1% & 0.1% Lows Explained!

[@JarrodsTech]

Just want to say the explanation here is still so simple and easy to understand that I end up linking people to this multiple times a day when asked what 1% low means 😁

[@BleachRush]

I came here just after watching your reviews lol

[@كاملالرمحي-ز2ز]

I just came here after watching the 80w vs 90w difference test.

[@twosad]

Kamel Alramahi which ones better?

[@jayjwin1178]

i came here just to like you.

[@Gabedavidd]

I can confirm😂 I'm here because of you

[@aegixxer1]

almost 3 years to the day and this video is still helping people!

[@Xenthetic]

I'm building my first PC and this really did help me!

[@stuff6585]

What did you think? It will be helpful for only 10 days?

[@aegixxer1]

@@stuff6585 I'm glad I don't see the world in a negative light like you do. Must be horrible walking around in all that gloom. Maybe therapy would help?

[@i_like_Peanuts]

Make it 5 lol

[@arnavsingh8385]

7 years

[@jasonb5944]

Everybody's hating because you used "as fast as possible". I still love you Kyle.

[@anassrhammar9126]

Jason Bayens I love you too Doge . Pepe bless .

[@jasonb5944]

oh my god BITWIT LOVED MY COMMENT JOY OF JOYS!!!

[@st0nedpenguin]

That's because "As fast as possible" is Kyle's bedroom motto. ( ͡° ͜ʖ ͡°)

[@dh66]

wut

[@vinniekiro]

wot

[@paulshardware]

Let's see how long it takes LTT to flag/copyright strike this one 😂

[@stycks32]

Not fast enough.

[@senseisadness9836]

Anass Rhammar Oh shit what did I miss.

[@umayrhussain2211]

Paul's Hardware why what happened

[@TheRealJackassJesse]

I was thinking the same thing! haha

[@77gestalt]

Elaborate

[@Hardwareunboxed]

This is why you take the average results from three separate runs. You will notice that the 1% and 0.1% results are actually lower than the 'minimum' frame per second figure. There should still be a very heavy emphasis of taking the average figure for the 0.1%, 1% and average frame rate from three runs. A single run can often still provide misleading data.

[@SamoMalo2]

Hi Steve :)

[@Geetinderrocks]

Hardware Unboxed i thought they were important in reviewing​ hardware ,but didn't know they were that important 😋

[@Hardwareunboxed]

+GYM BUDDIES - Honestly I often conduct hundreds if not thousands of benchmark runs per week testing GPUs, CPUs, system memory and so on. Almost never do I run into a situation where the .1% and 1% results tell me something the minimum doesn't. In fact it's been quite a few years since I found the frame time data from Fraps (and now PresentMon) necessary. Pretty much since AMD fixed the multi-GPU stuff ;) Guys in the comments often make a big deal over it, I think they are just trying to sound smart ;)

[@Geetinderrocks]

Hardware Unboxed 🤔 hmm , it all comed down to if the gameplay is smooth or not and when u do testes with so many hardware combination ,i can just sit back and watch your review were results are shown and at the background your reassuring voice is guiding which one is to choose ,and the point bitwit mentioned about user's getting used to stutter because they don't know the smoothness over time is true but he forgot one important aspect ,we all have friends and users who are watching videos on KZbin don't just play all day alone​ at their home , friends have different PC's but still we don't know the best combination to choose ,like you shut down all the blah blah comments by showing a dual core processor is strong enough to utilise the 1050ti fully (g4560) ,i read those comments on previously about when you used your i3 for testing gpus and people commenting​ that you are idot ,u r not using its full potential ,blah blah ,but the stutter issue will only happen who only play's at their own pc and don't have any friends ,no social life ,etc,even though i eas gonna upgrade to g4560 bye watching your video's but still waiting for ryzen r3 , because g4560 is amazing paired with 1050 or rx470 ,but the cpu usage is at it's full like 90% or more ,so when you gonna compare r3 with i3 or g4560 could you please add the cpu usage Percentage 🙏,even though i believe you ,but it will be quite reasuring seeing if the ryzen r3 is like using 70 % cpu usage than g4560 90% ,even if frame rate are little lower r3 , I will still buy that one because its more future reassuring , because anything can happen in future,🙏

