Blur Busters Forums

Alright, I did a retest. This time I counted the frame where the mouse was first depressed, and used the cut off end of a neon pink drinking straw (bent and jammed under the mouse button) to aid in spotting when the click began.

Equipment: Samsung Note 3's 720p 240 fps rear camera, my trusty index finger...and a straw...
Monitor Latency: 3-5ms (1440p Acer XB271HU)
Mouse Latency: 1ms reported (Deathadder Chroma 800 dpi, 1000hz)

Game: CS:GO
Settings: Lowest, Multi-Core disabled
Map: Dust II
Mode: Offline (no bots)
Weapon: P2000 Pistol
Samples: 13 shots per

144Hz Uncapped (600+) fps:
14.11ms

144Hz V-Sync (in-game double buffered) + Uncapped (600+) fps:
23.42ms

144Hz G-Sync + V-Sync (Nvidia Control Panel) + 300 fps in-game cap:
23.74ms

144Hz G-Sync + V-Sync (Nvidia Control Panel) + 142 (140 effective) fps in-game cap:
17.32ms

144Hz G-Sync + V-Sync (Nvidia Control Panel) + 120 fps in-game cap:
15.72ms

120 fps and 142 fps caps appear to be within margin of error (1.6ms difference), and very close to the uncapped number (barring sheer frametime difference). I'm not sure how accurate these results are (it's the latency test equivalent to a shoe box guitar), but it's something.

144Hz Uncapped (600+) fps:
14.11ms

That's 6ms higher than my CRT test data at fps_max 600, but it should be about that, considering the monitor and the fact you're using weapon flash instead of moving the mouse(have to wait till the scanline gets to the center of the screen).

Interesting. I'm sure my phone camera isn't able to pick up all the frames, as you said earlier, but, better than nothing. Forgot to add that I have Max pre-rendered frames set to "1" globally in the control panel.

Anyway, at this point, I feel a little more confident in claiming (as long as there isn't too much frame drift) that an in-game fps cap as high as 142 fps or so with G-Sync + V-Sync on seems sufficient to reduce latency enough that it effectively matches V-Sync off latency at the same framerate.

So, assuming those results are correct, there's virtually no noticable latency between using uncapped gsync/vsync off and 1-ms polling rate capped gsync/vsync on? This would mean we can enjoy having absolutely no tearing thanks to capping fps at framerate fullfilling the 1-ms polling rate time needed for gsync to work perfectly, AND have virtually no noticable input lag compared to normal, tearing gameplay? Now I'm excited to grab a 240Hz monitor with gsync and test it myself (if I had money to buy a high-speed camera).

Paul wrote:So, assuming those results are correct, there's virtually no noticable latency between using uncapped gsync/vsync off and 1-ms polling rate capped gsync/vsync on? This would mean we can enjoy having absolutely no tearing thanks to capping fps at framerate fullfilling the 1-ms polling rate time needed for gsync to work perfectly, AND have virtually no noticable input lag compared to normal, tearing gameplay? Now I'm excited to grab a 240Hz monitor with gsync and test it myself (if I had money to buy a high-speed camera).

My results should be taken with a grain of salt. It's a phone camera @240 fps, which means at worst, it's skipping frames, and at best, it isn't picking up nearly as much as a 1000 fps camera would. That, and I don't have an LED attached to my mouse, thus my frame counting starting points could vary a bit. Bottom line, those numbers could be off by multiple milliseconds.

However, I think it's safe to say the differences between each test number is closer to accurate. I've done several more tests to confirm my earlier results, and its safe to say there isn't a multiple frame difference between G-Sync + V-Sync on @140 fps and G-Sync off + V-Sync off @140 fps. It's likely within a couple of milliseconds, which is to be expected, as G-Sync (V-Sync on or off) adds a couple ms of input latency over uncapped.

For the record, if you're averaging 80 fps on a 144Hz display, input latency with G-Sync on + V-Sync on/off is identical. The only range in dispute has been that 120-144 fps area, and at what number in that range does the G-Sync ceiling actually begin.

G-Sync IS V-Sync by the way, so that V-Sync on/off argument is deceiving in the first place. What that V-Sync toggle actually does is decide when the variable refresh G-Sync behavior fades/switches off into fixed refresh V-Sync behavior. It appears the only thing the V-Sync off scenario does, is reduce G-Sync's effective range earlier than designed.

I just created a "G-Sync 101" thread that organizes all my findings into a single post:
http://forums.blurbusters.com/viewtopic.php?f=5&t=3073

Feel free to offer further corrections and input on the new thread, which I'll be updating if any new information comes to light.

Excellent post -- thank you for posting these findings.

Chief Blur Buster wrote:Excellent post -- thank you for posting these findings.

No, thank you for the acknowledgement!

I plan to maintain and update the new thread as time goes on and more G-Sync facts come to light. Too many new G-Sync users have been needlessly confused, and I was admittedly one of them until I began my (ongoing) research. I thought I'd do my part to clear some misinformation up, and document as much as I could on the subject.

No doubt the polling rate sections and fps limits are still speculative, so it would be great to get some more in-depth input latency tests across all possible G-Sync scenarios to further confirm and/or correct my findings.

Again, if you or anyone else has any further insights or corrections to offer on the subject, I'm all ears:
http://forums.blurbusters.com/viewtopic.php?f=5&t=3073

This is so awesome...thank you!!!!!

Now we can only hope that nVIDIA implements a proper frame rate limiter exposed in their control panel so we dont have to use 3rd party apps. Really...if G-Sync is enabled, it should globally limit the frame rate (correctly) so the end user doesnt have to do anything.

Samhain wrote:This is so awesome...thank you!!!!!

Now we can only hope that nVIDIA implements a proper frame rate limiter exposed in their control panel so we dont have to use 3rd party apps. Really...if G-Sync is enabled, it should globally limit the frame rate (correctly) so the end user doesnt have to do anything.

I'm glad you've found my information useful.

And for future reference, anyone is free to post replies in the new thread I linked.

Nvidia does actually have a built-in framerate limiter, but yes, you do have to use the third party Nvidia Inspector to expose it. Also, their limiting method isn't optimal, as it currently introduces up to 3 frames of input latency.