The AMD Mantle Thread

[OCGuy]

Some of these predictions sound like they are actually setting Mantle up for failure. 50% - 80% FPS gains in high end systems? Didnt an AMD employee at a demo already give his FPS predictions on another game?

I guess we shall see....

 

[PPB]

I don't think he was considering those who play at low settings with high end cards.

Then he is ignoring all the competitive players, the ones actually benefitting from 120 fps on a shooter.

 

[uclaLabrat]

Some of these predictions sound like they are actually setting Mantle up for failure. 50% - 80% FPS gains in high end systems? Didnt an AMD employee at a demo already give his FPS predictions on another game?

I guess we shall see....

If DX has a 20% overhead, isn't 20% the theoretical max improvement, with 10-15% more likely? Even that would be huge.

 

[sushiwarrior]

Even at GPU limited settings (eg. 2560x1600/1440, ultra, with a strong processor like an extreme edition), improvements should be 10-15% with a card like a 290 or 290x. So going from 40-45 to 50-55 FPS maybe, nothing insane. Give it a lower end processor (FX-8350, i5-4570, etc.) or a lower resolution/graphics setting, improvements can range from 15-35% (big range I know, depends on settings and maps and stuff). At absolutely CPU limited settings, 40-50% usually, but that's unrealistic considering you're probably at minimum settings and min res or something.

But there are still improvements even when you aren't draw call limited.

 

[McLovin42]

Realistic improvements with mantle will be 5-10% maximum.........and even that's not a guarantee.......and beyond that fact Mantle will be on very few games in the up coming 2-3 years and beyond even........it will make Nvidia's PhysX feature looking mainstream and widespread in comparison.......its a lot of talk about nothing in reality.

AMD did its PR job once again.....tons of hype and marketing to try and keep their customer base intact and excited............if only they put that much effort into their hardware and software divisions...........

 

[DamnedLife]

Realistic improvements with mantle will be 5-10% maximum.........and even that's not a guarantee.......and beyond that fact Mantle will be on very few games in the up coming 2-3 years and beyond even........it will make Nvidia's PhysX feature looking mainstream and widespread in comparison.......its a lot of talk about nothing in reality.

AMD did its just once again.....tons of PR and marketing to try and keep their customer base intact and excited............if only they put that much effort into their hardware and software divisions...........

Gosh! They do my 290x is 5.6 TFLOPS! With Maxwell (at least real ones) delayed it will be king of the jungle. Only Titan is better in double precision, but that's what Titan is designed for, graphical designers not gamers.

 

[smackababy]

If DX has a 20% overhead, isn't 20% the theoretical max improvement, with 10-15% more likely? Even that would be huge.

10% isn't that huge when we're talking about going from 40FPS to 44FPS.

 

[Imouto]

10% isn't that huge when we're talking about going from 40FPS to 44FPS.

That's the perf difference between a R9 290X and a GTX 780Ti.

 

[Phynaz]

I say from 10% (minimum possible gain, people playing high res with all the eyecandy and AA) to 70-80% (people like me, playing all low with no AA and currently having my GPU sitting at 40% load most of the time).

What CPU are you running?

 

[OCGuy]

Even at GPU limited settings (eg. 2560x1600/1440, ultra, with a strong processor like an extreme edition), improvements should be 10-15% with a card like a 290 or 290x. So going from 40-45 to 50-55 FPS maybe, nothing insane. Give it a lower end processor (FX-8350, i5-4570, etc.) or a lower resolution/graphics setting, improvements can range from 15-35% (big range I know, depends on settings and maps and stuff). At absolutely CPU limited settings, 40-50% usually, but that's unrealistic considering you're probably at minimum settings and min res or something.

But there are still improvements even when you aren't draw call limited.

Those are low-end processors?

 

[ASM-coder]

That's the perf difference between a R9 290X and a GTX 780Ti.

Exactly, and some are willing to pay big bucks for that. Mantle is free. Why complain about a free performance upgrade, no matter how good it is?

 

[McLovin42]

Gosh! They do my 290x is 5.6 TFLOPS! With Maxwell (at least real ones) delayed it will be king of the jungle. Only Titan is better in double precision, but that's what Titan is designed for, graphical designers not gamers.

I don't care what the 290x can put out for raw frames, the hardware quality and QC is still suspect. People think game fps #'s are what dictate a quality card from one that's suspect. Look at the bigger picture. The die quality, the power delivery components, the pcb, traces, the heatsink and its design etc etc.........its 2nd class hardware regardless if it outperforms a titan or a 780 or whatever else you can think of in a few benchmarks.

