Are people who buy AMD cards losing on a smooth gameplay experience?

[LigTasm]

Well thats nuts cos its so smooth here it cant be any better unless I have 120Hz.
I've not played any first person game that feels as smooth.

Can you describe how and when it isnt smooth?
Is it smooth some times and not others?

I'll be fair here and say I haven't bothered to install the game since around January so maybe they finally fixed something? I've played lots of the game, not saying Nvidia doesn't have its share of problems (SLI) in Skyrim but I've never had to cap the frame rate at 59 and run in fullscreen windowed mode just for it to not feel like a slideshow either. Normally what I would get is just every few seconds a serious stutter. This has been documented with fcat stuff by websites and discussed in any forum that has a dedicated Skyrim thread, I'm not just making stuff up to attack AMD.

Maybe I'll give it another shot when my 290 Vapor-X comes back from RMA in a few weeks. I'm seriously not trying to bash AMD here, the game engine is awful. The only cards I own right now are AMD, the 290, a 260X and an old faithful 5770. There are games that run like garbage on Nvidia too, although I am glad to hear you have no issues with the game now. I haven't re-installed it in a while, because I usually play the vanilla version until I get a decent save game and then spend hours adding in my 130-odd mods until the game is to my liking.

 

[Unknown-One]

I'll be fair here and say I haven't bothered to install the game since around January so maybe they finally fixed something?

Nah, I just fired it up on my secondary system with an HD 6970, still has horrible judder out-of-the-box.

Hacking the game to run in a border-less window fixes it for single-card users. Wont help crossfire users, though.

Don't need this fix when running the game on a GTX 780, so it still seems to be something specific to Skyrim + AMD.

 

[Nenu]

Maybe its fixed on Hawaii <shrug>
I havent had to apply any fixes, its incredibly smooth without any.
Occasionally when in a cave there will be a judder due to something in view.
Turning away stops it, and it doesnt happen often.

I can live with that, its a much better experience than my old GTX580, so smooooth.
And this is with memory and GPU clocks set to the lowest Afterburner will let me + load of mods running!

In case it matters, the rest of my hardware:
2500K @ 4.5GHz
Gigabyte P67a B3 motherboard
Kingston DDR3 2400 @ 2200MHz
Samsung 840 Pro, game runs from this as well as OS and swap file.
290 running 290x BIOS, AC Xtreme III cooler.

Oh yeah, I'm on Win 7.
I'm not an AMD fan btw, I'm royally pissed with AMD over quite a few things such that my next card will be NVidia.
Skyrim isnt one of my problems though, its an anti problem

 

[KazeoHin]

I've had tons of issues on-and-off with various AMD cards. Skyrim is a clear Nvidia win.

Though, it's really only when I try to tweak the game that issues pop up. When I sold a customer a 7770 to run vanilla Skyrim, it ran like a champ at ultra no AA (FXAA only)

 

[Blacklash]

AMD corrected the latency issues that were happening in outdoor areas on the 7950 in Skyrim around Cat 13.2.

My 7970 plays Skyrim very smoothly the overwhelming majority of the time and it does sometimes slow down in
certain areas in some dungeons. I'm not sure why it does it. I can actually have my character stand still and move
the camera in circles and "feel" the lag. It's not because those areas are "demanding", a regular GTX 660 will sail
though the same areas without issue. The OCed 7970 is capable of running Crysis 3 with VH settings and SMAA
@ 1920x fine, so it's not a matter of lacking muscle.

My 660 runs Skyrim fine with nVidia's 335.23 driver. Any of the latest drivers like 337.50 and 337.88 will cause
it to occasionally pause for up to 3 or 4 seconds when an outdoor area is loading. Any driver older than 335.23 is
fine too, as are modified Quadro drivers.

 

[ccityinstaller]

Nah, I just fired it up on my secondary system with an HD 6970, still has horrible judder out-of-the-box.

Hacking the game to run in a border-less window fixes it for single-card users. Wont help crossfire users, though.

