, who is a GPU driver developer AND has working experience on AMD's stack, raise your hands. Thats what I thought.
Anyway it may be, the why and why not speculation of AMD's Direct3D drivers has zero concern for this thread.
Just saying.

Apparently you didn't read any of the dozen or so Mantle reviews.. You don't have to be a GPU driver developer to look at benchmarks and come to a conclusion..

Then again, if people are willing to defend AMD on this, it's easy to see why they've gotten away with making subpar drivers for so long and why uninformed people believe DX is the problem, and not AMD.

Anyway, it doesn't really matter. The biggest hope Mantle can achieve is that Microsoft incorporates some it's aspects into future Direct3D versions.

 

[VulgarDisplay]

I think it's possible that you received such a large performance increase with Mantle seeing as you are completely CPU bound.

But the main reason why you got such a large increase with Mantle, is because AMD's Direct3D drivers perform so poorly to begin with.

Mantle does work very effectively for lower performing CPUs though admittedly..

So you are saying the same cpu would not bottleneck an nvidia card of the same performance tier? Give it a rest man.

The horse is a bloody puddle on the ground and any further amount of beating it is not going to miraculously prove this misguided point of yours that nvidia drivers are so much better at utilizing cpu power. They are better, but you are making it out to be a huge increase based on a benchmark for one 2 year old game.

I could easily show a benchmark where AMD has mysteriously more performance than nvidia to make the same point.

Its far more likely that nvidias larger dies brute force more performance than amd's smaller dies do, and both sides just have outliers that perform better on their hardware. No one claims nvidia drivers don't multithread based on sleeping dogs or dirt benchmarks. Should we start would you say that we are right?

 

[mindbomb]

Anyway, it doesn't really matter. The biggest hope Mantle can achieve is that Microsoft incorporates some it's aspects into future Direct3D versions.

they probably will be incorporated in the next directx, but it will end up requiring new graphics cards, new operating systems, and a new renderer in games.

 

[Carfax83]

So you are saying the same cpu would not bottleneck an nvidia card of the same performance tier? Give it a rest man.

I said no such thing. That's something YOU said, not me, so don't put words in my mouth.

I merely said that the reason for such a large increase (100% is massive) is because AMD's DirectX drivers start at such a low base level.

I've enumerated many times that Mantle is definitely more efficient than Direct3D. But then again, it's not really a fair comparison as Direct3D does WAY more than Mantle, is backward compatible and runs on various architectures.

Suffice to say though, that Direct3D already has a solution for CPU limitations, and it seems to be pretty effective.....not as effective as Mantle perhaps, but good enough.

Its far more likely that nvidias larger dies brute force more performance than amd's smaller dies do, and both sides just have outliers that perform better on their hardware. No one claims nvidia drivers don't multithread based on sleeping dogs or dirt benchmarks. Should we start would you say that we are right?

NVidia's multithreading enhancements relies on the game engine. The more threaded an engine is, the better it scales on NVidia hardware. That's why NVidia has the edge in the Frostbite 2 and 3 engine, plus CryEngine 3 as well.

I've never played Sleeping Dogs so I have no idea how many threads the engine supports.

 

[Gloomy]

If you earnestly believe Nvidia's driver would equal Mantle on an i7-920 then you are taking delusion to heights never before reached by man.

 

[kane15]

sounds like a little bit of 'moving the goalpost' going on here lmao... wow id expect as much from battlelog but i guess you cant escape it.

 

[wand3r3r]

So the D3D drivers are putting the bar "so low" yet they are very competitive between the tiers. What a bunch of fud.

 

[VulgarDisplay]

I said no such thing. That's something YOU said, not me, so don't put words in my mouth.

I merely said that the reason for such a large increase (100% is massive) is because AMD's DirectX drivers start at such a low base level.

I've enumerated many times that Mantle is definitely more efficient than Direct3D. But then again, it's not really a fair comparison as Direct3D does WAY more than Mantle, is backward compatible and runs on various architectures.

Suffice to say though, that Direct3D already has a solution for CPU limitations, and it seems to be pretty effective.....not as effective as Mantle perhaps, but good enough.