[@tiberseptim5103]

Hardware Unboxed No you don't have to do that. The 1% low is technically the 99th percentile in a frequency distribution. This way, you are using a sample set, get the 99th percentile and that's it. If you average it over three runs, then it's something different and not accurate. If you want to do three runs, get all frame-times on all three runs and set them as one bigger sample set. Then get the 99th percentile from those and that's it. You don't take percentiles of each run and average the three. That's just not how we do it with statistics. If you average the three then it's not the 99th percentile anymore because it's from different sample sets. Sure you can still average three runs but its not the 99th percentile (aka 1% low).

[@SilenceDestroier]

Great video Kyle, I actually didn't have a clue to what 1% and 0.1% meant in benchmarks, and you not only made it clear, but also entertaining to learn it

[@ZeroB4NG]

or in short: 1% and 0.1% lows tell you how badly a game will stutter. when you game at 144FPS, you do not want to see 0.1% lows that dip far below 100FPS or you will notice stuttering. getting average high FPS is easy, getting consistent high FPS is hard.

[@pegameplay7409]

So if i have a game running at 40 FPS and 0.1% at 30 FPS its good but at 25 FPS is bad?

[@rowdy1003]

@@pegameplay7409 Yes because even though you have a lower frame rate it's consistent and having a consistent frame rate is better even at lower frame rates than high but inconsistent frame rates.

[@Esspppiia]

@@rowdy1003 How do I improve my 1% and .01% lows?

[@namanhlehoang9999]

@@Esspppiia lower graphic quality, lower resolution, update driver.

[@LordOverTheLightOfRa]

So what is the cure for these low %s? Limiting framerate? Going back to 1080p?

[@drekoll]

Thanks for the explanation. Would give an extra thumbs up if i could for the shout out to Gamers Nexus, they are so underrated.

[@marinb_yt]

drekoll Totally agreed. They're amazing

[@PragmaticWisdom]

Thanks for clarifying what these measures are Kyle! And thanks to GN who is widely using it and to the bloke who discovered this gem of a method of frame testing. I've been watching GN for the last few months and wondered why I was seeing all of these values in their graphs, now I understand why and support it whole heartedly!

[@DevSeek]

aye i just found you're channel when i started working on building a computer so i just want to say thank you and keep the good videos up man

[@DrearierSpider1]

Was gonna call you out for taking Gamer's Nexus idea, but glad you included their content in the description. Good to see more creators adopting a much superior testing methodology, really.

[@blownfuse26]

DrearierSpider1 Yeah, I'd rather get smoother frames than higher frame rates tripping over themselves. After running RX 480's in crossfire, I can state that while some games were vastly better than others about keeping the stutters to a minimum, they were always there.

[@gravastar333]

Glad to see 1% and .1% framerates gaining more traction on my favorite tech channels. Really the only sensible way to do it. Great video. I had no idea that Wasson was the pioneer behind the concept, though I did know GN had been spreading the gospel of frametimes. Keep up the good work Kyle. Almost ready for ultra ...

[@drdyna]

Yes! I've wondered how to describe this effect for a while, it drove me crazy when it seemed like I had decent fps on the counter but it felt like it stuttered. Thanks for this!

[@HarlesBently]

you too? I get this all the time.

[@Ryzen1]

Ram speed ftw

[@buddyh13]

always wondered when i saw gamer nexus videos and others with 1% and 0.1% lows what they were. Never understood them really and always forgot to look them up. Glad you made this video, really easy to understand what it all means.

[@sudiptodeb4368]

great video, thanks for explaining so smoothly

[@frederikk.4187]

Good that you also adopted this method of analysis, cuz it DOES give a better picture. And respect for giving kudos to GamersNexus who really are front runners in the world of PC equipment and review thereof. ;-)

[@TheTolsonator]

Linus, you look a little....different.

