DX12 Performance Discussion And Analysis Thread

[Alessio1989]

I do have any identical GPUs pair to test, but it could be interesting to know what cross node sharing tier support the various architectures, because AFR become interesting in DX12 when the coupling of the resources is at the minimum..

 

[Genotypical]

Ultimately the cards are in the same price bracket so I don't see a problem.

its one way to look at it. But we all know its not the right way. We know a nano is much slower than a fury x and making such a comparison would be dishonest. The nano is not meant to compete with the 980ti. You're not buying a nano if you can fit a 980ti/Fury X in your case (most likely). That might as well have compared 980ti cards only.

We actually do not know how a fury x does against a 980ti above 1080p with the new drivers. One site tested at 1080p with new drivers, saw the 980ti barely ahead and all the other sites trumpeted some supposed switch in performance position. Then these guys test at 1440p, but oh... nano. Never mind they did all their other tests with a Fury X.

The clock rate the nano reached should at least be mentioned. I'd love to know why they used a nano. It stood out as really strange

 

[Deleted member 2197]

its things like these that make so many consumers suspicious of certain sites. I was wondering why they used a fury nano in those tests and went checking the other pages. On those pages they used a Fury X. So they had one. Maybe I missed something?

When the benchmark was released they were very selective who reviewed it ... Sometimes marketing finesse will later bite you in the ass. I don't think Guru3d received either software or hardware at the time the benchmark was released. Guru does now included the benchmark when testing GPU's and was included with their recent Nano test. Though I don't think it's been shown there any variation between AMD's cards in this particular benchmark. Below is a recent test of the R9 390x

 

[Genotypical]

So all they did was reuse old numbers from probably catalyst 15.8. Possibly old numbers for the fury x and the 980ti they aren't reviewing as well. Not malicious, just not thorough and liable to give people wrong ideas.

 

[Deleted member 2197]

I think you'll find this in most reviews, especially if you have to return the cards after the review. I have not yet heard of any tremendous performance increase when using newer drivers from any manufacturer. Most differences you'll likely see is reference vs factory OC'd models, with later reviews starting to use cards that the are inline with what most people purchase.

 

[Freemantle]

its one way to look at it. But we all know its not the right way. We know a nano is much slower than a fury x and making such a comparison would be dishonest. The nano is not meant to compete with the 980ti. You're not buying a nano if you can fit a 980ti/Fury X in your case (most likely). That might as well have compared 980ti cards only.

We actually do not know how a fury x does against a 980ti above 1080p with the new drivers. One site tested at 1080p with new drivers, saw the 980ti barely ahead and all the other sites trumpeted some supposed switch in performance position. Then these guys test at 1440p, but oh... nano. Never mind they did all their other tests with a Fury X.

Frankly it's hard to argue Fury X is any better value either, speculating on case size is just a weak excuse too... but hey ho, this is a single alpha benchmark so it doesn't have to be right. If consumers want to base their opinions on it as you indicate then that is their problem.

The clock rate the nano reached should at least be mentioned. I'd love to know why they used a nano. It stood out as really strange

Presumably it was "up to 1000Mhz". Joking aside.... odd but no great conspiracy I suspect.

As Pharma rightfully suggests, Fiji based cards are fairly mythical even now.

 

[Alessio1989]

I do NOT have any identical GPUs pair to test, but it could be interesting to know what cross node sharing tier support the various architectures, because AFR become interesting in DX12 when the coupling of the resources is at the minimum..

sorry, typo -.-

 

[gamervivek]

Frankly it's hard to argue Fury X is any better value either, speculating on case size is just a weak excuse too... but hey ho, this is a single alpha benchmark so it doesn't have to be right. If consumers want to base their opinions on it as you indicate then that is their problem.

Presumably it was "up to 1000Mhz". Joking aside.... odd but no great conspiracy I suspect.

As Pharma rightfully suggests, Fiji based cards are fairly mythical even now.

Fury X does well if you aren't taking overclocking into account. TPU's latest review puts it at par with a ref. 980Ti at 1440p and 5% faster at 4k.

