Impications of Mantle and AMD's slower per core but more core strategy...

[BallaTheFeared]

As a ostracized AMD fan at heart, this has caught my inner glee receptors.

If people weren't aware, because they haven't heard or avoid the GPU section of our forums...

http://www.techspot.com/news/54134-amd-launches-mantle-api-to-optimize-pc-gpu-performance.html

To summarize:

AMD claims Mantle enables nine times more draw calls per second than other APIs, which is a huge increase in performance.

I also believe I read it was designed for use on both consoles, and is specifically meant to help enable eight core cpu multithreading at previously unseen levels of utilization.



So while I in the past have said, paraphrasing... "Even if future games use eight cores, the 8350 will only reach similarity with the i5 from Intel. And this will likely only include cross platform titles and no PC exclusives where the real burden of AMDs per core performance is often seen"

Nothing I've said has really been changed, only confirmation that it's possible through a proprietary API that targets GCN based GPUs AMD CPU users of eight core processors could greatly benefit from many cross platform titles (assuming most devs use this).

So my thought/question for the CPU section, is will this influence your future purchase decisions, not for graphics cards, but for CPUs specifically? Or will this be of no benefit to you because you prefer Nvidia GPUs, or will you switch from nVidia with your already purchased AMD cpu to benefit from this?

 

[BrightCandle]

Need to see the benefits before I do anything. Most games do have a thread that is exclusively graphics and runs at high usage, but that does not mean reducing the utilisation of that thread will bring practical benefits. The game world updates will still be in the developers hands and they are still single threaded.

I am optimistically watching to see the impact, but I am sceptical because I suspect the reality is far away from the optimal conditions AMD often likes to market with. Let's see what the reviewers make of it all.

 

[cbn]

http://www.anandtech.com/show/7368/amd-gpu-product-showcase-live-blog

06:43PM EDT - Compared to consoles, on PCs developers could only do less with more, with Mantle now the PC developers can get more with more. More info on Mantle will be available at AMD's developer conference in November. Looks like we're finally done, the rest of the day is under NDA. We'll share more info as soon as we can!

06:40PM EDT - Mantle enables 9x more draw calls per second than other APIs by reducing CPU overhead.

06:36PM EDT - Battlefield 4 really looks great, I'm very curious to see how big of an impact the Mantle update makes on performance and things like CPU utilization

06:32PM EDT - BF4 will launch with a DX11.1 renderer in October, in December it'll launch with a Mantle renderer that'll let gamers get the most out of their PCs.

06:30PM EDT - Whats being described for Mantle sounds a great deal like Glide for the modern age with compute capabilities

06:29PM EDT - The proposal: Mantle. A low level high performance console style graphics API for PCs. It's being built by AMD. It's cross platform, but Windows initially. Battlefield 4 is the pilot project & first user of Mantle.

06:28PM EDT - The proposal: Mantle. A low level high performance console style graphics API for PCs. It's being built by AMD. It's cross platform, but Windows initially. Battlefield 4 is the pilot project & first user of Mantle.

06:24PM EDT - Talking about challenges for PC development - PC is easy to develop for on the CPU, GPU development is still stuck in a traditional model where the CPU is still feeding the GPU, and not all modern functionality on GPUs is exposed. Abstractions are nice but having low level access can be important (and obviously a benefit from a performance standpoint)

Sounds really interesting!

I wonder how cpu loads will be affected in Battlefield 4 vs. Battlefield 3?

EDIT: (My mistake) Actually CPU load comparison pre and post Mantle on BF4 would be more pertinent.

 

[SPBHM]

I thought the strategy of AMD for the PC was basically shifting to less cores, with 2-3 module APUs,

still, I don't know exactly what to think about it, the most interesting aspect would be, how the console ports will translate in terms of cpu/memory performance with unified memory vs what we have on the PC (PCI Express and so on)...

but as for CPUs, I still think there is a huge difference in terms of performance from the 8 core Jaguar vs a 4 core SB/IB/HW, and there is also a huge difference to Piledriver based CPUs, so I don't think it will make much of a difference!?

