AMD Ryzen vs. Intel System Latency Benchmark: Best Gaming CPUs for Fortnite, CSGO, etc.
video description
Date: 2020-08-28
Related videos
Comments and reviews: 9
Xeigren
I'm not sure how you would cover this but I'd be interested in some conversation about laptop thermals.
I've been looking into buying a gaming laptop and it seems even on high end laptops with seemingly well designed cooling you can get CPU/GPU temps that hit 100C while gaming.
I am the kind of guy that will use a computer for 10 years so this doesn't exactly sit well with me, considering at those temperatures you're going to be degrading the performance and/or lifespan of other components in the system. Let alone other temperature/heat cycling related failures and issues that can come with that (solder joint cracking, PCB delamination, warping, etc).
Personally I'm going to be replacing the thermal paste and pads in whatever laptop I end up buying along with doing any tricks I can pull to reduce thermals to be in a more reasonable range, even at the cost of gasp performance. Looks like that'll end up being a mix of using ThrottleStop, Afterburner, and NoteBook FanControl to do some undervolting and tuning to bring the temps down then control what the system does when it reaches the temperature ceiling I set.
Notes for other commentors, not really GN:
I am not the average user, I do more than gaming and will regularly max out the CPU and/or GPU for hours at a time. Yes a desktop/using a remote machine is a better solution. Yes just because the CPU/GPU is that hot doesn't mean the rest of the system is that hot, especially if the cooling solution works well. Millions of dollars of R&D do not change the laws of physics nor can they account for every possible use case, environment, supply chain issue, manufacturing issue/variance, or human error. I both repair and design electronics semi-professionally and have had to deal with heat related failures, so these are very real concerns to me.
Even though an electronic component may have an operating temperature range of say -40C to 105C (ambient and/or junction) its performance/characteristics can vary quite a bit over that range and of course prolonged exposure to the high or low end will decrease the overall lifespan. This includes passive components like resistors and things like capacitors are often rated for say 5000 hours at 105C. Ratings of course only mean something if they've been throughly tested and aren't a guess.
Thank you for coming to my TED Talk lol
reply
I'm not sure how you would cover this but I'd be interested in some conversation about laptop thermals.
I've been looking into buying a gaming laptop and it seems even on high end laptops with seemingly well designed cooling you can get CPU/GPU temps that hit 100C while gaming.
I am the kind of guy that will use a computer for 10 years so this doesn't exactly sit well with me, considering at those temperatures you're going to be degrading the performance and/or lifespan of other components in the system. Let alone other temperature/heat cycling related failures and issues that can come with that (solder joint cracking, PCB delamination, warping, etc).
Personally I'm going to be replacing the thermal paste and pads in whatever laptop I end up buying along with doing any tricks I can pull to reduce thermals to be in a more reasonable range, even at the cost of gasp performance. Looks like that'll end up being a mix of using ThrottleStop, Afterburner, and NoteBook FanControl to do some undervolting and tuning to bring the temps down then control what the system does when it reaches the temperature ceiling I set.
Notes for other commentors, not really GN:
I am not the average user, I do more than gaming and will regularly max out the CPU and/or GPU for hours at a time. Yes a desktop/using a remote machine is a better solution. Yes just because the CPU/GPU is that hot doesn't mean the rest of the system is that hot, especially if the cooling solution works well. Millions of dollars of R&D do not change the laws of physics nor can they account for every possible use case, environment, supply chain issue, manufacturing issue/variance, or human error. I both repair and design electronics semi-professionally and have had to deal with heat related failures, so these are very real concerns to me.
Even though an electronic component may have an operating temperature range of say -40C to 105C (ambient and/or junction) its performance/characteristics can vary quite a bit over that range and of course prolonged exposure to the high or low end will decrease the overall lifespan. This includes passive components like resistors and things like capacitors are often rated for say 5000 hours at 105C. Ratings of course only mean something if they've been throughly tested and aren't a guess.
