AMD Mantle Performance Preview in Battlefield 4 @ [H]

[FrgMstr]

I am done with the thread trolling. I have tried to be nice, but sometimes people will not listen.

PLEASE USE THE REPORT POST BUTTON AND REPORT THE TROLLS.

These posts will be reviewed and those dubbed to be trolling will lose access to this forum. This has already begun.

 

[Zarathustra[H]]

Hmm, so, it looks like Mantle is mostly a tool to make up for the lack of CPU power in AMD's chips, but will have little to no benefit at the resolutions and quality settings most of us play.

Fair assessment?

It should be a great tool for low end APU only systems

 

[Rizen]

By the way my info is an amalgamation of what the editors of the websites that got Xfire working did plus the steps of other users who reported Xfire working for them. This is what Joel from Extremetech shared with me:

Then I contacted others to nail down the cause of that variation, hence the steps about enabling Mantle and making changes to the graphics settings only while in the Test range level.

Unlike others, who are happy to spend their time spewing blanket criticisms and negativity because they have an axe to grind and fits their agenda, I'm more interested in going the extra mile to HELP the community.

Thanks. I'll give this a go this weekend. I'm reinstalling my system and upgrading to Z87/Haswell so perfect time to install the 14.1 set fr scratch.

 

[Rizen]

Hmm, so, it looks like Mantle is mostly a tool to make up for the lack of CPU power in AMD's chips, but will have little to no benefit at the resolutions and quality settings most of us play.

Fair assessment?

It should be a great tool for low end APU only systems

i don't think that's entirely accurate. Mantle helps by eliminating CPU overhead amongst other things. If you have a powerful CPU and GPU the gains are minimal. If you are CPU bottlenecked, either because you have a slow CPU or because you have multiple GPUs, it will improve performance. It really depends on your particular system.

I guess you are right that most people only have a single GPU, but I wouldn't categorize Mantle as only helping low end setups since it helps anywhere the CPU would otherwise be a bottleneck. I think games developed with Mantle in mind would be able to do a lot more with the extra CPU cycles that are freed up. That's where I am hoping this goes.

 

[maboblivion]

Hmm, so, it looks like Mantle is mostly a tool to make up for the lack of CPU power in AMD's chips, but will have little to no benefit at the resolutions and quality settings most of us play.

Fair assessment?

It should be a great tool for low end APU only systems

wrong, cpu bottlenecked area show the highest increase for sure but theres an increase everywhere. Even on highest resolution or strongest intel cpu. You haven't seen crossfire yet, when it works increase is noticeable.

Minimum frames are much higher and frame times are much more stable and lower.
You might have the same settings but it'll feel smoother. Kinda hard to tell until they iron out the crossfire hitching.

mid end cpus can get just as bottlenecked. lower intel i3 i5 , core2quads or even just running your cpu at stock instead of trying for 5 ghz. The same argument keeps cropping up intel cpu isn t overclocked from 3.5ghz to 4.9ghz , so mantle is useless.
You won't get the same amount % of frames boost from a 1ghz oc compared to the % mantle gives at stock frequency.

I m sure [h] will have a indebt evaluation once its done.

 

[chameleoneel]

Mantle currently only focuses on the CPU side. There has been little work yet, in improving the GPU side. That will come later and nobody knows what sort of improvements that might bring.

 

[magoo]

I have not installed Cat 14 due to the Crossfire issues with the driver/Mantle.

What I'm understanding is this software seems to be pointed directly at the consoles?
I mean they have a lesser CPU function, correct?

A top end computer wont benefit (nearly as much) at least currently, it seems.
I hope that changes over the long haul.

I enjoyed the detailed article however.

 

[Rizen]

I have not installed Cat 14 due to the Crossfire issues with the driver/Mantle.

What I'm understanding is this software seems to be pointed directly at the consoles?
I mean they have a lesser CPU function, correct?

A top end computer wont benefit (nearly as much) at least currently, it seems.
I hope that changes over the long haul.

I enjoyed the detailed article however.

