Special K - v 0.10.3 - [HDR stuff probably]

[Kaldaien]

What's New? (Current Test Build :: 0.10.3)

I've opted to add a new form of frametime analysis to Special K's framepacing graph.

The old data (which still remains), was based on higher-level statistics and standard deviation; an uncommon representation of data not easily understood or comparable to expected benchmark performance and limited to a very small sample window of 120 frames. In short, it was useful to me for performance optimization and analysis but those numbers had no meaning to anyone else!

Fun bit of trivia, however, the color of the framepacing graph is based entirely on the standard deviation.

What are Percentiles and Why Should I care?

To put it simply, when we speak of frametime percentiles, we mean either the bottom N-% or top N-% performing frames recorded. Since I don't care about high framerates and only care about stuttering, Special K's analysis focuses on the low percentiles.

When numbers like "10% Low FPS" are being thrown about, it's another way of saying 90% of all frames drawn were faster than xyz FPS.

Common numbers used for frametime analysis are 10% / 1% and 1% / 0.1%.

It is my strong belief that 1% / 0.1% is not appropriate for measuring and tweaking game performance. 1% / .1% should be reserved only for people publishing benchmark scores because that's really all they're good for :P

I would suggest configuring the framepacing widget manually and selecting something more reasonable such as 5% / 0.5% or even 10% / 1%, but it is pre-configured for 1% and 0.1% since this is the most common percentile ranking published in online reviews of games.

How large is the dataset used?

There are two different datasets for percentile analysis.

There's one that keeps track of the last 1,000 frames and then there's an even longer-term one that gets a summary of all 1,000 frames every time the short-term history fills up with data. Effectively this gives a 1,000,000 frame history and that ought to be enough for anybody :P

If you're trying to find optimal in-game settings, I suggest focusing on the smaller 1,000 frame history. That's at least 8 seconds worth of data even at 144 Hz.

Conclusion

Knowing what this new feature is designed to measure (how poorly the worst-performing part of the game is performing), you should be able to use it to find the perfect set of graphics settings that keeps the two percentile charts in the green. Failing the ability to find any settings in-game that stop stuttering, you can always fallback to Special K's framerate limiter.

[HolyDeath3000]

https://youtu.be/LAUtMF_wu_U

Any truth to this? And does the Epic Game store version have all the DLC's included?

Edit:
It seems guy basically proves that Denuvo removed in game doesn't really affect FPS at all and no noticeable issues in the game or any real improvement in fps.

[Erebus]

God is that one of those shitty comparison vids where some assclown does one run through and doesn't account for any variables while shouting SEEE DENUVOOOOO RAPES YOUR FRAMESSSSSSssssss?

Cause I'm getting really sick of those.

[HolyDeath3000]

Originally posted by Erebus:

God is that one of those shitty comparison vids where some assclown does one run through and doesn't account for any variables while shouting SEEE DENUVOOOOO RAPES YOUR FRAMESSSSSSssssss?

Cause I'm getting really sick of those.

Actually that video proves that Denuvo removed in that game didn't change performance by much at all. I mis read that title.

Reading thru comments it shows that Denuvo removed didn't have any noticeable impact on game performance for most part. So he is basically proven in this title denuvo wasn't really an issue at all.

[Erebus]

Explains why it's so few views. Reddit and a certain watch subreddit will bury it.

[Aemony]

Originally posted by HolyDeath3000:

And does the Epic Game store version have all the DLC's included?

Yes. It didn't have them on release date, but got all of the DLCs through a ~15 GB update that was released the next day.

[Erebus]

Originally posted by Aemony:

Originally posted by HolyDeath3000:

And does the Epic Game store version have all the DLC's included?

Yes. It didn't have them on release date, but got all of the DLCs through a ~15 GB update that was released the next day.

I wonder if it was something they overlooked, or if they saw a negative PR ball starting to move and changed their minds.

[HolyDeath3000]

Originally posted by Erebus:

Originally posted by Aemony:

Yes. It didn't have them on release date, but got all of the DLCs through a ~15 GB update that was released the next day.

