AMDs Mantle

[spaceelf]

...because we've talked about it enough in other threads

 

http://www.geeks3d.com/20131113/amd-mantle-first-interesting-slides-and-details-target-100k-draw-calls-per-frame/

 

The image code isn't working for me ATM so here's some choice bits

 

According to AMD, it will be possible to have up to 100K render calls per frames with Mantle, while today, optimized render codes achieve only 10k draw calls/frame.

 

Another interesting fact is that Mantle is not limited to AMD Radeon GCN architecture. The core of Mantle could be ported to other GPUs (NVIDIA for example, some NVIDIA GPU designers were present at the AMD Mantle talk #APU13).

[Brandon H Supervisor]

cool beans

 

should be interesting to see what comes of this

[Yusuf M. Veteran]

100K draw calls per frame vs. 10K. Damn. DICE needs to hurry up with the Mantle update for Battlefield 4.

[Andre S. Veteran]

If NVIDIA picks this up, which I'm hoping they will, this might become a serious competitor for Direct3D. Mantle on PC + SteamOS and a similar API on PS4, would make it the best API to optimize for in AAA titles. 

[Boo Berry]

AMD could leverage Mantle in exchange for G-Sync licensing. Everyone wins in that case. Can you imagine Mantle + G-Sync?

[The_Decryptor Veteran]

Yay for proprietary APIs and marketing nonsense! Nvidia have their own low level API but they don't market it as the be all and end all for performance issues in game engines.

 

100K draw calls per frame vs. 10K. Damn. DICE needs to hurry up with the Mantle update for Battlefield 4.

Considering Nvidia was seeing 40-50k calls a second back around 2004, I think that's pretty much just marketing talk.

Edit: The only company talking about how well consoles perform to PCs (That should be the first flag) is AMD, and AMDs solution is an API that only works for their hardware, so...

[Athernar]

Yay for proprietary APIs and marketing nonsense! Nvidia have their own low level API but they don't market it as the be all and end all for performance issues in game engines.

 

Considering Nvidia was seeing 40-50k calls a second back around 2004, I think that's pretty much just marketing talk.

Edit: The only company talking about how well consoles perform to PCs (That should be the first flag) is AMD, and AMDs solution is an API that only works for their hardware, so...

 

CUDA.

[Boo Berry]

NVAPI.

 

[spaceelf]

Yay for proprietary APIs and marketing nonsense! Nvidia have their own low level API but they don't market it as the be all and end all for performance issues in game engines.

 

Considering Nvidia was seeing 40-50k calls a second back around 2004, I think that's pretty much just marketing talk.

Edit: The only company talking about how well consoles perform to PCs (That should be the first flag) is AMD, and AMDs solution is an API that only works for their hardware, so...

Weren't we still using fixed transform and lighting back then?  Not too hard to be fast when you're not really doing anything.

 

As NVIDIA is very likely to be adding support I'm not really sure why you're so down on it.

[psionicinversion]

we'll have to se how it goes tbh

[spaceelf]

EA and other developers on the topic

http://techreport.com/news/25651/mantle-to-power-15-frostbite-games-dice-calls-for-multi-vendor-support

The Mantle version of Battlefield 4 is on track to be released as an update in late December. Andersson said creating a Mantle version of the Frostbite 3 engine took about two months of work. The Mantle release's core renderer is closer to the PlayStation 4 version than to the existing DirectX 11 one, and it includes both CPU and GPU optimizations. Andersson didn't bring up performance estimates, but other developers who discussed Mantle at APU13 did. Jorjen Katsman of Nixxes, the firm porting Thief to the PC, mentioned a reduction in API overhead from 40% with DirectX 11 to around 8% with Mantle. He added that it's "not unrealistic that you'd get 20% additional GPU performance" with Mantle.

[TyRidian]

I'm curious how well this will work for their entry level graphics cards R7 240 and R7 250, both in which offer Mantle out of the box.

 

Some people say that the R7 250 in particular is a downgrade in comparison to the HD 7700 series cards, but wouldn't Mantle change this, even with the lower specs that the R7 250 has?

[Slammers]

I'm curious how well this will work for their entry level graphics cards R7 240 and R7 250, both in which offer Mantle out of the box.

 

Some people say that the R7 250 in particular is a downgrade in comparison to the HD 7700 series cards, but wouldn't Mantle change this, even with the lower specs that the R7 250 has?

 

Id say certain areas of the low end cards would bottleneck mantle in this case.

[TyRidian]

Id say certain areas of the low end cards would bottleneck mantle in this case.

Kind of defeats the purpose of offering Mantle then, especially if they advertise the cards as supporting it.

[spaceelf]

Kind of defeats the purpose of offering Mantle then, especially if they advertise the cards as supporting it.

I'm sure it'd help them, but they're still going to be pretty weaksauce.

[TyRidian]

I was going to get the R7 250, Because I am cheap like that hahaha, but then realized that an HD 7790 or even an R7 260x would be a better fit in the $100 to $140 price range.

[The_Decryptor Veteran]

Weren't we still using fixed transform and lighting back then?  Not too hard to be fast when you're not really doing anything.

 

As NVIDIA is very likely to be adding support I'm not really sure why you're so down on it.

I seriously doubt Nvidia will redesign their GPU to resemble an AMD one anytime soon.

Because that's what Mantle is, a thin wrapper around the underlying GPU not, not a reference GPU or such. Even AMD say it's (currently) tied to GCN.

Of course Nvidia could change Mantle to work on their cards, but then it's not the same Mantle that works on AMD cards, which defeats the whole purpose.

