The AMD Mantle Thread

[psoomah]

You aren't looking at this correctly. nVidia will need to support Mantle's feature set, just like they now need to support DX and OpenGL feature sets.


What features are missing from nVidia GPU's? Is this known?

It's not a matter of Nvidia needing to support Mantle, if Mantle is widely adopted they either support it or steadily arterial bleed market share to AMD.

I don't know what the particular feature differences are. As Mantle is not tied to CGN because it supports a minimum feature set common to modern processors, the implication is there are particular features not common to modern processors. Plus common sense dictates different architects = differences in features.

 

[Lonbjerg]

They weren't talking vendor specific. He was using a hardware example during the demo talking about a 4core/8thread processor. They are talking what they are capable of getting with DX in absolute terms.

It's really worth the time to listen to the demo.

So they were talking singletrheaded river perfomance or not?

You cannot talk absolute when AMD's DX11 drivers isn't supporting the full feature set.

 

[bystander36]

From what I understand you are wrong. We'll have to wait and see for sure.

No point in making stuff up now unless you have a source for your position. Someone stating that the Mantle driver needs to be optimized per game or future versions will break old games?

You are taking things too literal. With lower level access, they can code closer to the hardware, meaning there is less mystery, and they don't have to rely on optimizations nearly as much, but there is still an abstraction layer, they are still not programming at the hardware level. The drivers still get altered and there will be new hardware that comes to exist.

There will still be changes to drivers, which may effect performance, after a game is released. They just are more in control, and likely won't require as much assistance in the driver department.

 

[Imouto]

What features are missing from nVidia GPU's? Is this known?

Different archs. The same way AMD destroys NV with mining they could have Mantle features way slower with NV GPUs and could take several generations to be included efficiently.

 

[blastingcap]

I was pointing out your link did not reference Maxwell GPUs containing ARM CPUs. Nor does the TH article you reference, the Nvidia CTO said 'sometime this decade we arealso going to start bringing integrated CPUs and GPUs together in the Tesla line' - which may or may not include Maxwell.

Yes, possibly it will happen with Maxwell and if so probably it will be included in the GeForce line, but there is nothing definitive from the links you've provided so far.

I don't think you are reading the articles linked to. Tom's is guessing 2015. You don't have to take the most extreme, anti-NV, negative reading of every quote, you know. Real gamers will want both sides to innovate as quickly as possible. I prefer to go with Tom's optimistic reading of the quote which is ~2015 for Tesla. My own reading is that 2015 may also see ARM cores on GeForce if NV feels threatened by Mantle--as someone else in the thread said earlier, it makes sense if Mantle gives significant gains like 10-20%+.

 

[3DVagabond]

So they were talking singletrheaded river perfomance or not?

You cannot talk absolute when AMD's DX11 drivers isn't supporting the full feature set.

They were strictly talking the max amt. of draw calls that are possible while designing a game with DX. It had nothing to do with AMD or nVidia. They were talking real world limitations with DX compared to Mantle. Mantle is approaching an order of magnitude better, and that's in the early alpha stages.

 

[Lonbjerg]

They were strictly talking the max amt. of draw calls that are possible while designing a game with DX. It had nothing to do with AMD or nVidia. They were talking real world limitations with DX compared to Mantle. Mantle is approaching an order of magnitude better, and that's in the early alpha stages.

Is your claim that there is no difference in the limitations between a singlethreaded and a multithreaded DX 11 driver?
Simple yes/no.

 

[3DVagabond]

Different archs. The same way AMD destroys NV with mining they could have Mantle features way slower with NV GPUs and could take several generations to be included efficiently.

Could be, but maybe not? My point is we have a lot of nay saying going on about what can and can't be done and how things are going to end up but nobody actually knows. It's different to question if, and to state that something is going to be a problem or issue.

Someone saying it might not run as well on nVidia hardware is correct. Stating that nVidia hardware is disadvantaged though... nobody actually knows that.

 

[3DVagabond]

Is your claim that there is no difference in the limitations between a singlethreaded and a multithreaded DX 11 driver?
Simple yes/no.

I'm not claiming anything. Try reading what I'm actually writing.

 

[Imouto]

Could be, but maybe not? My point is we have a lot of nay saying going on about what can and can't be done and how things are going to end up but nobody actually knows. It's different to question if, and to state that something is going to be a problem or issue.

Someone saying it might not run as well on nVidia hardware is correct. Stating that nVidia hardware is disadvantaged though... nobody actually knows that.

It was the only explanation I could find for AMD to offer Mantle to 3rd parties. A good 2 or 3 years advantage for Nvidia to catch up. Offering Mantle without the guarantee to be the best performer by a huge margin is simply the dumbest idea I've ever read.

 

[blackened23]

Obviously you just skipped around in the video. Spend some time and watch the whole thing.

It is pretty amazing, but i do agree. For someone looking for instant satisfaction it looks pretty bleh too.