No. This is aimed squarely at PCs. The consoles already have their own custom APIs that are closer to the hardware than DirectX is. They don't use Mantle.

 

[wonderfield]

Expecting announcement from Microsoft about a legacy stripped, tweaked up DirectX API in 5...4...3...

Considering Mantle is Windows-only at this time, they have no compelling reason to compete.

 

[octoberasian]

Hmm, so, it looks like Mantle is mostly a tool to make up for the lack of CPU power in AMD's chips, but will have little to no benefit at the resolutions and quality settings most of us play.

Fair assessment?

It should be a great tool for low end APU only systems

Fair assessment.

Just as I said in another thread: This is going to help a lot in areas where there are bound to be CPU limitations.

This is going to help a lot of people who have not only AMD APU and FX processors, but those running older processors with higher end video cards, where the game becomes CPU-bound due to a weaker processor or older processor.

It will also probably change the way gaming computers are built. There doesn't seem to be a necessity for a higher end, $300 or $500-plus processor to get the best frame rates under Mantle. You could go with a sub-$200 processor and pair it with a higher end video card, and still get very good frame rates. For games that necessitate extra CPU power and are also not Mantle-supported, then yes, a mid-range processor or better will be a better choice than a budget or mainstream one.

It's interesting how this dynamic works now. With Mantle able to shift out a lot of the CPU bottlenecks, developers using Mantle can use the extra CPU cycles for other things in the game such as physics and AI. That right there will probably benefit more higher end midrange and enthusiast processors that have oodles of extra CPU performance freed up to do something else besides processing DirectX calls.

 

[chameleoneel]

It's not just about low end hardware.

It's about when the GPU isn't limited. GPUs are so fast, that when they aren't bogged down by graphics settings/resolution, CPUs have a hard time feeding them fast enough, due to the overhead from how DirectX handles things and usually relying on singlethread performance for feeding the GPU. Even High end i7 CPUs have trouble, when the GPU isn't limited.

People that are lowering graphics settings and/or resoltion to maintain 120fps for their 120hz monitors, just got a boost.

 

[verylostindeed]

By the way my info is an amalgamation of what the editors of the websites that got Xfire working did plus the steps of other users who reported Xfire working for them.

Why the resolution scaling? Is that part required?

 

[Relayer]

Hmm, so, it looks like Mantle is mostly a tool to make up for the lack of CPU power in AMD's chips, but will have little to no benefit at the resolutions and quality settings most of us play.

Fair assessment?

It should be a great tool for low end APU only systems

No. Look at the benches provided in this thread using Intel CPU's.

 

[Druneau]

It's not just about low end hardware.

It's about when the GPU isn't limited. GPUs are so fast, that when they aren't bogged down by graphics settings/resolution, CPUs have a hard time feeding them fast enough, due to the overhead from how DirectX handles things and usually relying on singlethread performance for feeding the GPU. Even High end i7 CPUs have trouble, when the GPU isn't limited.

People that are lowering graphics settings and/or resoltion to maintain 120fps for their 120hz monitors, just got a boost.

And a very big boost it is!

 

[rinaldo00]

theres a lot of FUD in here, to which user(s) do you refer?

I do not want to get banned over something so trivial - I will PM you.

 

[rinaldo00]

I have not installed Cat 14 due to the Crossfire issues with the driver/Mantle.

What I'm understanding is this software seems to be pointed directly at the consoles?
I mean they have a lesser CPU function, correct?

A top end computer wont benefit (nearly as much) at least currently, it seems.
I hope that changes over the long haul.

I enjoyed the detailed article however.

This will give a big boost to midrange laptops/tablets - a huge market.

 

[Trimlock]

Kyle/Brent - this may be asking for too much, but if you guys have an old high end system (C2Q or i7 920) could you test with a 290? I think this would appeal to a decent amount of users that don't really upgrade because of life but may reconsider upgrading the gpu.

 

[funkydmunky]

This will give a big boost to midrange laptops/tablets - a huge market.

