FPS Benchmarks Are Flawed - Part 1, Ft. Scott Wasson


  • Am Vor Monat

    Gamers NexusGamers Nexus

    Dauer: 22:17

    FPS metrics are flawed, but they still stand as one of the most understood ways to measure performance. Scott Wasson talks about why.
    Ad: Buy Thermaltake's M360 Plus on Amazon geni.us/ZZoY1O7
    We have long talked about FPS in averages -- even 1% and 0.1% are ultimately averages -- and that is flawed. The trouble is, it's the easiest way to display a bunch of devices on the screen simultaneously, and frametime plots (like we show) can only reasonably fit so many cards/CPUs before they become illegible. There's no perfect solution, especially because FPS is a metric that, despite being flawed, is still good; people still understand FPS, it conveys a message reasonably well, and its flaws can be countered if reviewers take observational steps to catch poor performance behaviors. Scott Wasson, formerly of Tech Report and now of AMD, joins us to talk about solutions.
    Support GN on the store: store.gamersnexus.net/
    Like our content? Please consider becoming our Patron to support us: www.patreon.com/gamersnexus
    ** Please like, comment, and subscribe for more! **
    Follow us in these locations for more gaming and hardware updates:
    t: twitter.com/gamersnexus
    f: facebook.com/gamersnexus
    w: www.gamersnexus.net/
    Host: Steve Burke
    Guest: Scott Wasson
    Video: Andrew Coleman, Keegan Gallick

    gamersnexus  gamers nexus  computer hardware  scott wasson  scott wasson podcast  tech report podcast  scott wasson tech report  fps benchmarks  gpu benchmarks 2018  cpu benchmarks 2018  how to benchmark games  

Gamers Nexus
Gamers Nexus

Thanks to Scott for joining us! Find our previous discussion together here: ("What is GPU IPC?") http://www.youtube.com/watch?v=c6sDyJzu6m8 Check out his old "Inside the Second" article for more on this: http://techreport.com/review/21516/inside-the-second-a-new-look-at-game-benchmarking Support GN via the store: http://store.gamersnexus.net/

Vor Monat
Awesome Legend
Awesome Legend

One way of conceiving what X% lows can represent is; At 100FPS 1% lows can represent a stutter ever second and 01% lows can represent a stutter every 10 seconds. With that in mind one would come to the conclusion that X% lows that are as close to or the same as the average FPS would represent the "best" gaming experience, which is logically correct but does not tell the full story. The problem is that the aforementioned concept of X% lows is something of a worse case scenario (or at least it is IMHO). For example, I can't stand frequent stuttering in games, it absolutely ruins my immersion, but I have little to no problem ignoring massive frame time spikes associated with loading a new map/area/level if it is done tastefully (I.E. not during fast paced action or when seeing a breathtaking vista for the first time). So a large delta between the average FPS and the X% lows might represent a totally normal reasonable acceptable gaming experience, but without a frame time plot there is no way of knowing. Hence, without a frame time plot to refer to, logic dictates that we have no recourse other than resorting to the aforementioned worst case scenario. P.S. GN, there is no harm in displaying the frame time plots for a few seconds, so that people who are interested can pause on it...

Vor Monat
Roy Parrish
Roy Parrish

When you show a time per frame plot, it would be nice if you also provided a line(s) for where 60 fps (16.67 ms) and maybe a few other fps values (like 80 & 100) fall on it. This way, the spikes in the plot will have a reference to show how "noticeable" they might be to someone watching. If your plot is all below that line, then even though it might have a bunch of spikes, overall, it's still constantly "non-observable". As with everything when converting from Digital (Display) to Analog (Eyeball & Brain), there are individual differences in how this all is perceived. One thing that Graphics might want to start to work on it how they can deliver the display in such a way to were even if there are some glitches, they are not noticeable. Scoot talked about a movie frame (IIRC, ~24 fps) but that they also have a frame exposure time of a few milliseconds per frame that results in a slightly blurred object when they are in motion. We are able to take those blurs and fill in that space and it APPEARS to be seamless. Maybe the Card Makers should, with the ability to deliver such higher average FPS should instead make the frames that are delivered to be a slower, consistent rate while having each frame that is delivered to have a simulated exposure time. Just an idea for where Graphics might want to go in the future.