Edit: And no, we had shaders. No idea why AMD cards apparently suffer such low draw performance (10,000 draw calls a second at 60fps is about 167 per frame, which seems extremely limited)

[psionicinversion]

I seriously doubt Nvidia will redesign their GPU to resemble an AMD one anytime soon.

Because that's what Mantle is, a thin wrapper around the underlying GPU not, not a reference GPU or such. Even AMD say it's (currently) tied to GCN.

Of course Nvidia could change Mantle to work on their cards, but then it's not the same Mantle that works on AMD cards, which defeats the whole purpose.

Edit: And no, we had shaders. No idea why AMD cards apparently suffer such low draw performance (10,000 draw calls a second at 60fps is about 167 per frame, which seems extremely limited)

 

Im sorry mate but do you actually read anything?

 

Nvidia doesnt have to change there architecture what Mantle does is make full use of the GCN architecture by optimising it for the pipeline or whatever it is. For Nvidia to use it theyd just add optimisations for nvidias architecture to Mantle so it can be used by them. Theyd prolly add support for the newer Maxwell architecture first if they were going to go along with mantle which will be released in a few months. Mantle is designed to be a low level API to enable more performance over Directx which is a jack of all trades and isnt optimised to be used on specific architectures which is why its slower and doesnt make full use of it.

[spaceelf]

I seriously doubt Nvidia will redesign their GPU to resemble an AMD one anytime soon.

Because that's what Mantle is, a thin wrapper around the underlying GPU not, not a reference GPU or such. Even AMD say it's (currently) tied to GCN.

Of course Nvidia could change Mantle to work on their cards, but then it's not the same Mantle that works on AMD cards, which defeats the whole purpose.

Edit: And no, we had shaders. No idea why AMD cards apparently suffer such low draw performance (10,000 draw calls a second at 60fps is about 167 per frame, which seems extremely limited)

It's fair to say that's a bunch of assumptions with a complete lack of evidence.  NV had reps at the event, so they were obviously interested and likely to support it.

 

Also, it's draw calls per frame, not per second...as you can see in the article.

 

From the TechReport article I linked up there somewhere -

 

But the "pink elephant in the room," as he called it, is multi-vendor support. Andersson made it clear that, while it only supports GCN-based GPUs right now, Mantle provides enough abstraction to support other hardware?i.e. future AMD GPUs and competing offerings. In fact, Andersson said that most Mantle functionality can work on most modern GPUs out today. I presume he meant Nvidia ones, though Nvidia's name wasn't explicitly mentioned. In any event, he repeated multiple times that he'd like to see Mantle become a cross-vendor API supported on "all modern GPUs."

[Boo Berry]

[Pupik]

AMD could leverage Mantle in exchange for G-Sync licensing. Everyone wins in that case. Can you imagine Mantle + G-Sync?

G-Sync is going to be a waste with Mantle. G-Sync is only good for games that drop bellow 60fps to low numbers frequently.

Currently, there's not a lot of games that do that with the current hardware. And with Mantle improving the game's performance even more, there won't be any reason to buy a G-Sync monitor.

[spaceelf]

G-Sync is going to be a waste with Mantle. G-Sync is only good for games that drop bellow 60fps to low numbers frequently.

Currently, there's not a lot of games that do that with the current hardware. And with Mantle improving the game's performance even more, there won't be any reason to buy a G-Sync monitor.

I'd say you might want to read up on GS a bit more.  You can always push detail higher than your card can handle, and even if you have a 120/144hz mon it should be very nice tech.

 

But I'm going to sleep at 9am for some reason nobody actually knows.  See you fine ladies and gentlemen later. <insert picture of guy with monocle here>

 

I hope Mantle puts NVAPI/PhysX/Cuda to rest.  I don't need vendor locked garbage and NV has been all about that lately.

[Andre S. Veteran]

NVAPI.

NVAPI is not a graphics API, it's a bunch of utility functions about the state of the graphics card.

[Andre S. Veteran]

G-Sync is going to be a waste with Mantle. G-Sync is only good for games that drop bellow 60fps to low numbers frequently.

That's not true. G-Sync is a win at any framerate that's not perfectly synchronized to the display, which is to say, any game framerate; and even in the case of perfect synchronization (which never really happened and was hacked through V-sync), it will reduce latency to the absolute minimum response time of the monitor.

 

Even in case of a solid V-synced 60fps you still have to wait anywhere between 0 to 16.7ms between presentation and display, so even at 60fps you have this slight judder and irregular latency. Actually the nice thing about G-Sync is that 60fps ceases to be this magic number; any decently high framerate will feel responsive and smooth. 

 

Besides, the assumption that running a game on Mantle will automatically translate into 60+ frames per second regardless of video card specs and resolution is obviously wrong.

[Andre S. Veteran]

I seriously doubt Nvidia will redesign their GPU to resemble an AMD one anytime soon.

Because that's what Mantle is, a thin wrapper around the underlying GPU not, not a reference GPU or such. Even AMD say it's (currently) tied to GCN.

Of course Nvidia could change Mantle to work on their cards, but then it's not the same Mantle that works on AMD cards, which defeats the whole purpose.

repi at APU13 yesterday said exactly the opposite. He affirms Mantle is not tied to GCN and would be a very nice API for other vendors (i.e. NVIDIA) to implement. http://www.dsogaming.com/news/amds-mantle-does-not-require-gpus-with-gcn-architecture/

 

Not sure where the rumor got started that Mantle was GCN only anyway?