The AMD Mantle Thread

[ShintaiDK]

If you look at the Star Swarm result, the 770 and 780 Ti is faster using DirectX (suggesting that nv's driver handles the issue better) but Mantle more than makes up for it.

Remember to enable DC in swarm for DX in the ini file. Gives almost twice the performance. The swarm demo is simply rigged.

 

[sontin]

AMD is a multi-billion dollar company, any single game can be ported to Mantle for a few hundred thousand dollars. Let's be generous and say a million bucks a game to implement Mantle in the top 20 games this year then the top 50 games next year.

$70 million bucks in 2 years for what is basically guaranteeing them the top position in gaming? $70 million isn't even a single GPU design yet their entire stack would be worth at least 10% performance and probably hitting nearer 50% in some games just from Mantle optimizations.

They *have* to do this. If they don't push Mantle with everything they've got then they've just basically given up and are beyond hope.

If they needed to push Mantle why would any sane publisher take only peanuts?
If i were EA i would bleed them out. Especially when you see that EA need every money they can get.

 

[f1sherman]

The cost to benefit ratio is so massively in AMD's favour that they basically will be doing everything you have suggested here to ensure Mantle becomes what you want it to be.

AMD is a multi-billion dollar company, any single game can be ported to Mantle for a few hundred thousand dollars. Let's say a million a game to implement Mantle in the top 20 games this year then the top 50 games next year.

$70 million bucks in 2 years for what is basically guaranteeing them the top position in gaming? That's not even a single GPU design yet their entire stack would be worth at least 10% performance and probably hitting nearer 50% in some games just from Mantle optimizations.

They *have* to do this. If they don't push Mantle with everything they've got then they've just basically given up and are beyond hope.

Good games get published at snail pace. So Mantle revisiting already published games would be HUUGE.
If they do that, I'd become an AMD fanboi LOL.

Seriously though, frametime/smoothness wise there is NOTHING more important than having your GPU properly fed.
(well ok that + frames being pushed on display aka G-Sync/FreeSync)

Well played AMD, GJ DICE. But now we want moar. More Mantle ofc :biggrin:

 

[krumme]

After some hours playing mp i feel its a far better smooth experience nearly all the time vs dx. Thinks it gives better K/D ratio but definitively better gameplay feel.

Sometimes every 5 minute or so though there is a bug, and there is a very small pause giving a hickup in the game. Thats irritating but i prefer to game with Mantle because the smoothness all the time. Could have been the server i just jumped on. But the smoothness is basicly always there and its very nice. I guess it also highly depens how sensistive you are to smoothness and low min fps and how your setup are.

Played on varios maps with 60-64 man. i5 3570k at 4.1, and 7970 1ghz. I like to game consistently over 60fps.

Saying Mantle is not for BF4 is nonsense. BF4 is a MP game.

 

[3DVagabond]

SP is basically completely GPU bound. I dunno how AMD manages it so often but they couldn't have picked a worse game to showcase Mantle with. Obviously they didn't know this would be the case, but still.

The scaling is near perfectly in line with what we know about each card. 290X ~30% faster than the 280X which is ~20% faster than the 270X etc. Nvidia appears to have squeezed a few more percent out of their DX drivers and that's why the Ti has a 13% lead instead of say 6% or 7%, and why the 770 is 7% ahead of the 280X (in DX).

At the lower end 260X, 270X the improvements are basically non-existent, which imo is pretty solid evidence that there hasn't been a lot done on GPU optimization.

This is the only review I've seen with a basic gaming system

source

Benchmark Results: In Battlefield 4, using DICE’s new benchmarking tool we received a 15.8% performance boost once Mantle was enabled at a screen resolution of 1920×1080. This is not a bad performance uplift and we’d consider a system with an Intel Core i5-3570 processor along with a Radeon R7 260X a pretty good real world platform for gamers.

15% for a i5-3570/260X isn't too shabby.

 

[SiliconWars]

If they needed to push Mantle why would any sane publisher take only peanuts?
If i were EA i would bleed them out. Especially when you see that EA need every money they can get.

If AMD can basically get complete rights to BF4 - so much so that they can get a whole new API patched into the game at the same time when it's been riddled with bugs - for $8 million, including all those keys etc...then they can certainly get the next top 19 games this year for an average of $1 million a piece.

