Frame Time Testing: What It Means and Why We Care

The concept of frame time testing may seem complex and technical, but it's actually quite straightforward once you understand what it entails. In essence, frame time testing involves measuring the number of frames per second that are displayed on screen within a certain percentage of the total frame rate. This can be expressed as 1% or 0.01, and it represents the point at which the game starts to stutter and become unresponsive. For example, if a game is running at an average FPS of 60, but the frames below 60 are only reaching 30, then that's considered a high frame time. This can be damaging to the gameplay experience, as it disrupts the smoothness and immersion that players expect.

We test for frame times because they're actually very revealing about how well a GPU is performing in real-world scenarios. In the past, drivers were often poor performers when it came to testing, with some old drivers struggling to even reach 1% or 0.01 frame times. However, as drivers have improved, so has our ability to accurately measure and report on frame times. Now, we can get a more accurate picture of how well a GPU is performing in games that are running at high frame rates.

The impact of frame time testing cannot be overstated. If you're playing a game with a high frame rate, but the frames below that point are stuttering and dropping, it's not just your imagination - there's actually something wrong with the game. This can be frustrating for players who expect a smooth gaming experience, and it's something that we want to help mitigate as much as possible.

One of the key issues that frame time testing helps us address is the difference between benchmarking metrics and actual gameplay performance. In the past, games were often optimized with high frame rates in mind, but this would sometimes come at the cost of poor performance in real-world scenarios. By testing for frame times, we can get a more accurate picture of how well a game is performing in practice.

In terms of specific measurements, 1% or 0.01 frame time refers to the point at which the frames are dropping below that threshold. This is where the stuttering and frame drops become most noticeable, and it's here that our testing becomes most relevant. We can measure this using specialized software tools, such as those provided by NVIDIA or AMD, which help us identify areas of performance improvement.

For example, if a GPU is running at 60 FPS, but the frames below 60 are only reaching 30, then that's considered a high frame time of 0.5 seconds per second. This may not seem like a lot, but it can actually make a big difference in the overall gaming experience. In reality, these small variations in performance can add up to significant differences in terms of immersion and responsiveness.

One thing we've found is that some GPUs are more prone to frame time issues than others. For example, older drivers or lower-end GPUs may struggle with frame times, even at high frame rates. On the other hand, higher-end GPUs may be able to maintain a consistent performance level, even when pushing for high frame rates.

In terms of advice on how to improve your gaming experience, we would strongly recommend looking into frame time testing and understanding more about what it entails. There are some excellent resources out there, such as PC Perspective, which have done a lot of pioneering work in this area. By understanding more about frame times and how they affect performance, you can take steps to optimize your system and ensure that you're getting the best possible gaming experience.

One question we received from bu7 was whether or not it's worth getting a superclocked 980 Ti over a non-superclocked version. In short, the answer is no - unless there's a very small price difference, which would be less than $10. We've done tests in the past on superclocking and found that it doesn't make a huge difference in terms of performance, especially at lower frame rates.

That being said, if you do want an overclocked card, it's not hard to apply most moderate pre-overclocks using software tools like Precision or Afterburner. This can be done relatively easily and can provide a small boost to performance, but it's generally not worth the extra cost unless there's a significant price difference.

One final question came from Jan and Christopher, who asked whether superclocking is "worth getting". The answer, as we've discussed above, is no - unless there's a very small price difference. However, we did want to acknowledge that some people may still find it worth getting, especially if they're looking for the latest and greatest technology.

And finally, I'd like to thank everyone who has been commenting on our videos and supporting us on Patreon. We really appreciate your input and feedback - it helps us shape our content and ensure that we're producing something you enjoy. If you haven't already, please consider checking out our patreon page for more information on how to support us.