I wonder if it was something they overlooked, or if they saw a negative PR ball starting to move and changed their minds.

Yeah b/c the game listed doesn't say the Premium edition one but regular. So I was a bit skeptical of that.

[HolyDeath3000]

This video is really good at breaking down everything o.O
https://youtu.be/e6FWiVwq1fI

I advise using it to do graphic tweaks for BL3. Kaldaien was right about fog. Just switch it to Medium. You can't really see difference at all between ultra to medium. Drastic increase in FPS from Ultra to medium with barely any changes in quality.

[HolyDeath3000]

Ugghh You kidding me???
https://screenrant.com/epic-games-10-million-control-exclusive-remedy/

[[SK] Kaldaien]

Originally posted by HolyDeath3000:

This video is really good at breaking down everything o.O
https://youtu.be/e6FWiVwq1fI

I advise using it to do graphic tweaks for BL3. Kaldaien was right about fog. Just switch it to Medium. You can't really see difference at all between ultra to medium. Drastic increase in FPS from Ultra to medium with barely any changes in quality.

Speaking of which... I do not think many people realize this, but Steam's hardware survey does not accurately represent any particular market segment. The public data just summarizes everyone who participated for the month.

Publishers have access to statistics derived from just the users licensed for each of their products. I was always under the impression that developers were making educated guesses off the global data we all see, but that is wrong if their publisher is competent and shares sales statistics with their devs.

Individual games should probably ask for permission to have graphics settings sent back with the hardware survery information. If they see that everyone turns this stuff off, maybe do something about it? :)

[Erebus]

Originally posted by SK Kaldaien:

Originally posted by HolyDeath3000:

This video is really good at breaking down everything o.O
https://youtu.be/e6FWiVwq1fI

I advise using it to do graphic tweaks for BL3. Kaldaien was right about fog. Just switch it to Medium. You can't really see difference at all between ultra to medium. Drastic increase in FPS from Ultra to medium with barely any changes in quality.

Speaking of which... I do not think many people realize this, but Steam's hardware survey does not accurately represent any particular market segment. The public data just summarizes everyone who participated for the month.

Publishers have access to statistics derived from just the users licensed for each of their products. I was always under the impression that developers were making educated guesses off the global data we all see, but that is wrong if their publisher is competent and shares sales statistics with their devs.

Individual games should probably ask for permission to have graphics settings sent back with the hardware survery information. If they see that everyone turns this stuff off, maybe do something about it? :)

I've long questioned why even have the public hardware survey the way it is.

[[SK] Kaldaien]

Looks like I have my work cut out for me,

https://dotesports.com/dota-2/news/dota-2s-20-year-ban-wave-explained

I wonder how many of these system-wide bans it would take for the counter to get up to 20 for the community hub? So far it's been growing predictably: 1 week, 3 weeks, 1 month, 3 months.

I assume 1 year is next and then 3 years, but do not know what comes after that.

[Dr. Dro ඞ]

Originally posted by HolyDeath3000:

https://youtu.be/LAUtMF_wu_U

Any truth to this? And does the Epic Game store version have all the DLC's included?

Edit:
It seems guy basically proves that Denuvo removed in game doesn't really affect FPS at all and no noticeable issues in the game or any real improvement in fps.

Yes, the EGS version contains every single DLC released.

Oh, oof, for some reason I missed a full page of replies that contained the answer. Sorry lol

Originally posted by SK Kaldaien:

Looks like I have my work cut out for me,

https://dotesports.com/dota-2/news/dota-2s-20-year-ban-wave-explained

I wonder how many of these system-wide bans it would take for the counter to get up to 20 for the community hub? So far it's been growing predictably: 1 week, 3 weeks, 1 month, 3 months.

I assume 1 year is next and then 3 years, but do not know what comes after that.

Following the trend of x*3, 9 years?

[[SK] Kaldaien]

Wow, I forgot people have GPUs running at 85 C and CPUs at 6 0C, that video's OSD really surprised me. Mine cap out at ~48 C under full load now :)

[Wastelander121]

The EGS version also contains the Earth 2 skin which was exclusive to PS4. All other exclusives aren't there.