Don't need this fix when running the game on a GTX 780, so it still seems to be something specific to Skyrim + AMD.

Nearly every thread you post involves you mentioning that 6970 you have and how much you hate it blah blah..If the card is so bad, why not get rid of it? You purposely causing yourself stress or something? I seriously don't understand why you haven't ordered an Nvidia card to replace based on at least 20 posts I have read about that card causing you problems.

 

[ITSTHINKING]

I am on the market for a new graphics card guys. My 680 is still pretty solid for most demanding games but I am pretty sure the 2 GB RAM will soon be a bottleneck for most games even at 1080p. Proof: Watch Dogs. I am pretty sure the GPU will hold up but the RAM amount wont..In my country the R9 290X is much cheaper compared to a 780Ti but I am not so sure if I would want to go the AMD path. The reason being that most of the games nowadays and in the future tend to be "The way its meant to be played" titles. In E3 2014, Most of the games seems to integrate Gameworks Nvidia technologies into it. The Division, AC Unity and Arkham Knight are Nvidia titles. Any thoughts guys?

Going from a 290x (asus dcII) to the 780ti in my sig, the first thing I noticed was the smoothness of the gameplay across the board. Vsync on or off.

 

[Remon]

If you look at reviews that test for latency issues you will see AMD still lags behind in this area. This can lead to "hiccups" that will affect how smoothy the game runs.

http://techreport.com/review/26279/amd-radeon-r9-295-x2-graphics-card-reviewed/6




NVIDIA uses hardware frame metering, while AMD uses a software method. This may be why AMD has a latency issue in more games than NVIDIA.

I don't think that site has any credibility left.

 

[wonderfield]

Unless I'm mistaken, it doesn't appear as though AMD ever got around to supporting frame pacing in DX9. In 13.8, they added frame pacing support for a certain subset of cards and for DX10/11. I believe support was to be added in 14.1 in what they referred to as "Phase 2", but I'm not seeing where DX9 frame pacing was confirmed. It's certainly not mentioned in the 14.1 release notes.

 

[Unknown-One]

Nearly every thread you post involves you mentioning that 6970 you have and how much you hate it blah blah..If the card is so bad, why not get rid of it?

Excuse me for personally testing things on the most-recent AMD card I happen to have on-hand...

I even offered up a work-around for the problem rather than just whining about it, if you noticed.

You purposely causing yourself stress or something? I seriously don't understand why you haven't ordered an Nvidia card to replace based on at least 20 posts I have read about that card causing you problems.

For someone paying so close attention to my posts, you somehow missed the part where (even in this thread) I mention how that 6970 is in a secondary system?

Check my signature, my main rig has a GTX 780 in it (which replaced the HD 6970 ages ago).

I keep the 6970 around so I have some kind of AMD comparable I can fire up. It'll eventually get replaced with the GTX 780 I'm using now when I upgrade my main rig to something newer, though.

 

[RamonGTP]

Excuse me for personally testing things on the most-recent AMD card I happen to have on-hand...

I even offered up a work-around for the problem rather than just whining about it, if you noticed.


For someone paying so close attention to my posts, you somehow missed the part where (even in this thread) I mention how that 6970 is in a secondary system?

Check my signature, my main rig has a GTX 780 in it (which replaced the HD 6970 ages ago).

I keep the 6970 around so I have some kind of AMD comparable I can fire up. It'll eventually get replaced with the GTX 780 I'm using now when I upgrade my main rig to something newer, though.

It's amazing to me how misunderstood you are. This must be everyone else fault but your own too right?

 

[Unknown-One]

Seriously RamonGTP? Are you going to crusade for everyone who can't read in every thread now? Totally off-topic, unwarranted, and off-base.

Anyway...

Unless I'm mistaken, it doesn't appear as though AMD ever got around to supporting frame pacing in DX9. In 13.8, they added frame pacing support for a certain subset of cards and for DX10/11. I believe support was to be added in 14.1 in what they referred to as "Phase 2", but I'm not seeing where DX9 frame pacing was confirmed. It's certainly not mentioned in the 14.1 release notes.

