The AMD Mantle Thread

[smackababy]

Well if it measurable, probably means is more than 5%.

Measurable with FPS counters? Eyes?

Measurable simply means you can measure it. If the improvement is great than 1FPS, it is measurable.

Not a great PR speak word. Sounds like they are trying to not hype it. If it was a great increase, I'd wager they'd have used something like 'substantial'.

But, honestly, any real improvement could at least prove the Mantle 'concept' is correct.

 

[Arkadrel]

/BF4 shows Mantle quite strong/
/as a starting point/
/will improve/

/performance boost is measurable(290)/
/especially with X-Fire/

Is this AMD marketing having just another bad day, or is there a way to read this and be optimistic?

right?

why didnt he just say something like instead:

BF4 shows Mantle quite strong

performance boost is noteworthy "even" on a (290)
especially with X-Fire/

Basically the same thing.. and yet so much better worded.


"as a starting point" big no no.
"will improve" (could) sound like your makeing escuses.
"measurable" something that you have to measure... means its not perceptable with the naken eye.


Maybe this means it wont be as big a improvement as all of us think.

 

[sushiwarrior]

I believe your post will turn out to be true, I never thought the 45% claims were remotely close to true based on AMD's exaggerated marketing powerpoints of years past.

On the other hand, it will be good if Mantle is finally released this month. Because I think a lot of people are just sick and tired of seeing the marketing spin being put towards Mantle with non stop powerpoints, so benchmarks as tested by independent sources will be a good thing. To put AMD's claims to the test. To end the speculation. Finally. Will AMD's claims be met? It will be interesting to see. If it deserves props I will give it props. But the marketing powerpoints on non-existing software or hardware is much harder to respect. So I look forward to seeing how the real Mantle stacks up in BF4, the sooner the better (tomorrow?)

But it has to be asked. Based on your blanket statement here. What are your sources here? Citation ? Proof? Source? Did you see it first hand? Did you beta test Mantle already? Eagerly awaiting your reply. It's only a fair question since anyone downplaying the potential for Mantle in this thread is always asked for a source, so what's your source? Or did you see it first hand? Did you use it first hand? Your wording seems to indicate, that you did, in fact, use it. So i'm very curious about these GPU limited situations since that will apply to the vast majority of high end PC owners with Haswell or Ivy Bridge.

45% is a 100% real and accurate number, not at max resolution/settings with an i7 though. 45% is what you'd see with a low-mid range CPU (FX4350, i3's and i5's) at low, medium, high settings, at resolutions around 1080p. 45% is not what you'll see with an i7-4970x at 25x16 on ultra. Mantle gives a slight boost there, think 10%, but not 45%.

I don't need citations or proof, either believe me or not.

 

[blackened23]

You don't need citations or proof. Only everyone else does and they're not expected to fall back on a non answer. Anyway, so 10% is the expectation for high end PCs. Alrighty. That translates to about 5 frames on average for WHQD resolution. That answers at least one part of the question. That is good information, thank you.

I eagerly await independently tested Mantle BF4 benchmarks to verify this. Now if this is true, the real question is whether AMD's marketing will try to muddy the waters with testers required to use only low end hardware, which would of course not be representative of someone who bought a 290X for their 4770k. Which from what you're saying, will be a 10% and not a huge gain.

 

[smackababy]

You don't need citations or proof. Only everyone else does and they're not expected to fall back on a non answer. Anyway, so 10% is the expectation for high end PCs. Alrighty. That answers at least one part of the question. That is good information, thank you.

To be fair, the way he posts seems like his citations and proof could get him into a bit of hot water.

I think 10% at max isn't terrible, but definitely not something to make a giant spectacle of.

I wonder exactly how large the market for $400 GPU pair with a $40 CPU actually is that will see this giant improvements?

I suspect that most of the people who are CPU bound, will also be GPU bound very quickly afterwards.