[@BryantWhite86]

BT and his voice is a few octaves lower.

[@hassanjamil1099]

BT he hit puberty

[@Dirtyboxer1]

and he's speaking more slowly. Linus has some great content, other than when he's using cringeworthy construction techniques, but the speed with which he speaks can be annoying.

[@TheEragoon]

I don't get it.

[@ritvikbudhiraja2944]

the video title earlier had "as fast as possible" in it... and it is used by techQuickie, a channel of linus... so yea!

[@steven3893]

This clarifies so much. Was so lost until this video :)

[@hellowill]

And now this is a standard in benchmarks. Nice one bro.

[@pear7828]

Man!! U R an example of why I glad many KZbinrs review or explain the same products and things. I love GamersNexus, but I understood it a lot more after u explained it. ...or course it could be that that I'm idiot & I need things broken down to me....plus I think I could've been a little lit when I watch G. Nexus vid.... Thanks man

[@Mealz108]

Thanks, that does explain a lot! Keep throwing these tidbits of info from time to time.

[@NizarElZarif]

Bitwit, Paul's hardware and Gamernexus are the best tech KZbinrs out there. I love your channels guys, keep up the good work.

[@ameyasatpute3987]

Thank you kyle... i was pondering all over the net to get some simple info... you provided it... you are awesome

[@kristaps.mezavilks]

Awesome video - finally some1 explained this stuff. Greetings from Latvia.

[@Spillrag]

Happy to see you using this testing methodology. Even happier to see the shout out to GN and their work, seriously underrated channel.

[@erazorCTF]

Kyle I do strongly suggest that you link this explanation video unser EVERY video from now on. Otherwise the questions will never stop.

[@uberinfinitegaming]

Can't wait to see you in the next Scrapyard Wars with Linus!

[@B_Clauva]

Fuck yeah, thank you for tightening the 1% and .1% lowskis Kyle, also, Gamers Nexus is very in depth with their information which is awesome. Keep it up!!!🤘😎

[@Pax.YouTube]

Still the best 1% & 0.1% Lows Explanation Video on KZbin.

[@tahsinalam8256]

Thanks Kyle......I asked for this in the comments of previous videos.I am glad you did this.

[@ironhans3243]

Good on you for crediting Wasson as the originator and GN for reformatting and popularizing the 1 and .1 % method. It's the least you can do in a sea of derivative (me too) content that KZbin is awash in. Glad to see other reviewers picking it up, even "review lite" type channels like this one.

[@RobertMcDonald_trz]

I really appreciate this video, Kyle

[@eywhatswrong]

Thank you for this video. I have been looking for explanation to this for weeks.

[@___Robin___]

I'm really digging the flip book analogy.

[@saabgang]

Good explanation, gg Kyle, solid content.

[@nexos911]

Thanks Kyle ! This insight is highly apreciated

[@drewb242]

No offence but I did not expect you to explain it this well, good job. And I like that you mentioned Scott Watson and Steve Burke, it's great that more tech channels are implementing 0.1% and 1% low metrics. Good video.

[@arideout121]

the new method for bench marking with the %lows makes it way easier to see what matters. hopefully this becomes the standard for all bench markers very soon

[@captainseagull3792]

I literally googled this ten minutes ago and then Kyle uploads this video.... it's like he knows

[@wetere7203]

I've learned a lot of things about computers this week, great video🖤

[@Kenshin9977]

Don't know if you care but OCAT does that except for the 0.1% but provide you a CSV with the data. I developed a Python software that load and compute the 0.1%

[@SaturnusDK]

I wish reviewers would agree on a metric that basically takes the average FPS of 100%, lowest 1%, and lowest 0.1%. Add them together and call it a playability score. In the example in the video the system with 80 average FPS and lower 1% and 0.1% would have a score of 135 and the other system would have a score of 160. Although not showing the full data set it gives a single easy to compare metric that is actually based on what a real gaming experience would be on such a system. The full data set could then just be parts of that same one graph with colour coding to denote each segment of data.

[@SchmidtP89]

Great stuff! Glad to see this change.

