News - AMD Mantle support coming to GTA V and CoD: AW says report

Quote:

---

VideoCardz has published an exclusive support list with many upcoming AAA titles.

---

Read more.

That's impressive. Having the CryEngine, Frostbyte 3 and Unreal Engine 3 listed in there will add a lot of weight to the AMD campaign.

How much of a difference is Mantle making to people playing current titles at the moment?

GTA V is interesting. One of the biggest advantages to mantle (apparently) is when there is lots of small detailed AI controlled bots in the scene. Perhaps mantle could significantly improve draw distance?

Quote:

---

Originally Posted by KeyboardDemon

That's impressive. Having the CryEngine, Frostbyte 3 and Unreal Engine 3 listed in there will add a lot of weight to the AMD campaign.

How much of a difference is Mantle making to people playing current titles at the moment?

---

I don't have much experience with Mantle. But in BF4, it causes me more problems than anything :( I used it in Thief. But I used Mantle the whole time, so not sure what difference it made :P Not even sure I've used it in any other games.

But there is one big, potential problem I could see with Mantle. If it works out as a really good Api a little further down the road. I think it's very likely that we'll be getting lower spec graphics cards for the money - Less bang for buck (or equal bang for buck when using Mantle). Seeing as we can do more with the power we have. It will still be awesome having a more optimised Api. But while it may seem to squeeze a bit more life out of our current hardware, later on it'll probably just be like nothing much changed, as far as the performance we get for our money is concerned. Hope I'm wrong!

So, as it stands with relatively budget minded gaming ( being my mind set at the moment ) I pose this question.

If someone wanted to get the best performance from a CPU in a budget minded system what would peoples recommendations be:

1) Cheap AMD quadcore - More cores
2) The Intel Pentium anniversary edition overclocked super high. - More single thread performance

It's a single threaded performance vs multithreading. And with Mantle take up being decent it makes you wonder. Where is best to spend money?

Quote:

---

Originally Posted by Jowsey

So, as it stands with relatively budget minded gaming ( being my mind set at the moment ) I pose this question.

If someone wanted to get the best performance from a CPU in a budget minded system what would peoples recommendations be:

1) Cheap AMD quadcore - More cores
2) The Intel Pentium anniversary edition overclocked super high. - More single thread performance

It's a single threaded performance vs multithreading. And with Mantle take up being decent it makes you wonder. Where is best to spend money?

---

It depends on what games they want to play. Some games will be better on the Pentium and others will be better on the AMD chips.

If it were a general purpose gaming system it would hinge on one question:

"Would they be willing to upgrade the CPU at all and if so in the next 18 months??"

If not,then I would go with the AMD CPU,although people should not forget the Haswell Core i3 either.

The Pentium dual core is not well balanced IMHO,and it does not help a number of reviews,are benching it under favrourable(if not slightly interesting) conditions to make it look a tad better.
Even then it does not change the fact that even in many supposedly lightly threaded games a Haswell Core i3 still matches or exceeds an overclocked Pentium dual core in many instances,or simply produces enough framerates and has decent frametimes to give a good performance.

Yet the HT on the Core i3 gives it enough legs to help in more threaded games.

One benchmark which annoys me is Skyrim. Its a third person perspective single player game,yet the review E-PEEN when one CPU is producing 100FPS and one producing 80FPS is just hilarious.

Its even funnier when many players mod the game massively,meaning the graphics card is more a limitation in many cases,and many of the reviews don't test with the most popular mods enabled either.

I increasingly get the impression many reviewers have not actually played the games they benchmark. An example is Crysis3 - parts are lightly threaded in the SP campaign but others will push all 4 cores of a Core i5 to the maximum and even use HT effectively enough for a Core i7 to destroy a Core i5. MP is more of the latter too,and its why I got a Xeon E3 in the first place.

People as a result read these reviews,and get a disjointed view of actual performance in these games,and what is actually needed.

They might find simply upgrading their old Core2 or Phenom II rig with a new graphics card does the job.

Or get an older chip like the Xeon X5650 - intel Hex core, hits 4.2 GHz with all power saving still on and within intel spec voltage - for only £90, a socket 1366 m/board for about same and as big a gfx card as you can afford. Best of all worlds :P

Quote:

---

Originally Posted by zaph0d

Or get an older chip like the Xeon X5650 - intel Hex core, hits 4.2 GHz with all power saving still on and within intel spec voltage - for only £90, a socket 1366 m/board for about same and as big a gfx card as you can afford. Best of all worlds :P

---

That too!! :p

Only thing I would worry about is the motherboards OFC.