 

[TerryMathews]

I don't care what the 290x can put out for raw frames, the hardware quality and QC is still suspect. People think game fps #'s are what dictate a quality card from one that's suspect. Look at the bigger picture. The die quality, the power delivery, the heatsink design etc etc.........its 2nd class hardware regardless if it outperforms a titan or a 780 or whatever else you can think of in a few benchmarks.

"Look man, I dont care how fast your Camaro is it will never have the elegance of Mercedes Benz engineering. I mean, I have windshield wipers. For my headlights."

You're just going to have to make piece with the fact that a video card is intended to render frames. If a card does it faster and for less money, it is better. There may be reasons why that choice is contraindicated, like heat or power draw but that doesn't negate that original selection.

 

[Despoiler]

I invite everyone to watch the dev interviews where they explain that Mantle doesn't just increase fps. The draw call overhead is greatly reduced (DX is up to 40%) so your GPU isn't waiting on the CPU. Even if you aren't CPU limited your CPU is still not feeding your GPU as fast as the GPU can accept because DX sucks that bad. Mantle gives you full CPU scaling. You might go from 60 to 63-66fps (5-10% gain), but the graphics will be better with many many more objects on screen because of the increase in # draw calls and full cpu scaling. Better physics can now also happen. Artist also don't have to limit themselves because of performance concerns so environments and characters are going to be more vibrant.

I also find it hilarious that people are talking about # of games or lack there off that will support Mantle. Mantle isn't getting added to any games right now. It's being added to the ENGINES that they run on. BF4 is getting Mantle by virtue that Frostbite is getting it, not that they are adding it to BF4. When you add support to an engine the implementation to a game is a lot less intensive.

 

[ASM-coder]

Give it a lower end processor (FX-8350, i5-4570, etc.) or a lower resolution/graphics setting, improvements can range from 15-35% (big range I know, depends on settings and maps and stuff).

I run BF4 with a CPU half as powerful as the FX-8350, so if that does show
a +25% improvement, I'll certainly be looking for a new GCN GPU.

 

[ultimatebob]

Exactly, and some are willing to pay big bucks for that. Mantle is free. Why complain about a free performance upgrade, no matter how good it is?

I'll complain (and do so loudly) if the performance benefit isn't anywhere near the "Up to 45% faster in Battlefield 4!" or the "Up to 3X Faster on our newest processor!" claims that we keep seeing in the AMD marketing slides.

They simply shouldn't be allowed to advertise that crap if it isn't true.

I know that some posters here regard "Team Red" or "Team Green" with the same respect that they would have for a pair of fart stained underwear, but I think that we can all agree that we need more honesty in hardware advertising.

 

[96Firebird]

I can't wait to see the 290X "ridicule" the Titan.

 

[McLovin42]

"Look man, I dont care how fast your Camaro is it will never have the elegance of Mercedes Benz engineering. I mean, I have windshield wipers. For my headlights."

You're just going to have to make piece with the fact that a video card is intended to render frames. If a card does it faster and for less money, it is better. There may be reasons why that choice is contraindicated, like heat or power draw but that doesn't negate that original selection.

No, its not that simple. When a card acts like an oven heating up your other components or comes with a piss poor cooler or weak vr circuits, I could care less if the cheaper card is a bit faster or not. I will gladly PAY more for a card that may show up on a "benchmark" as being a bit slower, but in real world performance looks and plays the same, yet does so with greater reliability and stability long term due to better hardware component quality on the pcb and in the die itself.

There are plenty of Chevy Camaros that go faster from 0-60 than say a Ferrari..........but I'll still take the Ferrari......and will gladly pay more for it. There are more ways performance is measured that just plain raw fps on a counter.

Besides...... a 290x might be a few bucks cheaper than my 780 Ti............but it in this case and scenario it doesn't perform as fast as it in FPS, it doesn't have as good driver support, with less visual artifacts and games don't play nearly as smooth. So considering I paid that premium for some of the features the 780 has(like G sync, or a quality reference cooler that doesn't throttle the clock down constantly, or drivers that still have mediocre frame pacing etc etc)was my choice because I value OVERALL quality vs just a few benchmark PR #'s.

 

[BallaTheFeared]

Gosh! They do my 290x is 5.6 TFLOPS! With Maxwell (at least real ones) delayed it will be king of the jungle. Only Titan is better in double precision, but that's what Titan is designed for, graphical designers not gamers.

lol, 290x has less DP than the 7970, and far less than Titan.

