Battlefield 1 DX12 Benches: Nvidia still king, AMD gaining

P

psyclist

Gawd
Joined
Jan 25, 2005
Messages
844
Hey all,

Just thought id put THIS UP for discussion, Not wanting to start any shit. But there is a definite trend developing in the jump from DX11 to DX12, Nvidia is regressing and AMD progressing in relative terms. 390X beating the 980Ti in all but 4K Ultra! Loving my Hawaii (Bonaire) based cards right about now!
 
Last edited:
No reason why the 980ti is not much faster than the 480. Get off your ass nvidia and fix this
 
bnUFZgx.png
 
DX11 beats DX12...again.

It surprised us that DirectX 11 performance was better on average, especially where you would not expect it. DirectX 12 was supposed to decrease CPU-overhead on lower resolution and therefore improve performance. It is not clear why DirectX 12 performs less well in Battlefield 1. DirectX 12 may enable better and more demanding visual effects.
Click to expand...

All the DX12 hype, so little delivery. If it wasn't for the console version, they could have skipped DX12 entirely.
 
Last edited:
Im looking forward to a deep dive on the DX11 vs DX12 image quality. Its gotta be giving us something better to account for the loss in FPS. I was just surprised (and happy) the 3 year old Hawaii arch has such a strong showing still, that 512bit bus perhaps. I agree its premature, and id wait till a month or so into release to really judge...But still interesting to see.
 
Not sure what's up with the thread title, the DX12 build runs worse for NV so I don't know what exactly they are kings of. The only amusing thing here is that it seems like NV's DX11 performance is still higher than the competing AMD card's performance in DX12
 
So in game testing is showing some different?

or different parts of the level is showing different results even in DX11, which should be noted.
 
Ieldra said:
Not sure what's up with the thread title, the DX12 build runs worse for NV so I don't know what exactly they are kings of. The only amusing thing here is that it seems like NV's DX11 performance is still higher than the competing AMD card's performance in DX12
Click to expand...
Well you see the 1080 wins on pure rendering muscle, its lead is diminished comparatively by the stronger showing of AMD cards. Of course its a well established point that Nvidia are awesome at DX11 optimizations but have yet to show us similar DX12 performance. Both camps drop but Nvidia's is much more pronounced. We are still in the early days of DX12, and transitioning from uarch's built mostly for DX11, im wondering if this is something that will be fixed with drivers or hardware limitations. One thing I am noticing as we go back in Nvidia generations performance falls off much more precipitously relative to AMD with a 280X (7970 originally released in 2011) nipping at the heels of the 780 that was released 1.5 years later and the 970 now behind the 290. The 390X beating the 980Ti, where before it held a 25% lead etc... Thats what I meant by Nvidia is still king but AMD gaining. Im just hoping its not Nvidia artificially gimping older cards to entice their users to upgrade. We shall see how this all plays out in the upcoming months of DX12 releases and as drivers mature
 
Hi, I didn't benchmark DX12 (yet) because both on Nvida and AMD performance by and large is worse (at least in MP) and the (CPU-)frametimes spike. Also, both AMD and Nvidia are working on new drivers (and will hopefully fix some issues I reported) and with some certainty, the day-0-version of BF1 will get patched soon. So I actually did test the best mode possible - which is still DX11, regardless of AMD or Nvidia. I'm pretty sure, that's going to get better soon. Anyhow, if I had benchmarked DX12, I would probably have to bench it (and DX11 on top of that) yet again in the coming days, because new patches, new drivers, new performance...
Click to expand...

Half baked DX12 for now. Surprise!
 
He also tested an MP match while others are using solo, I think.
Sorry I don't know much about Battlefield. That seems to be the jist based on the other reviews.
 
psyclist said:
Well you see the 1080 wins on pure rendering muscle, its lead is diminished comparatively by the stronger showing of AMD cards. Of course its a well established point that Nvidia are awesome at DX11 optimizations but have yet to show us similar DX12 performance. Both camps drop but Nvidia's is much more pronounced. We are still in the early days of DX12, and transitioning from uarch's built mostly for DX11, im wondering if this is something that will be fixed with drivers or hardware limitations. One thing I am noticing as we go back in Nvidia generations performance falls off much more precipitously relative to AMD with a 280X (7970 originally released in 2011) nipping at the heels of the 780 that was released 1.5 years later and the 970 now behind the 290. The 390X beating the 980Ti, where before it held a 25% lead etc... Thats what I meant by Nvidia is still king but AMD gaining. Im just hoping its not Nvidia artificially gimping older cards to entice their users to upgrade. We shall see how this all plays out in the upcoming months of DX12 releases and as drivers mature
Click to expand...

