AMD Mantle Vs. Direct X Battlefield 4 Multiplayer Benchmarks

[LtMatt]

28% improvement at 1440P with a 6 core cpu. Expect bigger improvements with a lesser cpu.

You might recall I have been ranting on and on about how Mantle will work wonders on Multiplayer gaming because the situation is so CPU bound. Well, I have finally gotten my hands on standardized Battlefield 4 Multiplayer Benchmarks with Mantle. Since Multiplayer is an extremely tricky thing to benchmark because the conditions change with every run, these benches are quite unique.

AMD Mantle Vs Direct X Multiplayer


AMD Mantle Vs Direct X


AMD Mantle Vs Direct X Multiplayer Crossfire


AMD Mantle Vs Direct X Crossfire



Source
http://wccftech.com/amd-mantle-direct-battlefield-4-multiplayer-benchmarks/

 

[LtMatt]

http://forums.overclockers.co.uk/showthread.php?t=18577693

That's for user benchmarks/feedback. You should know that Rusty, otherwise Greg's thread would have been merged with it.

Did it need it's own thread?
This was posted back in the Mantle thread when it first got posted lol.

See my reply to Rusty, also this was posted today on wccftech.

 

[LtMatt]

Keep on topic please.

My figures are correct rusty. 27.8% improvement at 1440P and 25% at 4k.

 

[LtMatt]

No they aren't.

48.6 / 45 = 1.08

i.e. 8%

Or are you looking at something else? OP doesn't make it clear.

You're looking at the wrong results.

 

[LtMatt]

Yeah there are way too many threads on this now, when you consider it's one game lol. I linked it in the BF4 benchmark thread an hour ago

It was posted 15 days ago...

http://forums.overclockers.co.uk/showpost.php?p=25771448&postcount=4738

My bad. Wccftech made it sound like the results were brand new.

I want to know why old news is being regurgitated?

Tbh i had never seen this so must have missed it.

To be honest i think the results deserve their own discussion thread anyway. I can see the thread has blossomed since i posted it last night.

 

[LtMatt]

Yeah I get that. Sounds like AMD don't have good frame times / consistency in DX in BF4. Though that isn't necessarily a fault of DX and a benefit of Mantle because as I said before, for me my game can't be any smoother. But I have a different vendor GPU so perhaps Nvidia are better on this angle currently.

No. When i last looked AMD had better frame times than equivalent Nvidia cards. The point is with Mantle even near perfection can be improved upon. The link to HardOCP's findings has already been shown to you. Plenty of other sites out there have reported similar things, not to mention a ton of user feedback on this very forum. If you actually tried it yourself, you'd realise and feel the smoothness improvement. Its not something you can dismiss as oh well it must not be smooth for single card AMD users currently to come to that conclusion Rusty. If you could run Mantle on your 780, you'd be noticing a smoothness increase as well.






Look at your 780 Rusty. You'd get a nice smoothness increase with Mantle.

 

[LtMatt]

That's not what I'm hearing though. People say DX is choppy and Mantle improves this.

My DX isn't choppy in the slightest and cannot be improved upon in that aspect so I don't know where this spooky phantom increase in smoothness is coming from. HardOCP said frame times but if the frame times were already OK then

Well AMD cards have superior frame times in this title, so im afraid you can't write it off as well Nvidia cards must be smoother thus the Mantle smoothness benefit won't apply to me and my 780. You need to try it, then you'll see the difference. As you'll never get to try it, you'll never know and will just have to take our word for it. Trust me its better, its smoother and it would be smoother for you, if you had a supported card.

As someone who has the performance overlay running 24/7 while i play BF4 i notice these things. The rendering times for cpu+gpu are lower on Mantle. Hence you get a smoother gameplay experience.

 

[LtMatt]

I can because a) my DX experience is 100% smooth already as I average 100-110 FPS on ultra with 2x MSAA and b) frame times are only one angle; consistency is another factor which I stated in my first post.