Current Gaming Evolved and DX driver optimization will be costing them a lot more. GE becomes Mantle, seems simple enough.

They are already in the Frostbite games anyway. It's going to take time though, it's not like games are insta-developed.

 

[SiliconWars]

Good games get published at snail pace. So Mantle revisiting already published games would be HUUGE.
If they do that, I'd become an AMD fanboi LOL.

I agree, this would be absolutely massive and while I'd love to see them go back and do stuff like WoW, GW2 etc - I just don't see it happening. I think Nvidia probably would, which is quite frustrating.

 

[sontin]

So and nVidia would not pay more, yes? Sure.

 

[Erenhardt]

<shameless>
If someone is interested in seeing mantle improvements in BF4 with a CPU bottlenecked system (Phenom II 965 + 1.2GHz HD7870) send me a PM with BF4 key
</shameless>

Star swarm sees a huge shift from heavy CPU bound to mild GPU bottleneck.

 

[SiliconWars]

So and nVidia would not pay more, yes? Sure.

Pay more for what?

Anyway, Nvidia obviously thought $8 million was too much for BF4 yet AMD was prepared to go there. So yeah, I have my doubts that Nvidia will pay more.

 

[Silverforce11]

You guys are so late to the party, lots of sites test 0/4 servers, where they run around by themselves.. it really is indicative of what players would do, buying a full game to run around jerking by themselves.. not.

The only ones that so far test actual REAL multiplayer is [H], GameGPU.RU and others do simulated MP with empty servers. Weaksauce. A few even post a videos of their "run".

 

[SiliconWars]

Don't forget the real beauty of Mantle - it can just completely bypass any of Nvidia's optimizations that decrease AMD's performance.

Let's say Nvidia pays $5 million for a game - AMD can pay another $1-$2 million for Mantle support, completely bypassing GameWorks or whatever. Nvidia can't stop that, they can't do a thing about it without breaking the law as that would be exclusionary.

 

[coercitiv]

Don't forget the real beauty of Mantle

Just wait till people realize a mere piece of software can lower thermals and power consumption in mobile gaming.

 

[Dresdenboy]

I've skipped the last 10 pages, where the same 2 or 3 questions and 2 or 3 opinions where repeated using different wordings. It's always the same: more money usually buys more performance, and a bottlenecking GPU won't get faster with Mantle as I said before.

For a followup to a review the testers might compare same budget machines with different weight on CPU or GPU performance. And, as coercitiv mentioned, do some power consumption/battery life tests.

To give you an impression, what parallelism in graphics rendering and game code is all about, I have here a slide which I found in the backup of an older SIGGRAPH talk. It shows different tasks running on two GPUs (AFR) and four CPU cores:

Now imagine some tasks depending on the results of others (eg. the rosé one on CPU3). If there is enough CPU headroom (think of early BF4 CPU results), the whole thing works. But if that balance is lost (e.g. weaker cores, load spikes due to MP gaming in certain scenarios), the whole thing quickly hits a wall. And DX11 handling (either DC or IC) adds a burden, dictated by the way MS and likely also the GPU vendors implement it. But as I said, MS surely didn't only provide header files. So there is framework code too, which is the same for all GPUs.

 

[Lonyo]

This is the only review I've seen with a basic gaming system


15% for a i5-3570/260X isn't too shabby.

I would disagree that it makes sense as a basic gaming system.
You have a $200 CPU with a $120 GPU for a gaming system?

A 270X with a slower processor would make more sense for a gaming system, since that's a $200 GPU, which is where you want to allocate money for a gaming system really...
A 260X should go with a cheaper CPU, like a Core i3 or something from AMD.

 

[desprado]

Don't forget the real beauty of Mantle - it can just completely bypass any of Nvidia's optimizations that decrease AMD's performance.

Let's say Nvidia pays $5 million for a game - AMD can pay another $1-$2 million for Mantle support, completely bypassing GameWorks or whatever. Nvidia can't stop that, they can't do a thing about it without breaking the law as that would be exclusionary.