 

[frozentundra123456]

As a ostracized AMD fan at heart, this has caught my inner glee receptors.

If people weren't aware, because they haven't heard or avoid the GPU section of our forums...

http://www.techspot.com/news/54134-amd-launches-mantle-api-to-optimize-pc-gpu-performance.html

To summarize: I also believe I read it was designed for use on both consoles, and is specifically meant to help enable eight core cpu multithreading at previously unseen levels of utilization.



So while I in the past have said, paraphrasing... "Even if future games use eight cores, the 8350 will only reach similarity with the i5 from Intel. And this will likely only include cross platform titles and no PC exclusives where the real burden of AMDs per core performance is often seen"

Nothing I've said has really been changed, only confirmation that it's possible through a proprietary API that targets GCN based GPUs AMD CPU users of eight core processors could greatly benefit from many cross platform titles (assuming most devs use this).

So my thought/question for the CPU section, is will this influence your future purchase decisions, not for graphics cards, but for CPUs specifically? Or will this be of no benefit to you because you prefer Nvidia GPUs, or will you switch from nVidia with your already purchased AMD cpu to benefit from this?

Sounds promising, but again like HSA and AVX2, I would wait to see what the results are. I guess I do not really understand the interactions of cpu and gpu enough to know how much this is really going to affect performance and how much it would affect Intel cpus vs AMD cpus. It should also increase performance on Intel cpus as well, right? Only thing is if some other factor is limiting the performance, increasing the number of draw calls would have little effect. It also sounds like nVidia could incorporate this if they wish.

Would they also incorporate this into the integrated gpu part of their APUs? I would assume so, otherwise it is kind of negating their focus on APUs. Maybe finally there will be an IGP that is decent for more than 720p gaming.

But to answer your question, I an running an i5, so I might upgrade the gpu, but I dont go for max visuals so I would stick with whatever performance my current cpu gives. Would this affect older games as well, or would there have to be a patch to update the game to use this in a game not designed for it?

 

[monstercameron]

Sounds promising, but again like HSA and AVX2, I would wait to see what the results are. I guess I do not really understand the interactions of cpu and gpu enough to know how much this is really going to affect performance and how much it would affect Intel cpus vs AMD cpus. It should also increase performance on Intel cpus as well, right? Only thing is if some other factor is limiting the performance, increasing the number of draw calls would have little effect. It also sounds like nVidia could incorporate this if they wish.

Would they also incorporate this into the integrated gpu part of their APUs? I would assume so, otherwise it is kind of negating their focus on APUs. Maybe finally there will be an IGP that is decent for more than 720p gaming.

But to answer your question, I an running an i5, so I might upgrade the gpu, but I dont go for max visuals so I would stick with whatever performance my current cpu gives. Would this affect older games as well, or would there have to be a patch to update the game to use this in a game not designed for it?

afaik the whole gfx engine has to be tweaked like consoles to allow this to work. So maybe not just a patch but an engine rewrite.

 

[cbn]

http://www.techspot.com/news/54134-amd-launches-mantle-api-to-optimize-pc-gpu-performance.html

I find it very interesting indeed that Mantle will be used rather than DX11.

What effect would this have on porting to Linux?

 

[ZGR]

When the mantle update arrives in December for BF4, it would be interesting to see a comparison to the BF4 without the mantle update.

But right now, I need all the single-threaded performance I can get; which AMD is failing to do.

 

[SammichPG]

So my thought/question for the CPU section, is will this influence your future purchase decisions, not for graphics cards, but for CPUs specifically?

It's either a marketing gimnic or something that I don't like at all.
I remember well when games had to ship with different executables including different manifacturer extensions (glide, power vr and several others) and I'm not looking forward to it.

 

[erunion]

Nope. We've already seen the minimal gains brought by multithreaded driver calls in DX11.

Game threads will still need fast single threaded performance. That's where the bottleneck is.

 

[SPBHM]

http://www.techspot.com/news/54134-amd-launches-mantle-api-to-optimize-pc-gpu-performance.html

I find it very interesting indeed that Mantle will be used rather than DX11.

What effect would this have on porting to Linux?

