BF4 AMD Mantle Video Card Performance Review Part 1 @ [H]

[FrgMstr]

BF4 AMD Mantle Video Card Performance Review Part 1 - AMD's Mantle API has been with us for just over a month now, and we have strapped a variety of video cards to the test bench to see what real world differences are being delivered to gamers within Battlefield 4. We will compare D3D11, Mantle, on various GPUs, looking at highest playable settings, frame times, and discuss our experiences.

 

[Cool Vibrations]

Thanks for the review guys.

Those of you looking for smoother frame times in BF4 with CrossFire configurations look to be very much in luck though.

I can definitely confirm that those frame times are indeed very tight. The stuttering and choppiness I got playing with CF D3D is almost nonexistent with Mantle.

Still using my 7970s to tide me over until the R9 290x Lightning releases.

 

[mach1.9pants]

I cannot get BF4 to run in Crossfire, no matter what I do, crashes at load everytime. Single card it crashes at load often, but is playable. Mantle is much smoother and higher frames for me.

 

[Rizen]

As a CrossFire 290X owner I can say that in Battlefield 4, Mantle feels much, much smoother than Direct3D and I see a noticeable frame rate increase @ 1440p Ultra 4X MSAA. I have the same problem with resolution scaling, however. Enabling any amount of resolution scale results in unplayable frame rates (which appears to be VRAM usage). I hope they'll get that kink worked out, but otherwise I am pleased with the results so far. Especially given that BF4 was not designed with Mantle support in mind.

 

[FearTheCow]

Any plans to test mantle using a a less powerful processor?

 

[chameleoneel]

Hey Kyle I'm wondering about this:

During the course of testing and data collection, Catalyst 14.2 Beta driver was released by AMD. The driver primarily addressed bug fixes as well as fixed CrossFire running in Mantle. No other performance enhancements were made, thus for single cards, the 14.1 Beta driver we started evaluating with will provide the same performance as the 14.2 Beta driver.

With my 7870, I feel like overall FPS is a little lower with 14.2 VS. 14.1. Even though stuttering/hitching isn't quite as bad with 14.2 (VRAM usage is still way higher for same settings) and image quality is now exactly the same (after a BF4 patch for the Mantle render path). A couple of other users have also mentioned things about 14.2 possibly not giving quite as high of a boost to FPS.

So I guess I'm asking for a double confirmation that AMD's high end cards did not show consistent performance differences between 14.1 and 14.2?---whether it be in overall FPS or frame times. and if that's the case, did you notice a performance difference after the recent patch that fixed the image quality disparity? I'm wondering if the possible performance difference is due either to the new driver or maybe due to the fact that image quality is now in parity (see: maybe corrected render is actually more expensive?).


*also interesting: with the conclusion that current, patched BF4 + current driver = less DX11 performance than at BF4 launch; have you guys actually played BF4 recently with those older drivers? Becasue on my 7870, Since sometime in January, I have run into several graphical glitches in BF4, when running anything previous to 13.12whql


*it would also be nice if [H] would focus on the CPU side of Mantle. Because your articles so far have seemed focused mainly on the graphics side. Which is fine, but isn't the total picture. It is mentioned at the end of the article that Mantle doesn't bring the "30%" performance boosts, to the GPUs. But what about total system performance? what about people that aren't using an i7? what about people that aren't using Intel? On a Phenom II X6 and a 7870, I've at times recorded over a 25fps boost to average framerate (14.1).
With AMD consistently landing behind Intel in game performance, does Mantle mean AMD can punch with an i5 or i7? I've seen people report 100% performance increase, with core2quads in BF4.

The CPU side of the coin is something that I feel a lot of tech sites have somewhat ignored with Mantle. When AMD has been pretty vocal about that aspect. I mean, the dev journal videos/presentational videos for Starswarm have basically been all about the CPU. Citing no specific GPU optimization. All focus was on the CPU side. and while BF4 is certainly a graphics showcase and the single player is particularly GPU limited----the multiplayer side represents some tough stuff for CPUs. I think that should be explored.

 

[Brent_Justice]

I tested the new driver when it came out to verify there were no differences, and in my testing, there were no differences.
No differences were reported or documented from AMD either.
14.2 did change CrossFire performance, and because of that, we used it with 290X CrossFire.

