AMD Mantle: A Graphics API Tested In Depth

[TechyInAZ]

Very very good review!

I've never really thought about APIs since openGL and DirectX were the only ones until Mantle poped up.

This makes me seriously consider getting a amd based card since I've always been an NVidia fan.

It's going to be a war however, since even though mantle does show improvements to amd cards, we've got NVidia making drivers that use DirectX 11 a lot more efficiently and thus make their cards get more performance, I just wonder what will happen when directX 12 rolls in.

 

[cleeve]

Cleeve, do you know if the Performer API tech ended up going elsewhere?

Hey Ian,

Honestly I don't have any specific info on that. Looks like its pretty old tech, so if it worked Its probably safe to assume succeeding graphics devs used techniques learned there, but only an API/Driver developer might know exactly at this point.

Having said that, OpenGL still gets iterated and the target du jour is "close to the metal" efficiency, so it's a safe bet that we can expect this direction from all the APIs.

 

[vaughn2k]

Now I can stick to my AMD Phenom X4 965BE, and just upgrade my video card. This is nice improvement indeed!

 

[-Fran-]

I hear you. But in this case it's not about difficulty, it's about adding unnecessary variables.

The point of benchmarking is to target specific data. For this Mantle article, that data is the graphics API in BF4, not SP vs. Multiplayer performance in BF4.

In this particular case, testing MP wouldn't make sense. it would just muddle the data with a less-consistent test case.

That's why I said you could just describe what you saw. Personal perception is not as bad as you think, as long as you have some data to back it up.

Adding additional variables to testing, if I get what you're trying to say, is the pure nature of MP testing. Lab conditions is good when testing for lab results, but MP is as real conditions as it gets (meaning hectic, but good). The more data you get from that, the less variability you'll get. Meaning, more testing time, yes, but better and throughout results to share.

Like I said, you're testing a car in a drag strip, when the conception/idea of the car was track days (BRZ reviews much? haha). You know the added effort for testing MANTLE is when CPUs are pushed. You even said that was the idea behind MANTLE, to lower the CPU footprint

Anyway, it's a good article. Sorry for not mentioning that. The data you got from Thief really shows MANTLE is headed in the right direction so far.

Thanks for the efforts.

Cheers!

 

[cleeve]

Anyway, it's a good article. Sorry for not mentioning that. The data you got from Thief really shows MANTLE is headed in the right direction so far.

No need to apologize for mentioning a concern.

And I apologize if you feel my response was anything of an explanation of why we didn't go that route. I wasn't blasting you man, I'm just trying to explaining why I made that decision.

The problem with multiplayer is that it's almost impossible to recreate identical scenarios. People move, change, shoot differently every time. Those are the variables we can't control, and they'll affect the outcome of the test. In single player, it's 100% repeatable.

Kind regards,

- Cleeve

 

[filippi]

From page 4:
"Finally, with the High preset enabled, and a Radeon R9 290X/GeForce GTX 780 Ti installed, the results start getting strange. Nvidia's high-end gaming card averages about 80 FPS on a Core i7-4770K, and is matched by the Radeon R9 290X using Mantle and an FX-8350. Then, we swap the R9 290X into our Core i7-4770K-based system and observe dismal results with Mantle turned on."

Strange indeed: Dismal results with DirectX R9 290X + Core i7-4770K too!

 

[Traciatim]

The point of benchmarking is to target specific data. For this Mantle article, that data is the graphics API in BF4, not SP vs. Multiplayer performance in BF4.

In this particular case, testing MP wouldn't make sense. it would just muddle the data with a less-consistent test case.

I disagree here, in BF4 it would seem mostly single player isn't all that CPU heavy anyway, as shown by the terrible scaling with CPU power. However in multiplayer it suddenly becomes a huge issue when you have piles of players. This means that mantle, which is designed to alleviate slow CPU issues, should have a much greater impact on multiplayer games so the test should reflect the situations where it has a greater chance of a benefit.

It would be best if you could do a small article on it with at least two machines that are popular like an 8350 and a 4670/90k and just play some rounds with and without mantle in each scenario, at the very least to see if it scales better, worse, or the same. Preferably just using the same machines in the same games and follow each other around while playing. It's not perfect, but if you average out a few rounds it should be good enough to get some nice data.

 

[TechnoD]

Another well-written article.

Since it looks like Don put a lot of effort into this article and is responding to reader's comments, I found a couple minor errors:
Page 9: " Even though the gap is much closer, DirectX (<-Should be Mantle) doesn't offer an advantage in anything less than a Radeon R9 290.
Page 4: "Again, Thief appears to be an ideal proof point for Mantle. Under DirectX, the Core i7-4770K-based platform averages 57.7 FPS (Should be 56.7 FPS), while the FX-8350 registers 46.5 (Should be 46.0).

Again, I understand that everyone makes mistakes. Just thought that Don would appreciate the feedback. Thanks Don for another great read!

[Response by Cleeve]

Thanks for the proofread man, issues fixed!