 

[blackened23]

Yeah, that's what i'm thinking. I don't know of anyone who would be crazy enough to pair something like a 290 with an i3, that is ridiculous. And yes I know Mantle can be used on anything GCN, but AMD touted Mantle for their high end 290 cards. That was the basis of their entire 290X marketing launch campaign from what I remember, all the way back to that GPU14 event in Hawaii. And again...nobody is using that stuff with i3's, Pentiums or Celerons.

So i'm really interested in seeing if AMD marketing will try to manipulate the benchmarks that will be released (supposedly) in the next few days. Will they require benchmarks to be done only on low end hardware? I would hope not, but that would definitely muddy the waters and give users incorrect expectations if they do.

Either way, i'm personally glad that it is apparently being released soon - we can just end our speculation about Mantle (finally) and see real benchmarks. Hopefully benchmarks representative of what high end PC gamers are using and not PR twisted benchmarks, but nonetheless, a good thing. If it deserves props, like I said, I will do just that. Just glad the speculation is about to end, really.

 

[Despoiler]

45% is a 100% real and accurate number, not at max resolution/settings with an i7 though. 45% is what you'd see with a low-mid range CPU (FX4350, i3's and i5's) at low, medium, high settings, at resolutions around 1080p. 45% is not what you'll see with an i7-4970x at 25x16 on ultra. Mantle gives a slight boost there, think 10%, but not 45%.

I don't need citations or proof, either believe me or not.

the 45% is for a Kaveri APU with a discret 290x.

 

[Stuka87]

To be fair, the way he posts seems like his citations and proof could get him into a bit of hot water.

I think 10% at max isn't terrible, but definitely not something to make a giant spectacle of.

I wonder exactly how large the market for $400 GPU pair with a $40 CPU actually is that will see this giant improvements?

I suspect that most of the people who are CPU bound, will also be GPU bound very quickly afterwards.

Obviously most people won't be pairing a A-series chip with a 290/290X. But there are plenty of people like myself with a decent GPU and a CPU that just isn't able to quite hold up. In my case, my GPU averages 60% utilization at 1080P. My CPU averages 95%. So in theory, I could certainly see a nice boost.

Or at least, thats what I am hoping.

But it has to be asked. Based on your blanket statement here. What are your sources here? Citation ? Proof? Source?.

Blackened, you know why he cannot post that.

 

[Beavosaur]

Isn't that 45% using directx 11.1 in Windows 8.1?

What about the guys using Win 7 which has been
Shown to show less performance than Win 8.
I'm hoping to get better performance in Win 7 with Mantle.

 

[blackened23]

Blackened, you know why he cannot post that.

No, I certainly don't "know". I know what it SOUNDED like based on his wording, but I don't know. Maybe you know and you could fill us in. What I do know is that over the course of this thread, anyone questioning Mantle was asked for "source". In fact, any thread pertaining to unreleased hardware/software be it Maxwell or whatever is asked for "source" as the first course of business. So I ask for a source and there's no answer. Is that a double standard?

Asking for a source isn't unreasonable. And i'm afraid that's going to happen when making blanket statements about unreleased hardware or software. Period. And I don't know. But if you know, fill us in. Or don't. If you can't post it. (?) I really don't care at this point, what I do care about is this:

I'm just glad the Mantle speculation is about to come to a halt. Finally real benchmarks as tested by independent reviewers, so I do look forward to that. Again, I just hope that the waters aren't muddied by benchmarks done on ridiculously low end hardware, if sushiwarriors' statement is accurate. I want to know what BF4 Mantle does on a high end 4770k or any high end PC with a high end 290 card. It is reasonable to think that anyone who buys a 500$+ GPU will pair it with a high end PC, and AMD's marketing of the 290 cards had a strong basis in Mantle. Again, nobody is using that stuff with i3's or celerons. 500-700$ GPU with a 50$ CPU. I don't think so. Gimme a break.

 

[ASM-coder]

Is this AMD marketing having just another bad day, or is there a way to read this and be optimistic?

There seems to be some confusion about this Tweet.

