Codemasters builds DX12 support into its EGO Engine 4.0

[fs123]

5% yea Humbugs gtx970 is getting stomped at around average 40% between dx11 and 12.


DX12
Average FR 48.8, GPU TitanX @1409/1652, Normal FR 57.1, Medium FR 48.8, Heavy FR 42.6, CPU 3930k @4.4, Gregster Link
Average FR 47.3, GPU 980 Ti @1484/1974, Normal FR 58.7, Medium FR 46.2, Heavy FR 40.5, CPU 5820k @4.625, andybird123 Link
Average FR 43.4, GPU Fury X @1120/560, Normal FR 49.3, Medium FR 42.1, Heavy FR 39.9, CPU 3770k @4.2, Dygaza Link
Average FR 38.0, GPU Fury X @1050/500, Normal FR 47.2, Medium FR 37.9, Heavy FR 31.9, CPU 5930k @4.375, muziqaz Link
Average FR 36.0, GPU TitanX @1400/1866, Normal FR 45.5, Medium FR 35.6, Heavy FR 30.0, CPU 5930k @4.3, Streetlight Link
Average FR 33.6, GPU 290P @1285/1500, Normal FR 41.0, Medium FR 33.9, Heavy FR 28.3, CPU FX-9590 @5.0, Streetlight Link
Average FR 33.2, GPU 290P @1190/1300, Normal FR 41.7, Medium FR 33.3, Heavy FR 27.6, CPU 4790k @4.7, thebennyboy Link
Average FR 21.7, GPU 970 @1450/1900, Normal FR 27.2, Medium FR 20.8, Heavy FR 18.7, CPU FX-9590 @4.6, humbug Link


DX11
Average FR 39.0, GPU TitanX @1471/2002, Normal FR 53.6, Medium FR 39.2, Heavy FR 30.5, CPU 5960X @4.5, Kaapstad Link
Average FR 22.6, GPU 970 @1450/1900, Normal FR 29.7, Medium FR 22.3, Heavy FR 18.4, CPU FX-9590 @4.6, humbug Link

If this holds true only the top end Mxwell will compete in dx12. This makes no sense to me though but this result looks pretty bad for anything below the 980ti as he's overclocked and still getting wasted.

It still does not change the fact that Maxwell seems to be less effective when scaled down. Fury x can't keep up with the 980ti/Titan X yet the 290p curb stomps a gtx970.

Not sure why you think FuryX is beaten when a stock FuryX is faster than an overclocked titanX when the cpu is the same. In the top 3 results the Fury gets beaten but the discrepency is that the FuryX is in a 4 core/8 thread 3770K system and the Nvidia cards are in 6 core/12 thread systems.

 

[GoogalyMoogaly]

So we're saying that the lower Nvidia 900 series cards don't perform well because of benchmarks in a game written by one of AMD's pet developers?
Didn't they also write the StarSwarm demo to show off AMD tech?

If we based DX11 performance purely off Heaven 4 results you'd have to wonder how AMD got 20% of market share with what would look like a complete disaster of a GPU line-up. Are we just assuming that because this is written by one of AMD's loyal group that it's completely unbiased?
Or is it just that the Titan X and 980Ti do well that mean it's unbiased?

 

[D.P.]

Funny how everyone ignores the Fable legends DX12 benchmarks and look at a proven flawed benchmark that is slower than DX11.

 

[Stu]

While you are older than me i also grew up in the 80's with my first gaming console being the atari 2900. I have played games on just about every platform to exist. My first computer was an Amiga 1200 (waste of time due to no games using the extra power). PC gaming since 2000 and used to high end machine but unlike yourself i don't go the extra mile these days on hardware.

The main point i was making is PC hardware like ours and especially yours is not very common. On the average i would say the ps4 is much more powerful than what most people game on PC wise.

You're making me feel old!!! I started on an Atari 2600, and I started PC gaming in 1986 on an IBM XT 286.

You should check out a Xeon to go in that X58 board... might be a good match for DX12...

http://forums.overclockers.co.uk/showthread.php?t=18618052&page=65

 

[fs123]

So we're saying that the lower Nvidia 900 series cards don't perform well because of benchmarks in a game written by one of AMD's pet developers?
Didn't they also write the StarSwarm demo to show off AMD tech?