I don't know if it changes anything, since they probably still going to have to also port to OpenGL because of Intel and Nvidia (and 6900 series and lower)!?

Mantle will only be used for GCN GPUs, when the patch is released from what I understand, DX11 looks like the main target.

 

[AtenRa]

We will know more on November 11-14.

 

[cbn]

I don't know if it changes anything, since they probably still going to have to also port to OpenGL because of Intel and Nvidia (and 6900 series and lower)!?

Mantle will only be used for GCN GPUs, when the patch is released from what I understand, DX11 looks like the main target.

I would just love to see AAA games run natively on Linux.....even if it meant I was forced to use an AMD iGPU* or Discrete AMD GPU with this mantle graphics API.

(*This provided AMD improves their single thread cpu performance sufficiently on Kaveri, otherwise I will continue to use Intel CPU with AMD Discrete GPU.)

 

[ShintaiDK]

Isnt draw calls just getting the current hype. After someone found out the consoles got less overhead than the PC? Even if it was a 10000x more draw calls, it means nothing unless we know how much execution time it takes.

Its not gonna change CPU wise. Just as AMD itself killed anything above 4 cores for the near future.

"Multithreaded drivers" didnt change much either. But as always the hype is strong.

Also ask yourself this. Did they show a slide for how much this actually increase performance/lower execution overhead? If no, then it doesnt really matter much. Because it was deemed too insignificant to show contra saying 9x more as text only.

 

[Dribble]

Not convinced it'll mean much in the end. It only applies to fixed range of AMD gpus (too old and it won't work, future ones will be too advanced and it won't work). Hence the number of people that can take advantage of this will be limited - most use Intel or nvidia anyway, and say it only works with half the AMD gpu's out there that's a too small a % for devs to care much about.

It's used instead of direct X - MS not going to be impressed about that. They'll want AMD to make direct X to work efficiently for xbone, using an AMD api instead won't be a solution MS will accept. As direct X is already on PC and works for all vendors chances are the ports will still just be direct X.

The cull thing is a solution to specific console bottleneck (traditional cull uses one very fast cpu core but consoles have lots of low powered cpu cores), it's not a problem that most PC's have.

 

[GaiaHunter]

http://www.anandtech.com/show/7371/understanding-amds-mantle-a-lowlevel-graphics-api-for-gcn

At the highest level Mantle exists because high level API have drawbacks in exchange for their ability to support a wide variety of GPUs. Abstractions in these APIs hide what the hardware is capable of, and the code that holds those abstractions together comes with its own performance penalty. Of all of those performance issues the principle issue at hand is the matter of draw calls, which are the individual calls sent to the GPU to get objects rendered. A single frame can be composed of many draw calls, upwards of a hundred or more, and every one of those draw calls takes time to setup and submit.

Although the issue will receive renewed focus today with the announcement of Mantle, we have known for some time now that groups of developers on both the hardware and software side of game development have been dissatisfied with draw call performance. Microsoft and the rest of the Direct3D partners addressed this issue once with Direct3D 10, which significantly cut down on some forms of overhead.

But the issue was never entirely mitigated, and to this day the number of draw calls high-end GPUs can process is far greater than the number of draw calls high-end CPUs can submit in most instances. The interim solution has been to attempt to use as few draw calls as possible – GPU utilization takes a hit if the draw calls are too small – but there comes a point where a few large draw calls aren’t enough, and where the CPU penalty from generating more draw calls becomes notably expensive.

Ergo: Mantle. A low-level API that cuts the abstraction and in the process makes draw calls cheap (among other features).

What’s not being said, but what becomes increasingly hinted at as we read through AMD’s material is not just that Mantle is a low level API, but rather Mantle is the low level API. As in it’s either a direct copy or a very close derivative of the Xbox One’s low level graphics API. All of the pieces are there; AMD will tell you from the start that Mantle is designed to leverage the optimization work done for games on the next generation consoles, and furthermore Mantle can even use the Direct3D High Level Shader Language (HLSL), the high level shader language Xbox One shaders will be coded against in the first place. Let’s be very clear here: AMD will not discuss the matter let alone confirm it, so this is speculation on our part. But it’s speculation that we believe is well grounded. Based on what we know thus far, we believe Mantle is the Xbox One’s low level API brought to the PC.