 

[chameleoneel]

I tested the new driver when it came out to verify there were no differences, and in my testing, there were no differences.
No differences were reported or documented from AMD either.
14.2 did change CrossFire performance, and because of that, we used it with 290X CrossFire.

Ok, thanks for the quick reply!

*for now I'll have to chalk it up to the fact that 290 and Kevari seem to be the focus right now for Mantle optimizations!


**also, something that some of us have been curious about: did you guys happen to track VRAM usage and notice higher VRAM usage in Mantle VS. DX11, for the same settings? Some of us that have lesser AMD cards (I have a 7870) have been curious whether or not higher VRAM usage is intended behavior or if that's just one of the bugs on the cards that haven't been optimized for Mantle yet.

 

[Pieter3dnow]

The only thing that is missing is the part where Mantle absolutely destroys gaming on low/mid end cpu where you can not get any performance from traditional methods Mantle allows the now dated Phenom II x4 to work as well as some of the high end cpu.

 

[Enhanced Interrogator]

Thanks for the review guys. I hope you look deeper into the VRAM issue with part 2. My 2GB 7850 is getting hammered in that department. I have to reduce textures by one step and resolution by about 15% to avoid massive stuttering. That's a pretty significant sacrifice. Even then I have to stop every 2-3 games because VRAM continually increases over time.

That said, my minimums are improved significantly in a lot of areas that normally choked my Core2Quad. So if they can get a handle on the memory management, Mantle will basically be a game changer for mid-tier systems.

 

[NukeDukem]

I am at 1440p and with dual 290's and mantle I have the game at 150% on resolution scale, all ultra settings, no ambient occlusion or msaa, low fxaa. I rarely see minimums lower than my target 96 FPS. It is pure gaming bliss!

 

[Schro]

Hey Kyle I'm wondering about this:



With my 7870, I feel like overall FPS is a little lower with 14.2 VS. 14.1.

At least it is playable for you on the 7870. The R9 270X is basically a rebadged 7870, and quite frankly, the 270X was getting pounded. Per the article, AMD said this was a game issue that would need to be patched by DICE, but they have at least been able to reproduce the issue. Keep in mind that, only the 290X, 290 and 260X are officially optimized for Mantle right now (though, the 260X is quite broken as well).

*also interesting: with the conclusion that current, patched BF4 + current driver = less DX11 performance than at BF4 launch; have you guys actually played BF4 recently with those older drivers? Becasue on my 7870, Since sometime in January, I have run into several graphical glitches in BF4, when running anything previous to 13.12whql

Yes, we've played with the 13.12's quite extensively and performance with those drivers are on par with the launch performance. Right now, if you try to run the game with less than 13.12 installed, it will give you a warning dialog box to update your video drivers.

Thanks for the review guys. I hope you look deeper into the VRAM issue with part 2. My 2GB 7850 is getting hammered in that department. I have to reduce textures by one step and resolution by about 15% to avoid massive stuttering. That's a pretty significant sacrifice. Even then I have to stop every 2-3 games because VRAM continually increases over time.

That said, my minimums are improved significantly in a lot of areas that normally choked my Core2Quad. So if they can get a handle on the memory management, Mantle will basically be a game changer for mid-tier systems.

Without testing, I would assume that the issue that is facing the 270X and 260X also translates to the 7850 (as the 270 is basically a rebadged 7850). With the 270X and the 260X, the vram usage does show as a bit higher within the game, however, the performance problems started before the cards vram was exceeded, so I think there's a different root cause that needs to be addressed. Until it is patched, I'm not sure you'll have a great option...

 

[gersson]

Can we have some comments on the "Resolution Scale" option in BF4 vis-à-vis DX11 vs Mantle?
I feel like it is getting ignored even though it is definitely an enthusiast option.

It seems like Mantle uses more VRAM and that leaves less for using 150% scaling, etc.

My observation is that anything above 120% with Mantle will result in low FPS; my 290x Tri X OC can handle 150% under DX11.

Comments? I feel like this is the kind of thing most of us would normally geek out about.

 

[magoo]

Does this really seem worth it for a 10% improvement?

I realize this is very new and rather cobbled together so far.....

It will be interesting to see a game developed from the ground up with Mantle in mind.
Also, if I recall the hefty improvements originally were directed toward the lower performance end of the spectrum......that should be interesting.