Yeah, they mentioned a later driver would bring those fixes to DX9 games, but I never actually heard a release announcement for that feature.

They also never made mention of implementing frame-pacing for single-card users, which makes me wonder if the frame-pacing fixes only show up when using crossfire.

 

[gerardfraser]

R9 290 crossfire seems ok to me

So I tried new enbseries_skyrim_v0254 with SkyRealism - ENB Evolved (DOF turned off,to much blur.)

 

[Nauseous]

i think its kinda the luck of the draw really, either your happy with your setup or UMAD! Seems some people have problems while others do not. But i do tend to think nvidia is a bit better because of driver support.

 

[tonyftw]

Nearly every thread you post involves you mentioning that 6970 you have and how much you hate it blah blah..If the card is so bad, why not get rid of it? You purposely causing yourself stress or something? I seriously don't understand why you haven't ordered an Nvidia card to replace based on at least 20 posts I have read about that card causing you problems.

lol. Who doesn't love owning stressful things?

 

[sirmonkey1985]

Yeah, not sure why that's such a sticking point for them.

AMD implemented all kinds of latency improvements (including major crossfire-specific fixes for frame pacing), and yet I still can't get Skyrim to run without being jerky on my AMD cards

with all the stuff they've done and it still having issues, i'd begin to look at the game being the problem and not drivers.

Seriously RamonGTP? Are you going to crusade for everyone who can't read in every thread now? Totally off-topic, unwarranted, and off-base.

Anyway...

Yeah, they mentioned a later driver would bring those fixes to DX9 games, but I never actually heard a release announcement for that feature.

They also never made mention of implementing frame-pacing for single-card users, which makes me wonder if the frame-pacing fixes only show up when using crossfire.

if i'm not mistaken frame pacing is really only a crossfire/multi-card issue to begin with so i don't think theres a need for it to support single cards.

 

[chameleoneel]

I don't think that site has any credibility left.

Techreport's GPU team are one of, if not the primary reason AMD became aware of the frametime issues they were having. As such, Techreport was the first site to receive a beta driver with preliminary fixes, for frame time issues. * for the sake of completeness, however: the issue was quickly shared amongst staff from various sites in order to more quickly think about how best to approach/present the issues. and monitoring tools presented at a presentation by Nvidia were key in collecting relevant data.

They also never made mention of implementing frame-pacing for single-card users, which makes me wonder if the frame-pacing fixes only show up when using crossfire.

As mentioned, "frame pacing" is specifically for multi-card situations. It's to fix a specific variation of frame time issues, unique to multi-card setups (keeping more than one card in better sync with eachother). The general issue of frame times problems was already being addressed, a few months before AMD released their first frame pacing driver to the public. See: Single card setups were being fixed long before muti-cards were frame paced and the fixes for single cards also inherently helped multi-cards, as well.

 

[techtechie]

so after reading/skimming through this thread here's the deal SLI or CX are a problem, avoid if you can. then it turns into and amd vs. nvidia thing, heres my opinion both cards at their thresholded level will play the game but were dealings with computers here it not going to be perfect because people make them after all and nobody's perfect. so stop bitching about minor glitches and learn to live with some stuff...............

 

[The Mac]

Crossfire framepacing was fixed in hardware for Hawaii, thats why it does better than Tahiti.

Tahiti relies on a software fix in the drivers.

AMD announced this back in October, why are people still scratching their heads?

 

[Unknown-One]

As mentioned, "frame pacing" is specifically for multi-card situations. It's to fix a specific variation of frame time issues, unique to multi-card setups (keeping more than one card in better sync with eachother).

Sure, but you can pace the frames coming out of a single card as well (and I've seen instances of single-cards having variable-enough frame times for this to be worthwhile in specific games).