The 390X is beating a 980Ti ?

7bca4ec540698f0ace60bae68cb735c4fb033b2e76d1e51b8e576b07799a0523.jpg


Here is what the reviewer had to say about DX12
Hi, I didn't benchmark DX12 (yet) because both on Nvida and AMD performance by and large is worse (at least in MP) and the (CPU-)frametimes spike. Also, both AMD and Nvidia are working on new drivers (and will hopefully fix some issues I reported) and with some certainty, the day-0-version of BF1 will get patched soon. So I actually did test the best mode possible - which is still DX11, regardless of AMD or Nvidia. I'm pretty sure, that's going to get better soon. Anyhow, if I had benchmarked DX12, I would probably have to bench it (and DX11 on top o
Click to expand...
 
In the original post, read the review, 390X is beating the 980Ti in everything but 4K in DX12...Further down a second review is posted...I can see where this is going so, before this just turns into a shitshow. I was posting it as an interesting review(first one I read) on an upcoming game using a comprehensive list of GPUs both past a present and how they fair in DX12...Not getting into a pissing competition. Lets pull er together
 
DX12 is irrelevant as it currently causes a performance loss for everyone, it's just MUCH larger for Nvidia which puts the AMD cards ahead.
Looking at DX11 exclusively or even DX11 vs DX12, there's not much to see here. Just plenty of opportunity for people to cherry-pick data.

Like I said before, maybe in another thread, these results are weak across the board for AMD when you look at past DICE games (Battlefront, Battlefield 4). It's just been funny to see the classic "aging like fine wine" thrown around.
 
psyclist said:
In the original post, read the review, 390X is beating the 980Ti in everything but 4K in DX12...Further down a second review is posted...I can see where this is going so, before this just turns into a shitshow. I was posting it as an interesting review(first one I read) on an upcoming game using a comprehensive list of GPUs both past a present and how they fair in DX12...Not getting into a pissing competition. Lets pull er together
Click to expand...

Interesting results those. At 1080p they're equal, at 1440p 390x takes a slight lead, at 4K the 980ti is ahead.

Makes no sense to me, seems like DX12 performance is all over the place; seems like DX12 results in inferior CPU side performance - and its not the first time we see this. Deus Ex anyone?
 
I know that I got waaaay worse performance under dx12 on my system. stuttering and spikes/drops it was unplayable. dx11 ran perfect, 1080p ultra and held 60+. the complete opposite from most dx11 vs dx12 experiences ive had.
 
pendragon1 said:
I know that I got waaaay worse performance under dx12 on my system. stuttering and spikes/drops it was unplayable. dx11 ran perfect, 1080p ultra and held 60+. the complete opposite from most dx11 vs dx12 experiences ive had.
Click to expand...

To be fair frostbite was already really good with spreading the load over several cores, so they basically have to match the quality of years of iterations and improvements with a new build
 
Glad to see my 980 is still a viable option. If SLi works, my PC will crush this game. Think I might just sit out Pascal entirely. Only thing tempting me is VR performance.
 
GMcDougal said:
No reason why the 980ti is not much faster than the 480. Get off your ass nvidia and fix this
Click to expand...


Sure there is. Dice devs know what they are doing and how to optimize their games themselves. When that happens with a dev this is the result. When they don't have those skills it is left up to the path least resistance combined with whatever assistance amd or nvidia can provide.

These are the types of devs/games you look at to get the clearest picture of what the hardware can actually do, when the variable of how well the dev can optimize and code is set to hyper competent it's less about a game performing 70% of its potential instead of 85% due to hand holding from nvidia. This is a purer performance case. And not enough of them exist, even from other companies with plenty of resources.
 
Seems that for Nvidia use DX11. AMD just use DX12. We should know why by now.

 
Quite the loss for AMD.
They need DX12 just to hit MoE with Nvidia in DX11, plus their massive advantage from Battlefront has entirely evaporated. I guess we could chalk that up to more mature Nvidia drivers in Frostbite but as we know, Nvidia GPUs age like moldy cheese of course.

The one ray of hope I see here is that it feeds the "Nvidia sucks at DX12" meme even though DX12 doesn't offer anything particularly useful here. Not sure what's going on with the minimums, I guess that's a separate problem that needs to be examined.

I just keep getting reminded, over and over, about a post I made where I said DX12 was a blight for the majority of PC gamers as it leads to performance loss and other problems. I took a lot of criticism for that but after seeing ROTTR, Mankind Divided, Quantum Break, and now this, I feel a little bit vindicated.
 