Quote:

---

Originally Posted by ZaO

But there is one big, potential problem I could see with Mantle. If it works out as a really good Api a little further down the road. I think it's very likely that we'll be getting lower spec graphics cards for the money - Less bang for buck (or equal bang for buck when using Mantle). Seeing as we can do more with the power we have. It will still be awesome having a more optimised Api. But while it may seem to squeeze a bit more life out of our current hardware, later on it'll probably just be like nothing much changed, as far as the performance we get for our money is concerned. Hope I'm wrong!

---

No, I think you have that backwards.

While there certainly are some system specs where Mantle helps low-end systems, Mantle makes far more sense where the GPU is being held back by the CPU. Scenarios like these:

1. AMD quad core with R9-270X or R9-280
2. Intel i3 with R9-270X or R9-280
3. AMD FX6300 or FX8350 with R9-280X or R9-290X
4. Intel i5 non-K with R9-280X or R9-290X
5. Intel i5/i7 K with crossfired R9-280X, R9-290 or R9-290X
6. AMD FX7350 with crossfired R9-280X, R9-290 or R9-290X
7. i7 LGA2011 with tri-fire R9-290 or R9-290X

The point being: anywhere where the CPU might be holding back the GPUs potential. It does not make sense with something like a fast i7 paired with just one R9-280.

That's why in BF4 it is heavy load multiplayer (which is next to impossible to benchmark) where the big gains are to be found.

Mantle lower the CPU overhead, or rather the CPU spends less time organising the draw calls. It does not make the GPU faster unless the GPU spends time waiting for CPU. What it does do even in GPU limited scenarios is make the frame rates more consistent so better min frames (which are far more important than max frames - just a pity so few review sites realise this).

Of course, since the GPU spends less time waiting on the CPU power consumption may go up as the GPU doesn't get a chance to idle.

All well and good. Can AMD now focus on cooler cards that won't fry the rest of my system. That way there might be a chance I'd opt for them over Nvidia next time and maybe have half a chance of mantle becoming a purchase consideration.

Quote:

---

Originally Posted by kompukare

No, I think you have that backwards.

While there certainly are some system specs where Mantle helps low-end systems, Mantle makes far more sense where the GPU is being held back by the CPU. Scenarios like these:

1. AMD quad core with R9-270X or R9-280
2. Intel i3 with R9-270X or R9-280
3. AMD FX6300 or FX8350 with R9-280X or R9-290X
4. Intel i5 non-K with R9-280X or R9-290X
5. Intel i5/i7 K with crossfired R9-280X, R9-290 or R9-290X
6. AMD FX7350 with crossfired R9-280X, R9-290 or R9-290X
7. i7 LGA2011 with tri-fire R9-290 or R9-290X

The point being: anywhere where the CPU might be holding back the GPUs potential. It does not make sense with something like a fast i7 paired with just one R9-280.

That's why in BF4 it is heavy load multiplayer (which is next to impossible to benchmark) where the big gains are to be found.

Mantle lower the CPU overhead, or rather the CPU spends less time organising the draw calls. It does not make the GPU faster unless the GPU spends time waiting for CPU. What it does do even in GPU limited scenarios is make the frame rates more consistent so better min frames (which are far more important than max frames - just a pity so few review sites realise this).

Of course, since the GPU spends less time waiting on the CPU power consumption may go up as the GPU doesn't get a chance to idle.

---

I understand how mantle helps lower the the Cpu load (in an end user sort of a way), which in turn can benefit the Gpu. But are you saying it does nothing directly for the Gpu's capability?

CIV:BE and Elite: Dangerous are reasons to go AMD for me at least.

Edit -Star Citizen as well.

Quote:

---

Originally Posted by ZaO

I understand how mantle helps lower the the Cpu load (in an end user sort of a way), which in turn can benefit the Gpu. But are you saying it does nothing directly for the Gpu's capability?

---

Well, yes more or less. Mantle doesn't make the GPU faster unless in some scenarios where DirectX is the limit. Of course the current stuff (BF4, Thief, PvsZ) had Mantle added very late in the development. But games which were designed around Mantle are more likely to have more units onscreen, better AI etc. (that is things which use the CPU).

Quote:

---

Originally Posted by imshimaru

All well and good. Can AMD now focus on cooler cards that won't fry the rest of my system. That way there might be a chance I'd opt for them over Nvidia next time and maybe have half a chance of mantle becoming a purchase consideration.