A paltry 8% boost wouldn't let me run 4x with the same frame rates so it wouldn't apply to me unless you mean by diverting to a sup-par DX experience in order to see the 'benefits' of Mantle.

I get more frames than that with a single card Rusty and as has been shown AMD cards have better frame times and are more consistent. If i can notice a difference, you will as well. The trouble is you seem to be willing to write it off as, its so smooth it can't get any smoother. Yeah, i thought the same, till i tried it.

I can see you've made your mind up that it wouldn't improve your experience despite myself and others showing you it will. I guess theres no hope for you. Ignorance is bliss.

 

[LtMatt]

Mate you said EVERY driver felt smoother back in the BF3 days . Placebo much? You haven't linked to the article you graph grabbed from either.

I'm just going with the majority who say that their DX experience isn't stellar and Mantle improved this.

I've got one thing up my sleeve too. I've USED Mantle on a friends PC. Side by side with my PC we couldn't tell much difference if anything other than the fogging and draw distance difference. DX versions we could though.

Edit: haven't really made much of it because it's anecdotal of course. He brought his gear round for me to fix it so thought I'd give it a whirl.

This is not just me Rusty, this is everyone saying it.

As for your trial on Mantle, sounds like PEBKAC as you're on the minority there.

It simply comes down to AMD have better frame times on BF4, Mantle improves on cpu+gpu rendering times hence you get a thinner, more consistent rendering line hence you get smoother gameplay. When i play with a single card in DX11 the line is smooth and thin. When you use Mantle its smoother and thinner. You notice the difference. It is immediately noticeable. Everyone in our TS noticed it who owned 290 cards and numerous people notice it here, not to mention review sites. Seems like its only you and your mate, who don't, so its likely PEBKAC there.

 

[LtMatt]

I'm not disputing that people are saying it's smoother. What I'm saying is that the majority are also saying that their DX experience wasn't stellar too. So the improvements are obvious.

So they've all compared AMD DX and Mantle next to Nvidia DX next to each other?

All you're doing is explaining the effects of what you're seeing on the graph so it's still anecdotal. As I said you said every driver was smoother a couple of years ago so... this was why I said let the hype die down and analyse it properly.

Still no graph link.

If I can get PEBKAC then I'd be amazed as it was a fresh install of Windows with only drivers installed. Not sure how I can get PEBKAC double clicking an install file but I suppose there's a first for everything .

AMD have lower frame times Rusty, so if anything Mantle would potentially have more effect on lowering frame times on your 780 than it would on my 290 as AMD already have a slight lead in this area. That performance overlay, that measures the frame times of how long it takes the cpu and gpu to render the frames. When you have two cards, the gpu rendering time is doubled, hence it appears above the cpu line. So heres the thing, mantle decreases the rendering time significantly on the gpu and the cpu so its improving both (not just the gpu rendering) and that why is why everyone notices a smoothness improvement with both combined it makes for a superior experience.

A massive reduction in CPU dependent situations, which to be fair is almost never on a gaming rig. Still, even less reasons to buy an i7 for gaming now, at least until games with a lot more simulation come out.

Slight average fps improvements at GPU limited scenaries might not tell the whole story. Mantle is supposed to reduce lag spikes and micro-stuttering due to improved sheduling, e.g. no random shader recompilation lags. I look forward to seeing FCAT frame times (from techreport or otherwise).

Edit: And DICE showed exactly that: (frame times in ms, lower/less variation is better)


low: AMD A10-7850K (‘Kaveri’ APU), 4 cores @ 3.7 GHz



mid: AMD FX-8350, 8 cores @ 4 GHz, AMD Radeon 7970 3 GB



multi-GPU: Intel Core i7-3970x Extreme, 12 logical cores @ 3.5 GHz, 2x AMD Radeon R9 290x 4 GB

http://www.neogaf.com/forum/showpost.php?p=98939900&postcount=28


Those were pulled from Single player. The gains in multiplayer will be twice that, maybe more.