Never as bad as multi-card, but still there. Both of the single-card setups here still showed enough frame-to-frame variance to be measurable: http://techreport.com/review/21516/inside-the-second-a-new-look-at-game-benchmarking/5

Frame pacing could be used to hammer out that semi-regular sawtooth pattern.

The general issue of frame times problems was already being addressed, a few months before AMD released their first frame pacing driver to the public. See: Single card setups were being fixed long before muti-cards were frame paced and the fixes for single cards also inherently helped multi-cards, as well.

Yup, was already aware of the situation. I mentioned those very fixes a few posts back:

AMD implemented all kinds of latency improvements (including major crossfire-specific fixes for frame pacing)

 

[Relayer]

There is no frame pacing for single GPU from AMD. It is purely multi GPU.

Variations of a few milliseconds is typical with real time rendering. There is no type of frame pacing used with single GPU by AMD. Never has been.

 

[chameleoneel]

Sure, but you can pace the frames coming out of a single card as well (and I've seen instances of single-cards having variable-enough frame times for this to be worthwhile in specific games).

Never as bad as multi-card, but still there. Both of the single-card setups here still showed enough frame-to-frame variance to be measurable: http://techreport.com/review/21516/inside-the-second-a-new-look-at-game-benchmarking/5

Frame pacing could be used to hammer out that semi-regular sawtooth pattern.


Yup, was already aware of the situation. I mentioned those very fixes a few posts back:

I think you are splitting hairs over specific word usage/labels.

AMD made major improvements to frame times, with single cards. It is not called frame pacing, as that moniker implies keeping two cards in pace with eachother.

Additionally, we need to keep in mind that the graphs will look different, depending on what tool we use to guage the frame times. Each tool picks out the data from a different point in the pipeline. None is more correct than the other, so as with any benchmarking, it is useful to use a suite of tools, to guage the overall performance.

Case in point: FCAT (by Nvidia) generally produces more consistant frame times, with overall lower frame times---compared to Fraps (but not always!):

http://techreport.com/review/24553/inside-the-second-with-nvidia-frame-capture-tools/6

but you still see the same general trouble areas.
---------------

that image you linked is from 2011and is a good view at what best case scenario used to look like for AMD, with single cards. Not bad, but not great. They admitted to not paying much attention to frame times, for a long time. The problem went unchecked and got worse with the HD7 series. To the point that could be pretty darn bad.

But, they made huge improvements. I think the first public driver with improvements, was in March 2013. Tech Report had the first private beta with improvements, in January 2013.

Sleeping Dogs isn't the best way to show it, because the game is well coded and runs easily anyway. But even it shows noteable improvement:

Here's Sleeping dogs in February 2013, running on a driver that likely does not yet have the improvements working for Sleeping Dogs. or maybe just the first revision, with improvements:
Fraps data

Here's Sleeping Dogs a few months later (August 2013), with a driver that definitely includes single card improvements for many games (AMD said they initially made a lot of game specific improvements, but also at some point made core changes to the drivers, to benefit across the baord) shows not only much more consistant frames times, but overall lower frame times and MUCH less time above 50ms.

Fraps data

FCAT data

 

[Unknown-One]

I think you are splitting hairs over specific word usage/labels.

What do you mean?

I mentioned latency improvements (which help both single and multi-GPU), and then I also went on to mention the frame-pacing improvements for crossfire specifically. Exactly the same things you mentioned...

There is no frame pacing for single GPU from AMD. It is purely multi GPU.

Never said their was...

Was simply wishing they'd implement it or make it an option for Single-GPU. Not hard to delay a frame slightly so it's more in-line with the previous frame's timing, which is part of what frame-pacing does.

Variations of a few milliseconds is typical with real time rendering.

Right, and I'd rather those variations not be there.

As chameleoneel pointed out above, AMD has already been doing work that helps remove even those small variations. +1 to AMD for that.

There is no type of frame pacing used with single GPU by AMD. Never has been.

Again, never said there was, never said there ever has been.