Well the GPU's in mid-range laptops usually fall far bellow what the CPU is capable of, so I'm not sure if much of a boost is to be expected. Possibly high-end laptops with serious mobile AMD GPU's will put in quite a good show as the GPU can stretch its legs and not be held back by the CPU.
As far as tablets? I'm not sure what you expect Mantle to do there. Completely anemic gpu/cpu. They aren't designed for Mantle unless AMD unleashes some serious APU's down the road. Even then.

 

[Rizen]

Well the GPU's in mid-range laptops usually fall far bellow what the CPU is capable of, so I'm not sure if much of a boost is to be expected. Possibly high-end laptops with serious mobile AMD GPU's will put in quite a good show as the GPU can stretch its legs and not be held back by the CPU.
As far as tablets? I'm not sure what you expect Mantle to do there. Completely anemic gpu/cpu. They aren't designed for Mantle unless AMD unleashes some serious APU's down the road. Even then.

I believe the lowest end Kaveri part right now is 45w. Which is still too much for a tablet. Quite a bit too much. So there aren't any GCN tablets in the foreseeable future.

 

[octoberasian]

I believe the lowest end Kaveri part right now is 45w. Which is still too much for a tablet. Quite a bit too much. So there aren't any GCN tablets in the foreseeable future.

They could pull an Nvidia like they did with the Tegra K1-- take one GCN 1.1 Shader Engine and pair it with something like a single Jaguar or Kaveri module. The Tegra K1 is 1 SMX unit-- 192 CUDA cores if I recall, and a dual or quad core ARM SoC.

It may actually work for a tablet x86 SoC.

 

[Rizen]

They could pull an Nvidia like they did with the Tegra K1-- take one GCN 1.1 Shader Engine and pair it with something like a single Jaguar or Kaveri module. The Tegra K1 is 1 SMX unit-- 192 CUDA cores if I recall, and a dual or quad core ARM SoC.

It may actually work for a tablet x86 SoC.

They could certainly do this, but I don't know how feasible such a product would be. The only x86 "tablet" I am aware of is the Surface Pro, and I think that Intel has a lock on those for battery/power reasons. Intel has AMD beat pretty badly on the efficiency front right now. Although AMD APUs have much better integrated graphics than Intel for the most part so it would be very interesting to see what AMD could come up with.

 

[funkydmunky]

I believe the lowest end Kaveri part right now is 45w. Which is still too much for a tablet. Quite a bit too much. So there aren't any GCN tablets in the foreseeable future.

You need to read up on AMD hardware. Kavari isn't what AMD is targeting for ultra mobile. Check out what they were showing at recent electronic shows. Every AMD chip for laptops/tablets will have GCN.
But my point stands, Mantle will have a minor impact on these tablets. At this time Mantle is designed to enable performance on desktop level hardware. Not ultra efficient low power APU's that do not have the power to run AAA games. At least not yet

 

[rinaldo00]

Well the GPU's in mid-range laptops usually fall far bellow what the CPU is capable of, so I'm not sure if much of a boost is to be expected. Possibly high-end laptops with serious mobile AMD GPU's will put in quite a good show as the GPU can stretch its legs and not be held back by the CPU.
As far as tablets? I'm not sure what you expect Mantle to do there. Completely anemic gpu/cpu. They aren't designed for Mantle unless AMD unleashes some serious APU's down the road. Even then.

Yeah tablets were a stretch, but CPU limited scenarious are were Mantle shines.

 

[KickAssCop]

Well, that was underwhelming compared to all other benches I have seen. Seems like a good performance enhancer for mid range CPUs. For high end about 10% performance increase can be had by overclocking cards alone.

Waiting for full article and reserving judgement.

 

[funkydmunky]

Yeah tablets were a stretch, but CPU limited scenarious are were Mantle shines.

As long as the GPU is strong enough to take advantage. Otherwise it is moot.

 

[funkydmunky]

Well, that was underwhelming compared to all other benches I have seen. Seems like a good performance enhancer for mid range CPUs. For high end about 10% performance increase can be had by overclocking cards alone.