Thank you for coming to my TED Talk lol
reply
Alexander
reduce buffering in overwatch straightens the pipeline and synchronizes the cpu and gpu in making frames, making gpu wait for the cpu to prepare one, however long it takes. It lowers fps in some cpu-bound scenarios, of which the game has plenty when big teamfights are going on in some of the maps, however, if your cpu is able to maintain the cap that it s at and not go lower at all, it will actually reduce the buffering by AT MAX 1 FRAME. In case of stable, never-dipping at any times, 300 fps, that s just 3.33ms max.
In case you are limited by your gpu, reduce buffering not only isn t helping, it s actually making things worse. That s (and another bug with gpu usage / scheduler) is why you have to toggle reduce buffering off and back on if you tabbed out of fullscreen game and back into it.
The game itself is able to tell you whether or not it s cpu (or ram, actually, which matters a whole lot for 1% lows in this game even on minimal settings apart from render scale, going as far as disabling shadows) or gpu bottlenecked - if there are 3 dots near your fps counter - it s gpu bottlenecked. If there is only 1 - it s cpu/ram bottlenecked.
As someone who plays it to a somewhat high standard, I tested things myself and it s certainly better to have your fps capped at, say, 210 instead of 300 but never dip below that, and have reduce buffering on. Otherwise, in massive teamfights, especially on some maps (Oasis and Volskaya Industries in particular are known to lag more often) your frametime is increased, the delay between your actions happening irl and in-game is increased, and it can lead to weird feeling of the game and/or more mistakes, which, in turn, might cost you the match. Surely it s a very rare occasion, but it s better to protect yourself even from those.
About half the high-level-players streamers (and professional players) in the game prefer to cap the framerate at stable 250fps instead of 300, have reduce buffering on, and enjoy a steady experience.
P.S. if you are indeed gpu-bottlenecked, turn on Nvidia Ultra Low Latency mode (amd has a similar setting but it escapes my memory what the title of it is)
P.P.S. Huge thanks to GN team for all this work, great testing as always.
reply
reduce buffering in overwatch straightens the pipeline and synchronizes the cpu and gpu in making frames, making gpu wait for the cpu to prepare one, however long it takes. It lowers fps in some cpu-bound scenarios, of which the game has plenty when big teamfights are going on in some of the maps, however, if your cpu is able to maintain the cap that it s at and not go lower at all, it will actually reduce the buffering by AT MAX 1 FRAME. In case of stable, never-dipping at any times, 300 fps, that s just 3.33ms max.
In case you are limited by your gpu, reduce buffering not only isn t helping, it s actually making things worse. That s (and another bug with gpu usage / scheduler) is why you have to toggle reduce buffering off and back on if you tabbed out of fullscreen game and back into it.
The game itself is able to tell you whether or not it s cpu (or ram, actually, which matters a whole lot for 1% lows in this game even on minimal settings apart from render scale, going as far as disabling shadows) or gpu bottlenecked - if there are 3 dots near your fps counter - it s gpu bottlenecked. If there is only 1 - it s cpu/ram bottlenecked.
As someone who plays it to a somewhat high standard, I tested things myself and it s certainly better to have your fps capped at, say, 210 instead of 300 but never dip below that, and have reduce buffering on. Otherwise, in massive teamfights, especially on some maps (Oasis and Volskaya Industries in particular are known to lag more often) your frametime is increased, the delay between your actions happening irl and in-game is increased, and it can lead to weird feeling of the game and/or more mistakes, which, in turn, might cost you the match. Surely it s a very rare occasion, but it s better to protect yourself even from those.
About half the high-level-players streamers (and professional players) in the game prefer to cap the framerate at stable 250fps instead of 300, have reduce buffering on, and enjoy a steady experience.
P.S. if you are indeed gpu-bottlenecked, turn on Nvidia Ultra Low Latency mode (amd has a similar setting but it escapes my memory what the title of it is)
P.P.S. Huge thanks to GN team for all this work, great testing as always.
reply
Nick
Hello GN!
I very much like your in-depth performance reviews of gaming hardware, however the question that I was always wondering is not the absolute difference between different CPU s/GPUs that face little to no bottle-neck when paired with RTX 2080Ti/i9 9900K respectively but rather the difference in performance between systems of completely different classes.