Vor Monat
J Klauser
J Klauser

I want to argue the point you two made in the video "but we are the internet, we can see 300 fps" with the argument, "but is there a display which could display that amount of frames". For that, I would like to include a video 3kliksphilip did on how many frames you "need". http://www.youtube.com/watch?v=hjWSRTYV8e0 Granted its a few years old but i still think it adds value.

Vor Monat
Snake Snake
Snake Snake

Eazy trick: Locking FPS on ~minimum smooth FPS/frametime means less power usage, less heat, more OC, higher low FPS. Chek clip where Nexus OC AMD gpu to hell, basicly low fps are same like on NVidia HI-END cards.

Vor Monat
Iron Maiden
Iron Maiden

This is all great talking about it. But how to do you combat it?

Vor Monat
Daniel Quirant
Daniel Quirant

The most useful metric for smooth gameplay is highest frametime per second or over several seconds. If that number is anywhere under 20 ms the game will feel okay. If it's under 16 the game feels smooth to most people, even if it's a shooter. And if it's under 10 it's good enough for a genuinely fast (arena) shooter or for VR. You guys talk a lot about "percentiles" and other ways of crunching numbers - I wouldn't bother. Some information simply isn't conveyed by numbers very well. If you want more information than the worst frametime over X seconds the best way would be a graph with max. frametimes as data points. More seconds per data point make this scalable (for longer benchmarks). For my own games I wrote a framecounter in the UI that tells me the worst frametime of the past 5 seconds. If there's a hitch in performance that number will shoot up immediately as it will top anything within the last 5 seconds, and then it will be there long enough that you don't have to interrupt the flow of gameplay to take note of it. Maybe MSI afterburner and similar tools can be configured to work like this for game reviewers so they don't need to make an elaborate benchmark to report their experience with performance in a review title. Any way, "maxFT" is the proper way to measure smoothness, I hope it catches on and people stop talking about framerate.

Vor 10 Tage
Faf Dus
Faf Dus

Frame times if you have a 1080.....

Vor 20 Tage
The White Waffle
The White Waffle

All of your videos shot in this room are "grainy" lately as if it is overexposed or something. Not sure if anyone has told you this.

Vor 21 Tag
The Nose Plays
The Nose Plays

Part 2 please.

Vor 22 Tage
zxn6869
zxn6869

60 fps is a minimum must have

Vor 22 Tage
RunRoneilRun
RunRoneilRun

Should I not rely on FPS metrics in my decision making when buying or recommending new hardware? Especially when charts like the one at 2:29 shows the .1% and 1% lows with the average, I feel like this kind of metric highlights all I need to know to make a good decision/recommendation (like avoiding processors with a wide spread between the lows and the highs as they don't provide consistency in FPS for the game? Am I understanding it correctly?).

Vor 23 Tage
ParisLondonRoma
ParisLondonRoma

Nvidia been milking that fps bullspit a long time. Gaming hardware marketing is scammy in general. Just name something "Predator" and you're on your way

Vor 23 Tage
Nathan Gamble
Nathan Gamble

This is why we use percentiles in fps benchmarks. We already have a solution to the problem. It's a total non-issue.

Vor 27 Tage
Klumsy Gamer
Klumsy Gamer

This is a bit of an old topic, but it's good to have a dedicated video on it, now! Good benchmarking channels show 1% and .1% lows for this very reason, so it's been hard not to get informed on this topic

Vor 29 Tage
musikSkool
musikSkool

It's not that complicated, the 99th percentile is just a "lag spike" probably because of some sort of software bottleneck. The hardware is hardly ever at fault, software is what drives it.