EDIT

Heres your techspot link as well

http://www.techspot.com/review/762-gigabyte-radeon-r9-290-oc/page3.html

 

[LtMatt]

I think you're not getting my point or you're purposefully arguing a straw man because everything you're linking is AMD DX vs Mantle or multi-GPU or both. I'm not saying AMD haven't improved the situation. What I'm saying is that the place they have improved it to, to me, was not noticeably different from nvidia's DX version. Whereas we could tell a difference between AMD Mantle and AMD DX they (AMD Mantle and Nvidia DX) were roughly equally as smooth at 100-110 FPS (ish). We don't have multi GPU to test that angle hence no comment on it. The gains there look impressive.

Thanks for the link at last . The graph at 1080p (what I was looking at myself as only noobs play at 60 Hz ) is more interesting to me.

They all have low frame times but anyway this isn't the 14.1 driver so it's not a correct comparison point.

Not sure why they're suddenly a touch higher at higher res, no comments to explain.

Anyway these graphs aren't showing frame consistency just 99th percentile which is different. Also the difference at 1080p is a couple of ms so not that noticeable a difference.

They're higher at higher res because it takes longer to render the frame at highert res. Lower frames, longer frame times.

AMD have better frame times in this game. This is single gpu we're talking about here, not multi. There will not be any of this inconsistency you seem to be mentioning. 780 has longer frame times than a 290 on DX. Mantle lowers the AMD frame times further and it also lowers cpu rendering times meaning a smoother output. This will show itself in smoother performance via lower rendering times. If AMD have lower rendering times before Mantle, they will not have higher rendering times after it.

 

[LtMatt]

Heres an example of what frame times are going to look like for single cards on BF4 on DX. Assuming the card has enough grunt and vram, there will not be any inconsistency that Rusty mentions. If a R265/R270X can deliver smooth performance at 1080P Ultra settings, you can be sure a R290 will with vastly more fps.





I tried to find some other BF4 frame time results on pcper featuring a 290/780 but there are none. Only BF3 results so far (aside from the one above dated Fed 13th) so im guessing they've only just added it to their testing suit

 

[LtMatt]

Your graph is old drivers not the 14.1 drivers. That's what I was comparing to.

The difference in frame times at 1080p is 2-5ms which isn't going to translate to a perceivable difference. Additionally, it's a 99th percentile graph which is different to consistency.

You're still arguing a straw man - what is your actual point because I can summarise mine in a couple of sentences?

When I compared (allbeit briefly) AMD Mantle vs AMD DX in 14.1 I could tell a difference in smoothness. When I compared AMD Mantle vs nVidia DX I couldn't tell hardly any difference.

Is there a difference in pre Mantle DX driver and Mantle? Has anybody compared this side by side? Anyway I don't think there's much point continuing while you're attacking a straw man so if you want to discuss the point I'm raising then feel free to respond accordingly .

As i said earlier, probably a case of PEBKAC on that system you used. You admitted you only tested it briefly so i don't put much weight behind your claims. Especially when pretty much everyone else says the opposite.

When i benched 14.1 and 13.12 on DX and there was little difference. I think the fps were pretty much identical. Maybe 14.1 had a fractional improvement but nothing major.

I would expect a bigger reduction in 780 frame times and an improvement in smoothness if you were able to use Mantle. Probably more than i would see due to the slightly higher rendering times exhibited on your card. Its likely one of those things that you wouldn't realise until you were able to see and feel it. Clearly you've not been able to yet.

I didn't think you liked PCper Matt, but as you mentioned it and it is a graph war, let's have a look at this.

A very minimal gain on their testing.

Lol. Its not a graph war.