It'd be interesting if AMD keep up the advantage that they've had in Battlefront beta with Fury X doing 15-25% better than 980Ti and it turns out to be a popular game. Many purchases are made on big games, 7970 lost out to 680 big time in BF3 and that surely hurt its sales despite having more powerful hardware and same vram advantage that 980Ti has.

Newegg quick search seems to have many Fiji cards in stock. I guess the new refrain would be that they are not selling.

Anyway, DX12 on Frostbite would be even more interesting considering their close relationship with AMD.

 

[Razor1]

You mean Starwars Battlefront? Last I remember the ti and fury x was virtually a tie or small edge to the Fury X by like 5% at 4k.

The Frostbite engine is interesting but does it look like AMD has the chops to get engine companies to work for their cards more? So far Oxide hasn't been able to show any appreciable benefit from their marketing stunt.......

No company in their right minds will cut off 80% of their potential market for the possibility to work exclusively for an IHV. This is more so with an engine company because 80% lose of maybe 10 licenses a year is practically everything.

 

[Silent_Buddha]

Presumably next gen Nvidia cards would be better at this stuff. So if developers hold back in order to cater to IHV X or Y then future development gets stunted/delayed. Just because it's currently slow on Nvidia cards, doesn't mean developers shouldn't take advantage of it. Next gen it may end up faster than AMD at this stuff (see tessellation).

Similar to how Nvidia heavily discouraged the use of DX 10.1 in games because their cards didn't have it. Then discouraged use of Dx11 in games until their cards caught up to the 5870. Which just delayed how long it was until games started to use Dx11. Similar with ATI at the time discouraging things from one of the subsets of DX9

Regards,
SB

 

[gamervivek]

You mean Starwars Battlefront? Last I remember the ti and fury x was virtually a tie or small edge to the Fury X by like 5% at 4k.

The Frostbite engine is interesting but does it look like AMD has the chops to get engine companies to work for their cards more? So far Oxide hasn't been able to show any appreciable benefit from their marketing stunt.......

No company in their right minds will cut off 80% of their potential market for the possibility to work exclusively for an IHV. This is more so with an engine company because 80% lose of maybe 10 licenses a year is practically everything.

We had a thread on it.

Anyway, AMD seem to be running away with performance in this game for now, and Fury X seems to be doing rather well in computerbase and gamegpu reviews with a lead of 20% and 27% respectively at 4k over 980Ti. Even the nano is faster in gamegpu review.

https://forum.beyond3d.com/posts/1875582/

Different from other Frosbite games though, BF4 being the big one runs quite better on nvidia's.

As for Oxide, AMD were their usual self,

Saying that Multi-Engine (aka Async Compute) is the root of performance increases on Ashes between DX11 to DX12 on AMD is definitely not true. Most of the performance gains in AMDs case are due to CPU driver head reductions. Async is a modest perf increase relative to that. Weirdly, though there is a marketing deal on Ashes with AMD, they never did ask us to use async compute. Since it was part of D3D12, we just decided to give it a whirl.

http://www.overclock.net/t/1575638/...able-legends-dx12-benchmark/110#post_24475280

 

[Deleted member 2197]

You mean Starwars Battlefront? Last I remember the ti and fury x was virtually a tie or small edge to the Fury X by like 5% at 4k.

I believe you are correct and numbers correlate with the Guru3D had reviewing the game.
http://www.guru3d.com/articles-page...ta-vga-graphics-performance-benchmarks,1.html

 

[Razor1]

We had a thread on it.

Anyway, AMD seem to be running away with performance in this game for now, and Fury X seems to be doing rather well in computerbase and gamegpu reviews with a lead of 20% and 27% respectively at 4k over 980Ti. Even the nano is faster in gamegpu review.

https://forum.beyond3d.com/posts/1875582/

ComputerBase.de doesn't show that percentage at all, ok I see multiplayer 20% , which means nV has work to do on this game for Dx11 driver overhead, possibly but lets look more deeply at the numbers at different resolutions. Off the bat we should expect AMD to have better performance with their drivers at this stage because this is one of AMD's more important games. but even with a cursory look at these numbers something isn't adding up. something is bottle necking not just the 980 ti, or nV cards, they are also bottlenecking the Nano, only 7% faster than the 390X, during multiplayer. And at a lower resolution multiplayer seems to affect the Nano more than the other cards where its only 6% faster than a gtx 980.