If indeed Mantle is the Xbox One’s low level API, then this changes the frame of reference for Mantle dramatically. No longer is Mantle just a new low level API for AMD GCN cards, whose success is defined by whether AMD can get developers to create games specifically for it, but Mantle becomes the bridge for porting over Xbox One games to the PC. Developers who make extensive use of the Xbox One low level API would be able to directly bring over large pieces of their rendering code to the PC and reuse it, and in doing so maintain the benefits of using that low-level code in the first place. Mantle will not (and cannot) preclude the need for developers to also do a proper port to Direct3D – after all AMD is currently the minority party in the discrete PC graphics space – but it does provide the option of keeping that low level code, when in the past that would never be an option.

Also if AMD actually scores some of the bigger engines that will make sure a bunch of AAA titles will be able to access mantle, as this will be per Engine and not per Game.

 

[Ventanni]

I think we need to take a step back and remove some of the sensation out of AMD/DICE's announcements here. It sounds like a marketing major fresh out of college wrote some of these line.

Mantle won't benefit AMD cpus any more than it will Intel cpus. If that were the case, then we'd all be ditching out Core i7's and FX chips for cheap Jaguar modules, and that just isn't going to happen. The Mantle API is a low level API designed to help reduce CPU overhead by removing the abstract translation layers, not optimize current CPU architecture for a specific rendering path. That's just not how API's work. It's the engine that handles the threading, not the API.

Secondly, DICE's "perfect parallel rendering" statement where they claim to utilize all 8 cores is nothing but a marketing claim. Did DICE suddenly disprove Amdahl's law? No, of course not. The reason they're saying that is because the Frostbite engine has always been exceptionally well-threaded even before Mantle came around. And if I'm not mistaken, BF3 is one of the few games where an FX processor performs more closely to a Core i5/7.

Basically what they're saying is though, is because the Frostbite engine is already well designed to utilize up to 8 threads, and because Mantle simplifies the translation layer between the CPU and GPU, that we'll be getting a lot more out of our 4+ core CPUs. And hey, let's look at the positive side here, if this actually does start a trend where AMD CPUs start outperforming Intel CPUs (which I doubt, it's hard to overcome the enormous IPC difference between Piledriver and Haswell), and their sales are affected by it, then Intel will respond in kind by releasing more hexacore CPUs, and that's a win-win for us!

 

[frozentundra123456]

Hope this isnt too off topic, but is anyone else concerned that these new programming sets could cause fragmentation in the market and make programming/porting to PC (and choosing hardware) *more* complicated? I mean you have Mantle, but what about nVidia cards? Do you have to have a separate codepath? If it replaces DX11, how difficult will it be to program to use both? And you also have Valve pushing Valve OS. How many companies will program for that and how many more resources will that take. What about current and older games, will they run on Mantle, and if not will it still support DX9, DX10, etc so older games will run optimally? Apparently it will require a major rework to make then run on Steam OS.

And what about CPUs? If Mantle somehow pulls AMD ahead on future games, then you will likely be forced to compromise either performance in future games or older games, which I assume would still run better on intel.

Say what you will about Microsoft, at least you got a pretty standardized experience and generally good backwards compatibility.

 

[GaiaHunter]

I think we need to take a step back and remove some of the sensation out of AMD/DICE's announcements here. It sounds like a marketing major fresh out of college wrote some of these line.

Mantle won't benefit AMD cpus any more than it will Intel cpus. If that were the case, then we'd all be ditching out Core i7's and FX chips for cheap Jaguar modules, and that just isn't going to happen. The Mantle API is a low level API designed to help reduce CPU overhead by removing the abstract translation layers, not optimize current CPU architecture for a specific rendering path. That's just not how API's work. It's the engine that handles the threading, not the API.

What happens in situations where the GPU is the bottleneck for gaming?

Intel and AMD CPU perform virtually the same.
In situations where the CPU is bottleneck Intel has a big advantage over AMD CPUs.

