Lucid Virtual-VSync / Nvidia Adaptive V-Sync

[matrigs]

I'm in the midst of upgrading my PC, and while doing some research i stumbled across these two "ideas":

http://www.lucidlogix.com/technology-vi ... -sync.html

http://www.geforce.com/hardware/technol ... technology

Both don't seem to be very popular as i cannot really find any useful information or extensive tests on them.

The Lucid implementation is part of the "Virtu" Suite which adds a few gimmicks supposed to increase the performance of desktop graphics while using both the integrated motherboard GPU and a dedicated GPU. Sadly it is supposed to be very buggy and unreliable, and needs to be configured separately for every game and application, but the Lucid site shows frames of "Avatar" so i guess this works with movies as well.

The Nvidia implementation seems to not have any sophisticated settings for this option apart from just switching it on in Nvidia control panel.

I'm curious if this might have a benefit for Mame or watching movies. I have to admit that i have no idea how both of these work.

[Ed Oscuro]

Did you see the nVidia G-SYNC thread?

All this stuff is loaded up with buzzwords so I have no idea how it's actually working - but Adaptive V-Sync from nVidia just seems to be a stepping stone to G-SYNC.

I would stay away from the Lucid technology as they admit upfront it's proprietary, and since it doesn't seem to require any special monitor handling it can't offer the specific benefits of G-SYNC.

For G-SYNC I would look at a graphics card and something like one of these monitors.

[BuckoA51]

All adaptive v-sync does is enable v-sync while the PC can maintain 60fps, then turn it off automatically (and thereby allow frame-tear) when there's too much going on for v-synced 60fps to be maintained. It's quite useful since it allows you to have the occasional tear-frame rather than a sudden drop down to 30fps.

It's of no benefit at all for watching movies or anything like that. Movies are better played on a dedicated media streamer like a Dune in my experience.

[Ed Oscuro]

That sounds terrible, because it means you're going to get mush-mouse whenever the action slows down a bit.

G-SYNC is so much better than this.

[matrigs]

Did you see the nVidia G-SYNC thread?

All this stuff is loaded up with buzzwords so I have no idea how it's actually working - but Adaptive V-Sync from nVidia just seems to be a stepping stone to G-SYNC.

I would stay away from the Lucid technology as they admit upfront it's proprietary, and since it doesn't seem to require any special monitor handling it can't offer the specific benefits of G-SYNC.

For G-SYNC I would look at a graphics card and something like one of these monitors.

Dood. Srsly what the hell is wrong with you.

Of course i saw the topic as i have posted in it SEVERAL TIMES and you even APOLOGIZED TO ME at one point for the same stupidly aggressive manner you are showing with your post above AGAIN.

Your wizz-kid-i-know-f*ck-all-attitude is getting on my nerve A LOT.

Furthermore, your post (and several other of yours) adds absolutely nothing relevant to this topic as it is completely assumption based which you state in your very first sentence:

I have no idea how it's actually working

I opened up this new thread on purpose as this is completely unrelated to G-Sync and at least on paper seemed like a good idea.

[BuckoA51]

That sounds terrible, because it means you're going to get mush-mouse whenever the action slows down a bit.

I don't understand what you mean, all you get is a torn frame.

G-SYNC is so much better than this.

Duh, of course! G-sync also requires a special monitor though.

[matrigs]

All adaptive v-sync does is enable v-sync while the PC can maintain 60fps

That sounds a bit too simple. First of all if this enables "normal v-sync" then you would encounter all the lag that comes with it. If that's so, they would just blatantly lie saying that this system feels more responsive than "normal v-sync".

Furthermore, it would feel terrible if you are playing on a setup which plays at 50-70 FPS and v-sync would randomly switch on and off.

Actually, that makes me curious - does a higher framerate (e.g. 100+ FPS on a 60 HZ screen) create the same amount of lag with v-sync on, then a low framerate one (e.g. 50 FPS). Then this implementation BuckoA51 is mentioning above would make at least some sense.

G-SYNC is so much better than this.

Duh, of course! G-sync also requires a special monitor though.

It requires a special and most importantly - an expensive monitor, is unavaillable right now and as far as i understand the system might be game specific like the Lucid system, ruling out a Mame or multimedia use.

[BuckoA51]

What 'lag' do you mean? V-sync doesn't introduce lag by itself, it only lags when you miss the refresh window.