Ehh. I was looking for the "immediacy" factor. Like, when I saw crysis for the first time in 2007 it was jaw dropping. Far better than anything that proceeded it. You know? When I saw crysis 3 I was pretty darn impressed. I'm thinking to myself, if Mantle truly does have all this power at its disposal as AMD's marketing indicates, go all out. Give us something completely ridiculous in terms of pushing the limits.

So I look at this game and....I understand the draw calls and all, but I want something to translate that into better graphics. I don't see that in this game, I want that immediacy effect that something like Crysis 1 gave me in 2007. I mean draw calls are great and all, but it's hard to quantify how that actually benefits graphics. Don't get me wrong. Maybe it will be a great game. But I haven't heard of Oxide and their track record is unproven. They're basically unknown at this point. In terms graphics, on the surface though, I don't know. It doesn't strike me as super compelling, although it is still in development. I don't really care about what goes on in the background - I don't *really* care if the CPU usage is 90 or 99%. I'd really like to see Mantle doing something really compelling and groundbreaking in terms of end results. I don't feel that this is it. I understand what is happening in the background is compelling - yet i'm not so sure that translates into the more compelling visuals or gaming experience. We'll see whenever it's released.

I don't know when BF4 with Mantle will hit, but that could be the first example of a truly compelling example of Mantle. We'll see. Hopefully DICE won't delay that too much, because I am really interested to see how much BF4 benefits from Mantle.

 

[mikk]

Dice has repeatedly said DX11 multithreaded rendering isn't working for them, and Far Cry 3 even had it patched away. While it's clear that it helped performance in Civilization V, it's also clear that's it's of no use for other games.

Do you have a source? Never heard this. DX11MT implementation seems complicated.

 

[3DVagabond]

It was the only explanation I could find for AMD to offer Mantle to 3rd parties. A good 2 or 3 years advantage for Nvidia to catch up. Offering Mantle without the guarantee to be the best performer by a huge margin is simply the dumbest idea I've ever read.

AMD is known to cut their noses off to spite their faces in the name of open standards. Also you rarely see GE games exploiting AMD advantages that will make nVidia cards non competitive. Considering how much better AMD is in most openCL tasks I'm pretty sure that if they wanted to they could exploit that to make nVidia cards crawl in GE titles.

Anyway, I do agree that it's possible that it would take years for nVidia to be competitive in Mantle. It's also possible that all of the architectures nVidia has planned don't have the minimum Mantle feature set. Although, I think it's likely that's not the case.

 

[psoomah]

I don't think you are reading the articles linked to. Tom's is guessing 2015. You don't have to take the most extreme, anti-NV, negative reading of every quote, you know. Real gamers will want both sides to innovate as quickly as possible. I prefer to go with Tom's optimistic reading of the quote which is ~2015 for Tesla. My own reading is that 2015 may also see ARM cores on GeForce if NV feels threatened by Mantle--as someone else in the thread said earlier, it makes sense if Mantle gives significant gains like 10-20%+.

Strawman. I wasn't taking a position, 'extreme anti-NV negative' or otherwise, I was merely pointing out Nvidia's CTO said 'sometime this decade' and that you have yet to link to a source stating Maxwell will definitely incorporate ARM CPUs. Tom's Hardware's guess for when that will happen is just that, a guess.

 

[Skurge]

Is your claim that there is no difference in the limitations between a singlethreaded and a multithreaded DX 11 driver?
Simple yes/no.

I don't see a difference. Care to show me one?

Seems to me Devs aren't using Command lists not because AMD don't support it, but cause it does nothing and as FC3 proved it makes the game perform worse. DICE is using Mantle rather bothering with driver command lists.

The fact that there is only been 1 game with command list working and that there are already 4 games that have been confirmed to have mantle suppport proves the devs think there is more in mantle.

 

[Lonbjerg]

I'm not claiming anything. Try reading what I'm actually writing.

Are we talking AMD's limitaitons in DX11 or NVIDIA's limitaiton in DX 11?

Those two are NOT the same.

 

[mikk]

I don't see a difference. Care to show me one?

Seems to me Devs aren't using Command lists not because AMD don't support it, but cause it does nothing and as FC3 proved it makes the game perform worse. DICE is using Mantle rather bothering with driver command lists.

The fact that there is only been 1 game with command list working and that there are already 4 games that have been confirmed to have mantle suppport proves the devs think there is more in mantle.

DX11MT helps in CPU limited environments not in GPU bound cases. Your links are not helpful.

 

[3DVagabond]

Are we talking AMD's limitaitons in DX11 or NVIDIA's limitaiton in DX 11?

Those two are NOT the same.