 

[Remon]

Techreport's GPU team are one of, if not the primary reason AMD became aware of the frametime issues they were having. As such, Techreport was the first site to receive a beta driver with preliminary fixes, for frame time issues. * for the sake of completeness, however: the issue was quickly shared amongst staff from various sites in order to more quickly think about how best to approach/present the issues. and monitoring tools presented at a presentation by Nvidia were key in collecting relevant data.

And? Just go read the idiots that comment on the AMD/Nvidia articles. I can tell by the people that site attracts what kind of bias it has.

 

[chameleoneel]

What do you mean?

I mentioned latency improvements (which help both single and multi-GPU), and then I also went on to mention the frame-pacing improvements for crossfire specifically. Exactly the same things you mentioned...


Never said their was...

Was simply wishing they'd implement it or make it an option for Single-GPU. Not hard to delay a frame slightly so it's more in-line with the previous frame's timing, which is part of what frame-pacing does.

They have made huge strides with single cards. It's not called frame pacing, because with AMD's usage, that implies 2 or more cards. It's useful marketing. checkout techreport's several articles on the issue (or another site which presents frame time data) and you will see that AMD has been doing exactly what you are asking for, with single cards.

 

[chameleoneel]

And? Just go read the idiots that comment on the AMD/Nvidia articles. I can tell by the people that site attracts what kind of bias it has.

haha, what? since when is a site's credibility based on the internet dummies whom post comments? I've never even payed much attention to the comments on Tech Report. Their articles are substantial and well presented.

Tech Report is one of the few sites deeply investigating issues like this. Even using tools created by Nvidia, to test AMD's stuff. It's really great, because it 'checks' these companies and pushes them to improve. There's a reason why Techreport was among the very first to receive AMD's initial beta driver to fix frame time issues on single cards.

Tech Report was the only site that posted initial Mantle tests, which mixed more than one CPU, with more than one GPU. Even [H] only posted numbers with one CPU and one GPU. and their follow up article still only had one CPU. Something for which I criticized them.

Incredibly frustrating that a new API which is known to be focusing on the CPU side for now, was only tested with more than one CPU, by one site, for the debut tests.

They also have detailed articles investigating Nvidia's G-sync.

Techreport is an incredibly good site, as they are one of the few who take a deep, yet understandable look at these new types of tech. Not just posting number, but trying to quantify for us in words, diagrams, videos, etc what this stuff actually means to us in real world use.

 

[Unknown-One]

They have made huge strides with single cards.

Right... I already mentioned that

It's not called frame pacing, because with AMD's usage, that implies 2 or more cards

Frame rate metering (what AMD calls "frame pacing") is a general term that describes minimizing vast differences in frame-time from one frame to the next. Those principles apply to uneven output from a single card just as well as uneven output from multi-card.

Just because AMD has only created a metering implementation for crossfire doesn't mean it couldn't also be implemented for single-gpu systems.

It's useful marketing. checkout techreport's several articles on the issue (or another site which presents frame time data) and you will see that AMD has been doing exactly what you are asking for, with single cards.

I already linked such an article (more than one person in this thread has already linked such articles). I'm already well aware AMD has made latency improvements for single-card (I even made mention of this fact a while ago in this thread).

I'm simply wishing that frame pacing were added as a check-box option for single-card users who want to actively ENFORCE even frame times. I'm not allowed to want additional driver features?

 

[chameleoneel]

I already linked such an article (more than one person in this thread has already linked such articles). I'm already well aware AMD has made latency improvements for single-card (I even made mention of this fact a while ago in this thread).

The article you linked is from 2011, long before AMD's recent push to improve (first private beta driver was in early 2013). But it does support Techreport's pretty excellent investigation into frame times and generally new/different ways to gauge GPU performance.

......

Just because AMD has only created a metering implementation for crossfire doesn't mean it couldn't also be implemented for single-gpu systems.......I'm simply wishing that frame pacing were added as a check-box option for single-card users who want to actively ENFORCE even frame times. I'm not allowed to want additional driver features?