Vor Monat
musikSkool
musikSkool

Or you could say I got 0 fps for 1/5th of a second. Some games have reoccurring lag spikes, like when chunks load in Minecraft, others have the "lag spike of death" an inexplicable half second or more of solid nothing. Consoles are usually much smoother with their lag, like dipping to 25 fps in a thick fight, but almost never dropping to 0 fps like PC games. Software. PC Software just isn't as refined somehow. Maybe that is because of the enormous difference in hardware on the market, and consoles all use roughly the same hardware. Yes, your PS4 has slightly different chips as the PS4 did at launch.

Vor Monat
musikSkool
musikSkool

Oh, and for "lag spikes", I like to use a fraction of a second as the metric. "I was close to winning, but a 1/5th of a second lag spike right when I pressed the button, so the skill didn't activate, that's why we lost." One reason I don't competitive game anymore, it is way too expensive to remove all those lag spikes.

Vor Monat
musikSkool
musikSkool

I miss my old 85 Hz CRT. Can't beat that color quality without spending all the $$$

Vor Monat
musikSkool
musikSkool

A 120Hz monitor would insert frames and look not as clean as a rock solid 85 fps with each frame being where it should. Gaming at 120 fps is way more expensive than 85. Personally, for the money, 85 is where I want to be, 75 isn't good enough for me.

Vor Monat
zuhair shahid
zuhair shahid

very good steave good work

Vor Monat
aproudft
aproudft

I think that what we are actually looking for is what I'm going to call Frame Overrun - this is when the frame-time spikes are longer than the refresh rate of the monitor making the previous frame last longer to get the new frame on the screen, causing stutter. VR tries to fix it with Async timewarp and the like by estimating based upon the instruments (eg, it smooths out movement). The measure of Frame Overrun would be the # of frames that were displaced in that spike that causes stutter. this you can show in a graph much better as you can just show the Frame Overruns over time to give a measure of the real-time frequency that it occurs. A two frame overrun every 10 seconds wouldn't matter, but an overrun of 10 (Stutter lasting the length of 10 fps based upon monitor output) would suck on a 60Hz but not so much on a 240Hz.

Vor Monat
Hirens
Hirens

Stop putting that Far Cry 5 graph!! That issue don't appear on any other test done! I did not manage to reproduce that with exactly the same configuration. Please!

Vor Monat
Yamil Abugattas
Yamil Abugattas

"Let me channel my inner Gordon" Haha, that was awesome!

Vor Monat
mathburn1
mathburn1

Yes, Frame time all the way babyyyy. Start following them years ago since AMD 7000series have micro stuttering/frame pacing issues.

Vor Monat
Ray Lopez
Ray Lopez

@12:20 the dude is doing math on a calculator, wow, that's deep, man!

Vor Monat
krayzieka
krayzieka

if people have half the brain as gamers nexus the world is better. hopefully customer better educate.

Vor Monat
youmrdr
youmrdr

Is it possible that in the future there will gpus which optimpize Frame Timing, per game?

Vor Monat
Hastur Dagon
Hastur Dagon

AMD still going with Raja's "feels like" test... RX 580 "feels like" 2080 Ti...

Vor Monat
Ivan Luzyanin
Ivan Luzyanin

Maybe you should use non-linear scale for the FPS and compute difference in % based on a length of the bars on this scale. This would allow for better demonstration of diminishing returns of the higher FPS numbers.

Vor Monat
Dazouri
Dazouri

Steve, I love what you are doing with this channel. You're continuing to question convensional thinking about how to "make the case". This is different than most of your peers. You are making news, not just reporting on others news. You are often quoted and should expect that to continue and grow. Thanks for the work you and the GN Team do! DaZouRi

Vor Monat
Apathetic Millennial
Apathetic Millennial

I’m flawed.

Vor Monat
muttleydk
muttleydk

so what you are saying is to buy nvidia. k thx...

Vor Monat
Derek Swanson
Derek Swanson

FPS are just one metric that we should look it, but at more angles. Steve starts to capture some variance with 1% and 0.1% lows, but we need to take this even figure and do our best not to discount major outliers. In statistics, that's often the "right thing to do." In our visual perception world, it isn't, as it can break the immersive experience. Early on, FPS was an attempt to provide objective metrics on graphics performance and largely met that goal... but we can do better now.