If Developers are under Nvidia contract than neither developers or AMD cannot do any thing about it and if they do anything than they can both be sued big time because of contract ,terms and condition.

 

[GaiaHunter]

I would disagree that it makes sense as a basic gaming system.
You have a $200 CPU with a $120 GPU for a gaming system?

A 270X with a slower processor would make more sense for a gaming system, since that's a $200 GPU, which is where you want to allocate money for a gaming system really...
A 260X should go with a cheaper CPU, like a Core i3 or something from AMD.

My gaming rigs are generally GPU 1.5-2x more expensive than the CPU.

 

[parvadomus]

:whiste:

 

[cmdrdredd]

It's going to be used in all the upcoming EA Frostbite games and others.

http://www.amd.com/us/products/technologies/mantle/pages/mantle.aspx#2

Of course AMD would like more developers to join, and we're just going to have to wait and see what happens down the road.

That doesn't say "all upcoming FB3 games will use it" at all

 

[mikepembo]

Lag spikes with mantle. Any ideas anyone? Possibly slow hard drive? (lag spike happen when moving around the map especially in a fast vehicle - so possibly reading hdd for level-of-detail)
http://postimg.org/image/7is5qggff/

Phenom II X4 840 (OC to 3.5GHz)
HD 7850
6GB ddr3 ram

 

[cmdrdredd]

Nono... thats purely my conjecture.

But judging how reviews are done (NO NV CARD IN SIGHT) - not entirely baseless. What do you think?

I don't think it's that. I think that this was pushed out last minute and nobody really had time to prepare for their testing. It's not like everyone got their cards and a new driver a week ago and had all that time to get stuff together for their review NDA to lift. It's more like a quick and dirty test to show what people who already have a card that can take advantage of this can expect. It is only one game so far and when you have cards that retailers are pricing just too high because of the mining craze, it's not that sweet of a deal unless you already have a card anyhow. Looking for a new card now and want to go ultra high end? Unless BF4 is the only game that matters or Thief is going to be a life changing moment for you, it isn't all that special at the moment. Now the main problem I see is what has been said a bit now. You need a slower CPU matched with a fast GPU to get the maximum benefit or move up to crossfire. We will have to see if this can be "fixed". I say fixed because I don't consider this a realistic scenario...slow CPU with $700 GPU.

 

[biostud]

Lag spikes with mantle. Any ideas anyone? Possibly slow hard drive? (lag spike happen when moving around the map especially in a fast vehicle - so possibly reading hdd for level-of-detail)
http://postimg.org/image/7is5qggff/

Phenom II X4 840 (OC to 3.5GHz)
HD 7850
6GB ddr3 ram

From release notes:

Mantle performance for the AMD Radeon™ HD 7000/HD 8000 Series GPUs and AMD Radeon™ R9 280X and R9 270X GPUs will be optimized for BattleField 4™ in future AMD Catalyst™ releases. These products will see limited gains in BattleField 4™ and AMD is currently investigating optimizations for them.

 

[USER8000]

slow CPU with $700 GPU.

It is over time as it means systems might be able to survive another GPU upgrade, before the CPU needs to be replaced in Mantle enabled games.

 

[blackened23]

http://battlelog.battlefield.com/bf4/forum/threadview/2979150493821981180/1/

A lot of people on the battlelog forums stating that the Mantle renderer looks different with reduced draw distances or way more fog than DX11. Maybe the beta nature of the driver.

It's going to be used in all the upcoming EA Frostbite games and others.

http://www.amd.com/us/products/techn.../mantle.aspx#2

Of course AMD would like more developers to join, and we're just going to have to wait and see what happens down the road.

It is in fact not being used in all FB3 games. NFS Rivals specifically is not getting it despite using FB3. To my knowledge, Dragon Age: I is being released this year and the developers have not confirmed Mantle. PvZ Zombie Warfare is also unconfirmed for Mantle by the developers, although the bigger question is why would that game need it in the first place? Not a demanding game, more oriented to the super casual crowd. We'll see, I guess, if the developers confirm Mantle or not.

 

[cmdrdredd]

It is over time as it means systems might be able to survive another GPU upgrade, before the CPU needs to be replaced in Mantle enabled games.

If all you play is the couple games using it, sure.