Again, you are missing the point entirely. Frame pacing, as AMD has labeled it, is about getting two or more GPUs in better sync with eachother. Before AMD's frame pacing drivers, their crossfire setups would pretty much spit out frames the millisecond they were ready. Which not only resulted in uneven frame times, it was compounded when you have two cards spitting out parts of frames unevenly, that are supposed to match up for one whole screen frame. With two or more cards doing that together (failing to do that); frame tearing was absolutely horrible. Many frames were barely even being seen at all. Like, slivers of visual frame data. What AMD markets as frame pacing, is more about keeping the GPUs in sync, for more complete screen/frame data, without tearing. Yes, the frame times are being smoothed out as well. But that's already a goal anyway. "Pacing" is in consideration of two or more GPUs.

AMD has been doing exactly what you are wanting, with single cards. They just don't call it frame pacing or metering or anything like that. and it's always on, because a single card doesn't have as many potential questions marks. It's always better, there's no reason to turn it off. There's nothing to add to the driver, it is already added. As I already mentioned, I think you are hung up on specific terminology that AMD is taking some marketing liberties with. Read the newer articles from 2013 and onward and you will see data for what you want.

There might possibly be, with multiple cards, a reason to turn off frame pacing. Although even that is generally set and forget, they probably have the option toggle-able for the sake of what if. Because multiple card setups are a pain.

 

[Unknown-One]

The article you linked is from 2011, long before AMD's recent push to improve (first private beta driver was in early 2013). But it does support Techreport's pretty excellent investigation into frame times and generally new/different ways to gauge GPU performance.

Right, but as I mentioned, there are already other (newer) articles linked in this thread (which I've also read).

Like I said, I'm already aware of the improvements they've made for single-cards.

Again, you are missing the point entirely. Frame pacing, as AMD has labeled it, is specifically about getting two GPUs in better sync with eachother.

No, you missed my point entierly.

I don't care what AMD's current frame-pacing limitations are. I want an implementation made-available for single-card users as well. Simple as that.

Before AMD's frame pacing drivers, their crossfire setups would pretty much spit out frames the millisecond they were ready. Which not only resulted in on screen-tearing, but many frames were barely even being seen at all. Like, slivers of visual frame data.

You're preaching to the choir. I'm already aware of the frame pacing issues AMD has had to resolve for crossfire configurations.

AMD has been doing exactly what you are wanting, with single cards. They just don't call it frame pacing or metering or anything like that.

AMD made "latency improvements" for single cards that have helped, but it's still not what I'm asking for. They went to the root of the problem and made improvements that actually make the games run better and smoother. That's great...but it's still not what I'm asking for.

I'm asking for something in ADDITION to that. The other improvements they've made even make it easier for metering to do its job (because it has less to correct for). What's so hard to understand that I want the option to further smooth framerates by way of metering?

As I already mentioned, I think you are hung up on specific terminology that AMD is taking some marketing liberties with.

I'm not hung up on any terminology.

AMD meters the frames coming out of crossfire configurations.
AMD does not meter the frames coming out of single-card configurations.

I want the ability to enable metering on single-card (even though single-card doesn't need it as badly as multi-card). Simple as that.

There might possibly be, with multiple cards, a reason to turn off frame pacing. Although even that is generally set and forget, they probably have the option toggle-able for the sake of what if. Because multiple card setups are a pain.

I never said I would ever want to turn off frame pacing, so I'm not sure what this is in response to.

 

[dcho]

having owned a 5970 and sworn to never use AMD ever again due to mGPU microstutter, im now getting smoother gameplay on my 290x lightnings (CF) than with my 690 and 2x780's.

 

[Unknown-One]

im now getting smoother gameplay on my 290x lightnings (CF) than with my 690 and 2x780's.

All three of those solutions have hardware framerate metering. What makes the 290x CF config "smoother"?

 

[maboblivion]

290x uses no bridges and the latency is lower in most cases, pcper recent reviews clearly show it.
Even single card, latency have equaled or become better than nvidia and we're not talking mantle.