Vor Monat
SH18
SH18

Give me the minimum frame time percentage to have a smooth fps gaming experience with zero issues/flawless experience. 2ndly FPS should stay around because it clearly shows you easily which hardware you need for the type of gaming (hz) experience you want. 240hz, 165hz, 144hz, 60hz via FPS, it's an easy way to determine which hardware you'll need based on your specific game and how you want to view the game your playing.

Vor Monat
GregorysBrain
GregorysBrain

Steve, please never say 0. anything. Refer to the numbers as they are. I'll be discussing averages and latencies soon.

Vor Monat
Captain Transistor
Captain Transistor

Um, FPS-frametime relation is completely linear, idk what you people are talking about...

Vor Monat
Pierre Barreau
Pierre Barreau

Framerate is not an exact value because frame timing is the real thing that matters actually

Vor Monat
B14CK_H4WK
B14CK_H4WK

who else watched this and just thought of this as AMD's way to justify their lower FPS output from their cards?

Vor Monat
John-Paul
John-Paul

Quick math

Vor Monat
teekanne15
teekanne15

i have to say, that it seems to he makes the whole frametime thing bigger than it is. The reason why we compare fps in benchmarks is not to compare that percievable somoothnes but the power of the system. Being aware of the Frametime thing is good and all but so is beeing aware of the inputlag etc of your setup. But FPS are good to compare performance quickly.

Vor Monat
Alberto Martinez
Alberto Martinez

Keep digging in your brain there must be some way to summarize frame time graphs into bars... as in a 100% of frame time are between 30 and 3, 99.9% between 15 and 3 and 99% between 7 and 3... Maybe you need to contact a mathematician? 🤔

Vor Monat
B M
B M

Is there an advantage in running games above 144 fps on a 144hz monitor?

Vor Monat
Doug Chan
Doug Chan

For those have taken classes in Descriptive Statistics, FPS is a bad metric is because the underlying frequency distribution of frame time is skewed. So FPS as mean is biased and sensitive to outliners, which are the dips or spikes in FPS. Ideally the best way to present or compare data from skewed distribution is using boxplots. An alternative is using stacked bar charts to show the interquartile range of frame time from different graphic cards on the same game of the same duration.

Vor Monat
Picasso Pablo
Picasso Pablo

Just bring the Navi+ right now because people are hungry for you

Vor Monat
Markus W
Markus W

Too long. I bet the actual information in this video could be stated concisely in a relatively brief period of time. I can't pay attention to this guy drawing 30 fps at 200 fps being different from 30 fps at low fps to three minutes. Very annoying. If you actually have something to say just say it, stop wasting people's time. I watched over a dozen minutes of this video and I can honestly say I came with very little out of it.

Vor Monat
Deliverygirl
Deliverygirl

Intel stutter fires on suicide watch

Vor Monat
Lotus cola
Lotus cola

How do you explain it to the user? Forget fps forget % and even that 99% percentile, you use in the graphics industry. My answer, explain it like we do with Internet quality, do "ping" do "jitter" and then use the fps metric for the real world application of a gamer.

Vor Monat
Gottlieb M.
Gottlieb M.

RGB cables, when? RGB monitors for maximum distraction, when?!

Vor Monat
Lotus cola
Lotus cola

For all the haters, do this experiment. Watch a something at an increased speed, say 1.18x, not too fast but still a chunk faster, or use youtube 1.25x instead After 5 min reset speed to 1x. The higher the speed the more dramatic the effect is. Perception is very tricky.

Vor Monat
Lotus cola
Lotus cola

Always annoys me wheather it is audio or image, people syncing the human experience to the machines output. There's a few steps a few conversions just within what a monitor or a speaker or even processing that justify bigger numbers, let alone what humans perceive.

Vor Monat
Fuddy Duddy
Fuddy Duddy

3:59, I have a 32" 144hz panel at home and 2 x 60hz 24" at work. Let me tell you, that's aggrevating AF.