Very comprehensive article. I'm sure it was a ton of work.

 

[Pieter3dnow]

It seems like Mantle uses more VRAM and that leaves less for using 150% scaling, etc.

My observation is that anything above 120% with Mantle will result in low FPS; my 290x Tri X OC can handle 150% under DX11.
.

Yeah and you should know that the part you are describing is done by DICE , under Mantle developers are responsible for allocating memory on the gpu.

 

[Modred189]

Last page, just above the "rough edges" paragraph:

Many people were concerned about image quality differences as well. There is a known gamma bug between D3D and Mantle in this game. This will most likely take an appellation fix that will have to come from EA. Otherwise we did not observe any differences between the DirectX 11 API and the Mantle API in image quality.

Should be "application"?

Great review so far. I am very interested in how it all scales over time as new revisions and driver updates come out.

 

[Wierdo]

Thanks for the review guys. The smoother frame times chart was the most interesting to me.

I can see this is paying great dividends in multi-GPU setups already, so would be nice to see how this develops as Mantle and its utilization in games that support it matures.

Oh the other hand with news of MS getting off their asses and working on similar improvements in DX, as well as new of OGL doing something like that too, I have to wonder how long before this goes the way of Glide.

But perhaps it served its pupose in that sense, AMD is apparently "glad" to see DX/OGL get serious about improving in the areas Mantle is tackling finally.

 

[Agalloch]

Why does everyone review mantle with an oc'd 3770k? Isn't the whole idea of mantle to help slow/old cpu's in newer games or have I misunderstood the technology?

 

[jbltecnicspro]

Why does everyone review mantle with an oc'd 3770k? Isn't the whole idea of mantle to help slow/old cpu's in newer games or have I misunderstood the technology?

Good point, but I think the purpose of this article was to look at the video card performance enhancements that come to users who have these cards. Since we're interested in video card performance, then it would make sense to have a powerful CPU to eliminate possible bottle-necking with the CPU. I think a good next article would be to see what Mantle does for lower-powered CPU's. Hell - let's see what happens when we try to game on a lowly Celeron!

 

[MIghty Hooligan]

Why does everyone review mantle with an oc'd 3770k? Isn't the whole idea of mantle to help slow/old cpu's in newer games or have I misunderstood the technology?

I agree, I would like to see how it performs with less than stellar hardware, to get an idea of the difference when using say a R9270 or R7260 with mid range CPU. I also thought that the Mantle was more productive there but could be mistaken.

 

[mi7chy]

In-line with what I'm seeing. Significant benefit with R9 290X but a step back with 270X. It's also benefiting the Kaveri A10-7850K which I hope you will add to your testing.

 

[Hameedo]

Still, with all of this improvements, 290X alone is not enough to take 780Ti, even with Mantle. even the other cards are barely faster (4-5fps).

The real benefit here for Mantle is the Crossfire configuration. The rest are Meh. Also testing with a high end CPU is crucial, since no body in their right mind will couple a 280X card or above with a low end CPU just for one game while limping and crawling in all of the other games.

 

[Wierdo]

Why does everyone review mantle with an oc'd 3770k? Isn't the whole idea of mantle to help slow/old cpu's in newer games or have I misunderstood the technology?

High end PC gains are interesting as well, even if not the target exactly. And I imagine going dual/triple/quad GPU would highlight the lower driver overhead more, even on highend rigs.

 

[Schro]

Can we have some comments on the "Resolution Scale" option in BF4 vis-à-vis DX11 vs Mantle?
I feel like it is getting ignored even though it is definitely an enthusiast option.

It seems like Mantle uses more VRAM and that leaves less for using 150% scaling, etc.

My observation is that anything above 120% with Mantle will result in low FPS; my 290x Tri X OC can handle 150% under DX11.

Comments? I feel like this is the kind of thing most of us would normally geek out about.

Sure! Performance on a pair of 290X's under DX11 really isn't good enough to increase resolution scale by more than 10-15, and the amount of image quality that you gain really isn't noticeable, so at that point, you're just twisting dials in order to lower your FPS with no apparent impact..

 

[acerbic]

Still, with all of this improvements, 290X alone is not enough to take 780Ti, even with Mantle. even the other cards are barely faster (4-5fps).