I wouldn't take the multiplayer benchmarks as testing graphics performance at all, because many things change in multiplayer that aren't related to graphics but only related to the CPU.

Different from other Frosbite games though, BF4 being the big one runs quite better on nvidia's.

As for Oxide, AMD were their usual self



http://www.overclock.net/t/1575638/...able-legends-dx12-benchmark/110#post_24475280

Interesting finally they stated something relevant, I'm surprised, I would love to juxtapose his first statement to this own, he flip flopped lol. But I think I don't need it to its pretty obvious.

 

[gamervivek]

Multiplayer is the (only?)relevant part for this game. Perhaps nvidia could change it when the game releases. DX12 benches would look multiplayer improvements like Mantle.

As for that oxide dev's flip-flopping, I'm not sure what you think he did so on. In his very first reply, he stated that nvidia were more involved than AMD was.

Certainly I could see how one might see that we are working closer with one hardware vendor then the other, but the numbers don't really bare that out. Since we've started, I think we've had about 3 site visits from NVidia, 3 from AMD, and 2 from Intel ( and 0 from Microsoft, but they never come visit anyone ;(). Nvidia was actually a far more active collaborator over the summer then AMD was, If you judged from email traffic and code-checkins, you'd draw the conclusion we were working closer with Nvidia rather than AMD

http://www.overclock.net/t/1569897/...ingularity-dx12-benchmarks/1200#post_24356995

 

[Ext3h]

As for that oxide dev's flip-flopping, I'm not sure what you think he did so on. In his very first reply, he stated that nvidia were more involved than AMD was.

Better let the Oxide case rest. The full story hasn't been disclosed yet, and you are not going to guess it correctly based on what has been published.

Sorry, I've been asked not to spill the details either.

 

[Razor1]

Multiplayer is the (only?)relevant part for this game. Perhaps nvidia could change it when the game releases. DX12 benches would look multiplayer improvements like Mantle.

Err, how are you going to use scientific methodology when things aren't the same? Magic? Those benchmarks aren't even giving us reasonable results comparing same IHV hardware across different segments, lets add more vendors in there to complicate the matter and call it good? No that isn't acceptable. Its one thing when you are forcing number of players to play in the same server and map, that stresses many things beyond the GPU, its stresses, the cpu, network, server, etc....., those benchmarks show us as a relevant testing for the other parts of the system based on possibly what the graphics card and its drivers are doing, but nothing that we can quantify in a meaningful way "the engine performs better for a certain IHV". Its just not that cut and dry.

As for that oxide dev's flip-flopping, I'm not sure what you think he did so on. In his very first reply, he stated that nvidia were more involved than AMD was.



http://www.overclock.net/t/1569897/...ingularity-dx12-benchmarks/1200#post_24356995

AFAIK, Maxwell doesn't support Async Compute, at least not natively. We disabled it at the request of Nvidia, as it was much slower to try to use it then to not.

Weather or not Async Compute is better or not is subjective, but it definitely does buy some performance on AMD's hardware. Whether it is the right architectural decision for Maxwell, or is even relevant to it's scheduler is hard to say.

http://www.overclock.net/t/1569897/...ingularity-dx12-benchmarks/1210#post_24357053

Sorry it wasn't the first post but....lets not forgot his next post?

 

[CSI PC]

ComputerBase.de doesn't show that percentage at all, ok I see multiplayer 20% , which means nV has work to do on this game for Dx11 driver overhead, possibly but lets look more deeply at the numbers at different resolutions. Off the bat we should expect AMD to have better performance with their drivers at this stage because this is one of AMD's more important games. but even with a cursory look at these numbers something isn't adding up. something is bottle necking not just the 980 ti, or nV cards, they are also bottlenecking the Nano, only 7% faster than the 390X, during multiplayer. And at a lower resolution multiplayer seems to affect the Nano more than the other cards where its only 6% faster than a gtx 980.

I wouldn't take the multiplayer benchmarks as testing graphics performance at all, because many things change in multiplayer that aren't related to graphics but only related to the CPU.