That graph you posted is single player and he (Ryan at pcper) deliberately picked the most gpu bound part in the game to test. The OP contains multi player benchmarks and the title says as much. Also the graph you posted is FPS and not frame time so its irrelevant to what myself and Rusty are discussing.

 

[LtMatt]

How can I get PEBKAC on a new install of Windows with just the bare drivers installed for the things plugged in?

You benched it but did you compare frame times and consistency? Why is everyone ditching 14.1 for DX then if it's OK? Everyone I have spoken to says that DX is knackered in 14.1... so?

The point is this smoothness is just noticed going from AMD DX to AMD Mantle. AMD Mantle vs nVidia DX at similarly high frame rates I didn't really notice anything different between the two.

I don't know but something is clearly not right from what you've said.

Yes, the BF4 benchmarking command delivery outputs fps and frame times for cpu + gpu. The problem with 14.1 is the bugs Rusty, not the performance improvements it offers for single card users and SOME crossfire users in BF4.

As i say i don't think we can hold your brief testing with too much regard. I'm likely to go with the majority and my own findings on this having spent some time looking at it.

Single card AMD DX BF4 performance is stellar and plenty of reviews show that to be the case. However in your case there appears to be an exception here going against the grain.

 

[LtMatt]

I'd like to test this smoothness claim, in fact I may very well be able to soon if I get this 290x I've been offered. Both sides have got frame times down to a point where it's under most monitors response times.

It would settle this argument once and for all. The way i see it is currently like for like both sides frame times are similar, close, excellent. However AMD may be slightly ahead, assuming like for like gpu speed ala 290/780. Now if Mantle lowers not only the gpu rendering time, but the cpu rendering time as well, how can that now not equal anything but smoother gameplay online at Ultra settings?

 

[LtMatt]

Best experience is with frame cap using different programs or v sync with mantle that at least for me, it shaves off the input lag that this option usually comes with.

This whole discussion reminds me of the first FCAT tests. Users (let's call them "the green guys") come and say "look, on AMD it plays baaad, nVIDIA offers a smoother experience!". Some AMD users come and say "well, it's pretty good on my end, even in multi GPU". The first ones go "oh, but look at the graphs, look at the numbers, you sir(s) are wrong, wrong I tell you!".
Time passes by and with newer drivers and future testes show that AMD is actually better than nVIDIA in BF 4. Mantle comes along and it improves on that even more. Now, the "green guys" come and say "oh, no more graphs, it can't be smoother than what it is on my rig. It's perfect, this is just placebo".

With performance it's the same. Although users experience better overall frame rates and frame times (the better as the hardware is weaker), the final sentiment is "oh, under 10%, it's small, nVIDIA gave 10% by driver", although everyone, including review sites and users, are saying things look a lot better and can DIFFER greatly from one setup to the other (from 0% to 30-50% or more).

Why is so hard to accept that DX is not the best API at the moment? Because in the end THAT is the whole freaking point! We are using to much brute force to overcome old software ways of doing stuff.

Lol. Spot on my friend. I remember those dark days well.

 

[LtMatt]

Ahhhh the "suit my argument only debate". You deliberately picked the most CPU demanding part of the SP to demonstrate the CPU bottleneck to skew your results in your BF4 bench thread (another one that was started here). I don't care either way and the OP shows the possible gains but it isn't representative for all and another part of the game will show little to no improvement, so it is all relative.

frametimes are all well and good but like Rusty has stated and I have seen graphs of yours and other users that show AMD to have frametimes very similar to what we are seeing now, so placebo effect is indeed happening.

What argument?

I posted multi player benches in the OP. That was pretty clear. I think everyone here knows that's where you will see the biggest gains.

I benched the part Johan recommended as its the most demanding on cpu+gpu in the single player campaign. I don't see the point in benching a part that has no cpu usage requirement at all. What would be the point in that? You have to bench a demanding sequence to try and best replicate demanding online play.

Doing anything else is a waste. Its why it was no surprise to me that pcper did this.