The real benefit here for Mantle is the Crossfire configuration. The rest are Meh. Also testing with a high end CPU is crucial, since no body in their right mind will couple a 280X card or above with a low end CPU just for one game while limping and crawling in all of the other games.

Define "all other games". Because your usage situation is yours alone.

I know lots of people who play almost all their AAA titles on console. The PC is largely used for various indie titles and other games that all fall into the "small, light, cheap, quirky" category. With the rare exception of usually one big multiplayer game (like BF4) which then consumes all their MP time for years on end.

That usage scenario is pretty damn common as well. And if your one big PC MP time sink is a mantle enabled game, you can make a good argument that you should be buying an AMD GPU right there and then.

 

[B00nie]

Many people may be able to get a fast graphics card without being able to upgrade their processor at the same time, which is where mantle should benefit hugely. Saving 500+ bucks on a new cpu and motherboard is no small deal.

 

[Raap]

Still, with all of this improvements, 290X alone is not enough to take 780Ti, even with Mantle. even the other cards are barely faster (4-5fps).

The real benefit here for Mantle is the Crossfire configuration. The rest are Meh. Also testing with a high end CPU is crucial, since no body in their right mind will couple a 280X card or above with a low end CPU just for one game while limping and crawling in all of the other games.

I don't think this is entirely accurate. I agree with the general idea, but I suspect that if you looked at the average BF player he'd be running a quad-core clocked at around 3ghz. The overclocked monster [H] is using in testing is close to the quickest CPU-setup you can get for 'normal' usage, and it's over 50% faster than my guesstimate for the average.

 

[Tamlin_WSGF]

Still, with all of this improvements, 290X alone is not enough to take 780Ti, even with Mantle. even the other cards are barely faster (4-5fps).

The real benefit here for Mantle is the Crossfire configuration. The rest are Meh. Also testing with a high end CPU is crucial, since no body in their right mind will couple a 280X card or above with a low end CPU just for one game while limping and crawling in all of the other games.

I disagree. I'm using a i5-2500K @ 4.3 with a 7970 OC which is similar to a 280X. The framerate increase in BF4 seems to come from the minimum which is an important point for me. In addition, I find the framerate delivery to be much smoother under Mantle then under DX 11. The biggest benefit with Mantle in BF4 isn't in the high end IMO, but rather on the largest gaming segment which is the midrange to low-end. Overall, Mantle could be made a benefit in all games, since there are features not implented in BF4 which could make use of the increased efficiency of Mantle in other games like higher draw distance etc. What we call a "mid-range" CPU today was a "high-end" CPU not long ago. With Mantle, we have a "high-end" CPU from the future with the use of Mantle on our current CPU's. All we need, is for the games to take more advantage of them.

For now, high-end setups gets a framerate increase where it matters the most, in the minimums and they also get more even and faster frametimes.

I have a MSI GTX 780 Gaming OC card with an i5-4670K CPU on my HTPC. Its not remotely apples-to-apples, but I tried BF4 on it last weekend. BF4 is actually smoother with an i5-2500K and 7970 OC under Mantle, then with an i5-4670K with a GTX 780 OC under DX11.

It wasn't a scientific test. The Mantle machine ran under Windows 8.1 64-bit and the GTX 780 would have run better if I would have used 8.1 instead of win 7 64-bit. But I used the same screen, keyboard and mouse under 1080P/120hz resolution to make the gameplay as similar as possible at least.

Mantle "feels" faster.

 

[TomGreen321]

Would really like to see a game that isn't already optimized for AMD cards for comparisons soon. AMD said the 290X would destroy the 780Ti or Titan(or something to that effect) and that really isn't the case. Other Mantle reviews also show how much better nVidias drivers are in terms of CPU utilization across the board.

I also wonder about how the game looks and if that has a positive gain in performance for AMD as it looks like AO is missing or bugged or something. It seems like if OpenGL and D3D follow suit with the low level stuff AMD is doing with Mantle it wouldn't look very good for AMD.

I hope you guys do a review with i3s and AMD Quad cores as well. It seems like someone with a upper-mid range setup like mine for only goes for 60 fps, Mantle won't really benefit much, but Low end and super highend enthusiast builds will show decent gains from all of this.

 

[Modred189]

Would really like to see a game that isn't already optimized for AMD cards for comparisons soon.