Starswarm ran faster on a 980 in DX11 which means the engine was not biased in AMD's favour. It was comparing Mantle to DX at the time and showing off draw call performance.

The reason why a 970 does not perform as well as an AMD 290 is to do with the bandwidth, not because of developers favouring AMD as you Nvidia pet's like to think. AoTS uses lots of drawcalls so performs better on gpu's with more bandwidth, fairly simple to understand I would have thought.

 

[Mauller]

It is not to do with draw calls, if you look at the break down on the benchmark you will see that on the high draw call part it only reaches around 50k - 60k batches. Greater than what DX11 can handle but nothing compared to the API overhead test.

It is due to the cpu being used in the forum benches for the 970. Because AOTS is CPU heavy you can only have a fair comparisson between similar systems.

 

[fs123]

Funny how everyone ignores the Fable legends DX12 benchmarks and look at a proven flawed benchmark that is slower than DX11.

Ok let's have a look at Fable Legends shall we...

http://www.anandtech.com/show/9659/fable-legends-directx-12-benchmark-analysis/2

Apart from the FuryX being slower than 980ti we can see that the other AMD cards are handily beating the Nvidia equivalents. Quite amazing to see the old 7970 and 285 give the much newer gtx 960 a beating. Also don't forget that the gtx 680 was faster than the 7970 when they were released.


Some more resutls from Pcper:

http://www.pcper.com/reviews/Graphics-Cards/Fable-Legends-Benchmark-DX12-Performance-Testing-Continues/Results-1080p-Ultr

And here we see the 390X and R9 380 beating their Nvidia equivalents. The FuryX is lagging behind the 980ti which could be due to a number of reasons such as lower memory clockspeed, drivers, architecture, etc. The 390X doesn't seem to be hindered by having a slower clockspeed than the 980 though.

In the Anandtech benchmarks, The FuryX and the 290X beat the 980Ti in the 720P test so not quite sure whats going on to be honest.

 

[Mauller]

Funny how everyone ignores the Fable legends DX12 benchmarks and look at a proven flawed benchmark that is slower than DX11.

The funny part about that statement is that fable legends has no DX11 mode to compare performance to.

 

[andybird123]

5% yea Humbugs gtx970 is getting stomped at around average 40% between dx11 and 12..

Humbug only had the very old build which is why his results look so bad (which was already pointed out a couple of times in the thread), hence why I went back and looked for review sites that used the same Ashes build and the same CPU on 970 vs 390

 

[andybird123]

Not sure why you think FuryX is beaten when a stock FuryX is faster than an overclocked titanX when the cpu is the same. In the top 3 results the Fury gets beaten but the discrepency is that the FuryX is in a 4 core/8 thread 3770K system and the Nvidia cards are in 6 core/12 thread systems.

Different builds, some people re-ran their tests on the newer build and others didnt - that furyX run is on the new build and the lower TitanX score is on the older slower build

I've run that test on stock vs overclocked 5820k and it makes 1fps difference

 

[GoogalyMoogaly]

Starswarm ran faster on a 980 in DX11 which means the engine was not biased in AMD's favour. It was comparing Mantle to DX at the time and showing off draw call performance.

The reason why a 970 does not perform as well as an AMD 290 is to do with the bandwidth, not because of developers favouring AMD as you Nvidia pet's like to think. AoTS uses lots of drawcalls so performs better on gpu's with more bandwidth, fairly simple to understand I would have thought.

It was faster because of Nvidia's DX11 tweaks, if I recall, which I think was released around the same time (I can't remember exactly but I thought that benchmark was used when the Nvidia driver was released). So it may have been that without that it did favour AMD. Also it used an API that was only available to AMD users, so it seems to me like it favoured AMD. Didn't see any Nvidia tech being used in the demo. As I recall AMD never made Mantle open to everyone (as they said they would) so only the developers approved by AMD got to play with it, developers such as Oxide.

 

[tommybhoy]

So we're saying that the lower Nvidia 900 series cards don't perform well because of benchmarks in a game written by one of AMD's pet developers?