---

While Nvidia's card have a bit better perf/watt, the difference is hardly that great (aside from 750Ti). These are taken from TPU's power chart (TPU measure only the card, not the whole system):

660Ti	137W	vs	270X	122W
770	180W	vs	280X	208W
780	220W	vs	290	263W
780Ti	268W	vs	290X	294W

Some differences for sure, but nothing that would 'fry the rest of my system'. The telling thing is that if you look at TPU's other chart, their 'Performance per Watt' chart, prior to the 750Ti coming out the difference between the worst and the best card 28nm is 81% to 115%. So 115/81 = 1.42 which shows that there isn't any great magical difference. But yes, certainly AMD could improve and that probably means cutting out DP performance in their gaming card like Nvidia do, plus more power gating and some die binning (which AMD don't seem to do much).
http://www.techpowerup.com/reviews/S...i-X_OC/25.html

Those R9 280X figures are for an overclocked Gigabyte card against a stock GTX770. For instance TPU tested this MSI card:

http://www.techpowerup.com/reviews/M...g_6_GB/22.html

Quote:

---

Originally Posted by CAT-THE-FIFTH

Those R9 280X figures are for an overclocked Gigabyte card against a stock GTX770. For instance TPU tested this MSI card:

http://www.techpowerup.com/reviews/M...g_6_GB/22.html

---

Right, changed the figures. But I think a lot of people get confused that because the reference 290X was allowed to go up to 94°C it generated more heat.
Core temperature is not heat output.
A card which consumes 260W and has a core temp of 60°C and one which consumes 260W but has a core temp of 94°C will generate the same amount of heat (obviously different card design may dump more or less heat inside: reference blower type cards may be noisy but they tend to push heat outside the case unlike quieter aftermarket open-air coolers)

Quote:

---

Originally Posted by kompukare

Well, yes more or less. Mantle doesn't make the GPU faster unless in some scenarios where DirectX is the limit. Of course the current stuff (BF4, Thief, PvsZ) had Mantle added very late in the development. But games which were designed around Mantle are more likely to have more units onscreen, better AI etc. (that is things which use the CPU).

---

Hmm. I'm not sure about this. But I can't say I can prove you wrong either. We'll see what happens further down the road :P But either way - a better optimised Api for pc's is very much welcomed :)

I will say as I do a log of video editing I keep considering jumping to an AMD 8 core, but the potential loss on gaming worries me.

But with mantle and directx 12 both showing better utilisation of multiple processor cores jumping AMD might end up being a win win situation for me.

Quote:

---

Originally Posted by ZaO

Hmm. I'm not sure about this. But I can't say I can prove you wrong either. We'll see what happens further down the road :P But either way - a better optimised Api for pc's is very much welcomed :)

---

Well, ok saying it won't make it faster is probably not correct either. A new API could easily do certain things faster so for example a certain routine which DirectX11 has to be done in what might be rather inefficient might be far faster and more efficient in Mantle which after all starts with GCN as it's base, while DirectX11 might have lots legacy bloat.

Since the current Mantle games are probably all quick patches/hacks, maybe later stuff will perform faster. But for now getting rid of the CPU bottleneck is no bad thing.

A lot of the hostile discussion on certain forums seems to miss this. Rather suspect that if Mantle had been Nvidia's idea, a lot of that hostility would not exist. But for these PC Masterrace people don't seem to understand, is that one of the best showcases for Mantle at the moment is a OTT gaming machine 2-4 high-end GPUs.

So a system with an overclocked LGA2011 processor and 3-4 R9-290X playing BF4 is likely to outperform any similar non-Mantle system by quite a bit.

Quote:

---

Originally Posted by kompukare

Well, ok saying it won't make it faster is probably not correct either. A new API could easily do certain things faster so for example a certain routine which DirectX11 has to be done in what might be rather inefficient might be far faster and more efficient in Mantle which after all starts with GCN as it's base, while DirectX11 might have lots legacy bloat.

Since the current Mantle games are probably all quick patches/hacks, maybe later stuff will perform faster. But for now getting rid of the CPU bottleneck is no bad thing.

A lot of the hostile discussion on certain forums seems to miss this. Rather suspect that if Mantle had been Nvidia's idea, a lot of that hostility would not exist. But for these PC Masterrace people don't seem to understand, is that one of the best showcases for Mantle at the moment is a OTT gaming machine 2-4 high-end GPUs.

So a system with an overclocked LGA2011 processor and 3-4 R9-290X playing BF4 is likely to outperform any similar non-Mantle system by quite a bit.

---

Well I will test Bf4 again with my setup once the next expansion is out, and see if it helps. Because my Cpu does get maxed out at times while playing it. But last time I played Bf4 with mantle (and all the times before), it was giving me some terrible slow down issues (I think there was a memory leak issue with Vram). But in fairness, I couldn't launch the game at all after a 1GB patch a couple days ago. And I'd left it on the Direct3D/DirectX setting last time I'd played haha..

Quote:

---

Originally Posted by ZaO

Well I will test Bf4 again with my setup once the next expansion is out, and see if it helps. Because my Cpu does get maxed out at times while playing it. But last time I played Bf4 with mantle (and all the times before), it was giving me some terrible slow down issues (I think there was a memory leak issue with Vram). But in fairness, I couldn't launch the game at all after a 1GB patch a couple days ago. And I'd left it on the Direct3D/DirectX setting last time I'd played haha..

