The AMD Mantle Thread

[DamnedLife]

Andersson already said mantle is very much alike with ps4 api. DirectX HLSL is also similar in a way that they can reuse the code with mantle just as you say. If game developers dont include mantle in their games they lose gcn market share which is getting bigger everyday with the apus and 7000 series on bargain with 200 series introduced. If mantle is what the devs are saying it is then a lot of people in my opinion switch their cards or since nv users are fairly rich they mat buy a second build. But this doesnt change the fact that amd has a bigger and stronger low and med build base. They are making gcn "the architecture" to have and since amd is pulling the strings on mantle for a while they will postpone making mantle an open platform untill they make some profit off it which is their right. Dont forget nv refused the developers when they came knocking and for me thats a big no no. Gamers are limited with what they are given while asking for more the same thing goes for developers and companies. Devs ask for more but always have to contend with what they are given to work with. Dx x.1 and x.2 advancements and improvements always come from amd, they R&D and then microsoft includes it on the next version of OS they will be releasing as if they are behind it and also this way dx improvements come fairly slow with each os cycle and time schedule according to M$. You guys talk about how hard it will be for developing 2 or 3 or 4 apis as if you have a doctorate in programming. Pfff devs are excited about it and since some of them were asking for it I say they wont mind developing for it. Unreal engine even can be made to work for mantle but their developers are in business with nvidia so of course they will bad mouth it. Then it will be my duty for myself at least to protest these game developers and their games by not buying their games. Why? Nv users can play fairly with gaming evolved games/amd games and get the performance worthy of their cards. The way its meant to be played/Nvidia games on the other hand punishes all amd users high end or low end doesnt matter and they cant get performance levels worthy of their cards. Maybe in a month or two amd engineers do some optimisations and release a driver and only then they may or may not still get the levels they should. So my friends I hope mantle crushes every expectation with even better than advertised results. And I am not wanting this just for myself but for all the gamers red and green. But some green heads are secretly or openly wishing for mantle to fail and that is what I dont get, supporting dx for the sake of siding with assumed winner?

 

[Grooveriding]

paragraphs pl0x

 

[DamnedLife]

I am writing from mobile so its hard as it is but will be careful next time.

 

[smackababy]

Andersson already said mantle is very much alike with ps4 api. DirectX HLSL is also similar in a way that they can reuse the code with mantle just as you say. If game developers dont include mantle in their games they lose gcn market share which is getting bigger everyday with the apus and 7000 series on bargain with 200 series introduced.

This is where I stopped reading.

1. Reusing code doesn't mean anything without quantification. I have numerous APIs were I can reuse a good amount of code; however, that doesn't mean switching between them is any easier.

2. Not coding for Mantle means you lose 0% of the GCN marketshare, as they all run DX just fine. The argument is that developers will be spending increased amount (because no matter how little, it is an increase) of resources for what is currently an extremely small market segment. The exact same argument goes for Linux and Mac. At a certain point, the development cost for those platforms outweighs the resource allocation a publisher is willing to give you. Now, if Mantle turns out to be really good AND developers like using it, how much does it eat into the cost of DX development for those without Mantle access? Publishers aren't going to just say "Hey, here's an extra $50 million! Add Mantle for exactly the same amount of potential customers as juts doing DX!" That is the biggest real concern I've seen in this thread

I have a 290, I can use Mantle when it comes out. I like the idea that developers are for improvements. I also remember the entire fuss about Windows having an abstraction layer from the hardware that didn't exist in DOS and it "killing gaming!" Well, obviously, that didn't happen.

 

[RampantAndroid]

Youre forgetting GCN-based APUs

That just supports my arguement more: how many people are on Intel's HD4000,HD5000/HD5200? The point I'm getting at here is AMD doesn't have a majority of GCN hardware compared to Intel and Nvidia offerings. Without console support, I just can't see Mantle sticking around long. It'll be as successful (at best) as TressFX and PhysX are today: mostly proprietary (or works best on proprietary hardware) APIs with few games supporting them.

Are DICE, Bioware and so forth really willing to test two codepaths? They have to at least do BVT level testing on each. As it stands today, it's DirectX whatever version on the XBox (9.something), XBone being DX11, PS3 being OGL, PC being OGL or DirectX of a version independent of the consoles. It's already a large matrix, and adding Mantle just seems to add a lot to that matrix for little benefit from the developer's standpoint. AMD cannot throw money around to make dev's support Mantle like nvidia or Intel can.

At best I see Mantle being a cry to make DirectX and OpenGL more efficient, but I don't see Mantle changing the game any. Maybe I'm missing the big picture though...we can certainly re-evaluate Mantle's position in a year...

This is where I stopped reading.

1. Reusing code doesn't mean anything without quantification. I have numerous APIs were I can reuse a good amount of code; however, that doesn't mean switching between them is any easier.

