Mantle: Teething Problems (Anandtech R9 285 review showing the downsides of low level APIs)

[mmj_uk]

Here's the thing though. This isn't an issue on DX11 for example.

Problems like these are exactly the stuff that higher level, industry standard APIs for 3d graphics were made to solve. Having to provide individual support and patches for individual GPUs and CPUs is the mess that we got away from due to DX etc.

And if this continues to be a problem then why develop for mantle in the first place if (without future patches) your mantle implementation as a developer will only offer good performance for current GPUs, and not future ones? Why would I now develop a mantle game if it can't without patching for each new GPU offer good performance for anything but Tahiti, Bonaire, Hawaii etc.?

Well said, key point highlighted.

 

[shankly1985]

So AMD saying it isnt enough of a fact?

So maybe it's just a driver tweaking? We don't fully know.
Like I said, before remember when thief and bf4 got the Mantle patch, we still needed a newer driver before we could use Mantle. So the driver is still in some way being used.

And I do agree if this is true and that game devs will need to patch a game couple years back, it isn't good for Mantle. And it also won't be good for dx12 either.

Am sure this isn't the case, am sure bigger minded people I'm the gaming world would off pointed this out way before now.

Guess time will tell.

 

[ubersonic]

I don't see how anyone can blame the devs here, they were nice enough to add Mantle support to the game, why anyone would expect them to update the game every time AMD releases a new card I have no idea. I mean it's not like it had any effect on their sales, nobody said "I'm going to buy BF4 because it has Mantle support!".

It's like complaining because your free Haribo wasn't the flavour you wanted.

 

[GoogalyMoogaly]

How Mantle perform- good or bad has nothing to do with the fact that the "current dx" is lacking, and people and developers alike been crying out for something better or improvement. People were stuck with dx even if they weren't happy with the limitation, because they didn't have an alternative choice option.

Just think of dx and Mantle being the motorway and the toll...I'd rather have the option of toll available than not. If there's maintainence work on the toll leading to longer travel time than the motorway, I'd just use the motorway instead. It is just that plain and simple.

OpenGL is and has been an alternative for some time.

Maybe Mantle is nice for developers, unless they have to revisit every Mantle game they've made every time AMD release a new architecture. They might get bored with that.
But if they don't then 'we' suffer as our new cards won't be supported under Mantle in older games.
Is that what people have been crying out for?

 

[dangel]

Is that what people have been crying out for?

Oh yes, we all miss the days of Glide immensely.

People act like DX doesn't do a whole heap of GOOD for end users and developers alike which of course it does. Abstraction has huge benefits regardless of whether an implementation is skewed the wrong side of the compromise on performance. MS will simply update DX and nullify mantle completely, and nVidia will do absolutely jack to support mantle either. Mantle only serves one purpose - to push MS into providing better performance in DX and nVidia into closing the gap in their drivers. Great that it's working but once that's done it's a dead duck (and I hope so from the POV of an end user and developer alike).

I don't see how anyone can blame the devs here, they were nice enough to add Mantle support to the game, why anyone would expect them to update the game every time AMD releases a new card I have no idea. I mean it's not like it had any effect on their sales, nobody said "I'm going to buy BF4 because it has Mantle support!".

It's like complaining because your free Haribo wasn't the flavour you wanted.

Because for an end user they simply want stuff to work (well if possible but basically functional is a primary requirement) and they have no appreciation of how software works. Mantle sets itself up for failure in this regard obviously.

 

[andybird123]

nobody said "I'm going to buy BF4 because it has Mantle support!".

erm, I've seen people say exactly that about several mantle titles

it isn't the devs who are directly to blame, but they did sign up to a system that places a greater onus on them as gatekeepers of performance - which they said they wanted... if the same developers then want to shrug their shoulders of being responsible for updating their own software then they have to accept some criticism, alongside AMD where drivers (DX or mantle) are also causing issues

 

[humbug]

I don't dispute their findings, well, i don't have any information to contradict their findings, but at the same time i don't read much into what many of these reviewers do with Mantle, including Anand, it seems to me they just don't understand it, or are being deliberately dumb and for whatever reason don't like to show it in a good light.

If they understood it or wanted to test for what it exists for they would not be using the most powerful CPU they could find.

For that blindingly obvious flaw in their reviews i'm not sure i trust anything they do.

 

[LambChop]

I don't dispute their findings, but at the same time i don't read much into what many of these reviewers do with Mantle, it seems to me they just don't understand it, or are being deliberately dumb and for whatever reason don't like to show it in a good light.

If they understood it or wanted to test for what it exists for they would not be using the most powerful CPU they could find.

For that blindingly obvious flaw in their reviews i'm not sure i trust anything they do.

And on the other hand, had they been showing it in a good way, I bet you'd have believed it all.

 

[humbug]

And on the other hand, had they been showing it in a good way, I bet you'd have believed it all.

I know what Mantle does, and what it does not, i use it everyday with a CPU thats not an overclocked 12 Thread Intel CPU.

I can believe what i know is true, and disbelieve what i know is not, in this case i don't know, i don't have a Tonga GPU.
What i do know is the most incompetent way i could test Mantle is by using one of the most powerful CPU's you can get, which is exactly what Anand did.