NVidia's multithreading enhancements relies on the game engine. The more threaded an engine is, the better it scales on NVidia hardware. That's why NVidia has the edge in the Frostbite 2 and 3 engine, plus CryEngine 3 as well.

I've never played Sleeping Dogs so I have no idea how many threads the engine supports.

Don't take it personal man. That is what anyone reading your posts about nvidia vs. Amd multi threading is under the impression you are saying. That's what I took away from it all.

Like I said some games just work better on either vendor. Civ 5 is one such instance and it is in no way a reason to say that amd drivers are sub par. The same can just as easily be said if nvidia drivers in the right benchmark.

 

[cytg111]

Then again, if people are willing to defend AMD on this, it's easy to see why they've gotten away with making subpar drivers for so long and why uninformed people believe DX is the problem, and not AMD.

And that right there is the fail, I have no brand preference whatsoever (running haswell and 780), to me it is a pure numbers game. You are the one who puts brand preference into play - and to sum it up this "brand preference" is why this thread is a massive fail. I am sorry to say but you "come to a conclusion" way to easy.

I say we stick to what we know, discuss the numbers and leave EVERYTHING we dont know behind. This should be about mantle and nothing else, not my gut feeling about amd's sub par direct3d whatever - whatever. There is a difference between educated guesses(and stating just that) and making stuff up ..and this is making stuff up.

For now we dont have any new mantle data/reviews so IMO this thread is done for until new data pours in.

edit :
"Anyway, it doesn't really matter. The biggest hope Mantle can achieve is that Microsoft incorporates some it's aspects into future Direct3D versions."

- I dont know about "biggest hope" that is way too crystalballish for me, but if the endresult would be a directx-mantle-thing then that would problary be the biggest win I can imagine. So here is to hoping.

"Suffice to say though, that Direct3D already has a solution for CPU limitations, and it seems to be pretty effective.....not as effective as Mantle perhaps, but good enough."
- Game devs pushed for mantle, not AMD. You are pretty much telling the guy that uses a hammer what kind of hammer he should be using! I'd argue the hammer man(cant touch this) has a pretty good idea about what kind of hammer suits his job.

 

[Will Robinson]

If AMD's DX drivers were/are so bad how come in all previous generations the cards have been so competitive?
7970 GHzEdition V GTX680 springs to mind...where were the ultimate NV DX drivers then

From the numbers so far it looks as tho my Lynnfield i7 860 would perform much better with an R290/Mantle than it would running on DirectX

 

[cytg111]

If AMD's DX drivers were/are so bad how come in all previous generations the cards have been so competitive?

The only possible issue I can see with this is going forward is a conflict of interrest on AMD's side. Say that mantle is taking off, more and more games implement it, what incentive would AMD have to go above and beyond on their directx work? Sure they'll have to stay competetive but perhaps not so much as to shine a lesser light on mantle?
Disclaimer, 100% non fact here.

 

[Red Hawk]

Last time I remember seeing AMD pulling some increased power from drivers were back when the 5xxx series came out. I think it was around the 5th month after release they extracted a good bit of performance.

However this also caused some issues for crossfire and eyefinity users as a side effect .

Ok, I'm starting to think you're purposefully spreading misinformation and FUD now. There were several drivers for the 6xxx and 7xxx series that improved performance significantly.

http://www.anandtech.com/show/6393/amds-holiday-plans-cat1211-new-bundle

 

[cytg111]

Ok, I'm starting to think you're purposefully spreading misinformation and FUD now. There were several drivers for the 6xxx and 7xxx series that improved performance significantly.

And he specifically stated "Last time I remember..". -And you cant do better than that. Just not possible .

 

[Venomous]

Ok, I'm starting to think you're purposefully spreading misinformation and FUD now. There were several drivers for the 6xxx and 7xxx series that improved performance significantly.

http://www.anandtech.com/show/6393/amds-holiday-plans-cat1211-new-bundle

You seriously need to get a life and learn how to comprehend sentences because I specifically said 5xxx series. Throwing out random benchmarks from 2 generations ahead is elementary.