2. Not coding for Mantle means you lose 0% of the GCN marketshare, as they all run DX just fine.

On reusing code, yes - I agree. Also, reusing code can often lead to bugs being introduced. Copy/pasting code is often nasty - I catch copy paste errors in code reviews often.

On number 2, this is what I'm thinking. Not supporting Mantle gains you nothing over what you have presently, but you lose nothing.

The ONLY argument for Mantle is that it might make gaming on low-power/cheap hardware feasible at higher settings.

 

[moonbogg]

The suspense is killing me. When is this fancy doohickey being released?

 

[blackened23]

If we believe hardware.fr, the 45% from the slides are not on Kaveri (as some people might think) but on R9 290/x hardware.

Source: http://www.hardware.fr/news/13525/ces-45-version-mantle-bf4.html

It sounds like what i'm reading there is 45% in the best possible situation, but the the source also indicates the gain is around 10% per AMD's statements in GPU limited situations. That would imply anyone at 1080p or higher wouldn't get a 45% boost, as those resolutions are (usually) GPU limited? Anyone else thinking this? This is what my translation indicates from that URL:

For situations where the GPU is the limiting factor, the gain would turn around instead of 10% based on the information we collected. We will of course check it all as soon as the patch is available, hoping that the delay is not extended beyond January.

But yeah. A release would be welcome like moonbogg indicated, to end all of the speculation.

 

[monstercameron]

http://www.engadget.com/2014/01/14/...time-strategy-mantle-demo/?ncid=rss_truncated
was this already mentioned? just got in...

 

[Skurge]

It sounds like what i'm reading there is 45% in the best possible situation, but the the source also indicates the gain is around 10% per AMD's statements in GPU limited situations. That would imply anyone at 1080p or higher wouldn't get a 45% boost, as those resolutions are (usually) GPU limited? Anyone else thinking this? This is what my translation indicates from that URL:



But yeah. A release would be welcome like moonbogg indicated, to end all of the speculation.

I don't see where they say AMD stated it would be 10%.

From the AT Kaveri Review

Despite the release of Mantle being held back by the delay in the release of the Mantle patch for Battlefield 4 (Frostbite 3 engine), AMD was happy to claim a 2x boost in an API call limited scenario benchmark and 45% better frame rates with pre-release versions of Battlefield 4. We were told this number may rise by the time it reaches a public release.

That 45% being a A10-7850K with a 290X as stated on in the same review. Now this could mean. Either a 7850K will be as fast as a 4770/4670 in a game like BF4 and we can spend less on the CPU. OR it will mean a 7850 will be as fast as a 4770/4670 without Mantle, but because CPU usage it better, you will get added effects so getting a FX8350/4770/4930 is still worthwhile as you maybe able to increase draw distance and more detailed physics. We shall see though, but the fact that Oxide engine already gives a massive boost on both AMD and Intel CPUs makes me very hopeful.

 

[DarkKnightDude]

http://www.engadget.com/2014/01/14/...time-strategy-mantle-demo/?ncid=rss_truncated
was this already mentioned? just got in...

That demo looks sweet. Lots of marketingspeak but some interesting info in the video as well.

 

[Shivansps]

I don't see where they say AMD stated it would be 10%.

From the AT Kaveri Review



That 45% being a A10-7850K with a 290X as stated on in the same review. Now this could mean. Either a 7850K will be as fast as a 4770/4670 in a game like BF4 and we can spend less on the CPU. OR it will mean a 7850 will be as fast as a 4770/4670 without Mantle, but because CPU usage it better, you will get added effects so getting a FX8350/4770/4930 is still worthwhile as you maybe able to increase draw distance and more detailed physics. We shall see though, but the fact that Oxide engine already gives a massive boost on both AMD and Intel CPUs makes me very hopeful.

You just cant compare what happens in a game engine techdemo to any real situacion, Oxide is really pushing the limits, in a real world scenario, games in that engine will target way below what DX can do.

 

[mikk]

If we believe hardware.fr, the 45% from the slides are not on Kaveri (as some people might think) but on R9 290/x hardware.

Source: http://www.hardware.fr/news/13525/ces-45-version-mantle-bf4.html

Yes it's from Kaveri with R9 290x. Tested in Full HD 4xMSAA at least and more importantly compared to Windows 8.1. Impressive since Mantle can't make use of the better core scaling. On Windows 8.1 dX11.1 version of BF4 already scales close to perfect on a 4 core CPU.

 

[tential]

The fact that they used an intel processor when they launched the R9 series to show how good it was vs nvidia, but then used a kaveri processor when doing dx vs mantle makes me slightly suspicious.

Wouldn't it be better to use a top of the line intel processor to ensure the CPU isn't hobbling them?