 

[lostgamer_03]

Never had confidence in Mantle and probably never will, goodbye Mantle nice not to have known you.

 

[utengineer]

Is there a significant difference between the development and performance of the game in Campaign(single player) versus Multiplayer? I see a lot of synthetic benchmarks that appear to be using the graphics/textures of the single player game and not the multiplayer. I pose this question because it seems like the Single Player version of the game uses better graphic textures than what the Multiplayer offers (e.g. BF4).

 

[TheAshigaru]

Great read, Don! These are the kind of articles I like to see on Tom's. It's pretty hard to understand what AMD really means from their press releases and conference speeches. 1st page broke it down nicely, I thought.

 

[Traciatim]

Is there a significant difference between the development and performance of the game in Campaign(single player) versus Multiplayer? I see a lot of synthetic benchmarks that appear to be using the graphics/textures of the single player game and not the multiplayer. I pose this question because it seems like the Single Player version of the game uses better graphic textures than what the Multiplayer offers (e.g. BF4).

There was a link to a Polish website (I think it was polish) that showed CPU and GPU benchmarks for multiplayer. Most peopel didn't think it was all that reliable, but it did show that CPU scaling was much more pronounced in multiplayer. I didn't think they (EA) used a different texture set or models for multiplayer or anything though... not sure on that one.

The problem is that every single game is different so you can't reliably benchmark multiplayer rounds. In the olden days you could generally record a session of muliplayer on your own dedicated server and then play it back so you could get repeatable results, but since EA seems to hate PC gamers you can't do cool stuff with games anymore.

 

[damric]

So when is part 2? December? There should be a handful of more games to test by then.

 

[-Fran-]

The problem is that every single game is different so you can't reliably benchmark multiplayer rounds. In the olden days you could generally record a session of muliplayer on your own dedicated server and then play it back so you could get repeatable results, but since EA seems to hate PC gamers you can't do cool stuff with games anymore.

If EA was smart enough, they'd develop something to "replicate" MP matches and give that information to reviewers so they can replicate MP runs against that data.

There's no need to make the game be host-able like in the old days if they don't want to (although we all want that back, haha), but it would be nice to have such a tool for these problems when reviewing.

Cheers!

 

[ohim]

www.youtube.com/watch?v=rWSluVC6-v4 - This is how FX 8350 + R9-290 OC handles BF4 in Mantle vs DX 11.1 in multiplayer. Judge by yourselves. Including Air - land - water warfare in both runs in the same server with same settings. Mantle for me is a blessing.

 

[Gillerer]

You should have put the minimun and average values in different graphs, and have DirectX and Mantle in the red and black bars instead. That way it'd be much easier to visually compare the differences in different setups.

 

[cleeve]

You should have put the minimun and average values in different graphs, and have DirectX and Mantle in the red and black bars instead. That way it'd be much easier to visually compare the differences in different setups.

You can differentiate between Mantle and DirectX in the FPS charts. the mantle results are solid colors, while the DirectX results are gradients.

 

[photonboy]

Did you mean less than "2" GB of memory for the BF4 issue? You said BF4 does poorly with cards of less than 4GB which is over 99% of the people running BF4 with Mantle.

 

[cleeve]

Did you mean less than "2" GB of memory for the BF4 issue? You said BF4 does poorly with cards of less than 4GB which is over 99% of the people running BF4 with Mantle.

I meant what I said: less than 4GB. Check the results.

2 GB cards don't appear to show an advantage in BF4 when using Mantle.

 

[west7]

full mantle review i've been waiting for this for months

 

[Martell1977]

So with the memory bug test with the 270X, all the results are pretty much the same, regardless of CPU. Doesn't this indicate an overall GPU bottleneck? If the GPU is the bottleneck and the nature of the performance gains from Mantle, it's not surprising that all the results were pretty much identical.
So it's not that Mantle didn't give a performance boost, it's that the GPU was bottlenecking before the CPU load was a limiting factor, or am I missing something?

 

[memadmax]

DirectX was gonna die after Vista anyways because DirectX had become a "you must upgrade windows to upgrade directx" gimmick, scam, fraud, whatever you want to call it...

I'm glad MS is killing directx but their replacement had BETTER NOT F'ING PULL the same crap that they did with directx...

 

[The_Icon]

I agree, even if Mantle starts disappears slowly , it started the push for low overheard API.

 

[rmpumper]

What I see here is that Mantle is completely useless if you have a high end CPU.

 

[cleeve]

So with the memory bug test with the 270X, all the results are pretty much the same, regardless of CPU. Doesn't this indicate an overall GPU bottleneck

Thats a good point. But if mantle were working properly we should expect to see *something* in the direction we'd hope, but you're right, it just might not be the best scenario for it.

As more games come out we might revisit this issue. Like I mentioned, AMD and Dice have already made some improvements since we last tested, so hopefully it won't take too long before we'll see some obvious mantle advantages in anything other than the 290 series when it comes to BF4.