Is transmission time identical between 60 and 144 at same res?
The Future of Melee: How fast can Dolphin accept inputs? What is the refresh rate?
|
02-04-2015, 04:40 PM
02-04-2015, 05:19 PM
So if buffered, the screen takes 1 transmission time to be transmitted, then added lag on top of that.
If not buffered, the screen takes an average of 0.5 transmissions plus lag. (Do any LCDs display on-the-go without buffering? I doubt it.) Are usual response time figures from fully buffered to on-screen, or transmission start to on-screen? (I think that the more useful measurement is begin-to-end.) 02-04-2015, 07:31 PM
Oh guys, too much smattering in here... Why don't you focus on the latency parts which we can touch?
We can't change both the USB polling and the monitor latency (but buying useful hardware), but we can touch the emulation time. Did anybody of you try to hack in a "sleep" after every frame swap? So the emulated 16ms can be reduced a lot, if the CPU is fast enough. I also haven't seen anything about GPU latency: With a perfect fast CPU, the GPU will start to render when the CPU queues the swapping call. So higher IRs always have a higher latency, also if still on fullspeed. About the LCD delay: Do you really think the GPU will stream the data continuously at exactly 60Hz? If the HDMI/DP connection supports faster modes, there is no need to slow it down. The same apply for the monitor itself. Did anyone test if reducing the monitor resolution also decrease the latency? btw2: LCD controllers are built to be cheap, caching is expensive. So they try really hard to flush as fast as they can. 02-04-2015, 08:54 PM
(02-04-2015, 07:31 PM)degasus Wrote: Oh guys, too much smattering in here... Why don't you focus on the latency parts which we can touch? Where does exclusive fullscreen fall into this? When JMC said: [color=#000000]"With exclusive fullscreen, we push to the monitor without dropping/losing any frames. The lower latency the monitor, the sooner you see the frame, no ifs, ands, or buts."[/color] [color=#000000]Is this relevant to what you were asking? [/color] As far as resolution goes, wouldn't it only be a marginal reduction because the gpu is doing a little less work? 02-04-2015, 10:19 PM
Without exclusive fullscreen, the GPU won't use the final image of dolphin directly for scanout. Instead it would be pushed to your window manager which use it with all other images from other application to generate the final image. So it may be a very small overhead if there is no other application and if the window manager is explicitly optimized for this case. But I think you can just add another frame of latency.
02-09-2015, 04:09 AM
So I have a question.
Basically when the Wii/Cube hardware renders an image, there is no screen tearing whatsoever when FPS drops. It makes me assume there is some sort of wait for vsync going on in the back. Also, there are definitely some frame buffering going on, fully evident by reflection and mirror effects in Pokemon Stadium and Fountain of Dreams in Melee. My question is, how much latency is present on the actual Cube/Wii hardware by use of vsync and the framebuffer as opposed to having said features disabled in Dolphin. Is it a full frame or just memory copy time? 02-09-2015, 07:31 AM
There is no memcpy as this "copy" is done by the GPU. But it is still a waiting for the GPU. As it's also very hard to predict ow long the frame will take, so it'll likeky start as early as possible.
But the question is a bit wrong, I guess you're talking about the latency between the CPU start rendering the frame to it's displayed on the monitor. So the rendering time itself does also matter. I haven't verified it, but so I guess it's a complete frame. Idling the GPU to reduce the latency is just too wasted. 02-16-2015, 10:09 AM
Some people in the CSGO reddit likes 120 Hz lightboosted over 144 Hz and I was wondering if it would be an option for melee.
http://www.blurbusters.com/wp-content/up...-graph.png The chart above shows that 144 Hz has 6.9ms of motion blur whereas with 120 Hz lightboost could range from 2.4ms to 1.4ms depending on the brightness. http://www.blurbusters.com/zero-motion-blur/lightboost/ This page says "LightBoost provides gaming with a CRT-like clarity, with zero motion blur" which I think might be desirable for melee. 02-24-2015, 11:54 AM
So does this mean that development of the adapter will continue?
Windows 7 x64 (SP1)
nVidia GeForce GTX 760 (Gigabyte GV-N760OC-2GD REV2.0) 2GB GDDR5 @ 1100 MHz (6020 MHz memory) Intel Core i7-4770 @ 3.90GHz 8GB G.Skill Sniper Low Voltage edition, Dual Channel @ 1600MHz 9-9-9-24 |
« Next Oldest | Next Newest »
|
Users browsing this thread: 1 Guest(s)