Also, when in gods name will this be released with a full review? I swear I had heard mid/late december and it's now mid january and all I see are lame press slides.

 

[Enigmoid]

The Mantle hype machine continues... it looks like AMD is now promising "Up To" 3X performance on their next generation of processors:

Mind you that 8 fps is basically unplayable for basically any game in existance, but whatever.

Man... I hope this API works at least half as well as promised, or nobody will ever take AMD seriously again.

I thought maybe just maybe I'd take them seriously. Then today they released kaveri.

0-10% CPU gain (AT review shows basically 0 gain in CPU performance over richland on average). Minimal GPU gain, most greatly seen at high settings where everything is unplayable. And from what techreport and other sites are showing fudging on the TDP numbers (cause if those A8 SKU's are getting that power consumption at that performance level then we are looking at more than ivy/haswell efficiency).

This morning I thought that while high 45% was okay reasonable (more expecting 30-40%). Now, after the marketing crap AMD pulled with Kaveri (30% IPC increase = biggest lie ever), do I believe 45%. No way in hell I do.

 

[mikk]

The fact that they used an intel processor when they launched the R9 series to show how good it was vs nvidia, but then used a kaveri processor when doing dx vs mantle makes me slightly suspicious.

It's coming from a Kaveri presentation, so it makes sense to use Kaveri. And for Mantle presentations Intel wouldn't be choice number one, a slow CPU with many cores should deliver the biggest gains from Mantle. FX-8120 for example.

 

[blackened23]

I don't see where they say AMD stated it would be 10%.

Its at hardware.fr. Apparently they have inside sources at AMD who gave them the scoop, and they said performance would be around 10% in gpu limited situations. I'm sure you can go there and read about it via translation. I'll help:

For situations where the GPU is the limiting factor, the gain would turn around instead of 10% based on the information we collected. We will of course check it all as soon as the patch is available, hoping that the delay is not extended beyond January.

In any case, 45% doesn't seem realistic at all - i'd agree with enigmoid. Given Kaveri's hype (and prior products, I could name quite a few) what happens generally doesn't align with what marketing slides state when it comes to AMD. Basically the marketing slides are purposed to generate hype and usually aren't overly indicative of reality. And 45% is from an AMD marketing slide. I'd expect real world performance gains to be less.

 

[mikk]

A minor gain in GPU bound situations is expected. The big gains comes from CPU bound cases. Real world performance gains are highly dependent on the CPU and game.

 

[blackened23]

A minor gain in GPU bound situations is expected. The big gains comes from CPU bound cases. Real world performance gains are highly dependent on the CPU and game.

Well one could argue that the games that need the most performance increase are the ones that are GPU bound. Most of the super demanding AAA DX11 titles are gpu limited at high resolutions, while CPU bound games tend to be the less demanding ones such as MOBA, MMO and RTS games. Most of the CPU bound titles are (usually) running at high framerates anyway, although there are exceptions.

In any case, we'll see what happens. I just don't expect 45%. I should add, it was AMD who told hardware.fr 10% in GPU limited situations. So it isn't exactly speculation on that part.

 

[Skurge]

I thought maybe just maybe I'd take them seriously. Then today they released kaveri.

0-10% CPU gain (AT review shows basically 0 gain in CPU performance over richland on average). Minimal GPU gain, most greatly seen at high settings where everything is unplayable. And from what techreport and other sites are showing fudging on the TDP numbers (cause if those A8 SKU's are getting that power consumption at that performance level then we are looking at more than ivy/haswell efficiency).

This morning I thought that while high 45% was okay reasonable (more expecting 30-40%). Now, after the marketing crap AMD pulled with Kaveri (30% IPC increase = biggest lie ever), do I believe 45%. No way in hell I do.

You do that Richland is clocked higher than Kaveri right? The fact that they are even close with a 400mhz deficit to Kaveri shows that IPC has gone up quite a bit. 2nd thing. Looking at TR. Going from 65W to 45W does drop consumption by 20W as it should AND with performance doesn't drop much if at all as shown in AMDs slides for Kaveri. So I don't see any lies here.

So Hardware.fr has sources inside AMD like charlie D. has sources inside AMD, Nvidia and Intel? Sure.

 

[Skurge]

Well one could argue that the games that need the most performance increase are the ones that are GPU bound. Most of the super demanding AAA DX11 titles are gpu limited at high resolutions, while CPU bound games tend to be the less demanding ones such as MOBA, MMO and RTS games. Most of the CPU bound titles are (usually) running at high framerates anyway, although there are exceptions.

In any case, we'll see what happens. I just don't expect 45%. I should add, it was AMD who told hardware.fr 10% in GPU limited situations. So it isn't exactly speculation on that part.

You keep saying this and it isn't what hardware.fr said at all. No where in the link does it say AMD told them it would be 10% and no where in the article does it say they have sources inside AMD.