[@hellowalkman4506]

Scott Wasson and Gamers' Nexus = Legend !!!!

[@kelkiiii]

Thanks so much for this, I was watching one of your videos cause I was trying to learn more about GPUs and wasn't completely sure what they meant. Now I'm golden

[@Diff3RentBreed]

I was wondering what those were. Thank you! Love the videos. :)

[@TheoHiggins]

"Explained fast as possible" = "Efap"

[@FartGamer39]

E-Fap

[@Getyourwishh]

@@FartGamer39 boxxy faps

[@a_b_c]

@@FartGamer39 HMMMMMM

[@jeotep]

This is great as its as better and more consistent way to test the quality of the gaming experience, and that's all that matters

[@Harzexe]

great video! we needed this explanation

[@buffaloxp]

This was actually super helpful! I haven't heard any arguments against using the minimum framerate other than this, but I think you're right. This seems like a much better metric.

[@MarioAwe]

great video and good move to point out the source of the method.

[@ctownskier]

Have you considered showing the actual distribution rather than just the 50%, 1%, and 0.1% lows? You could pack a lot more information while being able to quickly point out mean, median, and the lows that you already display. Or maybe continue using the bar graphs but throw in standard deviation?

[@DanielADickey]

Finally!!!!!!!!!!!!! So many quested answered! Thanks Kyle!

[@SolLignum]

Awesome methodology, far more informative! Thank you very much for switching to it!

[@ieatmails]

Thanks to Kyle from Awesomesauce News, i finally understand this "a bunch of crappy frames" stuff. Thanks for making this video. I'd give you a dollar if i had one.

[@MrFinkFixes]

Thanks for including these metrics!

[@dasPartyhut]

I would love it if you also included a frametime graph like MindBlank Tech

[@Sanju007ism]

Thank you for doing this, I was confused for a long time about this.

[@JonesJr_art]

Great video Kyle! I really like this method over min frame rate. It is obviously a better metric for playability and that's what it's all about!

[@dazai9727]

my favourite tech channel

[@ajdinalshareef2370]

best vid on this topic, thx, got yourself a new sub

[@coltauto22]

Great video. Thanks for the information!

[@MrBernardoMetal]

This is pretty damn nice, Kyle!

[@jameo3606]

Thanks for explaining well, actually understood everything instead of getting lost 5 seconds in like I usually do in every other video.

[@SimonLeBonbonbon]

This has become my new reference video for people that ask what the 0.1% lows or 99 percentile is. THANXS Kyle!

[@w1dowtaker]

Great stuff. Really informative video!

[@bigDmtb622]

I never asked what the %1 and %.1 fps meant because I knew this video was coming.

[@qwazyd0gg]

Dig the love towards GN! Great explanation of a great methodology

[@miguelagueda3928]

I'm glad you've started using this method, it's much more accurate. Good job

[@PaulRoneClarke]

Steve from GN often spends 20 minutes explaining unique testing methodology and not just explaining them but justifying them. I used to think he was a beatnik, but now I have respect for the hairy dude.

[@jasonrummery6316]

That makes sense. Thanks man 👍

[@TheManicGeek]

Excellent job explaining the material, bruh! I think I killed, like, at LEAST 24 minimum frame rates last night. It was strangely liberating...

[@marinb_yt]

Also, thank you a LOT for everything you do, Kyle. Will definitely sub to Ultra when I start getting my own income ❤️

[@johnconnor2402]

Great explanation Kyle! This is why I'm going with Ryzen for my next build!

[@ceeme03]

Ok. I'm sold, this makes perfect sense

[@GeoSebastians]

This was very helpful! Thank you, Kyle :)

[@n1kobg]

Thank you for this video.Many people are ignorant and argue about things they dont understand.

[@jacquesgouws]

Kyle you are awesome.. Thanks for this

[@christopherdonald9001]

Kyle, just a suggestion. I understand what the 1% and 0.1% lows mean (huge GamersNexus fan), but for those that may not see this video you should put in each graph something like "less spread between average FPS, 1% low and 0.1% low is better" like you explained at 3:22 in the video. Similar to how you do "lower is better" or "higher is better" in older videos.