"WEBVTTKind: captionsLanguage: enhey everyone I'm Steve from gamers Nexis Donna and today we're doing something a bit different we are doing an ask GN video so this is something I wanted to do for a while basically we're taking your comments on the YouTube channel on Twitter and stuff like that and we're just gonna address some of them here for the comments that need a little more depth or are perhaps just interesting comments in general and I'm sure you all know what I'm talking about so that's what we're doing here this is the first ask GN video if you have any questions for the next video leave a comment below or on any of the other videos on the channel I do read almost all of them for the most part and we're gonna start right off here with a comment from blue blur 50h posted on the AMD a10 7870 K ap review that we post posted recently and blue blur says good lord what is that now I'm not really sure to what blue blur is referring with this comment but I would assume that to answer your question the obvious choice is either going to be hair or the AMD APU that we were reviewing so that hopefully answers that question we've been out of something a bit more in-depth there is a good question from excuse me on the pronunciation here subic Chakraborty who posted on the intel i7 6700 k review we did recently that would be the skylake platform the first exposure to sky like that we had and Subic asks does skylake support ddr3 this is actually a really good question and one that does need to be addressed so skylake is an interesting platform it supports two different types of memory it Sports ddr3l and ddr4 and that L indicator is actually very important it's not as simple as just tagging an L at the end of ddr3l actually means low voltage and ddr3l memory chips have 208 pins which is a fair bit fewer than ddr3 as a base platform and a lot fewer than ddr4 which is something like 280 or to 88 I'll out my head and ddr3l is something that has been around for a little while it's used very commonly in certain types of laptops and other mobile devices or ultra portable computers and ddr3l draws less voltage which means that it's clock rates generally going to be lower across the board it also means that it has less overclocking Headroom depending on what you're looking to do with the memory so ddr3l and ddr4 are quite a bit different but ddr3l and ddr3 are also different so yes skylake does support a form of ddr3 but it is the low-voltage form so you have to buy special memory for it and to that end it is important that I address something here a lot of review outlets ours included have mentioned the option of purchasing motherboards for skylake with ddr3l support rather than ddr4 as a means of reducing build cost and that is true you can buy cheaper cheaper ddr3l memory and some of the platforms that supported our cheaper as well but the one item of note here is that if you have an existing system with ddr3 memory for Haswell or similar you can't rip that memory out and just move it over to skylake because it's not the correct low-voltage form of ddr3 so that is an important item that is worth pointing out and a good question by Subic the next question is sort of interesting it's from 21 EC and this was posted on one of our other recent videos so what I want to see asks coming from aq 9550 should I pick up the i7 4790k or the i7 6700 K and then goes on to add parenthetically I'm a gamer but I'm also working with 3ds max so that's something that actually changes the answer to the question question quite a bit the first and shortest bit of information I should say here is that I have not specifically tested the 4790k or 6700 K in 3ds max I have not compared their performance against one another in that particular piece of rendering software so I don't have first-hand knowledge of the exact gain from the 6700 K if there is and 3ds max but we can talk theory first of all coming from IQ 9550 really anything you buy either of those two CPUs is going to be so profoundly advanced in comparison that you really are probably not gonna be unhappy with your purchase and the q95 50 was a very good cpu that was one of the first quad-core CPUs it came out around the time the Q 6600 came out which was actually my first quad-core CPU and the same for a lot of other people I believe so cuteee 9550 was good but it's definitely dated the thing with the the sky link and the has well platforms for 3ds max and other professional software is the memory support so ddr4 if you buy a skylake platform with ddr4 is going to actually output a much greater performance for software like 3ds max maya Adobe Premiere Adobe Photoshop when you're working with very large files or you're doing batch processing of files and the reason for that is because ddr4 has a much greater memory bandwidth part of which is fueled by its increased frequency so ddr4 you can find memory in the 31:33 range pretty easily and there's some stuff that's listed at 3800 I haven't really tested that yet I haven't validated that it's stable but 31:33 definitely is if frequency you can buy a ddr4 pretty safely so the increased memory bandwidth will actually be advantageous to you if you're a very serious 3dsmax user or really if you're a serious production user in general and that is something that should be taken into account but when I say serious I mean this is your job you're doing this all day every day you're sitting there rendering