When I first read this post, it was on the same day that AMD released their Radeon Software Crimson Edition 16.10.1 drivers, although it was apparently posted the day prior. As a result, I thought that their reference to 16.9.1 was a typo, but it apparently wasn't. These changes have been in the driver for a month, at least internally, but it's unclear how much it was enabled until today. (The Scott Wasson video suggests 16.10.1.) It would have been nice to see it on their release notes as a new feature, but at least they made up for it with a blog post and a video.
If you don't recognize him, Scott Wasson used to run The Tech Report, and he shared notes with Ryan while we were developing our Frame Rating testing methodology. He was focused on benchmarking GPUs by frame time, rather than frame rate, because the number of frames that the user sees means less than how smooth the animation they present is. Our sites diverged on implementation, though, as The Tech Report focused on software, while Ryan determined that capturing and analyzing output frames, intercepted between the GPU and the monitor, would tell a more complete story. Regardless, Scott Wasson left his site to work for AMD last year, with the intent to lead User Experience.
We're now seeing AMD announce frame pacing for DirectX 12 Multi-GPU.
This feature particularly interesting, because, depending on the multi-adapter mode, a lot of that control should be in the hands of the game developers. It seems like the three titles they announced, 3D Mark: Time Spy, Rise of the Tomb Raider, and Total War: Warhammer, would be using implicit linked multi-adapter, which basically maps to CrossFire. I'd be interested to see if they can affect this in explicit mode via driver updates as well, but we'll need to wait and see for that (and there isn't many explicit mode titles anyway — basically just Ashes of the Singularity for now).
If you're interested to see how multi-GPU load-balancing works, we published an animation a little over a month ago that explains three different algorithms, and how explicit APIs differ from OpenGL and DirectX 11. It is also embedded above.
Very happy to see this. Now
Very happy to see this. Now all we need is for more DX12 titles to get launched. It is good to see AMD fixing the seemingly minor issues to improve customer experience.
My biggest hope is that by doing this and creating a competitive high-end product (vega 10 and zen) during the first quarter next year, both Intel and Nvidia will drop their prices.
The pricing on high end GPUs has become ridiculous…
Scott, what are the chances of something like this happening for DX11 titles?
Hopefully no DX12 titles
Hopefully no DX12 titles launch. More vendor lockin isn’t something I hope for.
True, but how long is AMD
True, but how long is AMD expected to last if they try to make everything open source or non-vendor specific? Nvidia surely hasn’t let up (Gameworks, G-sync, and others I don’t even know about). The way I see it is that if both companies have exclusive features, one of the major determining factors of which card to buy will be price and if this type of competition drives prices down, is it really such as bad thing?
And besides, isn’t this what Nvidia have been doing all along? If it works so well for Nvidia then I hope AMD finds more advantages such as DX12.
Frame pacing in DirectX 11
Frame pacing in DirectX 11 titles was implemented years ago, after we published our Frame Rating benchmark.
Cool
Cool
NVidia actually has HARDWARE
NVidia actually has HARDWARE inside their GPU’s to help with frame time analysis.
Smooth gaming is a very complicated issue, which requires all parties involved to have things working well (game engine, game, video drivers, DX11/12/Vulkan, OS, GPU and other hardware).
Both hardware and software
Both hardware and software capture is valid. Hardware captures real frame times which software may fail at, but hardware capture can fail at telling you if the information in the frame is smooth.
For example you render two frames within 1ms. If you have frame 1 pop up at 1ms and frame 2 pop up at 2ms on the monitor but frame three shows up in 50ms that is crap. But having the game render frame 1 and 2 within 1ms but then having the GPU simply delay frame 2 by 30ms is just as bad because then it’s just old information and the animation is jerky.
If you want a complete picture you’re going to have to look at both ends.
We get around this specific
We get around this specific issue by turning VSync off and measuring how much of a frame any given frame takes up. This gives us sub-frame timing precision, because it tells us precisely when the flip occured within the time a frame scans out. This also helped us find issues where drivers were temporarily drawing old frames under certain weird situations.
(As you'd expect, variable refresh features screw this hack up big time.)
That said, software could help if we could see within the game engine, to see what game time each frame corresponds to. That cannot be implemented at the driver (or API hook) level, though. It needs to be given actual information from the game engine.
Speaking of Multi GPU set
Speaking of Multi GPU set ups, I guess not enough people paid attention to the rx 480 in crossfire vs gtx 1080 presentation AMD did, as it doesn’t seem like a lot of people are buying the RX 400 series relative to the new pascal cards that nvidia have released. If you have a quick look at the latest steam hardware survey http://store.steampowered.com/hwsurvey/videocard/ the RX 400 series isn’t even appearing on the charts yet. Even the 1060 which was released after the RX 480 launched is already on there!
Oh don’t be petty…
Oh don't be petty…
Speaking of petty, Scott can
Speaking of petty, Scott can you ask Josh when his review of the CSL Elite racing wheel is going to be posted? He wrote “My review of the Fanatec setup will be posted here this next week” on the 27/08/16
Scott W is a fucking faggot,
Scott W is a fucking faggot, I’ve known him for years fuck that loser.