 

[chameleoneel]

AMD meters the frames coming out of crossfire configurations.
AMD does not meter the frames coming out of single-card configurations.

AMD meters the frames for multi-GPU configurations because each card is rendering a partial frame, which is meant to combine with the other parts from the other cards. What you called "root" improvements make sure each card is performing better and more smoothly than before and then the metering makes sure that all of the parts are delivered at the same time. Because part A might actually take less time to render than part b. So the output is metered, to make sure you get a complete frame. Not a tear, not a sliver of a frame.

Redundant clarity: Each piece of each complete frame is metered with the other pieces. But each complete frame is not strictly metered with the other complete frames, as you are describing. that would actually give a lower overall framerate, which would be seen by some as worse performance. AMD has struck a balance where they have dramatically improved consistency of whole frame times, while delivering the same or sometimes better overall framerates. Yes, this means not every whole frame is delivered at exactly the same time. and on multiple cards, they go a step further and meter the cards to eachother, so the partial frames sync and complete frames are properly displayed on your screen as often as possible.

With single cards, there is no real reason to meter and you can't do perfect, strict metering anyway. It sounds great, but due to the variance in what a GPU might have to render in any given game, it's not really practicle for performance to have perfectly strict metering.

 

[Unknown-One]

290x uses no bridges and the latency is lower in most cases, pcper recent reviews clearly show it.
Even single card, latency have equaled or become better than nvidia and we're not talking mantle.

Link that backs up these claims?

AMD meters the frames for multi-GPU configurations because each card is rendering a partial frame, which is meant to combine with the other parts from the other cards. The "root" improvements make sure each card is performing better and more smoothly than before and then the metering makes sure that all of the parts are delivered at the same time. Because part A might actually take less time to render than part b. So the output is metered, to make sure you get a complete frame. Not a tear, not a sliver of a frame.

Once again, preaching to the choir. I already know how this works on multi-card.

So each piece of each frame is metered with the other pieces. But each complete frame is not metered with the other complete frames, as you are describing.

Already told you, don't care about the limitations of their current setup. I'm asking for a new feature here.

that would actually give a lower overall framerate, which would be seen by some as worse performance.

I was already aware of this implication. There are instances where I already have more-than-enough performance (so the hit wont be noticeable), and I'd prefer to have smooth motion over reduced input latency. Metering would improve smooth motion.

With single cards, there is no reason to metter

What do you mean? Unless single-card can produce perfectly-even frame times, there's reason to meter. Single-card is good, but it's still not perfect.

I personally want this feature to smooth out a few problem-titles (who knows, might even fix Skyrim). I'd also like to have it as an option so that screen-captured recordings look better.

 

[chameleoneel]

I'd also like to have it as an option so that screen-captured recordings look better.

While I'm sure strictly metered frames could help capturing: in my experience, some capturing solutions just capture way better than others.

 

[chameleoneel]

I'd also like to have it as an option so that screen-captured recordings look better.

While I'm sure strictly metered frames could help capturing: in my experience, some capturing solutions just capture way better than others.

Already told you, don't care about the limitations of their current setup. I'm asking for a new feature here.

the point here is that asking for frame pacing on a single card, does not compute. You said that AMD meters frames in crossfire, but not in single GPU setups. So I am again clarifying the terminology and how it is used/marketed. they don't meter whole frames. They meter partial frames. Then, complete frame has the best chance to hit your screen. They "pace" each card with the others. This is not applicable to single GPUs, because there's nothing to pace.

I understand that you want a new feature, to strictly meter whole frames. But your previous request for multi GPU style frame pacing doesn't make sense in AMD context. Because the multi-gpu environment is required for their pacing to exist.

 

[Unknown-One]

the point here is that asking for frame pacing on a single card, does not compute.

What don't you understand?

The frames currently spew out of a single-card un-metered. I want the option to enforce metering on them. Simple as that.