Interesting finally they stated something relevant, I'm surprised, I would love to juxtapose his first statement to this own, he flip flopped lol. But I think I don't need it to its pretty obvious.

Yeah and regarding Battlefront Nvidia 980ti does seem closer and in real world probably a little bit faster than Fury X because for some reason most of these sites still use standard reference NVIDIA cards.
Personally this is not an apple-to-apple comparison with AMD as it is pretty clear NVIDIA provides a greater scope of performance improvement for AIBs with improved coolers-clocking-etc.
PCGamers use AIB cards for their tests showing their clocks and this is my preference as a user will buy a good AIB from either NVIDIA or AMD when they are available.
In their test NVIDIA is very competitive when one starts to use AIB: http://www.pcgameshardware.de/Star-...950/Specials/Beta-Technik-Benchmarks-1173656/
The 390X seems to be the one to watch when comparing though with 980.

Yeah funny how AoTS is now about AMD reduced driver overhead on DX12, where originally NVIDIA had massive performance gains in this area going from DX11 to DX12 with the similar engine in Star Swarm and yet went backwards with AoTS.
Ah well different discussion though and will be interesting just what happened with the development in this area by NVIDIA and the game developers between Star Swarm and AoTS.
Cheers

 

[gamervivek]

Err, how are you going to use scientific methodology when things aren't the same? Magic? Those benchmarks aren't even giving us reasonable results comparing same IHV hardware across different segments, lets add more vendors in there to complicate the matter and call it good? No that isn't acceptable. Its one thing when you are forcing number of players to play in the same server and map, that stresses many things beyond the GPU, its stresses, the cpu, network, server, etc....., those benchmarks show us as a relevant testing for the other parts of the system based on possibly what the graphics card and its drivers are doing, but nothing that we can quantify in a meaningful way "the engine performs better for a certain IHV". Its just not that cut and dry.

Take your concerns to these reviewers and get their responses. It'd be quite the magic you'd see considering these benchmarks are being put out for quite some time and are affecting those who buy the cards.

It's quite acceptable.

http://www.overclock.net/t/1569897/...ingularity-dx12-benchmarks/1210#post_24357053

Sorry it wasn't the first post but....lets not forgot his next post?

And? They don't use async compute on nvidia because it didn't work.In one of his later posts he implied that Maxwell doesn't even do poor man's async compute,

No, we aren't switching between compute to graphics jobs rapidly. The async compute pipe has only a handful of very big jobs. Usually there is no reason to swap between graphics and compute like this, however on GCN I'm told (by console guys) that you can use this as poor-mans async compute. That is, if you do Draw(), Dispatch(), Draw(), Dispatch() the scheduler will start executing things in parallel very efficiently, but I've never tried it. Due to the fact that the fences in D3D12 require a GPU pipeline flush, it's possible that this is a good method to do things on GCN vs async compute in some cases where you'd want fine grained dependencies.

http://www.overclock.net/t/1575062/...rectx-12-tips-for-developers/40#post_24453908

So again, what flip flopping do you think he's doing?

Yeah and regarding Battlefront Nvidia 980ti does seem closer and in real world probably a little bit faster than Fury X because for some reason most of these sites still use standard reference NVIDIA cards.
Personally this is not an apple-to-apple comparison with AMD as it is pretty clear NVIDIA provides a greater scope of performance improvement for AIBs with improved coolers-clocking-etc.
PCGamers use AIB cards for their tests showing their clocks and this is my preference as a user will buy a good AIB from either NVIDIA or AMD when they are available.
In their test NVIDIA is very competitive when one starts to use AIB: http://www.pcgameshardware.de/Star-...950/Specials/Beta-Technik-Benchmarks-1173656/
The 390X seems to be the one to watch when comparing though with 980.

The pcgh review does look better, but it seems to be not multiplayer and at 4k, the overclocked 980Ti is barely faster than a Fury which means that it'd still be slower than a Fury X.

Barring 980Ti lightning I don't see many AIBs 980Tis jumping across a 20% gulf in performance.

 

[Razor1]

Take your concerns to these reviewers and get their responses. It'd be quite the magic you'd see considering these benchmarks are being put out for quite some time and are affecting those who buy the cards.