Last edited:
TaintedSquirrel said:
Quite the loss for AMD.
They need DX12 just to hit MoE with Nvidia in DX11, plus their massive advantage from Battlefront has entirely evaporated. I guess we could chalk that up to more mature Nvidia drivers in Frostbite but as we know, Nvidia GPUs age like moldy cheese of course.

The one ray of hope I see here is that it feeds the "Nvidia sucks at DX12" meme even though DX12 doesn't offer anything particularly useful here. Not sure what's going on with the minimums, I guess that's a separate problem that needs to be examined.

I just keep getting reminded, over and over, about a post I made where I said DX12 was a blight for the majority of PC gamers as it leads to performance loss and other problems. I took a lot of criticism for that but after seeing ROTTR, Mankind Divided, and now this, I feel a little bit vindicated.
Click to expand...


Well I wouldn't say its a blight its just going to take time for developers to learn what it all entails.
 
TaintedSquirrel said:
Quite the loss for AMD.
They need DX12 just to hit MoE with Nvidia in DX11, plus their massive advantage from Battlefront has entirely evaporated. I guess we could chalk that up to more mature Nvidia drivers in Frostbite but as we know, Nvidia GPUs age like moldy cheese of course.

The one ray of hope I see here is that it feeds the "Nvidia sucks at DX12" meme even though DX12 doesn't offer anything particularly useful here. Not sure what's going on with the minimums, I guess that's a separate problem that needs to be examined.

I just keep getting reminded, over and over, about a post I made where I said DX12 was a blight for the majority of PC gamers as it leads to performance loss and other problems. I took a lot of criticism for that but after seeing ROTTR, Mankind Divided, and now this, I feel a little bit vindicated.
Click to expand...

Does digital foundry actually post averages for their benchmark runs? Watching a video is nice and all but at the end of the day I have no guarantee I'm looking at the same exact scene being rendered on both cards when I pause the video.

I find it just odd that some websites report large gains (up to 15%) for amd cards and PCGH, gamegpu, oc3d all report basically the same performance
 
Ieldra said:
I find it just odd that some websites report large gains (up to 15%) for amd cards and PCGH, gamegpu, oc3d all report basically the same performance
Click to expand...
Probably depends on test scene. Maybe CPU. The game doesn't have an in-game benchmark.

I pulled some GameGPU benches to illustrate my previous post:

Iprj3Se.png
 
I would assume that it depends on the CPU used also. DX12 works fine for me in games. Maybe it is broken on other systems?

/shrug
 
TaintedSquirrel said:
Quite the loss for AMD.
They need DX12 just to hit MoE with Nvidia in DX11, plus their massive advantage from Battlefront has entirely evaporated. I guess we could chalk that up to more mature Nvidia drivers in Frostbite but as we know, Nvidia GPUs age like moldy cheese of course.

The one ray of hope I see here is that it feeds the "Nvidia sucks at DX12" meme even though DX12 doesn't offer anything particularly useful here. Not sure what's going on with the minimums, I guess that's a separate problem that needs to be examined.

I just keep getting reminded, over and over, about a post I made where I said DX12 was a blight for the majority of PC gamers as it leads to performance loss and other problems. I took a lot of criticism for that but after seeing ROTTR, Mankind Divided, Quantum Break, and now this, I feel a little bit vindicated.
Click to expand...

This is pretty much exactly how I feel with DX12 vs DX11 argument. It's a moot argument.

It's well and good to say whether A is better at DX12 or B is at DX11, at the end of the day, the only actual useful comparison is 'best' vs 'best', because that's just what people will use if they want the best performance out of their system.

As of now, the best vs best is Green with DX11 vs Red with DX12, and more often than not, with similar cards, the two sides are still trading blows with one another, but by and large, the actual hierarchy is still retained (EG what is better is still better).

Even with GoW4, so far the only game we have seen that actively require Windows 10 to run (I can't remember if it also runs on DX12), the Green/red gap is still going strong.

While AMD cards are better equipped for the 'future', the 'future' isn't here yet.

As a certain commander quipped, "Your ... future needs to hurry up an' get here or there's gonna be no one left to see it".
 
cageymaru said:
Seems that for Nvidia use DX11. AMD just use DX12. We should know why by now.

Click to expand...

Nice performance of the RX 480 over the 1060 on a rather big title. From my experience DX 12 like in Deus Ex MD is better on the Nano then DX 11 while the other way around for the 1070. Tomb Raider the opposite with the 1070 running better with DX 12 and DX 11 for the Nano. It will be interesting as more titles hit that have big developers behind it how this shakes out.
 
You must log in or register to reply here.
Back
Top