New standard when testing Mantle games?

C

CombatChrisNC

[H]ard|Gawd
Joined
Apr 3, 2013
Messages
1,140
Will results show enough of a difference, and will testers change their test-bed (or at least make a secondary) when Kaveri and Mantle are here?

What's going to be the fastest Mantle rig?
Kaveri + 290x
Piledriver + 290x
Intel + 290x

With Kaveri bringing another 512 cng SP's to the table I've got to think that's going to make a difference.

Will we ever find out from [H]? I haven't seen them do any recent APU's.
 
Let me just put on my wizard hat and get my crystal ball...

We don't know yet. And I doubt we will know until BF4 gets the Mantle patch, and then more Mantle-enabled games come out. So we'll just have to wait and see, really.
 
Kaveri does 856 Gigaflops :) . Does not mean that Mantle is using it all by default it depends on the engine which does the scheduling.
It also depends on what the engine is made for , not everything will be just really fast some of the graphic engines might actually also do more graphics.
The idea is that you can run more batches on the gpu because the cpu is now doing "nothing" (Mantle). You can not say that every engine will go for speed or just do more graphics so it is hard to put a number on Mantle setups.

If it is just passing data faster and not do much more (graphics) then the DX11.2 engine does it could very well be Kaveri. Does it do way more batches then the normal DX engine then more cores will pay of.
 
Last edited:
NaroonGTX said:
Let me just put on my wizard hat and get my crystal ball...

We don't know yet. And I doubt we will know until BF4 gets the Mantle patch, and then more Mantle-enabled games come out. So we'll just have to wait and see, really.
Click to expand...

Exactly. BF4 is the only known game to be using Mantle anytime soon. And with the current state of BF4 and the developers scrambling to put out patch after patch to cure stability issues, Mantle should be of lowest priority to the developers right now since most of its problems are not graphics related. Hell, DICE introduced game breaking SLI and Xfire bugs with the last patch, so I have zero faith that they are capable of launching Mantle on BF4 anytime soon. Mantle won't fix the problems with BF4, only exacerbate them due to DICE/EA issues, not so much AMD issues.
 
Last edited:
CaptNumbNutz said:
Exactly. BF4 is the only known game to be using Mantle anytime soon. And with the current state of BF4 and the developers scrambling to put out patch after patch to cure stability issues, Mantle should be of lowest priority to the developers right now since most of its problems are not graphics related. Hell, DICE introduced game breaking SLI and Xfire bugs with the last patch, so I have zero faith that they are capable of launching Mantle on BF4 anytime soon. Mantle won't fix the problems with BF4, only exacerbate them due to DICE/EA issues, not so much AMD issues.
Click to expand...

Mantle is more then likely done by people not working on BF4, it is the team that works on the Frostbite 3 engine . That should be the team doing most if not all of the work.
 
Yeah, but it sucks that BF4 is such a terrible mess right now. It's not even just PC, the game is equally-fucked on all platforms it got released on. They clearly rushed this game out to compete again CoD: Ghosts.

As JC Denton would say, what a shame.
 
Pieter3dnow said:
Mantle is more then likely done by people not working on BF4, it is the team that works on the Frostbite 3 engine . That should be the team doing most if not all of the work.
Click to expand...

What makes you think that some of the problems with the game aren't related to the game engine itself? DICE fucking broke SLI and XFire with the last patch, that right there is just as related to graphics and game engine as adding Mantle would be.

There is zero need to add Mantle now, it would only add complexity. The whole key to troubleshooting and fixing problems is to eliminate possible variables. Once the game is finally stable, then it will be worth considering. Why add more gears to the drive train when they can't even get the ones they have now meshing right?
 
where the hell is Mantle?...they only have 10 days left to meet their 'December' release date...I feel sorry for all those people who rushed out to buy the new 200 series cards for BF4...not only is the game riddled with issues but the magic API has also gone AWOL
 
polonyc2 said:
where the hell is Mantle?...they only have 10 days left to meet their 'December' release date...I feel sorry for all those people who rushed out to buy the new 200 series cards for BF4...not only is the game riddled with issues but the magic API has also gone AWOL
Click to expand...