Nope.

It all started when I mentioned it's looking like DX12 drivers aren't ready since the Nv partnered Ark dev stated back in October that they weren't releasing DX12 build until drivers were better.

My logic is based off AMD are seeing huge DX11 to 12 gains, it's looking to me that Nvidia is having problems, considering their shouting from the rooftops that they are fully DX12 compliant and been ready to go on DX12 for ages now.

Then we got this clanger:

Not really. Have you tried AoTS? DX11 is better for Nvidia users over DX12 and DX12 on Nvidia is better than DX12 on AMD. Sorry but those are the facts.

980Ti/TX are faster than FX and that's where the upper hand stops, the rest of the stack it's AMD.

And then it turned into the usual OcUK way.

it used an API that was only available to AMD users, so it seems to me like it favoured AMD. Didn't see any Nvidia tech being used in the demo. As I recall AMD never made Mantle open to everyone (as they said they would) so only the developers approved by AMD got to play with it, developers such as Oxide.

It was 100% designed to show AMD favour, it wasn't a BM, it was an api stress test designed to show the advantage of Mantle so of course it was to shine light on AMD.

 

[fs123]

It was faster because of Nvidia's DX11 tweaks, if I recall, which I think was released around the same time (I can't remember exactly but I thought that benchmark was used when the Nvidia driver was released). So it may have been that without that it did favour AMD. Also it used an API that was only available to AMD users, so it seems to me like it favoured AMD. Didn't see any Nvidia tech being used in the demo. As I recall AMD never made Mantle open to everyone (as they said they would) so only the developers approved by AMD got to play with it, developers such as Oxide.

Why do Nvidia fans get so worked up about Mantle I don't know

I daresay there are many more games that favour NVidia than AMD so why get upset if one or two devs work with AMD? Shall we call every dev who uses Gameworks a pet of Nvidia?

AMD wanted Mantle to be used by everyone so I somehow doubt if they were picking and choosing which devs could use it. Would have been shooting themselves in the foot. The devs who did support it are well known proponents of low level api access. DICE supported it yet I don't see the Battlefield games being slower on Nvidia.Cards.

It was 100% designed to show AMD favour, it wasn't a BM, it was an api stress test designed to show the advantage of Mantle so of course it was to shine light on AMD.

I would say it was more to shine a light on the advantages of a low level api. AMD worked with Oxide to develop a test to highlight the strengths of a low level api which they saw as way to get rid of the api overhead their cards suffer in DX11. The more games used Mantle or similar, the better for AMD in the long run. I'm sure they would still have been pushing Mantle if DX12 hadn't come out.

 

[GoogalyMoogaly]

It was 100% designed to show AMD favour, it wasn't a BM, it was an api stress test designed to show the advantage of Mantle so of course it was to shine light on AMD.

So entirely possible that some/all games, such as this one, by this developer will also favour AMD.

Why do Nvidia fans get so worked up about Mantle I don't know

I daresay there are many more games that favour NVidia than AMD so why get upset if one or two devs work with AMD? Shall we call every dev who uses Gameworks a pet of Nvidia?

AMD wanted Mantle to be used by everyone so I somehow doubt if they were picking and choosing which devs could use it. Would have been shooting themselves in the foot. The devs who did support it are well known proponents of low level api access. DICE supported it yet I don't see the Battlefield games being slower on Nvidia.Cards.

AMD fans get worked up over Gameworks, same thing I guess, except at least AMD users can choose whether they use GameWorks or not in most cases.

Not upset if developers work with AMD, but we're now judging performance of Nvidia cards based on a benchmark by one of these developers. If Ubisoft released a game with a benchmark in that favoured Nvidia cards would that mean AMD cards weren't any good? I suspect there would be a few people that would call into question the validity of the benchmark.
It seems like pretty much any time Nvidia wins a benchmark or does better in a game there's accusations of Nvidia paying off the developer or the review sites. Or GameWorks is specifically coded to hurt AMD performance or the engine over uses tessellation, etc.