out files if you're working on video files because that's what I'm most familiar with then the kind of person who will benefit from ddr4 is the kind of person who is syncing entire days into rendering or producing content so that's really where you see your disparity there at the end of the day the 4790k is a very good cpu you can get pretty fast ddr3 and coming from aq 9550 there's really no wrong choice if you're working in a professional environment then definitely investigate the option of the 6700 K and ddr4 I'm not a source of authority on that because I haven't tested 3ds max personally I would recommend websites like Puget systems they are the system integrator so keep that in mind grain of salt etc but they do some pretty cool tests and we actually use their tests as a source for what we put in our render rig for work you know Adobe Premiere and there are other sites similar to that that work with the creative tools we're more of a gaming site so next question here from Morningstar says I don't understand what it does mean hi 1% low performance and 0.1% performance things what does it mean so this is actually a good question as well Morningstar is asking about our charts that we published in almost every review for CPUs and GPUs anyone who's seen them on the channel knows that we publish average FPS 1% low FPS and 0.1% low fps and these three numbers mean different things but all of them are very important so traditionally a lot of sites especially in days past will publish the minimum and maximum FPS numbers with their benchmarking we don't do this because we feel that minimums and maximums are outliers and are unimportant and really not indicative of overall performance because it's an outlier so the problem with minimum is you might hit the benchmarking key the recording key a quarter second too early and for example shadow of mordor and if you do that in shadow of mordor what happens is you could actually hit a serious very brief spike I have a hair in my face and the problem with this is it's not representative of in gameplay it's popping up because something to do with the shadow of mordor benchmark is loading an extra overlay so that's an example of where you would want to one refine test methodology but to not record the minimum FPS because it's not actually happening now in actual gameplay like with Skyrim if you have a very brief less than half a second period of loading a ton of data or dumping previous data you might have a one FPS output for a fraction of a sec so that's another place where minimum is not great to measure and it it's important to note that point 1 percent and 1 percent actually do take some of the lower times but they average them so in the case of 1% low it's the 99th percentile performance of the game so we take the slowest one percent of all frames delivered in the thirty second test period all three of the thirty-second test periods we take the slowest chunk of those the 1% chunk and then we average that chunk and that is our 1% below so in the case of a lot of games maybe you have a 60 FPS average output but you have a 30% 1% if I have a 31% low one of the best examples for why we measure these metrics is GTA 5 so we recently published a CPU benchmark looking at the g3 2 5 8 and the 7 6 DK and a number of other low-end CPUs and the thing here is that the g3 2 5 8 outperforms the 760 K Richland the APU in terms of average FPS but the 760 K had a better point 1 percent low the GC 2 5 8 had a for FPS point 1 percent low but a faster average FPS so this is important because when you're playing the game and you get really fierce stuttering or framedrops or whatever you want to call it colloquially colloquially the the slow frames that you're seeing can be measured in point 1 percent or 1 percent lows and that is actually very damaging to the gameplay experience and it's something that we want to reflect in our testing and convey to the readers what's actually going on here because with some drivers a.m. these old drivers and some of Nvidia's past drivers before this became a thin frame time testing became a thing they were actually very poor performing metrics for certain GPUs when looking at the 1% in point 1 for some Louis but their average FPS was very high in benchmarking that looks great in real-world gameplay you you see and pick up on those stutters and frame drops and it's really a lot harder to get in the immersion of the experience so that is what those numbers represent I'll try and get a full video and article op online in the future talking more in depth about all of that and how it all works and how we measure it but that's sort of a brief overview for those who are unfamiliar with what it means and why we test it I would strongly recommend or suggest looking at tax report and PC perspective for more information on frame time testing they innovated a lot of it and they've done a pretty good job reporting on it one of the last questions here is it's from bu7 saying kill illuminati killuminati something with the illuminati asks is 980 superclocked worth getting so we've done a video on superclocked versus non superclocked cards in the past I'm assuming Illuminati here is referring to the EVGA 980 because that is generally who calls their pre overclocking super clocking so with super clock and all that means is that they have pre overclocked the video card or the GPU beyond reference and just because the 980ti is something