Not sure that's going to happen. Why would a game NOT so optimized opt for Mantle? Wouldn't you expect any game that sports Mantle to be a AMD-focused game (as opposed to THWIMTBP) like BF4?

 

[Terpfen]

Oh the other hand with news of MS getting off their asses and working on similar improvements in DX, as well as new of OGL doing something like that too, I have to wonder how long before this goes the way of Glide.

What news is this?

 

[jbltecnicspro]

What news is this?

Microsoft is apparently trying to allow for even lower-level access in DirectX, ala Mantle. OpenGL is apparently trying to do the same, though I always thought that lower-level GPU access was always available via extensions in OpenGL. So I guess we'll see what happens.

 

[GoodBoy]

Thanks for the hard work put into this review.

Wondering what your thoughts are on the DX performance being worse than at launch? Is AMD trying to make DX11 look a bit worse to give the improvements mantle can make look all the better? Could those perf differences be something in the game engine itself? (maybe making the game mantle friendly results in DX performance compromises). Did those dx performance declines since launch only affect the AMD cards?

 

[Terpfen]

Microsoft is apparently trying to allow for even lower-level access in DirectX, ala Mantle. OpenGL is apparently trying to do the same, though I always thought that lower-level GPU access was always available via extensions in OpenGL. So I guess we'll see what happens.

Where was this reported, though?

 

[Modred189]

FOUND it.

 

[Terpfen]

FOUND it.

People still do this?

 

[Schro]

Thanks for the hard work put into this review.

Wondering what your thoughts are on the DX performance being worse than at launch? Is AMD trying to make DX11 look a bit worse to give the improvements mantle can make look all the better? Could those perf differences be something in the game itself? Did those dx performance declines since launch only affect the AMD cards?

I'm not exactly sure what happened there. It does not seems to be a patch level difference as the 13.12 WHQLs are performing on par with launch performance on the AMD side. I didn't put my tin foil hat on to speculate why there was a DirectX performance decrease though...

For NVIDIA cards average frame rate performance seems to be about the same, but many of the major dips have been ironed out. Overall, its just as playable now as it was back then on each card.

 

[Modred189]

People still do this?

OH YEAH

 

[GoodBoy]

It's about time Microsoft did something with DirectX. The last redist was released June 2010. And up until then, there were usually 2 to 4 redist releases per year. Anything that makes it more efficient will be a good thing.

In fact, as some of you may recall, an AMD executive publicly stated a year ago that there was no “DirectX 12″ on the Microsoft roadmap. Microsoft responded to those comments by affirming that it remained committed to evolving the DirectX standard — and then said nothing more on the topic. Then AMD launched Mantle, with significant support from multiple developers and a bevy of games launching this year — and apparently someone at Microsoft decided to pay attention.

Apparently AMD called Microsoft's bluff.

With as much GPU engine changes that there have been since 2010, Dx likely does need some work done to get it up to speed.

 

[sirmonkey1985]

Thanks for the review guys. The smoother frame times chart was the most interesting to me.


I can see this is paying great dividends in multi-GPU setups already, so would be nice to see how this develops as Mantle and its utilization in games that support it matures.

Oh the other hand with news of MS getting off their asses and working on similar improvements in DX, as well as new of OGL doing something like that too, I have to wonder how long before this goes the way of Glide.

But perhaps it served its pupose in that sense, AMD is apparently "glad" to see DX/OGL get serious about improving in the areas Mantle is tackling finally.

i think it's win win for AMD either way, their intention was to improve the gaming experience for people and in the end it worked, whether it's through their technology or some one elses. i mean you saw what happened with their heavy investment in GDDR5 development(which has now become the standard) when Nvidia was still screwing around with GDDR3. MLAA which was eventually improved on by others and became open source. if microsoft does actually get off their asses adds support in DX11 the only people that win are gamer's but we all know the reputation of microsoft and what they do with directx. (inb4 it requires windows 9 to actually use the features)

It's about time Microsoft did something with DirectX. The last redist was released June 2010. And up until then, there were usually 2 to 4 redist releases per year. Anything that makes it more efficient will be a good thing.



Apparently AMD called Microsoft's bluff.

With as much GPU engine changes that there have been since 2010, Dx likely does need some work done to get it up to speed.

doesn't surprise me at all, thats pretty much the way microsoft has treated directx for a while now.