Vor Monat
Rainer Riegler
Rainer Riegler

Why not using the number of frames (or percentage) above 16.7ms, or 33.3ms. I think that is quite eays to understand and measures the problem directly. So basically I would keep the fps average (or frametimes) and add the number of "problematic frames"

Vor Monat
Aerial Nguyen
Aerial Nguyen

So 27" 1080p 144hz or 27" 1440p 60hz, I play RPG, moba and office works?

Vor Monat
Dudu I.
Dudu I.

Around five years ago, frame pacing was the reason why Nvidia "launched" FCAT... back then, AMD drivers/GPUs were afwul at that, and Nvidia gave consistent frames. AMD GPUs showed more FPS, yet frametimes were inconsistent, hence Nvidia launching the tool so you could realize their GPUs were "actually better". And now I see this video, and find this... very funny.

Vor Monat
DestroyerFather
DestroyerFather

GOOD VIDEO GUYS

Vor Monat
Ben The Shrubber
Ben The Shrubber

the point with the 270 vs 300 as a percentage, is that while there might be an 11%(or 12) difference between those 2, the difference between that percentage different is not equivalent for any other framerate comparison, ie that 11% difference for 270-300 was 0.37ms ish but 11% for 60fps would be 1.67ms ish. but a lot of reviewers/ publications/ etc, use the time specific differences between 2 framerates are in the same sort of examples with other framerates (if that makes sense). so saying 11% for 270 is only 0.37ms which i think no human could detect, but 11% for 59 is still a very small amount of time, but its still significantly different at 1.67~1.7, and other similar suggested equivalencies from framerate ratios are all spurious the higher they go, besides the min/max hiding and other issues they are just false on their own even if best/ worst case frametime issues didnt exist. ed: besides you can still do bar graphs with frametimes, its just now '1mm' at the begining will be the same quantity as '1mm' at the end, ie '1mm' of change on the begining of the scale is orders of magnitude greater difference than '1mm' of change at the end of the bar. so seeing 1 card might be '1cm behind' another but a different card is '1cm ahead' well that card that is behind has a greater actual performance difference between that middle card than the card that is ahead of it, but there is no indication to people who arnt aware- that this is the case. and no ones going to manually alter a graph they find to 'scale' it every time. so a problem is that people think that say equivalent average framerate differences are equivalent performance difference, and theres still no effort to even have a notice with framerate graphs that this is the case.

Vor Monat
roax206
roax206

paraphrasing a uni lecture, different metrics are better at showing different information or different "views" of a subject. You want to pick the best graphs/metrics to show just the information you want to portray. In terms of gathering information, It shouldn't be a point of which one is better but rather which one will show good relevant information. then there is IPC. for a CISC processor each instruction can have a different CPI which depending on the complexity of the instruction could easily give a 100% difference in relative IPC depending on the biases of the program. For example if one architecture was biased towards integer based math it could get much lower IPC on floating point heavy programs. Not to mention RAM access is often implemented at a sub instruction level causing RAM speed and latency to affect IPC.

Vor Monat
Benjamin Jesuiter
Benjamin Jesuiter

Could you not put max values to the side of. 1 percent? Or saying like: how many frames per some time unit have needed more time to render than being acceptable to the eye?

Vor Monat
Crabash
Crabash

ploy to make diff look smaller, the point is still that a faster frame time if not perceivable can be used to improve visuals, this guy is using small numbers to mask bigger real performance differences.

Vor Monat
GamersAnonymous4me
GamersAnonymous4me

Does freesync make the experience more smooth because the only thing I notice is when the gameplay changes and becomes not smooth or not what I am used to. It doesn't matter whether its 30 fps or 90 fps either will feel ok although 30 will appear slightly slower in reactions like turning quickly but...if they are smooth i dont notice anything wrong but the minute my 90 fps has a jerky movement issue (even though its still reporting 60fps) it still feels much worse than a smooth 30 fps where i get no jerky movement.