I've memorized the reference clock of the 980 TI is a thousand megahertz if you by for example a 980 TI hybrid by EVGA I think it's 12 something maybe 12 28 megahertz so that's a pretty substantial overclock and it does actually have a big impact on gaming performance as we've seen in our hybrid or the liquid cooled 980ti testing and that's definitely an important factor but it's a bit of an outlier because normally when you're talking about overclocked and superclocked cards they're not a gap as big as as 200 plus megahertz they don't produce that big of an FPS output so looking at a 750 Ti that we tested ages ago the conclusion of our test was that there is very little impact on the super clocked cards performance with its pre overclocked applied rushes the reference 750ti which we also had we had both cards and that made it really not worth getting the superclass card unless the price difference was so small that it didn't matter so if you had like a 750 Ti superclocked card for $10 more sure get it whatever not a big loss and you might even gain a couple FPS we're talking one to four fps and the average metrics normally for super clocking at the higher end the 980 and similar cards it really it just depends on which model you're talking about and how big the pre overclock is in general if there's a big price difference it's not worth getting it and you can also normally apply that overclocked or pre overclocked setting very easily on your own using precision or afterburner or any number of other software Suites it's pretty easy to apply most of those moderate pre overclock so that is a certainly a back-up plan in the event that you do want an overclocked card hopefully that helps with that question we've tested a few of these things in the past search the channel for super clock or super clock in and you should find it and then finally a comment from Jan and Christopher s tlemen who said first and to that I say don't you or not you're actually the second comment on the video but good try try again on this one that is all for this ask GN video please leave questions in the comments below and in the other videos as always thank you for watching we really appreciate that you guys have been into the content it's a lot of fun to produce and I want to keep producing it the team has been expanding so things are going well over here at GN and of course check out our patreon page if you haven't yet I know I probably sound like a broken record but it does actually help quite a bit to get any amount of support on patreon because we're trying to shift away a bit from the traditional advertising model and that helps a good deal in that endeavor so that's all for this time I will see you all next time easyhey everyone I'm Steve from gamers Nexis Donna and today we're doing something a bit different we are doing an ask GN video so this is something I wanted to do for a while basically we're taking your comments on the YouTube channel on Twitter and stuff like that and we're just gonna address some of them here for the comments that need a little more depth or are perhaps just interesting comments in general and I'm sure you all know what I'm talking about so that's what we're doing here this is the first ask GN video if you have any questions for the next video leave a comment below or on any of the other videos on the channel I do read almost all of them for the most part and we're gonna start right off here with a comment from blue blur 50h posted on the AMD a10 7870 K ap review that we post posted recently and blue blur says good lord what is that now I'm not really sure to what blue blur is referring with this comment but I would assume that to answer your question the obvious choice is either going to be hair or the AMD APU that we were reviewing so that hopefully answers that question we've been out of something a bit more in-depth there is a good question from excuse me on the pronunciation here subic Chakraborty who posted on the intel i7 6700 k review we did recently that would be the skylake platform the first exposure to sky like that we had and Subic asks does skylake support ddr3 this is actually a really good question and one that does need to be addressed so skylake is an interesting platform it supports two different types of memory it Sports ddr3l and ddr4 and that L indicator is actually very important it's not as simple as just tagging an L at the end of ddr3l actually means low voltage and ddr3l memory chips have 208 pins which is a fair bit fewer than ddr3 as a base platform and a lot fewer than ddr4 which is something like 280 or to 88 I'll out my head and ddr3l is something that has been around for a little while it's used very commonly in certain types of laptops and other mobile devices or ultra portable computers and ddr3l draws less voltage which means that it's clock rates generally going to be lower across the board it also means that it has less overclocking Headroom depending on what you're looking to do with the memory so ddr3l and ddr4 are quite a bit different but ddr3l and ddr3 are also different so yes skylake does support a form of ddr3 but it is the low-voltage form so you have to buy special memory for it and to that end it is important that I address something here a lot of review outlets ours included have mentioned the option of purchasing motherboards for skylake with ddr3l support rather than ddr4 as a means of reducing build cost