Furthermore, it would feel terrible if you are playing on a setup which plays at 50-70 FPS and v-sync would randomly switch on and off.

Again I don't understand what you mean. If you v-sync to 60fps you will never go to 70fps, if you drop to 50 you will then get tearing until the GPU can push out enough frames to v-sync again.

[matrigs]

V-sync doesn't introduce lag by itself

Now i'm confused. I though that the general consensus is that "v-sync introduces lag".

You will then get tearing until the GPU can push out enough frames to v-sync again.

Again i'm confused. With V-sync on i don't experience any tearing at all when the framerate is below the refresh rate of my monitor. Mame with Dodonpachi @ 56 FPS is a classic example. V-sync on makes the output smooth but introduces lag.

[BuckoA51]

Right, it just twigged to me you might mean input lag after I made the post, let me see if I can explain this a bit better.

V-Sync came about to solve the problem of torn frames when the GPU was drawing them faster than the monitor could display them. If you have a game running at 80fps and your monitor only refreshes at 60, you will get very obvious, torn frames because the GPU will draw a frame while the monitor is in the middle of a refresh.

So, if you vertical sync, you only draw frames at the rate the monitor can refresh, so no torn frames.

However, with this system if the GPU can't keep up with the monitor refresh, you get some quite sharp frame-rate decreases as it tries to stay in sync. Since you can't render half a frame, you drop right down from 60fps (assuming a 60fps display here) to 30, basically the framerate is locked to multiples of 60 to stay in sync.

Now, the other scenario is that your GPU can only render frames at less than the rate of the monitors refresh, so for instance, 50 frames a second rather than 60. In this case, you /might/ see a torn frame, but you are much less likely to notice it. in this case of course you don't want to v-sync since you'd be cutting your max frame rate down to 30fps.

So now do you see the ideal situation is to not let the GPU go above 60fps but do away with the limitation that refresh can only be a multiple of 60?

If you mean input lag with v-sync, then yes, v-sync can cause some input lag in some situations because the GPU is basically holding back the frame until the monitor is ready. When sites talk about how adaptive v-sync makes the game smoother, they mean in games where the GPU can't keep up with the 60pfs requirement at all times.

Hope that clears it up.

[matrigs]

Thanks, that was highly informative.

If you mean input lag with v-sync, then yes, v-sync can cause some input lag in some situations because the GPU is basically holding back the frame until the monitor is ready. When sites talk about how adaptive v-sync makes the game smoother, they mean in games where the GPU can't keep up with the 60pfs requirement at all times.

Well they use specifically "more responsive" instead of "smoother". But that's marketing, i guess.

[Ed Oscuro]

@ Bucko: To clear up my comment, "action slows down" is meant as "the game is rendering at or faster than the monitor's refresh rate." HARDOCP claims that nVidia's technology fixes mouse lag and does that by essentially switching itself off below the monitor's refresh rate, and by limiting frame rendering to the refresh rate (eliminating tear from multiple frames being sent out to the monitor), doing exactly what you describe.

Just to be picky, 30 isn't a multiple of 60. It gets tricky to explain in a mathematically correct and easily understandable way, but I'll try: V-Sync forces the frame output to wait for sync with the refresh rate - so it only outputs at a multiple of "1/your refresh rate." If your refresh rate is 60Hz, that means it only outputs in time steps that are multiples of 1/60th of a second. If the GPU can't create a new frame every 1/60 second, then the frame rate will drop down to however many refresh rate sync opportunities the GPU needs to draw a frame * 1/your refresh rate. i.e., if the GPU could render a frame every 1/59th of a second, then it is taking two frames to sync. If it takes three sync opportunities, then the refresh rate drops down to 1/20th of a second; if it takes four opportunities, then the refresh rate drops to 1/15th of a second.

@ matrigs: I see where you're coming from, but you're still way outta line here. Basically, you have this wrong.

1 You getting mad at me because I honestly forgot you saw the G-Sync topic and just mentioned it is your problem, not mine. There is nothing aggressive there, I promise.

It's also strange that you harbor such deep resentment over this post:

It's true that it won't go below 30 Hz, but I think you missed the point if you're saying that this won't help 24Hz sources.