---

Mantle is a bit of double-sided sword. Not unexpectedly, direct-to-the-metal requires more or different work from the developer. That's why all those who were expecting a driver update from AMD to fix crossfire might have been somewhat optimistic.

I guess at some stage the Mantle API will get released and someone will write an article about; I wouldn't be surprised if it turns out that the developers have to do a lot of work they didn't used to have to do under DirectX. Maybe not the Crossfire stuff, but the memory leaks or vram allocation might be something which is nothing to do with the driver.

GTA V could really use Mantle. Getting rid of the CPU bottleneck would be astounding.

Quote:

---

Originally Posted by VijaySathyan

GTA V could really use Mantle. Getting rid of the CPU bottleneck would be astounding.

---

Are you Beta testing it or something?

The amount of engines that are now supporting mantle is nice however I personally plan on waiting a year for mantle to really kick in before I consider a move from Nvidia to AMD. However if mantle starts taking hold it should push Nvidia to incorporate it or push to make a similar technology itself.

It's nice to see mantle being used more in place of direct X, even as a devout Nvidia GPU user. Any API to improve games can only benefit game design and performance across the board.

I hope AMD do bring Mantle over to Linux though - especially for SteamOS. (Ubuntu is my primary platform, Windows is really only used for games that doesn't run on Linux that's it)

ZaO my friend is in the same boat as you are with mantle but I have had no problems since the 14.4 drivers came out a few months back its been smooth as silk @ 1440p.

This might interst people (something I did a couple of months ago to show my clan mantle):

Thought I would see if there is any difference between mantle and dx 11 on my rig using the built in Thief benchmark.

Results

DX 11

Min 27.4
Max 48.8
Avg 33.2

Mantle

Min 38.6
Max 64.7
Avg 46.6

Using the average fps Mantle on my rig is roughly 40% faster than DX. Also bare in mind if you have thief im running 1440p maxed out (very high preset 8* AA settings and AMD 14.4 WQHL drivers before trying to compare results. Rest of my pc specs are in my profile:)

Quote:

---

Originally Posted by dfour

ZaO my friend is in the same boat as you are with mantle but I have had no problems since the 14.4 drivers came out a few months back its been smooth as silk @ 1440p.

This might interst people (something I did a couple of months ago to show my clan mantle):

Thought I would see if there is any difference between mantle and dx 11 on my rig using the built in Thief benchmark.

Results

DX 11

Min 27.4
Max 48.8
Avg 33.2

Mantle

Min 38.6
Max 64.7
Avg 46.6

Using the average fps Mantle on my rig is roughly 40% faster than DX. Also bare in mind if you have thief im running 1440p maxed out (very high preset 8* AA settings and AMD 14.4 WQHL drivers before trying to compare results. Rest of my pc specs are in my profile:)

---

I just read this lastnight: http://www.tomshardware.co.uk/amd-ma...iew-32995.html
Those tests were done in June. So they may not be as relevant now. But the situation has not been so good for BF4. The main benefit has been for those like yourself, with high end graphics cards. And at least at the time of their tests, only those that have at least 4GB of Vram. But Mantle seemed to be working a lot better for me since the newest maps in bf4. Still had some lockups on occasion. But I definitely noticed a difference. Well compared to how it ran on any of the maps before this dlc. Maybe these new maps just run better anyway. I don't know.. It's early days still. But it's pretty cool having an Api designed by Amd. I'm optimistic about it :)

Very interesting article there Zao.

I feel bf4 is hard to benchmark anyway as to get a real feel you would need to have a scenario online with say 32 players that is repeatable constantly to get an accurate take on frame rates etc not just a scripted part of the single player. I dont have a lesser graphics card here capable of playing bf4/thief so cant do any other tests on either myself. I could run in medium settings no/2*AA on thief to see how that changes things if that would be any interest to people.

Quote:

---

Originally Posted by dfour

Very interesting article there Zao.

I feel bf4 is hard to benchmark anyway as to get a real feel you would need to have a scenario online with say 32 players that is repeatable constantly to get an accurate take on frame rates etc not just a scripted part of the single player. I dont have a lesser graphics card here capable of playing bf4/thief so cant do any other tests on either myself. I could run in medium settings no/2*AA on thief to see how that changes things if that would be any interest to people.

---

That's very true. It's the multiplayer part of bf that matters. And that's hard to get reliable benchmarks for. But the issues I've had with mantle seem pretty common place for those that don't have the top tier cards. But it is getting there :)

I'll reserve judgement until I see how Mantle performs in the real world. Somehow I can't see Nvidia getting left behind.