and that is true you can buy cheaper cheaper ddr3l memory and some of the platforms that supported our cheaper as well but the one item of note here is that if you have an existing system with ddr3 memory for Haswell or similar you can't rip that memory out and just move it over to skylake because it's not the correct low-voltage form of ddr3 so that is an important item that is worth pointing out and a good question by Subic the next question is sort of interesting it's from 21 EC and this was posted on one of our other recent videos so what I want to see asks coming from aq 9550 should I pick up the i7 4790k or the i7 6700 K and then goes on to add parenthetically I'm a gamer but I'm also working with 3ds max so that's something that actually changes the answer to the question question quite a bit the first and shortest bit of information I should say here is that I have not specifically tested the 4790k or 6700 K in 3ds max I have not compared their performance against one another in that particular piece of rendering software so I don't have first-hand knowledge of the exact gain from the 6700 K if there is and 3ds max but we can talk theory first of all coming from IQ 9550 really anything you buy either of those two CPUs is going to be so profoundly advanced in comparison that you really are probably not gonna be unhappy with your purchase and the q95 50 was a very good cpu that was one of the first quad-core CPUs it came out around the time the Q 6600 came out which was actually my first quad-core CPU and the same for a lot of other people I believe so cuteee 9550 was good but it's definitely dated the thing with the the sky link and the has well platforms for 3ds max and other professional software is the memory support so ddr4 if you buy a skylake platform with ddr4 is going to actually output a much greater performance for software like 3ds max maya Adobe Premiere Adobe Photoshop when you're working with very large files or you're doing batch processing of files and the reason for that is because ddr4 has a much greater memory bandwidth part of which is fueled by its increased frequency so ddr4 you can find memory in the 31:33 range pretty easily and there's some stuff that's listed at 3800 I haven't really tested that yet I haven't validated that it's stable but 31:33 definitely is if frequency you can buy a ddr4 pretty safely so the increased memory bandwidth will actually be advantageous to you if you're a very serious 3dsmax user or really if you're a serious production user in general and that is something that should be taken into account but when I say serious I mean this is your job you're doing this all day every day you're sitting there rendering out files if you're working on video files because that's what I'm most familiar with then the kind of person who will benefit from ddr4 is the kind of person who is syncing entire days into rendering or producing content so that's really where you see your disparity there at the end of the day the 4790k is a very good cpu you can get pretty fast ddr3 and coming from aq 9550 there's really no wrong choice if you're working in a professional environment then definitely investigate the option of the 6700 K and ddr4 I'm not a source of authority on that because I haven't tested 3ds max personally I would recommend websites like Puget systems they are the system integrator so keep that in mind grain of salt etc but they do some pretty cool tests and we actually use their tests as a source for what we put in our render rig for work you know Adobe Premiere and there are other sites similar to that that work with the creative tools we're more of a gaming site so next question here from Morningstar says I don't understand what it does mean hi 1% low performance and 0.1% performance things what does it mean so this is actually a good question as well Morningstar is asking about our charts that we published in almost every review for CPUs and GPUs anyone who's seen them on the channel knows that we publish average FPS 1% low FPS and 0.1% low fps and these three numbers mean different things but all of them are very important so traditionally a lot of sites especially in days past will publish the minimum and maximum FPS numbers with their benchmarking we don't do this because we feel that minimums and maximums are outliers and are unimportant and really not indicative of overall performance because it's an outlier so the problem with minimum is you might hit the benchmarking key the recording key a quarter second too early and for example shadow of mordor and if you do that in shadow of mordor what happens is you could actually hit a serious very brief spike I have a hair in my face and the problem with this is it's not representative of in gameplay it's popping up because something to do with the shadow of mordor benchmark is loading an extra overlay so that's an example of where you would want to one refine test methodology but to not record the minimum FPS because it's not actually happening now in actual gameplay like with Skyrim if you have a very brief less than half a second period of loading a ton of data or dumping previous data you might have a one FPS output for a fraction of a sec so that's another place where minimum is not great to measure and it it's important to note that point 1 percent and 1 percent actually do take some of the lower times but they average them so in the case of 1% low it's the 