Waiting for full article and reserving judgement.

I don't understand this logic. So O/C the cards then add your Mantle performance boost %. O/C'ing doesn't nullify Mantles performance enhancement. And if you are going to run a very expensive CPU and whine about %10, then get a second GPU and let Mantle do its job

 

[Rizen]

You need to read up on AMD hardware. Kavari isn't what AMD is targeting for ultra mobile. Check out what they were showing at recent electronic shows. Every AMD chip for laptops/tablets will have GCN.
But my point stands, Mantle will have a minor impact on these tablets. At this time Mantle is designed to enable performance on desktop level hardware. Not ultra efficient low power APU's that do not have the power to run AAA games. At least not yet

I'll be honest, I haven't followed the low power stuff at all besides reading a bit on Bay Trail. I'll check it out.

Most of the low power game is ARM and I know AMD is building ARM Opterons. It would be neat to see an ARM+GCN design.

 

[funkydmunky]

To the [H]ard OCP crew, I haven't seen any AMD CPU love for quite a while, and usually for good reason, but I would be extremely interested in you throwing an 8350 or such into the mix as far as Mantle benches go. Does having 8 cores provide any advantage advantage to how Mantle software works compared to an i7 with HT on/off 4/8 threads? Is there a choke point to quad cores, or can a fast quad overtake CPU's with 8 threads where Mantle is concerned?
Just curious as we have all heard how well Mantle is supposed to handle more thread/cores. Is it true?
Hope you can oblige.

 

[Relayer]

Thanks, Brent. Good review, as usual. I was waiting to see what [H] had to say.

 

[technogiant]

if u go by that logic, because some users aren't able to play bf4 cause it keeps crashing on launch/map load means that no one is playing bf4 on mantle.

I've found that it crashes on Mantle if you have MSI afterburner running....I just apply the overclock and then exit out of it......think there are a lot of things to be sorted yet....that's why we are getting such varied and different accounts from different users....but this is just the first stage of a process, and as such it is very promising.

 

[Meeho]

To the [H]ard OCP crew, I haven't seen any AMD CPU love for quite a while, and usually for good reason, but I would be extremely interested in you throwing an 8350 or such into the mix as far as Mantle benches go.

Yes, it would be interesting to see what Mantle does for the AMD vs Intel difference in gaming.

 

[ecktt]

Good "Preview"

Mantle performance is exactly as I suspected; reduced latency and mostly helps CPU bound stuff. I really didn't understand why people would think it was anything else. If anything, it highlights that AMD still hasn't gotten control of their driver performance problems as the latency spikes in mantle mode clearly shows. What would happen if that were to occur during a multi-player match? Missed head shot or you getting fragged? AMD driver history is exactly why my system does not have their product anymore.

If anything Mantle is step in wrong direction for consumers imho. You have an API (DX and OpenGL AFAIK) that provide standardizes interface for developers to use that give opportunity for market competition.

The old folks would remember when you either had to have either a Creative Sound blaster or compatible sound card. Or what about 3dfx and their OpenGlide. People without a Voodoo card were stuck with ugly software rendering even though you had a better video card. Do we really want to go back to that?

Kudos to AMD for opening up their hardware at this level to get even more performance. But industry wide adoption would detrimental for PC gaming. Devs would now have to make a whole new application path to take advantage of AMD hardware instead of spending that time making a better game.

Maybe Mantle would come in handy for Bitcoin mining.

 

[MrLonghair]

Mantle is painted in a better light when tested in its intended situations, on weak CPUs (preferably that latest APU) with these supported GPUs. I guess this is a way to say it gives a tiny bit of a boost on hardcore equipment..?

 

[The Mac]

10% is hardly tiny.

add another 10% for a mild OC, and that 20%.

