Answers and Thoughts
Well, did you write down your answers? Here is the key.
-
60 FPS vs 30 FPS comparison
- 60 FPS is on the left, 30 FPS on the right
-
60 FPS vs Vsync
- 60 FPS is on the left, Vsync is on the right
-
30 FPS vs Vsync
- Vsync is on the left, 30 FPS is on the right
How did you do? I think most users will be able to get the first two comparison without much work and only the third one that compares a static 30 FPS animation and a variable frame rate animation will be difficult.
What does this really mean based on our previous Frame Rating data sets, AMD's CrossFire issues and user experiences? First, not everyone can tell the difference between the videos we have created here and even some that can will prefer the Vsync options over the 30 FPS reference video we created. Because of that, there is no single fixed answer to the problems we are working on. At least not yet.
Some reps from AMD told me at a meeting during GDC that there were no animation issues once you enabled Vsync and that the constant shift from 16 ms to 33 ms times were imperceptible. I think our videos here have quite definitely discounted that theory. There can still be a debate of how much that difference affects gamers, but there shouldn't be any more debate on if it exists.
And of course, enabling Vsync results in a situation where the frame rate does NOT dip below 60 FPS, as we did with our reference videos used in this article, will not result in animation inconsistencies. Gamers, gaming PCs and game settings that result in a frame rate under 60 Hz will see frame rate bouncing like we are seeing in our Battlefield 3 and Sleeping Dogs demonstrations today. In those cases, the debate about the value of Vertical Sync will continue.
I am staunch believer that a smooth animation is just as important as the frame rate of said animation, and that is why I have invested so much time and money in the idea of capture-based performance testing and the Frame Rating articles up to this point. We have created a great discussion amongst gamers and enthusiasts alike, many of whom where not aware that problems like this even existed.
Other Vsync Considerations
The second big concern after animation consistency when enabling Vsync is the problem of input latency. Input latency is the time between an input being given by the user (mouse movement, click, keyboard press) and when the result of that action actually displays on the screen. There are lots of factors that come into play here including the game engine time, the rendering time and display time. With Vsync enabled on your game you are often forcing the result of your actions to "wait" to be displayed by a complete frame time, 16 ms.
Image source: Anandtech
Many articles have been written about testing input latency (input lag sometimes) over the years and this one from Derek Wilson at Anandtech comes to mind as being a solid explanation. The effect of Vsync on input latency will vary from game to game, but in many cases the addition of Vsync adds at least 1 full frame or latency to the configuration.
It is for that reason that we many gamers choose to play PC titles without Vsync enabled – regardless of their stance on the horizontal tearing seen without it.
We are working on another way to test input latency in a more scientific way, hopefully by utilizing portions of the Frame Rating system we have created. Intercepting specific mouse and keyboard DirectInput commands while also measuring frame placements seems like it could offer a compelling and reliable way to measure input delays. More on that soon!
Final Thoughts
Today's article was meant to address the theories revolving around Vertical Sync and animation smoothness, and I think with our videos and descriptions most users will be able to see how experiences are affected. Our goal today is only to compare a typical Vsync situation from either vendor to a reference result at 60 FPS and at 30 FPS; not to compare AMD against NVIDIA!! But keep in mind that we are engaging in a discussion, not telling you how you should feel! Let me know in the comments, or in an email, what you think about the results we showed and if you agree or disagree with our statements on the experiences they provide!
^^^…lol
^^^…lol
Ryan, thanks for doing this
Ryan, thanks for doing this work.
I’m not sure you will belive me, but when I have vsync disabled, I can tell with my eyes when my games are hitting exactly 60fps, plus or minus 3 or 4. I can do it every time, all the time, and I use fraps to check. The motion “clicks” into smootthness, often for only a second or two, but long enough for me to glance at my fraps readout and see that the number is between 57-63 every time. Is this because the tear lines are showing up at the very top or bottom of the screen (where I’m not looking?.
Please remove the 50% and 20%
Please remove the 50% and 20% from the blind test. The whole point is to see if you can see a difference between 60 and 30 FPS, so slowing it down changes the scenario completely. It just makes the file size larger for no benefit. The constant 60 FPS was best, the 30 FPS was worst and the v-sync was in between. But who runs v-sync if you cannot maintain >60 FPS anyway?