Vor Monat
NeutronicalGaming
NeutronicalGaming

Most things are flawed, perfection is somewhat hard.

Vor Monat
jonnolags
jonnolags

Very interesting

Vor Monat
banana junior 9000
banana junior 9000

One advantage of being an old fart...I can't tell the difference.

Vor Monat
TB Inc.
TB Inc.

It matters more how the framerate is FELT. All of cpu,gpu and the GAME contribute to this. So in every product review there should be this frame-time graph and special remarks from the reviewer on how the game FELT, ie., they should also PLAY it themselves rather than just benchmark it. Recent games at the time of launch should be chosen btw with a couple of stable titles as a common reference between all the products released at different times.

Vor Monat
D Jaquith
D Jaquith

If you think the Radeon doesn't stutter then you're smoking crack.

Vor Monat
David Voronin
David Voronin

Scott looks so scared ((

Vor Monat
Clive H
Clive H

I'm not certain you guys are coming at this from the right angle, Ive noticed this micro-hitching for years and through trial and error of changing hardware to resolve the problem it feels more like a combined timing issue between the cpu, chipset, memory and gpu often changing one component resolves the problem - perhaps you should run that i5-9600k test again with a different motherboard to see if it gives the same results.

Vor Monat
ToastyHere
ToastyHere

Meatspace?

Vor Monat
Wraith
Wraith

The Tom's Hardware review of the Radeon RX 590 included not just the usual averages and frame plots, but *frame time percentile distribution bar graphs* as well. These provide a comprehensive, detailed and easy to understand view of the performance of a GPU in a condensed, video/slide friendly format. They also summarised frame time data into simpler 'unevenness' percentile charts that categorise the data according to the perceptual fluidity of gameplay.

Vor Monat
evilwoodchuck
evilwoodchuck

I think another reason FPS is used is because if you use frametime, obviously a 2 second frametime does not appear to be a serious advancement in performance, where as 240 FPS instead of 120 FPS appears to be a much more significant improvement. So I think that advertising plays into that as well.

Vor Monat
Kevin H
Kevin H

I like your interviews with Scott. They are real intelligent discussions which are the GN standard. Which unfortunately, are scarce in the YouTube space. But...you need to get your senior AMD analyst Snowflake to sit in on your talks with Scott. The people demand more Scott and Snowflake!

Vor Monat
The Nihl
The Nihl

This is what I like about AMD. they actually start sending their guys to media instead of shoving dirty marketing and "dont" emails

Vor Monat
Bjørn Helge Hilseth
Bjørn Helge Hilseth

that water cooler looks soooo fine, but 500 grand is way to much....

Vor Monat
DJJOOLZDE
DJJOOLZDE

I use Afterburner to monitor both my framerate and frametimes, and have been for many, many years. Are people only getting on this now?... Also, anything below 120fps is unacceptable imho.

Vor Monat
John Terpack
John Terpack

Isn't there a way to show the frame time AND some sort of outlier index? Right now it's standard practice to show average, high, and low FPS in a simple bar graph. Why couldn't the same be done with frame time? Have the average frame time, the slowest, and some number which indicates what percentage of the times fall within a certain range of the average. I imagine standard deviation could work, it just might take some time for people to get accustomed to the new info. But then it would be fairly easy to compare the stability of cards just by looking at their standard deviations. Although that might not show the difference between a card that had a few large spikes and a card that had a lot of small spikes.

Vor Monat
Faux Sho
Faux Sho

Great discussion. I enjoyed it a lot.

Vor Monat
stolenlaptop
stolenlaptop

Been telling people how stupid a raw FPS number is forever and YOU, YES YOU, and your ilk has been hammering NVIDIA's superior FPS even at the cost of quality and smoothness forever, shame on you and all your benchmark happy friends who never understand a damn thing. I had an AMD 8 core 8350 that gave a beautiful desktop experience, I had Intel users go "wow your desktop is so smooth and responsive". Keep seeling intel and Nvidia and then appologize like this after. "Oh Intel did better in this one game!!! BEST CPU"... no, that's not how reality works that's how you sell something that if you applied the security patches it should have wouldn't even come close to those numbers. Shame on you Steve, and all like you for being the mouthpieces of EVIL! If you're still reading, this is mostly a joke I really don't care what people like Steve do I've always had great experiences by being more in the know than the cattle that make up most of the market, just like my phones have always been great because I don't blindly buy APPLE or SAMSUNG. Educate yourselves people!