Have to fix the base game before you start adding in features that are 100% going to cause some form of issues also.
 
CaptNumbNutz said:
What makes you think that some of the problems with the game aren't related to the game engine itself? DICE fucking broke SLI and XFire with the last patch, that right there is just as related to graphics and game engine as adding Mantle would be.

There is zero need to add Mantle now, it would only add complexity. The whole key to troubleshooting and fixing problems is to eliminate possible variables. Once the game is finally stable, then it will be worth considering. Why add more gears to the drive train when they can't even get the ones they have now meshing right?
Click to expand...
Netcode has nothing to do with the graphics engine it is pure tcp/ip.
In reality they are not adding anything to BF4. It is the engine which is getting an overhaul to be able to do Mantle.
As far as i know they fixed SLI/crossfire in the latest patch.

polonyc2 said:
where the hell is Mantle?...they only have 10 days left to meet their 'December' release date...I feel sorry for all those people who rushed out to buy the new 200 series cards for BF4...not only is the game riddled with issues but the magic API has also gone AWOL
Click to expand...

If they were late producing the Mantle version as opposed to rush it and have serious problems with it i don't mind the delay.
 
IMO they didn't fix SLI/Xfire in BF4, they simply hard disabled it. As it stands I don't see any difference in performance on my rig with or without SLI enabled. It's like the game is only using one card. In fact I've seen a couple instances where having SLI enabled produced lower frame rates. So, I just leave it disabled for now until they truly fix it.
 
Pieter3dnow said:
Netcode has nothing to do with the graphics engine it is pure tcp/ip.
In reality they are not adding anything to BF4. It is the engine which is getting an overhaul to be able to do Mantle.
As far as i know they fixed SLI/crossfire in the latest patch.



If they were late producing the Mantle version as opposed to rush it and have serious problems with it i don't mind the delay.
Click to expand...
Where did I mention netcode? That's right...nowhere. Why did you bring it up and why are you sidetracking the conversation? Netcode is only one of many problems BF4 has. In fact, it may be one of the lower priority items to fix. Making the game stable to play without crashing constantly is priority one. The game is mostly stable now (moreso with a few workarounds), but there are still some game crashing bugs.

I will say this again. DICE broke SLI and XFire with the patch last week when it had been working up to this point. That IS game engine and graphics related. They broke SLI and XFire with a patch that was to have ZERO to do with fixing graphical glitches (at least there were none listed in the patch notes). Their fix? Issue another patch that for all intents and purposes is the same as disabling multi-gpu rendering while they go back and try and figure it out. Many people are reporting that SLI and Xfire performance post-patch is actually worse than before and for the most part no better than using a single card.

So if DICE can somehow unintentionally break SLI and XFire capability by introducing graphical anomalies such as flashing textures AND still have an unstable game, what makes you so confident DICE is making Mantle a priority? I for one hope they re-evaluate their quality control and testing procedures for patching and get the game much more stable than it is now before delving into Mantle. And by "stable" I mean the few game crashing bugs left get squashed, no new bugs introduced, and no silly workarounds are required to make the game stable 3/4 of the time like it is now (e.g disabling core parking to avoid lag and CPU usage spikes or turning off surround sound to avoid the sound loop bug). If that means they delay Mantle's introduction so they can make the game more stable and make sure they don't fuck something else up by introducing Mantle, then they should do it.
 
Last edited:
I'm so fed up with the patches and now I have problems with Sli, on top of that when I'm flying the ground is littered with flashing textures. Now I'm noticing the foliage isn't moving like it used to when I'm hovering in the chopper. Wtf is wrong with these guys. I wish I could have my money back, I'm going back to BF3.
 
CaptNumbNutz said:
Where did I mention netcode? That's right...nowhere. Why did you bring it up and why are you sidetracking the conversation? Netcode is only one of many problems BF4 has. In fact, it may be one of the lower priority items to fix. Making the game stable to play without crashing constantly is priority one. The game is mostly stable now (moreso with a few workarounds), but there are still some game crashing bugs.