 

[Meeho]

Good "Preview"

Mantle performance is exactly as I suspected; reduced latency and mostly helps CPU bound stuff. I really didn't understand why people would think it was anything else. If anything, it highlights that AMD still hasn't gotten control of their driver performance problems as the latency spikes in mantle mode clearly shows. What would happen if that were to occur during a multi-player match? Missed head shot or you getting fragged? AMD driver history is exactly why my system does not have their product anymore.

If anything Mantle is step in wrong direction for consumers imho. You have an API (DX and OpenGL AFAIK) that provide standardizes interface for developers to use that give opportunity for market competition.

The old folks would remember when you either had to have either a Creative Sound blaster or compatible sound card. Or what about 3dfx and their OpenGlide. People without a Voodoo card were stuck with ugly software rendering even though you had a better video card. Do we really want to go back to that?

Kudos to AMD for opening up their hardware at this level to get even more performance. But industry wide adoption would detrimental for PC gaming. Devs would now have to make a whole new application path to take advantage of AMD hardware instead of spending that time making a better game.

Maybe Mantle would come in handy for Bitcoin mining.

Except that then proprietary APIs held the industry back, now it's the DirectX.

 

[Rizen]

Good "Preview"

Mantle performance is exactly as I suspected; reduced latency and mostly helps CPU bound stuff. I really didn't understand why people would think it was anything else. If anything, it highlights that AMD still hasn't gotten control of their driver performance problems as the latency spikes in mantle mode clearly shows. What would happen if that were to occur during a multi-player match? Missed head shot or you getting fragged? AMD driver history is exactly why my system does not have their product anymore.

If anything Mantle is step in wrong direction for consumers imho. You have an API (DX and OpenGL AFAIK) that provide standardizes interface for developers to use that give opportunity for market competition.

The old folks would remember when you either had to have either a Creative Sound blaster or compatible sound card. Or what about 3dfx and their OpenGlide. People without a Voodoo card were stuck with ugly software rendering even though you had a better video card. Do we really want to go back to that?

Kudos to AMD for opening up their hardware at this level to get even more performance. But industry wide adoption would detrimental for PC gaming. Devs would now have to make a whole new application path to take advantage of AMD hardware instead of spending that time making a better game.

Maybe Mantle would come in handy for Bitcoin mining.

The difference between Mantle and earlier APIs such as S3's Metal, 3dfx's Glide, etc is that those APIs were all closed and locked to only those vendors specific hardware designs.

Per AMD, Mantle is a lower level API but it is not completely locked to the GCN architecture - NVIDIA could support Mantle if they chose to implement it. It's unclear at this point if that means hardware specific changes on NVIDIA's side or just writing a Mantle driver, but I will give AMD the benefit of the doubt in that they recognize that having the API locked only to AMD hardware will prevent it from becoming an industry wide standard, which it sounds like they want.

 

[wonderfield]

10% is hardly tiny. add another 10% for a mild OC, and that 20%.

10% is the kind of per-app performance gain I would expect from a merely ample driver release. It's not all that uncommon to see per-app improvements twice as great from 'simple' driver optimizations.

 

[pcjunkie]

10% is the kind of per-app performance gain I would expect from a merely ample driver release. It's not all that uncommon to see per-app improvements twice as great from 'simple' driver optimizations.

Correct, wait until you see Mantle really optimized after a couple driver optimizations. It should really fly then! Being so new we should really see some gains...

 

[chameleoneel]

I know these are the [H]ardforums; but sometimes the narrow focus on the high end is a bit perplexing.

 

[Stoly]

The difference between Mantle and earlier APIs such as S3's Metal, 3dfx's Glide, etc is that those APIs were all closed and locked to only those vendors specific hardware designs.

Per AMD, Mantle is a lower level API but it is not completely locked to the GCN architecture - NVIDIA could support Mantle if they chose to implement it. It's unclear at this point if that means hardware specific changes on NVIDIA's side or just writing a Mantle driver, but I will give AMD the benefit of the doubt in that they recognize that having the API locked only to AMD hardware will prevent it from becoming an industry wide standard, which it sounds like they want.

I think AMD will pull a physx, Claiming other vendors are free to use and them blaming them for not supporting it.