Vor Monat
TCPUDPATM PORTS
TCPUDPATM PORTS

"Rough math" explained: (1f/270fps)-(1f/300fps) * 1000(ms/s) = 0.37ms

Vor Monat
TehObLiVioUs
TehObLiVioUs

SPEAKING OF WHICH, my battlefield 5 on an i5-3470 and 1070 FTW OC'd 61 fps avg 16 (1%) 0 (0.1%) and that was over like 3 hours in break through (64 players) yeaaaaaaaaaaah it's shit ("feels like 24 fps")

Vor Monat
Jean-Thomas Arcadio
Jean-Thomas Arcadio

Excellent, excellent discussion. Most interesting topic. Thx for making this video guys!

Vor Monat
Roman Depowski
Roman Depowski

AS USUALL, WHY I BELIVED THAT THEY WILL FIX SOMETHING THAT EVERYONE ELSE CAN DO NORMALLY, I MEAN FAN SPEED, SERIOUSLY ? HOW MANY MORE YEARS AMD NEEDS TO DO BASIC STUFF FOR THEIR DRIVERS? ohh, and best tip ever, it crashes system with MSI Afterburner!!! didnt belive will live time when GN will do crap like that

Vor Monat
screwyluie
screwyluie

Talk to the guys at digital foundry, they do fps and frame time in a legible way to digest way

Vor Monat
ZinoAmare
ZinoAmare

That classic benchmark amazing games :3 GTA Ridge racer damn I miss playing ridge racer

Vor Monat
Sina Madani
Sina Madani

This is exactly why i won't get a high refresh rate monitor. The difference between 60 and 144 is too small perceptually and much less stable at high framerates. I'll take the best graphics I can get at a stable 60 than a highly variable and ugly 144

Vor Monat
loak3d kirk
loak3d kirk

Another excellent Video GN! Not many Gamers are very well educated in this department. Bring in the Console Fanatics and they'll probably argue proven science LoL.

Vor Monat
Rob 'Emrys' Brown
Rob 'Emrys' Brown

I'm of the opinion that Far Cry 5 was a bad experience because it was a damned bad game.

Vor Monat
Cris
Cris

I wish i knew this when I built my pc, I bought my cpu and gpu based off of fps benchmarks but little did I know that in 90% of the games I play there would be random hitches for a second or two (always at the worst possible times too) that completely ruin the experience.

Vor Monat
tshinharR
tshinharR

Maybe I got it wrong but as I see it looking at % fps difference completely fixes the problem here As the difference form 200 to 230 is 15% gain but 30 to 60 is 100% gain So I don't really get the point here.... if people think that difference is the same its not the FPS that is flawed they just don't understand it

Vor Monat
Chromodynamix
Chromodynamix

Proof that most of the internet forgets their school statistics classes

Vor Monat
PITBULL
PITBULL

Man this is badass!

Vor Monat
SMGJohn
SMGJohn

120Hz to 170Hz seem to be really the best spot for gaming not just for visually smoothness but also performance and energy, in fact 100Hz might actually be the best option with energy consumption in mind.

Vor Monat
StopaskingformynameYouTube
StopaskingformynameYouTube

FPS counters aren't really doing much other than to show at what rate your GPU pushes frames. If you are suffering from issues like stuttering and such these frames will still be pushed, but they will contain the same image as the last one. I've got a constant 60fps in games while watching netflix, but it stutters like crazy and makes it at times unplayable. This is very game dependant tho, but for games like FS19 it's noticable. in more intensive games it's very anoying.

Vor Monat
Lee Fall
Lee Fall

The irony that someone from AMD is telling us FPS is flawed.... OK AMD. Maybe spend less money on this research and more money on competing with your competitor.