[@christopherdonald9001]

I made a quick example for you in photoshop: i.imgur.com/KBDEGDd.jpg

[@InsaneDire]

tighter is better.. oh wait...

[@Hardwareunboxed]

The spread really doesn’t matter, as long as all data sets show the same or at least similar trends. It just means you have benchmarked a section of the game that offers a varied workload.

[@kirigayakazuto1997]

Hardware Unboxed i still cant understand actually.. hhaha.. based on the example, can u tell me which is better n which is not so i can study how to differentiate them..

[@christopherdonald9001]

My point is not the technicality of what 1% and 0.1% lows mean. Just put in the graph what is "better" so everyone knows what they are looking at when they see a comparison.

[@justinpatterson5291]

I've been paying more attention to the 1pc lows these days. Only cos I noticed texture pop in and high particle scenes caused stutters in my old PC.

[@slaughterround643]

Thank you so much for this video!

[@neatoelectro3687]

great video Kyle!

[@shayson1357]

yes yes BUT what causes the frame rates to dip towards these 1% and 0.1% lows ?, in your yellow dots chart 3:02 it shows the 1% as being very few frames with high latency very far off from the rest of the normal frames, anyone know the cause of that ?

[@VJDrover]

Looking at benchmarks from the 1%/0.1% low standpoint is a refreshing way to view gaming performance. Thanks for the informative video Kyle :)

[@klyded8523]

Glad you've finally learned what the pros have know for years kyle. Frametime is al that matters.

[@JohnPurcell]

you should move to q-plots/cdf. would allow you to show cards against eachother more fairly while not muddying the vis with stats that might seem unfair

[@troyp467]

Thanks, I had no idea what that meant when I kept seeing it everywhere. Good video

[@deviroz]

these high frame times average fps are a very clever metric, they are easier to reason about than the variance and reflect better on user experience. I like it, always found silly the min max and average reports everyone gives.

[@bryanbruebaker1507]

The big question I have is, is how do you go about fixing these drops? It happens every now and then in several games I play and is really irritating when you pay so much for a PC that is supposed to provide a smooth experience.

[@frighteningspoon]

Indeed, Keep them into your benchmarks, please. Thank you for explaining this!

[@behimezio]

thank you so much sir . now i understand what it means

[@GrizzlyAmish]

Thanks for the explanation - this helped out a lot!

[@cortbelmont]

A consistent frametime is what gives a game a perfect sense of motion, it is more important than average framerate. I think this new methodology to measure performance is good but it's still better to do it by eye on games you know. my favorite benchmarks today are DOOM 2016 because I know it runs perfectly, it has a consistent frametime all the time. Rise of the Tomb Raider in the Geothermal Valley because I know that game drops framerate at least 30% in every graphics card while you roam that area, and MGSV-TPP it is capped at 60fps but the graphics engine is very light so you can push higher resolutions and it should run smoothly at 16.6ms frametime all the time, no drops

[@r4gequitter389]

Steve Burke is a awesome guy. Love his content. Arguably my favorite tech youtuber.

[@Tyranidex]

Oooooooh, thanks dude for that clarification.

[@Roch_C]

thank you for explaining this!

[@DHettNeck]

Now you only need to start show the results as one colour marked bar instead of 3 different ones ;) That would save much space and makes charts with more than 3 comparisons much easier to read.

[@thomass647]

This video was really helpful. Thanks :)

[@crazyleoz2318]

Great video! So I have a new xfx Merc 319 6800xt and it has terrible low end frames. Benchmarked on horizon zero dawn I had: 78 avg FPS, 164 max fps, and 2 min FPS. When it hits those 2 FPS it’s like hitting a wall, so does that mean I have a defective gpu?

[@GoodWoIf]

Frametimes would be best displayed in a bellcurve with average in the centre. Between the average and the spread, you'd have a comprehensive understanding of the typical performance of whatever game in whatever config you're looking at.