But didn't AMD say that Mantle wasn't open to everyone and that people had to apply (or something) to get access? How do developers do that if Mantle isn't openly available?

 

[fs123]

So entirely possible that some/all games, such as this one, by this developer will also favour AMD.



AMD fans get worked up over Gameworks, same thing I guess, except at least AMD users can choose whether they use GameWorks or not in most cases.

Not upset if developers work with AMD, but we're now judging performance of Nvidia cards based on a benchmark by one of these developers. If Ubisoft released a game with a benchmark in that favoured Nvidia cards would that mean AMD cards weren't any good? I suspect there would be a few people that would call into question the validity of the benchmark.
It seems like pretty much any time Nvidia wins a benchmark or does better in a game there's accusations of Nvidia paying off the developer or the review sites. Or GameWorks is specifically coded to hurt AMD performance or the engine over uses tessellation, etc.

But didn't AMD say that Mantle wasn't open to everyone and that people had to apply (or something) to get access? How do developers do that if Mantle isn't openly available?

Mantle is gone so why bother wasting time on it?

DX12 is where it's at now and AoTS shows the same pattern as the Nvidia approved Fable Legends...we can clearly see that the 980TI beats FuryX but we also can see that the rest of the AMD cards basically destroy the Nvidia equivalents.

I don't care if AoTS is disliked or discredited by the Nvidia crew but you cannot deny that Fable Legends shows exactly the same pattern. Essentially old AMD architecture (Hawaii, Tahiti) is beating Nvidia's Maxwell architecture in DX12. Fiji with HBM is lagging for some reason which may or may not be rectified with drivers but it was a stop gap architecture until the die shrink anyway.

But didn't AMD say that Mantle wasn't open to everyone and that people had to apply (or something) to get access? How do developers do that if Mantle isn't openly available?

Now you are just making stuff up. AMD said it was in beta and once it's finished it will be open to anyone. However after DX12 was announced, they ended it instead and gave the remains to Khronos for use in the Vulkan API.

 

[TheRealDeal]

Humbug only had the very old build which is why his results look so bad (which was already pointed out a couple of times in the thread), hence why I went back and looked for review sites that used the same Ashes build and the same CPU on 970 vs 390

In my following posts I did say that the newer version boosted scores. It gave around a 20% boost which would still leave the gtx970 well behind. It looks like the 290P at 1280 core was also run on the older bench as it scores around the same as the 1180 core 290p. Anyhow Nvidia look weak here below the high end and the fable legends bench shows the same to a lesser degree.

 

[beany_bot]

By the time DX12 is in full flow, most will have moved on to newer GPUs, so it is all rather moot anyways, and those who want to keep their current cards will have a better idea of how their GPUs cope

Eh?? Two of my games are already full fledged DX12 while another 6 of them are set to get DX12 patches in the near future. So that makes 80% my steam games DX12 or nearly DX12 already. You make it sound like it's years off? Not only is it not far away. It's already here. Hell my windows is DX12.

 

[tommybhoy]

Why do Nvidia fans get so worked up about Mantle I don't know

If aimed at Googaly, he isn't a Nvidia fan.

So entirely possible that some/all games, such as this one, by this developer will also favour AMD.

It clearly does as the dev refused at request of Nvidia to disable Async for all, why should they make their partner slower by not using DX12 features?

It won't be a one off with Oxide either, I think we are going to find some partnered DX12 titles are going to be finely tuned for AMD/Nvidia's strong areas giving an even bigger advantage.

DX12 changes the landscape, DX11 dev made title, AMD/Nvidia optimise at driver level, DX12 the onus on the devs who tune titles-AMD/Nvidia have less control.

If we are going off todays tie ins, Nvidia are running riot and AMD titles only release on a full moon.

AMD will be ******* themselves.

 

[tommybhoy]

Eh?? Two of my games are already full fledged DX12 while another 6 of them are set to get DX12 patches in the near future.

AOTS and Caffeine?

Which 6?

 

[beany_bot]

AOTS and Caffeine?

Which 6?

AOTS and Squad

Sorry the others are 5 due to get dx12 I believe are. Ark, DayZ, Arma3, Just Cause 3, Star citizen (not played yet)