It doesn't come from an AMD employee, it comes from a Dice employee.
And he is not from marketing, he is just making some comments.

 

[bepo]

the 45% is for a Kaveri APU with a discret 290x.

Is anyone ever going to be using that configuration? I would think you could get way more than a 45% increase by better distributing your funds between CPU and GPU. I would bet mantle will look the best when you combine a GPU that is a step higher than your CPU. If the GPU outclassed the CPU by too much you'd probably be better off narrowing the gap. I'm eager to see what the numbers for more realistic setups that are both CPU and GPU bound.

This release should be very interesting. The hype is at full steam and it seems like so many people are holding out on upgrades for this release. If the real world benchmarks show 5% that would be bad news for AMD at the current prices and probably piss off quite a few people who held out through the delays.

 

[smackababy]

Is anyone ever going to be using that configuration? I would think you could get way more than a 45% increase by better distributing your funds between CPU and GPU. I would bet mantle will look the best when you combine a GPU that is a step higher than your CPU. If the GPU outclassed the CPU by too much you'd probably be better off narrowing the gap. I'm eager to see what the numbers for more realistic setups that are both CPU and GPU bound.

This release should be very interesting. The hype is at full steam and it seems like so many people are holding out on upgrades for this release. If the real world benchmarks show 5% that would be bad news for AMD at the current prices and probably piss off quite a few people who held out through the delays.

Exactly, the A10-7850k is going to cost you ~$175. If you diverted the funds from a downgrade 290 instead of a 290x, you can easily get almost any i7 processor.

So, what is the point of such an incredibly silly test?

 

[Enigmoid]

the 45% is for a Kaveri APU with a discret 290x.

So 38 fps for kaveri a10-7850k on ultra? Does that seem really low to anyone?

 

[7stars]

45% is a 100% real and accurate number, not at max resolution/settings with an i7 though. 45% is what you'd see with a low-mid range CPU (FX4350, i3's and i5's) at low, medium, high settings, at resolutions around 1080p. 45% is not what you'll see with an i7-4970x at 25x16 on ultra. Mantle gives a slight boost there, think 10%, but not 45%.

I don't need citations or proof, either believe me or not.

but according to which consolidated theory? apart from the fact that if it efficiently scales to all the available cores and has better crossfire latency i'm not interested in a big fps boost anymore...but only in the total smoothness of the animation ;-) and possibly bigger minimum fps and better frametime... it's already a big win

but if the parallelism in DX is 2-3 cores max and with Mantle and eg. my fx-8320 scales to the 8 cores...and they say that the parallelism will be almost uniform, how is possible that i'll see just a 10% increase?
in my opinion they have focused on kaveri because it's their latest product and above all cheap... it's completely normal if they highlight that product.
Anyway, ATM these are only conjectures, we have to see it in action...and finally talk about real data :biggrin:

 

[moonbogg]

So 38 fps for kaveri a10-7850k on ultra? Does that seem really low to anyone?

I had to delete my true comment regarding gaming on Kaveri and its use for Mantle comparisons before posting this. An unearthly amount of restraint has been injected into my post here.
That said, I don't see how using Kaveri is relevant for Mantle comparisons or for gaming in general other than to bolster test results.

 

[Stuka87]

So 38 fps for kaveri a10-7850k on ultra? Does that seem really low to anyone?

Not at all. BF4 is by far the most CPU hungry game I have ever played (Talking multi-player here, 64 man servers). Like I mentioned above, my CPU sits at 95% utilization (average, it hits 100% at times). Even a 4.5Ghz 2500K will sit at 80-90% CPU.

Thats why mantle has less of an effect on faster CPU's.

 

[ASM-coder]

So, what is the point of such an incredibly silly test?

I might buy an R7 270(GDDR5) for under $100 to replace a 6670(DDR3). If Mantle improves performance something like 20% on an R7 with a stressed CPU, I am in. Cheap and simple upgrade for a machine that runs everything else fine.