Now with mantle there is the possibility of removing some of the CPU bottleneck by allowing the engine to talk closer to the hardware, bypassing the high level API that requires the OS and the CPU.

What do you think it will happen to AMD CPU performance vs Intel CPU performance if the CPU has to do less work?

 

[GaiaHunter]

Hope this isnt too off topic, but is anyone else concerned that these new programming sets could cause fragmentation in the market and make programming/porting to PC (and choosing hardware) *more* complicated? I mean you have Mantle, but what about nVidia cards? Do you have to have a separate codepath? If it replaces DX11, how difficult will it be to program to use both? And you also have Valve pushing Valve OS. How many companies will program for that and how many more resources will that take. What about current and older games, will they run on Mantle, and if not will it still support DX9, DX10, etc so older games will run optimally? Apparently it will require a major rework to make then run on Steam OS.

And what about CPUs? If Mantle somehow pulls AMD ahead on future games, then you will likely be forced to compromise either performance in future games or older games, which I assume would still run better on intel.

Say what you will about Microsoft, at least you got a pretty standardized experience and generally good backwards compatibility.

First this should make it no harder to port to the PC and maybe in some situations even easier.
Big engines and companies could probably have wrappers for both a NVIDIA and an AMD low-level API, although I expect NVIDIA having to invest significant resources to convince developers, while AMD due to their leverage in consoles will have an easier time convincing those coding for multi platform.

AMD mantle should't make Intel CPUs perform any worse, just make AMD CPUs less of a penalty.

We will have to see more about the Valve initiatives, but there is nothing preventing mantle to run alongside opengl (or whatever will be the high level API of it) like it will run with DX. Consoles already run low level API and a high level API.

 

[frozentundra123456]

First this should make it no harder to port to the PC and maybe in some situations even easier.

AMD mantle should't make Intel CPUs perform any worse, just make AMD CPUs less of a penalty.

That is the key I think. As long as one set of hardware improves and the other stays the same, of course it is a net gain. My concern is that optimizing for a vendor specific platform could cause less development resources and optimization of the competitors product. Kind of like physX, although that is not really a game changer.

 

[GaiaHunter]

That is the key I think. As long as one set of hardware improves and the other stays the same, of course it is a net gain. My concern is that optimizing for a vendor specific platform could cause less development resources and optimization of the competitors product. Kind of like physX, although that is not really a game changer.

Will have to wait and see, but that isn't much different than what happens for Intel products vs AMD products in the CPU arena, unless you believe optimizations for CPU architectures don't cause performance differences.

 

[24601]

I think it will be a great boon to people who were stupid enough to purchase a horribad CPU to pair with a 7970 or 7950.

Also, I believe that the reduction of CPU overhead is the most believable claim made in that mantle presentation.

Hilariously, I think the people who will benefit the most are the people with 4 core and 6 core pre-sandy bridge Intel processor owners that happen to have a 7970 or 7950.

These people will be less bottlenecked by their passable CPUs in a game that is predicted by many to be very important for a significant amount of people.

 

[beginner99]

Mantle will only be used for GCN GPUs, when the patch is released from what I understand, DX11 looks like the main target.

Yes but would you buy NV or AMD if AMD performs much better?

Not convinced it'll mean much in the end. It only applies to fixed range of AMD gpus Hence the number of people that can take advantage of this will be limited - most use Intel or nvidia anyway, and say it only works with half the AMD gpu's out there that's a too small a % for devs to care much about.

Also if AMD actually scores some of the bigger engines that will make sure a bunch of AAA titles will be able to access mantle, as this will be per Engine and not per Game.

Exactly. If not only BF4 uses this and it actually has a noticeable impact AMD could benefit greatly because I bet more people would buy an ADM card because it will also perform better with older CPUs and for all we know thats an issue with BF4.
I just hope it's not limited to the 290x BF4 special edition card...

 

[MisterMac]

I find it odd - that it's AMD has to push innovations like this.

Some that might enable better control and more threading in super optimal conditions.



....and Intel+AMD GPUs will most likely gain the most it.
Ironic.