Again, I didn't take notes, so I don't remember the exact figures. What was said is that when running a game on DX you are limited to "X" amount of draw calls per frame, "in theory" (10k-15k, IIRC). In actuality most of the time it was far less than that (3k-5k, again IIRC). They were running ~70k, IIRC in that demo. The only thing I remember being mentioned about MTR, and I think it was Johan that said it, was that implementation didn't offer any reliable/worthwhile gains that were worth pursuing. The problem with MTR is that they can not assign tasks to cores themselves. DX makes all those calls. They could put as many threads in as they wanted to but often they were still being put into ques and the CPU was being severely under utilized. Already they are better able to make use of the CPU with Mantle than DX and it's early days and they are no where near maxing out the CPU.

 

[psoomah]

AMD is known to cut their noses off to spite their faces in the name of open standards. Also you rarely see GE games exploiting AMD advantages that will make nVidia cards non competitive. Considering how much better AMD is in most openCL tasks I'm pretty sure that if they wanted to they could exploit that to make nVidia cards crawl in GE titles.

Anyway, I do agree that it's possible that it would take years for nVidia to be competitive in Mantle. It's also possible that all of the architectures nVidia has planned don't have the minimum Mantle feature set. Although, I think it's likely that's not the case.

If I recall correctly, during the Hawaii Mantle event it was revealed Mantle would provide access to GCN hardware features that could not be accessed through DX. Earlier in the thread it was revealed AMD planned an end run around DX at least 2 1/2 years ago.

Apparently AMD included features in their GCN architecture they knew wouldn't be fully utilized until Mantle was introduced some two years after the first GCN GPUs were released.

AMD has been loading the deck in Mantle's favor in various ways for years.

What Mantle compatible features Nvidia had the foresight to include in Maxwell remains to be seen.

 

[Skurge]

DX11MT helps in CPU limited environments not in GPU bound cases. Your links are not helpful.

Then explain the texture decompression tests. Also this.

Seems to me like the difference was cause of 5 and 6 series radeons poor gpgpu performance.

 

[mikk]

Then explain the texture decompression tests. Also this.

Texture decompression is meanignless for the game. What it needs are game tests from a battle and not some theoretical texture tests, I'm not even sure why and if DX11MT can boost a texture decompression test. The other two links are clearly GPU bound, otherwise faster GPUs wouldn't scale up so well. You last link the same. 720p noAA on a Titan+290x in a heavy battle scene test would be interesting and please with a sixcore CPU or more.

 

[SiliconWars]

Civ 5 late-game "multi-threading" was just a benchmarking exercise and not indicative of actual gameplay. Just ask your hero, SKYMTL.

In addition, most sites use the Civ benchmarking tool which I find IS NOT representative of the in-game performance. So we don't bench Civ either.

http://www.xtremesystems.org/forums...-590-reviews&p=4791582&viewfull=1#post4791582

I haven't been following the discussion of late but I thought AMD caught up and passed Nvidia in Civ 5 anyway?

 

[Lonbjerg]

Civ 5 late-game "multi-threading" was just a benchmarking exercise and not indicative of actual gameplay. Just ask your hero, SKYMTL.

http://www.xtremesystems.org/forums...-590-reviews&p=4791582&viewfull=1#post4791582

I haven't been following the discussion of late but I thought AMD caught up and passed Nvidia in Civ 5 anyway?

Some sites don' use the buildt in benchamrk, but use actually gameplay.

And could you provide links to AMD doing better than NVIDIA in CIV 5?

I mean...you are the guy saying SLI'ed Titans would lose to an APU in BF4 right?

 

[chimaxi83]

Clicked the Mantle thread after seeing a bunch of new replies...... if only they had something to do with Mantle, and not Nvidia multi threaded drivers that nobody cares about.

 

[SiliconWars]

Some sites don' use the buildt in benchamrk, but use actually gameplay.

What sites? Links with proof?

And could you provide links to AMD doing better than NVIDIA in CIV 5?

Yeah that's really no problem. I'll even give you quotes.

http://www.anandtech.com/show/5261/amd-radeon-hd-7970-review/24

Because of the fact that Civilization V uses driver command lists, we were originally not going to include it in this benchmark suite as a gaming benchmark. If it were solely a DCL test it would do a good job highlighting the fact that AMD doesn’t currently support the feature, but a poor job of actually explaining any hardware/architectural differences. It was only after we saw AMD’s reviewer’s guide that we decided to go ahead and include it, because quite frankly we didn’t believe the numbers AMD had published.
With the GTX 580 and the 6970, the 6970 routinely lost to the GTX 580 by large margins. We had long assumed this was solely due to NVIDIA’s inclusion of DCLs, as we’ve seen a moderate single-GPU performance deficit and equally moderate multi-GPU lead for AMD melt away when NVIDIA added DCL support. The 7970 required that we rethink this.
If Civilization V was solely a DCL test, then our 2560 results would be impossible – the 7970 is winning by 12% in a game NVIDIA previous won by a massive margin.

I mean...you are the guy saying SLI'ed Titans would lose to an APU in BF4 right?

I'm pretty sure I never said that or would ever make a statement anything like as ridiculous as that. By all means prove me wrong with some links/proof, otherwise you're full of it.