 

[ASM-coder]

That said, I don't see how using Kaveri is relevant for Mantle comparisons or for gaming in general other than to bolster test results.

Are you saying nobody would play BF4 using a Kaveri APU?

 

[f1sherman]

I might buy an R7 270(GDDR5) for under $100 to replace a 6670(DDR3).

There is no R7 270, but you can get R9 270.

The cheapest one on newegg is $209.99 LOL

 

[Despoiler]

I definitely think that setup is going to be rare to non-existent, but I think AMD is trying to mention their new APU as much as possible. It's the only thing I can think of that makes sense to someone in a company aka marketing. On the chip side, the only thing they have been talking about are their new APUs. Most people would be going with low to mid range FX or Intel line to paired with a 270 or 280. Ultimately the setup does show that in CPU limited situations you can gain a lot with Mantle.

That being said, Mantle isn't just to help with CPU limited situations. When you get rid of the draw call overhead artists no longer have to make technical decisions on what to draw on the screen. That means more objects onscreen or better looking objects, which means more vibrant environments. Freeing up the CPU and having true multicore scaling means you can do better cpu physics. Mantle will allow the GPU to be fed as much as possible. That means you can just do more with the GPU at any given time. Again, these things may not translate to raw fps, but it's definitely going to make games better. Look what they can do with consoles with hardware specs that PC gamers laugh at. Now we can do that on a PC hardware? This is going to be good.

 

[ASM-coder]

There is no R7 270, but you can get R9 270.

The cheapest one on newegg is $209.99 LOL

oops, meant R7 250

 

[bystander36]

That being said, Mantle isn't just to help with CPU limited situations. When you get rid of the draw call overhead artists no longer have to make technical decisions on what to draw on the screen. That means more objects onscreen or better looking objects, which means more vibrant environments. Freeing up the CPU and having true multicore scaling means you can do better cpu physics. Mantle will allow the GPU to be fed as much as possible. That means you can just do more with the GPU at any given time. Again, these things may not translate to raw fps, but it's definitely going to make games better. Look what they can do with consoles with hardware specs that PC gamers laugh at. Now we can do that on a PC hardware? This is going to be good.

As true as that "could" be, the reality is that game decisions will still have to include DX, which means it isn't likely going to happen. I find it unlikely that they will have two game designs for the PC. They are far more likely going to pick something that both DX and Mantle can handle. They could add a little sparkles to Mantle, not unlike PhysX, but I don't see them having different AI, base physics or more objects on the screen.

 

[Despoiler]

As true as that "could" be, the reality is that game decisions will still have to include DX, which means it isn't likely going to happen. I find it unlikely that they will have two game designs for the PC. They are far more likely going to pick something that both DX and Mantle can handle. They could add a little sparkles to Mantle, not unlike PhysX, but I don't see them having different AI, base physics or more objects on the screen.

They already let the gamer decide how much their computer can handle in DX and OGL in many different categories including AI, physics, and # of objects on screen. I'm thinking you are either going to get the same options that go faster in Mantle for a given quality level(as shown in the Swarm demo) or the ultra setting that won't show up unless Mantle is enabled. In any implementation DX will be relegated to the baseline experience.

 

[bystander36]

They already let the gamer decide how much their computer can handle in DX and OGL in many different categories including AI, physics, and # of objects on screen. I'm thinking you are either going to get the same options that go faster in Mantle for a given quality level(as shown in the Swarm demo) or the ultra setting that won't show up unless Mantle is enabled. In any implementation DX will be relegated to the baseline experience.

If by # of objects, you mean like explosion fragments and what not, or other trivial stuff, yeah. That would be like PhysX stuff. If you mean more units on an RTS game, that would change the game, and simply wouldn't work.

I did say base physics btw. If you think the baseline physics will be changeable, that would prevent multiplayer. If you mean like flowing capes, that would be what I meant by PhysX like sparkle.

I don't see Swarm demo like objects in a real game, as those things can't simply be removed and be the same game.