Worst Case Scenario

Our standard level and section used for testing Battlefield 4 is the opening portions of the single player SP_Dam stage. As it turns out, this section is quite GPU limited (as we would normally want for our GPU testing...) but this doesn't fair well for AMD's Mantle out of the gate. Here are results with the Core i7-3960X and R9 290X at 2560x1600 with the Ultra preset.

Even in their worst case scenario look at the benefits Mantle offers over DX11. Worst case!!

Rusty will be disappointed to see excellent consistency on frame times, even on DX11. Thanks Greg for helping me find these, didn't realise they existed. So there goes Rusty's consistency argument out of the window. Although both gpu rendering lines are smooth, Mantle is marginally better on worst case scenario. You go online for a better case scenario, maybe even with lesser cpu hardware and those improvements will climb much higher.

 

[LtMatt]

Not really. I'm laughing because you're still attacking the straw man . I suggested frame times and frame consistency as a reason not the absolute reason.

I don't know why Mantle is appearing so much smoother over DX. HardOCP said frame times, not me. I just suggested this may be the reason.

But everyone knows you're not allowed to say anything other than glowing positivities about Mantle in this sub forum...

Laugh away my friend. I'm just explaining to you why Mantle will give improved smoothness. Mantle is appearing smoother than DX11 because the API is a bottleneck. Just because you have a 780 does not mean you're excluded from the drawbacks of the API itself. Mantle is smoother and faster than DX. A 290 on Mantle will show smoother performance (lower gpu+cpu render times) than a 780 on DX. A 290 is currently showing similar or better frame times than a 780 with similar or better frame variance.

I don't think anyone cares if people say bad things about Mantle. That's what the feedback thread is for. I've criticised the drivers for not having crossfire working, but thats the problem when you have a new beta api coupled with beta drivers.

Its not frame times, frame times are very good, compared to Nvidia or not, the frame times don't actually improve, i don't think they can from what they are in DX, its something else.

That line graph in BF4, its pencil line straight in Mantle.

Yes gpu render times are already very good on DX for both sides. However Mantle is more efficient so its gpu render times will still be lower than DX overall. In that worst case scenario above you can see that even then the Mantle line is considerably thinner which means time spent by the gpu rendering each frame.

 

[LtMatt]

So are they testing DX on 14.1 or testing DX on 13.12?

Everyone knows that DX on 14.1 is gimped and frames look very low to me for a 290X? Another point you made a while ago Matt is MP is uesless, as there is so many variables that can take place, but these tests from Hardwareluxx look good so now it is ok?

14.1 worked fine for me on DX. I went back to 13.12 WHQL though because crossfire did not work at all on 14.1.

Single player is best for consistency yes of course. But in this case its better to try as best as possible to test multi player because thats where the biggest performance improvements are and that's what everyone plays the game for.

Yeah they look good there. Honestly though my DX experience in BF4 is flawless. It's probably my set up though which is fully GPU limited.

For anyone else: don't confuse that with me not thinking improvements aren't possible or that Mantle isn't doing the right things... I'm just sceptical about what benefits are being provided currently over a similar high end nVidia set up. I think that's largely the problem. We all know people can't understand somebody's point unless they're grouped in a red or green camp. The backwards and forwards defending is as a result of this mindset.

If i get gains and can notice a difference on a single card, you would as well Rusty.

 

[LtMatt]

Not being funny mate but again, you said every driver from about 12.3 onwards was smoother in BF3 so I'm just saying that it's important to let the hype down and analyse it objectively.

I have and ive looked at the gpu+cpu render times for my own outputs, so i know its smoother. Then i see these pcper graphs, i see the testing Dice did and published all over battlelog. The most important thing? I run the performance overlay 24/7. If i see a change in rendering behaviour and a smoothness/performance increase, i notice it immediately. This is not just being made up by everyone Rusty to counter your argument.