Warning issued for inflammatory language.
-- stahlhart

 

[The Alias]

I think it's possible that you received such a large performance increase with Mantle seeing as you are completely CPU bound.

But the main reason why you got such a large increase with Mantle, is because AMD's Direct3D drivers perform so poorly to begin with.

Mantle does work very effectively for lower performing CPUs though admittedly..

so you are saying he would get those same results with a gk110 ? yes or no question .

 

[Red Hawk]

You seriously need to get a life and learn how to comprehend sentences because I specifically said 5xxx series. Throwing out random benchmarks from 2 generations ahead is elementary.

You said "last time you remember". That meant you didn't remember the 6xxx and 7xxx series having performance improvements from drivers. Should I just conclude you have poor memory?

Warning issued for personal attack.
-- stahlhart

 

[Carfax83]

If you earnestly believe Nvidia's driver would equal Mantle on an i7-920 then you are taking delusion to heights never before reached by man.

I'll send you 100 USD on paypal if you can prove to me where I said that I believe NVidia's driver would equal Mantle on an i7 920...

Your reading comprehension must be extremely poor to come to such a conclusion based on what I actually said.

 

[Carfax83]

So the D3D drivers are putting the bar "so low" yet they are very competitive between the tiers. What a bunch of fud.

Two of the main reasons AMD has been competitive is because most 3D engines up till now haven't been very multithreaded, and most games tend to be heavily GPU bound.

Thats starting to change now with the next gen consoles. Developers have no choice but to bite the bullet and thread the hell out of their engines to extract performance from the weak CPUs (relatively speaking) that are in the PS4 and the Xbox One.

Also, games are becoming bigger and more open, which means the CPU will become more important than ever before. Games like Watch Dogs are an indication of whats coming, with quad core processors being necessary..

 

[Carfax83]

Don't take it personal man. That is what anyone reading your posts about nvidia vs. Amd multi threading is under the impression you are saying. That's what I took away from it all.

I didn't take it personal, it's just annoying to see my position constantly being misrepresented. I don't know how you got that impression from what I said, because I never even said the word "Nvidia."

And it's a fact that AMD's D3D drivers are subpar, as many reviews have highlighted.

On the 7850K D3D path, NVidia has a 35% lead on AMD.. NVidia is definitely exploiting the DX11.1 optimizations for extra CPU performance, much more than AMD..

Like I said some games just work better on either vendor. Civ 5 is one such instance and it is in no way a reason to say that amd drivers are sub par. The same can just as easily be said if nvidia drivers in the right benchmark.

Civ 5 runs faster on NVidia because of a certain DX11 feature......which AMD still doesn't support, despite DX11 having been out for nearly 5 years now..

 

[SPBHM]

the graphic on the previous post is quite interesting, Nvidia have a much nicer lead on D3D with slower CPUs, most VGA reviews (even low end VGAs) are made with overclocked high end CPUs,

but CIV not working as good on AMD hardware (it was MUCH worse if pre GCN stuff) I don't know, AMD have their own optimized stuff like Dirt Showdown with "advanced lighting",

 

[Carfax83]

I say we stick to what we know, discuss the numbers and leave EVERYTHING we dont know behind. This should be about mantle and nothing else, not my gut feeling about amd's sub par direct3d whatever - whatever. There is a difference between educated guesses(and stating just that) and making stuff up ..and this is making stuff up.

Then how do you explain things like this:

AMD has a huge 35% deficit on the D3D 7850K, compared to NVidia. Both GPUs are clearly CPU bound by that weak sauce CPU, but NVidia comes out looking much better.

The question is why?

Game devs pushed for mantle, not AMD. You are pretty much telling the guy that uses a hammer what kind of hammer he should be using! I'd argue the hammer man(cant touch this) has a pretty good idea about what kind of hammer suits his job.

Correction, SOME game developers pushed for Mantle......many others are against it for various reasons, such as increased platform fragmentation.

Also, and this is just my theory, I think that AMD developed Mantle because they knew they were in trouble when it came to D3D performance.

They are obviously having problems, as seen by their inability to implement DX11 multithreading in their drivers, and their lower relative D3D performance in CPU bound or highly multithreaded games like BF4 and Crysis 3.