 

[andybird123]

I don't dispute their findings, well, i don't have any information to contradict their findings, but at the same time i don't read much into what many of these reviewers do with Mantle, including Anand, it seems to me they just don't understand it, or are being deliberately dumb and for whatever reason don't like to show it in a good light.

If they understood it or wanted to test for what it exists for they would not be using the most powerful CPU they could find.

For that blindingly obvious flaw in their reviews i'm not sure i trust anything they do.

Surely, at best for Dx with a powerful Cpu should be equal to mantle, because there is no cpu bottleneck and the gpu is performing to its maximum capability?
Mantle being a good percentage behind DX, regardless of the cpu being used, is a curious result and shows an issue with mantle's Gpu performance, which is irrelevant to what cpu is being used

Why should they be expected to create an artificial DX bottleneck in order to test mantle's GPU performance?

 

[LambChop]

I know what Mantle does, and what it does not, i use it everyday with a CPU thats not an overclocked 12 Thread Intel CPU.

I can believe what i know is true, and disbelieve what i know is not, in this case i don't know, i don't have a Tonga GPU.
What i do know is the most incompetent way i could test Mantle is by using one of the most powerful CPU's you can get, which is exactly what Anand did.

So having a powerful cpu harms Mantle ?

 

[humbug]

Surely, at best for Dx with a powerful Cpu should be equal to mantle, because there is no cpu bottleneck and the gpu is performing to its maximum capability?
Mantle being a good percentage behind DX, regardless of the cpu being used, is a curious result and shows an issue with mantle's Gpu performance, which is irrelevant to what cpu is being used

Why should they be expected to create an artificial DX bottleneck in order to test mantle's GPU performance?

Just to reiterate i don't have a Tonga GPU so i don't know if there is a problem with that architecture on Mantle, so i do not dispute it.

On a broader sense if they are testing Mantle then yes they do need to test it for its purpose, otherwise at best its useless information. at worst misinformation, depending of how they present their findings. if they overclock the most power CPU they can find and test it on a mid level card and use that to conclude Mantle does little or nothing, for example, then that IS blatant misinformation.

 

[_Alatar_]

The findings aren't about the lowered overhead. They're about how games apparently need additional patches to properly support new GPUs from AMD.

Whether you hide the lost GPU performance with CPU overhead improvements (by using a low end CPU) or not the lost GPU performance is still there.

With a lower end CPU this would show as Tonga chips not seeing as big gains percentage wise as GCN GPUs that were launched before the game came out.

Using a high end CPU it's easier to illustrate the need for extra support. And that need for extra support/patches from game devs isn't going anywhere, intel extreme edition or not.

 

[Sin_Chase]

Oh my oh my oh my!

This thread makes for PAINFUL reading.

Basically people are defending the supposed findings with Mantle on new architectures with ludicrousy such as "need to test it for it's purpose" AKA "Should have tested it with a budget scrub CPU" I mean, really? Since when did Mantle change from a low level API that pushed the performance boundaries to only applicable in weak hardware situations? When the reality does not fit it seems?

So Mantle, The Best API since sliced bread is now - Mantle, for people with weak CPUS. DON'T USE IT ON A DECENT ONE!!!!!

Please, ignoring the benefit of the CPU overhead improvements Mantle brings is a good way of illustrating in raw terms how it is performing in "Balls To The Wall" mode. IE - Not on scrubtastic budget hardware.

- The official line from AMD is "Updates needed"
- The apparent results corroborate this.
- DX is now a viable alternative for when The Best API since sliced bread...is not. (oh the irony)

Mantle still needs lot's of work - Period. Don't do it a disservice by pitching weak excuses and just accept that.

 

[Orangey]

...people have always been saying Mantle makes the most sense on low CPUs.

 

[Sin_Chase]

...people have always been saying Mantle makes the most sense on low CPUs.

Ok, then everyone who wants that out of Mantle as an API can also hide behind the veil that such a testing scenario introduces.

That being - intentionally bottlenecking DX with a naff CPU to show Mantle in a good light. Basically, running better than DX because Mantle handles CPU overhead better. That's not a superior API, that's an API which handles CPU overhead better.

Regressive patching and driver updates for old games - All to boost some mid to low range hardware over DX in bottlenecked situations.

AWESOME!

Mantle making the most sense on budget hardware does not excuse it from failures where budget hardware is not a factor.

 

[mmj_uk]

...people have always been saying Mantle makes the most sense on low CPUs.

Not really, it started out with all of the 10x more draw calls, 200% faster than DX etc hype... it's only when the actual results came in and the only significant improvement was on slow CPU's it then became the API for slow CPU's.

 

[Silent_Scone]

Not really, it started out with all of the 10x more draw calls, 200% faster than DX etc hype... it's only when the actual results came in and the only significant improvement was on slow CPU's it then became the API for slow CPU's.

Unfortunately this is the truth as AMD were basing all their hopes on how DICE managed to optimise Frostbite for it. That is when the target audience changed dramatically

 

[Final8y]

When its supported i will use it when its not i will use DX,OpenGL, its as simple as that.

 

[Rroff]

Not really, it started out with all of the 10x more draw calls, 200% faster than DX etc hype... it's only when the actual results came in and the only significant improvement was on slow CPU's it then became the API for slow CPU's.

Game shifted a bit when nVidia started writing driver level function overrides for DX layer functions that worked around some of the API bottlenecks.