Just fuck of pcper and end
Just fuck of pcper and end the witch hunt against AMD. You don’t even know how to use a fucking computer. You can create frame latency issues in bf3 with Nvidia or AMD cards but not knowing how to use Vsync or in the case of knowing what you are doing reduce frame latency issues with it. If you fucking morons don’t know how just stop writing these articles seriously. The AMD witch hunt is bullshit and needs to end.
I can assure you, your frame rating results for BF3 are wrong and you don’t know how to use Vsync.
Stupid troll will troll.
Stupid troll will troll.
I still don’t get the point
I still don’t get the point of all of this?
I have a GTX 690, how do I set it up? What is the “proper” pcper method of setting this up so it runs the way you want / expect it to?
Same goes for 7970CFX…
I am not trying to be weird, just saying, you are puking out all of this info, please give me something I can use, tell me how to best setup my rig.
This was a subjective test.
This was a subjective test. It is to show you what different refresh rates look like with v-sync on. Do you notice stuttering when you are not maintaining a constant FPS? Everyone would likely agree that solid 60 FPS with v-sync on looks the best, but do you find solid 30 FPS with v-sync to be better than 40-50 FPS with v-sync? Some people do, some people don’t, but it does show the weakness of v-sync regardless.
So, I think something’s wrong
So, I think something’s wrong with the BF3 videos, or at least BF3-60v30.mp4 that I downloaded.
The phenomenon I’m referring to is easily seen between 2:03 and 2:10 of said video, when looking at the tile around the door that you’re turning into on the left side and the rug on the floor in the kitchen.
On the left side of the video, from the 60 FPS test, both the tile and the rug stay crystal clear with every frame. On the right side, the 30 FPS part, there clearly is some form of motion blur applied. The tile, the rug, in fact everything but the gun is blurred while the player is turning left.
Just pause the video at 2:08~2:09 and have a direct comparison – clear tile and textures on the left, motion-blur on the right:
http://imgur.com/2jaftwP
So, was there a mix-up with different BF3 settings, is this a artifact of the video pipeline? And is that video still a valid basis for comparison?
The download site Mega has
The download site Mega has changed it’s terms of service. Two days ago, I was able to download the Battlefield 3 comparison video files. Today, the download site is claiming this:
“Please update your browser. Warning: You are using an outdated browser that is not supported by MEGA. Please update your browser and make sure that you keep the default settings.”
This is completely untrue since I have done nothing in the last two days in regards to my browser which is IE 10. Mega only gives me the option of downloading and installing Google chrome. I will not be forced to install that browser just to see the Sleeping Dogs comparison videos. Please inform Mega that this is completely unacceptable behavior to viewers of PC Perspective. Thank you.
60 vs 30 60 wins.
30 vs vsync
60 vs 30 60 wins.
30 vs vsync 30 wins.
60 vs vsync tough call but 60 wins BARELY. I can only tell the difference when you turned the camera at 20%.
Good article but what I miss
Good article but what I miss is a game test with v-sync enabled with triple buffering build in game. World of Warcraft is good example of this. It runs smoothly without tearing even if v-sync is enabled thanks to triple buffering.
Most games do not have a way
Most games do not have a way to force triple-buffering, but it is pretty safe bet that if a game runs in the 40-50 FPS range with v-sync on, triple buffering is being used.
Triple buffering does not fix the problem v-sync causes, in which some frames take 16.7ms to display, and others take 33.3ms of time to display. It can’t. It is not possible. Triple buffering makes it possible for frames to be rendered while the current frame is waiting to be sent to the frame buffer. Without triple buffering, you end up forcing the GPU to wait until the previous frame is displayed, before it can create a new frame, resulting in a constant 30 FPS, even if 50 is possible.
The comparison you want is
The comparison you want is the constant 30 FPS example, compared to normal v-sync on the 7970. The constant 30 FPS with v-sync is like not having triple buffering, and the normal v-sync on the 7970 is like having triple buffering, because it does.
Assuming it’s 33ms and not
Assuming it’s 33ms and not 33.33ms, Isn’t the best solution to this is to simply cap at a divisible of 16.5ms? That’s what I do and it feels smooth as butter.
16.5×4 – cap at 66
8.25×16 – cap at 132
Which will obviously give us an obvious multiple for each step. If you wanted to play at 30fps you’d cap at 33 – play double that frame rate, you need to cap at 66 – on a 120hz monitor you’d need to cap at 132 – and so on.
The game engine would be sending you frame-per-frame. It seems to improve hit-detection and all that other crap too.