Since pcper insists on the
Since pcper insists on the use of ‘anonymous,’ you are free to be as a big of a dickhead as you are with impunity (without punishment). When I am a dickhead, I take credit for it at least.
You’re much better at it as
You're much better at it as well, and put more effort into it when you do. 😉
Thanks Jeremy. No matter what
Thanks Jeremy. No matter what I say, I will always adore you. If that makes me a ‘faggot’ as ‘anonymous’ says above, so be it.
Scott, the only inappropriate stuff going on here is attributable to ‘anonymous,’ don’t you think?
Jesus you’re a faggot. I bet
Jesus you’re a faggot. I bet your wife is fat and your kids are gay. Like legit homos.
… and here boys and girls,
… and here boys and girls, we have a wonderful example of why parental involvment in a child's life is very important. It might not help them grow into better people but at least they would pick up more interesting insults.
There is nothing wrong with
There is nothing wrong with being gay. No one will like you any less than they already do. So stop fighting it and come out of the closet. As in most cases of extreme homophobia, the homophobic person is gay. Often they don’t even realize it until they realize they are projecting their hate of themselves onto others. If you are gay, I hope you have something to bring to the party. Most gays are fun and colorful people who help to make this world less boring. You are probably just a white trash gay person with nothing to offer, but we can hope. Let us know when you come out. BTW, I did my second year of residency in Psychiatry at the U of New Mexico, so I have some experience in these matters.
no, you are still anonymous.
no, you are still anonymous. None of the readers have any idea of what your real name is. Anyone of us, including you, could create loads of "registered accounts" and we would all still not know what anyone's real name is.
New to the internet, eh?
New to the internet, eh?
You are not me and using my
You are not me and using my screen name is dishonest and unethical. Guess it is the fault of the system for allowing this. You are just exploiting a bug. Please, pcper, fix this bug.
This guy must be hurting
This guy must be hurting maybe someone used a barbwire condom on him filled with iodine sure is sour in almost everything posted.
AMD is a big joke some time.
AMD is a big joke some time. They try to keep up with Nvidia but they can`t. I use HD 6970 and R9 290 just to see how AMD cards work not to be apart of this. Comparative with Nvidia cards, AMD works very hot and very power hunger and GPUs throttle a lot. Neither with the RX 480 they could not get rid of these problems. But they are on the right way comparative with R9 200/300/Fury series. But the drivers are much behind versus Nvidia drivers.
Firstly, what the hell are
Firstly, what the hell are you rambling on about?
Secondly, please have the courtesy to post under a decent pseudonym and not hide behind Anonymous and thirdly, AMDs drivers have been at least on par if not better recently, than Nvidia’s have.
Since Crimson they have been on top form with game ready drivers (The latest is already out for Mafia and GOW4 which release on 4th and 11th Oct respectively) and have improved massively with patches and fixes. Bad AMD drivers are a myth perpetuated by Nvidia users for too many years. Nvidia and AMD have released their fair share of poor drivers and have also released good ones on both sides of the graphic divide. Both are no better or worse than each other.
No.
AMD has never improved
No.
AMD has never improved their DX11 drivers to compare with NVidia in terms of CPU usage.
It is very significant in some games particularly when the CPU is relatively weak.
Most benchmarks use a high-end Intel CPU so this issue goes largely unnoticed.
AMD continues to shine the spotlight on DX12/Vulkan where they can perform well, but frankly that’s not much of the current gaming market.
I hate to say this, but it
I hate to say this, but it seems a marketing item, I feel AMD will release dual chip cards to be competitive, so this needs to be “fixed” to help move them.
I would like to see how these new drivers of AFR compare to a single gpu, in frametime and latency.
Will never be better than
Will never be better than single high performance GPU.
Unless AMD really go to work
Unless AMD really go to work on drivers for MGPU, which would do them the world of good in the lower segment (i.e. 2 x RX480 would be good if the drivers were top notch as they could beat a 1070 and get within a whisker of a 1080 and probably beat a 1080 in DX12/Vulkan).
I would like to think that this would be a good target for AMD to aim at and if they get it right could be a winner in the long term.
Mind you this would only target a very small percentage of owners and that is why both AMD and Nvidia do not put more work into SLI/Crossfire drivers.
Exactly
Exactly
Multi-GPU will be ideal but
Multi-GPU will be ideal but not for several years. Specifically VULKAN with SFR (Split Frame Rendering) with both GPU’s working on the same frame.
Explicit mode multi-adaptor
Explicit mode multi-adaptor is the way to go. As with explicit mode multi-adaptor any and all GPU plugged in to a PC will become available for gaming graphics/physics. With explicit multi-adaptor the Integrated GPUs can be of some use for gaming compute acceleration while the discrete GPU/s do mostly the graphics. Vulkan should be using this also so the games developers can get at the GPU/s as a resource for both gaming graphics and gaming compute. No more will that available GPU power go to waste. It will take time but the games developers will be the ones to do this and take advantage of the GPU for all sorts of gaming acceleration!
how it was with old times ATi
how it was with old times ATi Hercules [3D prophet] drivers? they where just repacked or did they modified them?