It's quite acceptable.

Yeah a person that doesn't know the merits of the scientific method and how to read those benchmarks can twist those numbers to mean just about anything, just like you are doing right now.

By no means the benchmarks numbers are incorrect, but to use them as backing to what you have said, hell NO that is not reasonable thinking. Its inaccurate as using raw statistics to prove a point. Doesn't work.

Oh yeah many reviewers are quite aware of my stance on their reviews, just ask Kyle at H, I blasted him here when he went on a triad of how one way of benchmarks is the best way..... that was well before you where posting here I think.....so you might not know that.

And? They don't use async compute on nvidia because it didn't work.In one of his later posts he implied that Maxwell doesn't even do poor man's async compute,

Right the first part that wasn't the part I was talking about . Pass go and collect $200 ring a bell?

It wasn't implied, what I quoted was what he stated. Pretty stupid move, when what was stated was false. I think you really should read this entire thread to see what was really going on.

http://www.overclock.net/t/1575062/...rectx-12-tips-for-developers/40#post_24453908

So again, what flip flopping do you think he's doing?

I want you to read that document again and split it up based on what will work best on different IHV hardware and post that, and lets see where that goes, because I can tell you right now it won't go anywhere.

 

[madyasiwi]

Unreal Engine Explicit Multiadapter:
https://channel9.msdn.com/Blogs/DirectX-Developer-Blog/DirectX-12-Multiadapter-Unreal-Engine-4

This benchmark uses Intel iGPU to do the the post processing (NVIDIA discrete GPU renders everything else). No AFR, so no micro-stuttering or other problems. Pretty much everybody has Intel iGPU in their CPUs right now. Getting that 11% perf boost is a nice addition.

But how much additional input latency? Would gamers trade 10% more FPS for 30% more latency?

If they don't see some visible transitions with SFR, it would interesting to hear the explanation. Whether there is a visual discrepancy between single-card and AFR might be worth investigating, as there shouldn't be pixel-perfect consistency between hardware implementations, much less hardware vendors.

Reviewers need to take note on this. Potential slight flicker/shimmer due to discrepancy in the implementation algorithm of AF, transparency AA, etc. Though AoTS does not like a good example to expose this kind of problems due to its gameplay PoV.

Yeah funny how AoTS is now about AMD reduced driver overhead on DX12, where originally NVIDIA had massive performance gains in this area going from DX11 to DX12 with the similar engine in Star Swarm and yet went backwards with AoTS.
Ah well different discussion though and will be interesting just what happened with the development in this area by NVIDIA and the game developers between Star Swarm and AoTS.
Cheers

Yep, as a benchmark, AoTS is basically nothing much more than glorified CPU overhead test. It is probably what happened when 3DMark API overhad test give AMD+Oxide some ideas. On their blog, Oxide even bragging about having null renderer feature built in it. 3DMark API tests show that NVidia still have some small overhead in DX12 due to the way its driver handles scheduling as opposed to AMD's hardware scheduler. AMD performs better in this pure API overhead tests even without the much hyped Async Compute involved at all. AoTS secret sauce is just to pound the CPU heavily, even on the crop of the creme CPUs, which chokes NVidia driver sheduler and in turns reduce its GPU performance.

StarSwarm on the other hands, had nothing much going on the CPU, thus it is more of a pure high drawcalls bechmarks -- in which current NVidia hardware dominates.

The multi adapter tests further demonstrate that dual Maxwell is obviously severely CPU bottlenecked. It has tighter frame-time delta when compared to the more micro-stutter prone Fury pairs, indicating that its performance is largely governed by how much the CPU can keep up. Also when the resolution increased from 1440p to 4K, the relative FPS performance of Maxwell pair jumps from trailing to matching Furys -- not something anyone would expect in normal situation (i.e. it's usually the other way around).

Which might also explain why mixed Fury X + 980 Ti is faster than 980 Ti + TX, AoTS left the CPU with only enough grunts to somewhat properly feed one NVidia GPU.

This should left a question whether AoTS is a true representation of future DX 12 games, as it totally runs counter to all the DX 12 premises about giving more power to lesser CPUs. Though of course, Intel wouldn't mind.