Special K can drastically reduce latency

[axaro1]

The frametime graph with RTSS is perfectly flat. So how can SK have 10x lower deviation if it's already 0 with RTSS?

https://www.reddit.com/r/allbenchmarks/ ... x11_based/

It is not 10x let me correct myself
The adaptive standard deviation in the CapFrameX framerate analysis of the three DX11 games tested is anything between 7x-5.5x lower compared to to RTSS/NVCP.
0.2% lows in particular are 4-14% higher.

[Darktalon]

Special K seems incredibly consistent, the frametime deviation is 10x lower than RTSS or Nvidia's built in fps limiter

The frametime graph with RTSS is perfectly flat. So how can SK have 10x lower deviation if it's already 0 with RTSS?

The frametime graph of RTSS is not actually very accurate. You can use CapFrameX to see this.

[Darktalon]

The frametime graph with RTSS is perfectly flat. So how can SK have 10x lower deviation if it's already 0 with RTSS?

https://www.reddit.com/r/allbenchmarks/ ... x11_based/

It is not 10x let me correct myself
The adaptive standard deviation in the CapFrameX framerate analysis of the three DX11 games tested is anything between 7x-5.5x lower compared to to RTSS/NVCP.
0.2% lows in particular are 4-14% higher.

Yep, and the data from that reddit post was used using special K as it existed 4 months ago. In just the last month alone, Kaldaien has massively improved latency and stability even further.

[jorimt]

Using the system to measure itself has its limits. This is why many professionals/enthusiasts use tools such as FCAT for frametime performance measurements, and photodiode or high-speed setups for input lag measurements.

Until Special K's framerate limiter is tested in such a fashion, we aren't going to see the whole picture.

CapFrameX and the like produce approximated results. What actually reaches the display can be another thing entirely, and something system-side tools can't fully expose (including RTSS).

It would be interesting to see results for this limiter captured via external methods nonetheless (whomever ends up ultimately performing them)...

[Darktalon]

Using the system to measure itself has its limits. This is why many professionals/enthusiasts use tools such as FCAT for frametime performance measurements, and photodiode or high-speed setups for input lag measurements.

Until Special K's framerate limiter is tested in such a fashion, we aren't going to see the whole picture.

CapFrameX and the like produce approximated results. What actually reaches the display can be another thing entirely, and something system-side tools can't fully expose (including RTSS).

It would be interesting to see results for this limiter captured via external methods nonetheless (whomever ends up ultimately performing them)...

Nvidia is sending Kaldaien an LDAT.

[Chief Blur Buster]

Special K seems incredibly consistent, the frametime deviation is 10x lower than RTSS or Nvidia's built in fps limiter

The frametime graph with RTSS is perfectly flat. So how can SK have 10x lower deviation if it's already 0 with RTSS?

Well, there's many metrics. Just because RTSS graph may be flat, may not be representative of how flat the graph is for Present()-to-photons, since RTSS only has ontrol up to the timing of Present(), and may timestamp based off that.

You need a photodiode to get best Present()-to-photons which can jitter even with flat RTSS graph...that is indeed correct. Even if you measure two stopwatch points of a latency chain that is glassfloor, the part immediately preceding/succeeding may be jittering -- especially the black-box beyond Present() that the computer has difficulty figuring out.

I'd need to look into things further, but I definitely think further improvements are possible in framerate capping software. I have long considered building an in-house framerate capping utility that is optimized for Blur Busters worthy stuff -- but it's too much wheel reinventing and there are too many projects competing for Blur Busters attention.

This month I'm unable to compare framerate capping software -- but let me ask around if one of my writers is able to volunteer the time to test out the multiple framerate capping software.

Speical K looks very promising looking and has some quite real promise.

[jorimt]

Nvidia is sending Kaldaien an LDAT.

Nice, I look forward to his results (if shared).

I have not been offered the same. In fact, anyone familiar with my background at Blur Busters here might be surprised to hear Nvidia has never made direct contact with me on anything .

Beyond indirectly acknowledging my material in their driver feature releases (like implementing auto fps limiting below the refresh rate or finally recommending V-SYNC with G-SYNC or creating a CPU-level limiter, for instance) and the occassional replies to user comments in some of their articles linking to mine for answers in passing, they've been silent as the grave all these years. Not sure whether to take that as a compliment or a slight, but ah well, *shrugs.*

It's too bad, because I could make good use of a device like that myself...

[Darktalon]

Nvidia is sending Kaldaien an LDAT.

Nice, I look forward to his results (if shared).

I have not been offered the same. In fact, anyone familiar with my background at Blur Busters here might be surprised to hear Nvidia has never made direct contact with me on anything .

Beyond indirectly acknowledging my material in their driver feature releases (like implementing auto fps limiting below the refresh rate or finally recommending V-SYNC with G-SYNC or creating a CPU-level limiter, for instance) and the occassional replies to user comments in some of their articles linking to mine for answers in passing, they've been silent as the grave all these years. Not sure whether to take that as a compliment or a slight, but ah well, *shrugs.*

It's too bad, because I could make good use of a device like that myself...

Reach out to AndyBNV on ResetEra, use the start a conversation tool. Nvidia didn't contact Kaldaien first either.

[jorimt]

Reach out to AndyBNV on ResetEra, use the start a conversation tool. Nvidia didn't contact Kaldaien first either.

Giving them out like hardware candy on special request then, eh? No wonder Linus got one (I kid)

Don't know how much luck I'll have, but I may give that a shot (already have a ResetEra account anyway).

Thanks for the heads up

[diakou]

Reach out to AndyBNV on ResetEra, use the start a conversation tool. Nvidia didn't contact Kaldaien first either.

Giving them out like hardware candy on special request then, eh? No wonder Linus got one (I kid)

Don't know how much luck I'll have, but I may give that a shot (already have a ResetEra account anyway).

Thanks for the heads up

For the future, always reach out rather than waiting to be reached out to (unless you hope to get some sort of monetary benefit, waiting can land you better deals, in the case of a job based on previous accomplishments, reaching out is better.)

A lot of these things, especially if we are considering latency work, is beyond niche. Finding info on latency or latency related topics in general is very hard and the true gems are usually more hidden (a lot of misleadig info!) In your case it seems a bit different with the case that they pretty much know who you are, it probably stings a bit. But I assume that the people who linked to your past works are not in the correct department for marketing and collaboration, thus they most likely didn't have the power (or thought) to do these things. The people in the correct departments for collab work/marketing most likely just do not even know you exist as the people who do and the people who don't in NVIDIA most likely never overlapped information.

So all I'm saying is without dragging this too off-topic, if you don't really expect anything special and genuinely would like hardware/software etc that is within your niche and helps your own niches, just reach out. I have close to 0 programming/game dev knowledge yet I've talked to, gotten help and been able to give good info/advice on topics for low latency with many devs including AAA studios (ubisoft) on things like correct flip model implementations, improving rendering pipelines by testing v sync off / v sync on scenarios instead of just one, etc.