Vor Monat
scroat emm
scroat emm

Shut up and keep taking the BLUE pill

Vor Monat
Rakly3
Rakly3

You could have just said Exponential or Hyperbolic instead of trying to explain how framerate relates to frametime non-linear ) I happen to have had this arguement yesterday with some one who claimed "eyes can't see faster than 30fps" lol I told him to get a 30Hz screen then if he's so sure!

Vor Monat
Mark Kuusk
Mark Kuusk

Could you do a test on what frame gets displayed? If you ate running sth like 200fps on a 60fps monitor....are images displayed realtime or not? Sports and gaming have visual trigger response execises, so if you get 60 frames per sec, then getting new information faster is a big deal, since it will reduce user response time, but what frame is displayed? running 120 on a 60fps montor - what frames are skipped, do you get a information gain by running higher fps? (60pfs monitor example because the difference in information would be largest)

Vor Monat
Stu Bur
Stu Bur

Military studies of the limits of the ability of pilots to perceive objects within very short time frames found that most could notice some thing shown for only several hundredths of a second. Some have extrapolated that to mean that fps of several hundred would then have practical benefit. This does not quite follow from that data however. What can be concluded on the other hand is that a disruptive frame image of just a few milliseconds in length is likely to be noticeable. The way that visual perception works in humans is that we fill in the blanks between frames to blend them together and this is why a video of less than 30fps can appear to be smooth. But if a disruptive frame is inserted that does not match the flow of adjacent frame images it can be easily noticed. If the timing between frames is not consistent this can also be noticeable. When a person has direct control of the camera in real time disruptions are also much more noticeable. For a metric of quality that matches human visual perception then we should find some way to measure and quantify these disruptive effects with the highest scores given primarily to the least disruption and only secondarily to the apparent smooth motion from more frames in a second.

Vor Monat
GooglePlusSucks
GooglePlusSucks

So, you calculate average frame time, average frame deviation, and then put the spikes beyond one sigma or whatever into bins (logarithmic works best because some might be infinintely long, base 10 bins are too far apart, I like base 2-5) and count them. :)

Vor Monat
GooglePlusSucks
GooglePlusSucks

How do you do it? Give it an acronym! AFT - Average Frame Time, AFI - Average Frame Interval FID - Frame Interval Deviation. If you can make the acronym cutesy it will sell better

Vor Monat
GooglePlusSucks
GooglePlusSucks

why don't FPS metrics also calculate deviation?

Vor Monat
Crok
Crok

the difference between 165fps and 240fps is the same as the difference between 30fps and 32fps. i will accept that people can tell the difference and have a preference. but man, i no longer feel bad about not seeing the value

Vor Monat
GooglePlusSucks
GooglePlusSucks

This! I don't care much about frame rates past 70, it's the individual glitches or short periods where it sags below 40 that are the problem. The average is/has always been useless. Who cares what it averages out to?

Vor Monat
Big Baba Ganoush
Big Baba Ganoush

Agh! Im a hardcore NVIDIA fanboy... but this man makes such good sense... And i can tell Steve is thinking the same. Man i wish AMD made better graphics cards.... AND software like Hairworks.

Vor Monat
pronewname
pronewname

Where my photgraphets @ looking @ 1 frame @gamernexus awsome

Vor Monat
T C
T C

Wow...well that answers that! amazing content on this channel!...time for he golden button on this one..u need way more subs Sir

Vor Monat
whirlybird n
whirlybird n

Can someone explain to me the the pixel fluctuation on Steve's shirt while supposedly being shot @ 1080p... Wow, it got worse when I upped the resolution in the settings.

Vor Monat
Dsirius 4
Dsirius 4

Great topic man. This frame to frame rate becomes more relevant for a many people. Thank you for inviting Scott which points this famous and technically topic with a pleasant common sense. Keep up the good work .

Vor Monat

Nächstes Video

Dreaming Purple Neon | Treehouse of H season 05 episode 05 | Чу Джи-хун