And it's very likely AMD would continue to call the feature "frame pacing," even for single-card users. They never clearly defined that "pacing" refers to pacing the cards rather than pacing the frames (the feature is called "FRAME pacing" after all, not "card-pacing" or "GPU-pacing"...). The frame-metering component of AMD's frame pacing implementation is what I'm after.

You said that AMD meters frames in crossfire, but not in single GPU setups. So I am again clarifying the terminology and how it is used/marketed. they don't meter whole frames.

You can re-explain how it works in crossfire all you want, but I told you, I already know how it works. Doesn't matter here, because I'm not talking about crossfire.

Once again: this is a NEW feature I'm asking for.

 

[maboblivion]

This is the last review pcper for amd 290x lightning. Look at the Frame variance graph

http://www.pcper.com/reviews/Graphi...-Lightning-Graphics-Card-Review/Battlefield-4

http://www.pcper.com/reviews/Graphics-Cards/MSI-Radeon-R9-290X-Lightning-Graphics-Card-Review/Skyrim dx9 doesn t show the stutter fest that crossfire does. No need for metering.

Titan review for Comparison sakes
http://www.pcper.com/reviews/Graphics-Cards/NVIDIA-GeForce-GTX-TITAN-Z-Review/Bioshock-Infinite Look at those Frame times oh lord!

If you have trouble remembering what crossfire looked liked before frame pacing , every article as a reminder of what a mess it was in test setup (only in crossfire/sli articles)
http://www.pcper.com/reviews/Graphics-Cards/NVIDIA-GeForce-GTX-TITAN-Z-Review/Test-Setup


The only other thing that can give better frame times on single card is either a more efficient api, add to that a gsync/free sync monitor that isn't limited between 30~60 hz

 

[Unknown-One]

This is the last review pcper for amd 290x lightning. Look at the Frame variance graph

http://www.pcper.com/reviews/Graphi...-Lightning-Graphics-Card-Review/Battlefield-4

Looks about the same as the 780Ti. Would like to see metering as an option to make it even better.

http://www.pcper.com/reviews/Graphics-Cards/MSI-Radeon-R9-290X-Lightning-Graphics-Card-Review/Skyrim dx9 doesn t show the stutter fest that crossfire does. No need for metering.

Still FAR from perfect. Would like to see metering as an option to make it even better.

And all of my recent experience with Skyrim tells me that it needs metering. It's still terrible for me on my single-card AMD system.

Titan review for Comparison sakes
http://www.pcper.com/reviews/Graphics-Cards/NVIDIA-GeForce-GTX-TITAN-Z-Review/Bioshock-Infinite Look at those Frame times oh lord!

That review only covers multi-card. I'm talking about single-card.

If you have trouble remembering what crossfire looked liked before frame pacing , every article as a reminder of what a mess it was in test setup (only in crossfire/sli articles)
http://www.pcper.com/reviews/Graphics-Cards/NVIDIA-GeForce-GTX-TITAN-Z-Review/Test-Setup

Don't care about crossfire, talking about single-card.

The only other thing that can give better frame times on single card is either a more efficient api, add to that a gsync/free sync monitor that isn't limited between 30~60 hz

That depends on how you define "better"

I'm going for less frame-to-frame variance. Right now, the only way to achieve that on a single card is with an FPS limiter that hammers frame-rate to a specific value. This is a poor solution, though, as it means you have to pick a fairly low value to make sure your frame-rate never drops below the limiter to ensure consistent frame-times.

For example, a FPS limiter actually fixes the horrible stutter I still get in Skyrim by strictly enforcing consistent frame times... unless I get into a spot where my framerate falls below the limiter, in which case, the stutter-fest comes right back. Driver-level metering would solve this.

 

[maboblivion]

On release ,Skyrim didn't have pc optimization flags when compiled.Physics engine closely tied to fps... If stuttering on specific hardware hasn t been fixed by bethesda yet it probably never will.
Other than running the game on ramdisk or modding the config , using ENBoost or capping the fps like you did not awhole lot more to do.

Getting off topic OP.... oops