99th percentile performance of the game so we take the slowest one percent of all frames delivered in the thirty second test period all three of the thirty-second test periods we take the slowest chunk of those the 1% chunk and then we average that chunk and that is our 1% below so in the case of a lot of games maybe you have a 60 FPS average output but you have a 30% 1% if I have a 31% low one of the best examples for why we measure these metrics is GTA 5 so we recently published a CPU benchmark looking at the g3 2 5 8 and the 7 6 DK and a number of other low-end CPUs and the thing here is that the g3 2 5 8 outperforms the 760 K Richland the APU in terms of average FPS but the 760 K had a better point 1 percent low the GC 2 5 8 had a for FPS point 1 percent low but a faster average FPS so this is important because when you're playing the game and you get really fierce stuttering or framedrops or whatever you want to call it colloquially colloquially the the slow frames that you're seeing can be measured in point 1 percent or 1 percent lows and that is actually very damaging to the gameplay experience and it's something that we want to reflect in our testing and convey to the readers what's actually going on here because with some drivers a.m. these old drivers and some of Nvidia's past drivers before this became a thin frame time testing became a thing they were actually very poor performing metrics for certain GPUs when looking at the 1% in point 1 for some Louis but their average FPS was very high in benchmarking that looks great in real-world gameplay you you see and pick up on those stutters and frame drops and it's really a lot harder to get in the immersion of the experience so that is what those numbers represent I'll try and get a full video and article op online in the future talking more in depth about all of that and how it all works and how we measure it but that's sort of a brief overview for those who are unfamiliar with what it means and why we test it I would strongly recommend or suggest looking at tax report and PC perspective for more information on frame time testing they innovated a lot of it and they've done a pretty good job reporting on it one of the last questions here is it's from bu7 saying kill illuminati killuminati something with the illuminati asks is 980 superclocked worth getting so we've done a video on superclocked versus non superclocked cards in the past I'm assuming Illuminati here is referring to the EVGA 980 because that is generally who calls their pre overclocking super clocking so with super clock and all that means is that they have pre overclocked the video card or the GPU beyond reference and just because the 980ti is something I've memorized the reference clock of the 980 TI is a thousand megahertz if you by for example a 980 TI hybrid by EVGA I think it's 12 something maybe 12 28 megahertz so that's a pretty substantial overclock and it does actually have a big impact on gaming performance as we've seen in our hybrid or the liquid cooled 980ti testing and that's definitely an important factor but it's a bit of an outlier because normally when you're talking about overclocked and superclocked cards they're not a gap as big as as 200 plus megahertz they don't produce that big of an FPS output so looking at a 750 Ti that we tested ages ago the conclusion of our test was that there is very little impact on the super clocked cards performance with its pre overclocked applied rushes the reference 750ti which we also had we had both cards and that made it really not worth getting the superclass card unless the price difference was so small that it didn't matter so if you had like a 750 Ti superclocked card for $10 more sure get it whatever not a big loss and you might even gain a couple FPS we're talking one to four fps and the average metrics normally for super clocking at the higher end the 980 and similar cards it really it just depends on which model you're talking about and how big the pre overclock is in general if there's a big price difference it's not worth getting it and you can also normally apply that overclocked or pre overclocked setting very easily on your own using precision or afterburner or any number of other software Suites it's pretty easy to apply most of those moderate pre overclock so that is a certainly a back-up plan in the event that you do want an overclocked card hopefully that helps with that question we've tested a few of these things in the past search the channel for super clock or super clock in and you should find it and then finally a comment from Jan and Christopher s tlemen who said first and to that I say don't you or not you're actually the second comment on the video but good try try again on this one that is all for this ask GN video please leave questions in the comments below and in the other videos as always thank you for watching we really appreciate that you guys have been into the content it's a lot of fun to produce and I want to keep producing it the team has been expanding so things are going well over here at GN and of course check out our patreon page if you haven't yet I know I probably sound like a broken record but it does actually help quite a bit to get any amount of support on patreon because we're trying to shift away a bit from the traditional advertising model and that helps a good deal in that endeavor so that's all for this time I will see you all next time easy\n"