I will say this again. DICE broke SLI and XFire with the patch last week when it had been working up to this point. That IS game engine and graphics related. They broke SLI and XFire with a patch that was to have ZERO to do with fixing graphical glitches (at least there were none listed in the patch notes). Their fix? Issue another patch that for all intents and purposes is the same as disabling multi-gpu rendering while they go back and try and figure it out. Many people are reporting that SLI and Xfire performance post-patch is actually worse than before and for the most part no better than using a single card.

So if DICE can somehow unintentionally break SLI and XFire capability by introducing graphical anomalies such as flashing textures AND still have an unstable game, what makes you so confident DICE is making Mantle a priority? I for one hope they re-evaluate their quality control and testing procedures for patching and get the game much more stable than it is now before delving into Mantle. And by "stable" I mean the few game crashing bugs left get squashed, no new bugs introduced, and no silly workarounds are required to make the game stable 3/4 of the time like it is now (e.g disabling core parking to avoid lag and CPU usage spikes or turning off surround sound to avoid the sound loop bug). If that means they delay Mantle's introduction so they can make the game more stable and make sure they don't fuck something else up by introducing Mantle, then they should do it.
Click to expand...

http://battlelog.battlefield.com/bf4/forum/threadview/2955065221669468231/

You are talking about DICE if it is a single entity. DICE has several developing teams beside the one BF4 there is a separate one for the engine Frostbite.

There are some overlapping things but they are still doing different stuff. the BF4 live team does not work on the engine, yes they may track and fix bugs but in itself they work on BF4.

Mantle requires the graphics engine Frostbite to be ported that is what is going on.

Btw I play BF4 and no constant crashing for me and that is why I was talking about the netcode.( I do crash once in a while usually task manager saves me)
 
Mantle is the name of this thread, not BF4! Personally, I am looking forward too what mantle will do and do not expect everything to happen overnight. For those with NVIDIA cards who have issues with Mantle coming out, go cry up a river somewhere else. New tech is cool and drives competition and in addition to that, I am really enjoying my XFX R9 290 and consider Mantle to be a nice extra for when it comes out.
 
CombatChrisNC said:
Will results show enough of a difference, and will testers change their test-bed (or at least make a secondary) when Kaveri and Mantle are here?

What's going to be the fastest Mantle rig?
Kaveri + 290x
Piledriver + 290x
Intel + 290x

With Kaveri bringing another 512 cng SP's to the table I've got to think that's going to make a difference.

Will we ever find out from [H]? I haven't seen them do any recent APU's.
Click to expand...

One of the guys talking during the oxide demo seemed to imply that mantle scales better with physical cores vs virtual cores. So if a game was constrained by core counts, I would expect the following performance hierarchy:

Piledriver + 290x
Intel + 290x
Kaveri + 290x


If the game was more gpu intensive, where the bottleneck was less on the number of cores and more on the raw gpu power, it might look more like this:

Kaveri + 290x
Piledriver + 290x
Intel + 290x

Since Kaveri has gcn cores, it can use the asynchronous gpu scaling in mantle to lessen the gpu bottlenecks.

So basically, games with tons of units, piledriver probably wins
games with fewer draw calls and more raw gpu power needed, kaveri probably wins...


interestingly... under what situation does the intel i7 win? not the hex core parts, the standard i7-4xxx series?
 
CombatChrisNC said:
Will results show enough of a difference, and will testers change their test-bed (or at least make a secondary) when Kaveri and Mantle are here?

What's going to be the fastest Mantle rig?
Kaveri + 290x
Piledriver + 290x
Intel + 290x

With Kaveri bringing another 512 cng SP's to the table I've got to think that's going to make a difference.

Will we ever find out from [H]? I haven't seen them do any recent APU's.
Click to expand...

The systems will all perform the same. He says in the video that a 8350 down clocked to 2ghz will not bottleneck a 290. So if you're gpu limited, the CPU won't matter.