 

[ASM-coder]

I'll complain (and do so loudly) if the performance benefit isn't anywhere near the "Up to 45% faster in Battlefield 4!" or the "Up to 3X Faster on our newest processor!" claims that we keep seeing in the AMD marketing slides.

They simply shouldn't be allowed to advertise that crap if it isn't true.

Did you buy a new GCN card based on these claims? If so, then maybe you have a right to complain. Otherwise, I have to agree with the naysayers on at least one point. "Up to 45%" covers a lot of ground.

 

[TerryMathews]

No, its not that simple. When a card acts like an oven heating up your other components or comes with a piss poor cooler or weak vr circuits, I could care less if the cheaper card is a bit faster or not. I will gladly PAY more for a card that may show up on a "benchmark" as being a bit slower, but in real world performance looks and plays the same, yet does so with greater reliability and stability long term.

There are plenty of Chevy Camaros that go faster from 0-60 than say a Ferrari..........but I'll still take the Ferrari......and will gladly pay more for it.

Besides...... a 290x might be a few bucks cheaper than my 780 Ti............but it in this case and scenario it doesn't perform as fast as it in FPS, it doesn't have as good driver support, with less visual artifacts and games don't play nearly as smooth. So considering I paid that premium for some of the features the 780 has(like G sync, or a quality reference cooler that doesn't throttle the clock down constantly, or drivers that still have mediocre frame pacing etc etc)was my choice because I value OVERALL quality vs just a few benchmark PR #'s.

The words you are looking for are "total value proposition"...

 

[Venomous]

Realistic improvements with mantle will be 5-10% maximum.........and even that's not a guarantee.......and beyond that fact Mantle will be on very few games in the up coming 2-3 years and beyond even........it will make Nvidia's PhysX feature looking mainstream and widespread in comparison.......its a lot of talk about nothing in reality.

AMD did its PR job once again.....tons of hype and marketing to try and keep their customer base intact and excited............if only they put that much effort into their hardware and software divisions...........

Willing to put that obtuse statement against some money via PayPal? If you think Dice and other devs are going to take their time out for this mininal bump in performance, you are surely mistaken. Take your pom poms off and come back to reality kid.

I think denial has begun to set in for you and you will become the laughing stock on this forum with such erroneous predictions that obviously are nothing more than pure hatred for AMD.

 

[McLovin42]

Willing to put that obtuse statement against some money via PayPal? If you think Dice and other devs are going to take their time out for this mininal bump in performance, you are surely mistaken. Take your pom poms off and come back to reality kid.

I think denial has begun to set in for you and you will become the laughing stock on this forum with such erroneous predictions that obviously are nothing more than pure hatred for AMD.

AMD is what it is. A competitor that tries to undercut Intel/Nvidia with cheaper corner cutting hardware designs that try and approach the performance of their products. Its been that way for many years now. They overall have always been a step behind both Nvidia and Intel with their CPU/GPU designs in one fashion or another.

As it stands right now. Im typing this on a laptop with a pair of AMD 6990m GPU's.................Ive had this current set for nearly a year.............which is a shocker............Ive had these cards replaced though in the first year nearly a half dozen times. My previous 7950 was RMA'd out twice. The PC shop were I used to be a tech at for many years still has a significantly higher number of QC/RMA issues with AMD products.........over 3:1 in comparison to Nvidia/Intel.

It is what it is.

When Mantle is implemented into a few games and the results of its performance improvement are underwhelming, be sure to get back to this post of mine.......

 

[bystander36]

I run BF4 with a CPU half as powerful as the FX-8350, so if that does show
a +25% improvement, I'll certainly be looking for a new GCN GPU.

The cost of a CPU upgrade may be more useful, depending on your whole system. Sure, if Mantle does improve performance by 25%, it will only do so on Mantle games, which don't exist yet, and will not be taking over the gaming world in the near future. A CPU upgrade will net you positive results on all games that are CPU bound.

 

[smackababy]

Exactly, and some are willing to pay big bucks for that. Mantle is free. Why complain about a free performance upgrade, no matter how good it is?

Except Mantle isn't free. AMD might not be charging us, the end user, to have it included in games, but the development cost of those games goes up, even if it is only slightly. So, they either have to do more (include Mantle) with less budget or have a larger budget to include it. What is more likely to happen in the real world? A publisher saying "here's more money, go right ahead" or "here is your budget and your release date; get it done with what you have." There is a huge possibly we might be getting less total product to accomplish DX and Mantle versions. Now, the big question we can't answer is how much less of a product.

Mantle is not free.