What it does say is they the gain would be 10% from data they have collected. That could mean anything really. But you keep claiming AMD told them this. They don't even say this. I'm not sure what your problem is. Sure AMD have lied before (That is usually the CPU division) and we should question anything this say until proven by independent sources, but that doesn't mean you can go around spreading false info. You are ofcourse interested in the facts are you not?

 

[Silverforce11]

Whats interesting with Kaveri is not on AT's benching at 1080p with stupid settings to give a slideshow.. what most of us who consider a HTPC gaming rig consider, is 1080p at medium settings in newer games, and can it sustain >30 fps for that? Output to big 1080p TV.

Obviously it crushes other iGPU at 1080p, but at unplayable settings so the review was entirely useless.

As for Mantle, it can't come soon enough, too much hype it better lives up to expectations.

 

[blackened23]

You keep saying this and it isn't what hardware.fr said at all. No where in the link does it say AMD told them it would be 10% and no where in the article does it say they have sources inside AMD.

What it does say is they the gain would be 10% from data they have collected. That could mean anything really. But you keep claiming AMD told them this. They don't even say this. I'm not sure what your problem is. Sure AMD have lied before (That is usually the CPU division) and we should question anything this say until proven by independent sources, but that doesn't mean you can go around spreading false info. You are ofcourse interested in the facts are you not?

They didn't test mantle themselves. So they were told by AMD. Where else would they get the information, AMD told them 45% and then they say they were told it was more in line with 10% in GPU limited situations. Of course you can twist the words in the article however you want, much like some AMD fans indicated that EA/DICE wasn't paid 8 million dollars for Mantle When 2700+ google search results seem to state that they did.

You can hope for 45%. That's cool. But the chances of that happening, except in excessively cherry picked situations which won't happen in real world gaming? I don't know. Kaveri's pre-launch hype doesn't exactly inspire optimism either. Like you said. AMD have lied before. Right? AMD marketing slides are purposed for hype. Not truth.

Anyway, i'm just saying, 45% is probably not real world. If it's 10 % performance gain for free, if that's what happens as per hardware.fr, that is still respectable.

 

[blastingcap]

They didn't test mantle themselves. So they were told by AMD. Where else would they get the information, AMD told them 45% and then they say they were told it was more in line with 10% in GPU limited situations. Of course you can twist the words in the article however you want, much like some AMD fans indicated that EA/DICE wasn't paid 8 million dollars When 2700+ google search results seem to state that they did.

You can hope for 45%. That's cool. But the chances of that happening, except in excessively cherry picked situations which won't happen in real world gaming? I don't know. Kaveri's pre-launch hype doesn't exactly inspire optimism either. Like you said. AMD have lied before. Right? AMD marketing slides are purposed for hype. Not truth.

AMD's slides say "up to" 45%, that's not lying. It's ambiguous and does not really claim anything at all. Just wait and see, I guess. No new info. Plus up to 45% is with Kaveri, CPU-bound. Not sure how relevant it is for those running 4770s and the like.

 

[Enigmoid]

You do that Richland is clocked higher than Kaveri right? The fact that they are even close with a 400mhz deficit to Kaveri shows that IPC has gone up quite a bit. 2nd thing. Looking at TR. Going from 65W to 45W does drop consumption by 20W as it should AND with performance doesn't drop much if at all as shown in AMDs slides for Kaveri. So I don't see any lies here.

So Hardware.fr has sources inside AMD like charlie D. has sources inside AMD, Nvidia and Intel? Sure.

Check TR.....

Bout the same idle power.

Hmm. Much higher power than richland. But 79-35 = 44 and 98 -35 = 64. Seems legit.....Wait.... where is the igp? Wait... is that a 84W 4770k using less power than a supposed 65W SKU?

(and though TR labels it as peak, it clearly is not, when you look at their energy over time graphs peak is higher)

Using TR's test and AMD's TDP definition then the i3 would be a 32W part and the i7 a 55W part.

Here is efficiency. 65W moved nowhere. 45W got worse.

This is why I do not believe AMD with their mantle claims. But I also do not believe that nvidia's K1 will deliver as promised at a tablet power level.

(This post is a little OT and really belongs in the kaveri review thread but the primary goal was to show that AMD PR = not reliable, I won't post anything like this in this thread anymore).

 

[Spjut]

They didn't test mantle themselves. So they were told by AMD. Where else would they get the information, AMD told them 45% and then they say they were told it was more in line with 10% in GPU limited situations. Of course you can twist the words in the article however you want, much like some AMD fans indicated that EA/DICE wasn't paid 8 million dollars for Mantle When 2700+ google search results seem to state that they did.

Unless you work for AMD, you don't know where those 8 million dollars went. I'd guess that the sum covers the game bundles, marketing rights, as well as R&D on Mantle.