And using the gcn cores of Kaveri, I don't see that being likely. I'm sure 1 Kaveri gcn core is not equal to 1 290 gcn core. So coding to used two different cores (at different speeds) at the same time doesn't seem likely to me. I could be wrong, but I haven't seen anywhere that says you can combine 290 cores with apu cores.
 
I doubt much will change in cpu gaming power relation between AMD and Intel. Maybe in few "amd" games and that is it. There will still be tons of games that will love Intel more and games that will not even support Mantle. Game developers will have to make new patches for their games for Mantle and I dont know how will that work and who will do it? When Mantle will work in good form we will have all new gen of CPUs and GPUs and new games. I think so.
 
trick0502 said:
The systems will all perform the same. He says in the video that a 8350 down clocked to 2ghz will not bottleneck a 290. So if you're gpu limited, the CPU won't matter.

And using the gcn cores of Kaveri, I don't see that being likely. I'm sure 1 Kaveri gcn core is not equal to 1 290 gcn core. So coding to used two different cores (at different speeds) at the same time doesn't seem likely to me. I could be wrong, but I haven't seen anywhere that says you can combine 290 cores with apu cores.
Click to expand...

http://www.youtube.com/watch?v=tDPgJB2x7dQ#t=33m20s


The guy here specifically talks about asymmetric gpu scaling being much better with mantle where the developer can adjust the workloads for the apu graphics and the discreet graphics for better effects.
 
tybert7 said:
http://www.youtube.com/watch?v=tDPgJB2x7dQ#t=33m20s


The guy here specifically talks about asymmetric gpu scaling being much better with mantle where the developer can adjust the workloads for the apu graphics and the discreet graphics for better effects.
Click to expand...

That requires a TON of development for a system configuration that would represent less than 1% of the total market. This is a 'in theory' quote, saying it's possible. I highly doubt ANY developer will persue such a design.
 
trick0502 said:
Can you tell at what point they say this? That video is 45 minutes long.
Click to expand...

If they're already making the game for mantle, how much more intensive would this option end up being? If cng apu resent, offload these tasks to it based on cng count. If not, continue as usual.

But then again, I have no idea. I've seen some of the people in the mantle discussions who really sound like they know what they're talking about. I wish they would drop in on this thread.
 
polonyc2 said:
where the hell is Mantle?...they only have 10 days left to meet their 'December' release date...I feel sorry for all those people who rushed out to buy the new 200 series cards for BF4...not only is the game riddled with issues but the magic API has also gone AWOL
Click to expand...

Why? they bought a badass card with or without mantle.
 
trick0502 said:
Can you tell at what point they say this? That video is 45 minutes long.
Click to expand...

33m:20s

It's linked in the link, you might have one of those forum mods installed that show the actual video within the forum, that loses the tracking, better to turn that stuff off so you can click the link and go right to the spot.
 
Both AMD and the DICE said(between the lines) that the game/application has to be multi-GPU aware to use multiple GPU's under Mantle. So what does "multiGPU aware" means in real life? The game/app will decide during runtime which task will get executed where. So, if in the future you will have AMD&Nvidia dGPU with Intel iGPU in your computer, and all of them are Mantle enabled, a game/application CAN use ALL of them to generate the output.
 
In a sense it doesn't see GPU0 GPU1 it sees the available compute units and it can schedule tasks between any of them.
 
Zolishoru said:
Both AMD and the DICE said(between the lines) that the game/application has to be multi-GPU aware to use multiple GPU's under Mantle.
Click to expand...

So another key question is how much work or effort does it take to enable the multi-GPU aware feature? As simple as clicking an option? If so, why wouldn't everyone us it? If not, how much effort and how many devs will select it?
 
CombatChrisNC said:
So another key question is how much work or effort does it take to enable the multi-GPU aware feature? As simple as clicking an option? If so, why wouldn't everyone us it? If not, how much effort and how many devs will select it?
Click to expand...

I would estimate the effort needed to make a game/application multiGPU aware similar(but less) to the current DX/OGL efforts, so no, it isn't a simple checkbox feature during the engine development.
 
You must log in or register to reply here.
Back
Top