That shouldn't offend anyone. You seemed to agree with me that you hadn't considered what I mentioned (in the rest of the post), and Endymion also backed up what I said. People say things like that to each other all the time and it's not a problem - please grow a spine. I will try my best not to be offensive but there's only so much I can do to walk on eggshells.

Please also keep in mind that it is annoying when people write things only to have them ignored - you went and directly contradicted Endymion based on your ignorance. How is that not an instance of the same damn behavior you're protesting?

2 My statement that Lucid's technology is proprietary and apparently a losing scheme isn't based on terrible assumptions. I would take that into account when buying any system. My comment about not having enough information to cut through the buzzwords...I don't know why being honest to help the discussion along is a problem.

What wasn't clear at the time we started talking about this is whether this is actually expensive or problematic to set up. I still haven't seen anything that makes Lucid's tech look particularly good, but HARDOCP's article states that it's available for GeForce cards via the 301.24 drivers (and newer), and elsewhere I read that there may be similar features available for AMD graphics.

I would definitely check out Adaptive V-Sync for nVidia cards, and I will also see what I can find about AMD and adaptive V-Sync. For AMD cards, I've read that http://www.radeonpro.info/ has a feature called Dynamic V-Sync which accomplishes the same thing as Adaptive V-Sync. I haven't found any references yet to AMD developing Adaptive V-Sync internally, and I don't know how good this situation is.

nVidia is looking better and better this year. The one thing that AMD has which I don't think nVidia has is getting support for pinned memory in the Dolphin emulator - but that's way off topic.

[BuckoA51]

Just to be picky, 30 isn't a multiple of 60.

Yes you're correct, wrong word, is "factor" the correct word? I never was good at maths.

[Ed Oscuro]

I think so - I'm not up to speed on math terms either! Anyway, the way I've written it there is probably one of the easier to understand ways.

I adapted a post from TechPowerUp forums' administrator W1zzard for what I wrote above (to try to capture the math behind it); in its original form it's even shorter:

basically the card waits for a flag from the display output logic "image redraw complete", and then sends the next frame. if it misses it due to longer rendering time it has to wait until the next "complete" which is one frame later (1/60th + 1/60th = 30 FPS). if it misses that too it has to wait again (1/60th + 1/60th + 1/60th = 20 FPS)

The "flag" he refers to should be just the VBLANK. (It would be interesting to know if it's still present in G-SYNC but that's off-topic for this thread.)

[matrigs]

You went and directly contradicted Endymion based on your ignorance

Hola hola. Ignorance?

Please also keep in mind that it is annoying when people write things only to have them ignored

I was unaware of the obvious fact that you can drive a display with multiplied framerates, which i admitted in the very same post. This hasn't been mentioned anywhere above.

Yours and Endymions consecutive posts further cleared underlined the discussion and the topic was done with.

Back to the discussion:

I read a bit more about the Lucid implementation. It is supposed to make the GPU render as much frames as possible, then discards "unnecessary" ones to match up the screen refresh rate. Now if i understand correctly, that is more or less what triple-buffering is supposed to do?

One thing for sure, my initial question about these patents being useful with mame (games with < 60 fps) / movies has been pretty much answered. They don't, as both require the GPU to be able to render more frames than the screen refresh rate.

[BuckoA51]

I'm by no means an expert but I believe that is how a triple buffering setup could work, and actual implementations vary.

As far as I understand it basic triple buffering adds an extra back buffer so that the PC always has something to draw. Say you have a scene and you render to the back buffer, swap it out to the screen, then render the next frame and it is ready before it's time to refresh. On double buffer setups you would just have to wait, if you have a triple buffer you simply go draw to the other back buffer, when the refresh is ready the PC will then swap in whichever buffer is most recently updated. This helps cut down on input lag since there's always a frame to be refreshing.

Disclaimer - I could have this all wrong

[Ed Oscuro]

Best way to understand triple buffering is just to read this. Where it says "Another method of triple buffering involves synchronizing with the monitor frame rate," it's referring to v-sync + triple buffering.

The one (potential) downside to triple buffering is that it can use much more processing power (if the GPU draws a lot of frames which aren't copied to the front buffer) than other methods. Using a framerate limiter should help this.

Logically you would expect that triple buffering would have one additional frame of latency over double buffering, but this is just rendering ahead. Depending on who you ask it can be called triple buffering, but generally the term is reserved for the DirectX method (to make things less confusing for gamers).