In other words I m wondering: what a system that costs 150/ 200/ 500 more would achieve?
Such a review would
- builds a better combined picture of hardware performance
- new, exciting and refreshing for the audience
- can be done with existing tech while the there s a drought in new tech releases until Autumn
- I suppose it would be easy to do since every system is a different chassis and there s no need to swap components around
reply
Hello GN!
I very much like your in-depth performance reviews of gaming hardware, however the question that I was always wondering is not the absolute difference between different CPU s/GPUs that face little to no bottle-neck when paired with RTX 2080Ti/i9 9900K respectively but rather the difference in performance between systems of completely different classes.
In other words I m wondering: what a system that costs 150/ 200/ 500 more would achieve?
Such a review would
- builds a better combined picture of hardware performance
- new, exciting and refreshing for the audience
- can be done with existing tech while the there s a drought in new tech releases until Autumn
- I suppose it would be easy to do since every system is a different chassis and there s no need to swap components around
reply
Igor
You took the least cpu-intense scenarios and ran tests on these because they produce results. Mkay. But ultimately those are useless as you've mentioned yourself in the CS part - you didn't want to have a player in sight as it greatly increases fps and reduces latency, and yet, ultimately your tests were doing exactly this - reduce all load on cpu.
I realize it would be difficult to do your tests in real conditions, but achieved results are meaningless. The only thing they show is that you have methodology to test what you've tested, but has no implications to real world and for the choosing a cpu. I've played at least half a dozen cybersport games which had excellent fps in test maps, but tanked like hell in real matches, so your conclusion couldn't have been farther from truth.
reply
You took the least cpu-intense scenarios and ran tests on these because they produce results. Mkay. But ultimately those are useless as you've mentioned yourself in the CS part - you didn't want to have a player in sight as it greatly increases fps and reduces latency, and yet, ultimately your tests were doing exactly this - reduce all load on cpu.
I realize it would be difficult to do your tests in real conditions, but achieved results are meaningless. The only thing they show is that you have methodology to test what you've tested, but has no implications to real world and for the choosing a cpu. I've played at least half a dozen cybersport games which had excellent fps in test maps, but tanked like hell in real matches, so your conclusion couldn't have been farther from truth.
reply
Addy
Thanks for the hard work man. But I noticed on this video and thermal compound video, u saying that the difference is negligible. I think to be qualified to say a difference is negligible u need to prove its not consistent eg. alternating or fluctuating. Ideally do statistics. If I'm gonna run the same processor for the next 5 years playing fps 5 hours a day then even 1ms is a BIG big difference IF consistent throughout my 5 years . If u don't want to do statistics u could alternatively do the global average latency between all the games and provide just one average and one Stdev for each cpu. Could you please publish this in description or next video? That would be so awesome. Anyways keep up the good work and thanks
reply
Thanks for the hard work man. But I noticed on this video and thermal compound video, u saying that the difference is negligible. I think to be qualified to say a difference is negligible u need to prove its not consistent eg. alternating or fluctuating. Ideally do statistics. If I'm gonna run the same processor for the next 5 years playing fps 5 hours a day then even 1ms is a BIG big difference IF consistent throughout my 5 years . If u don't want to do statistics u could alternatively do the global average latency between all the games and provide just one average and one Stdev for each cpu. Could you please publish this in description or next video? That would be so awesome. Anyways keep up the good work and thanks
reply
pino
Shouldn't ram speed and cpu clock frequency influence this? My cpu latency goes down significantly when i oc my ram from 2666 cl16 (stock) to 4000 cl16 and again a little when i oc then 9900k to 5.1 from 4.7. Or are they at base already much faster then the input from the mouse with a 1000 hz polling rate?
Also what mouse was used and was this confirmed to be set at 1000 hz polling rate? My 1000 polling rate mouse has a default of 200 Hz for some reason which gave me serious grief before i figured that out. Also why are your latencys so much lower then those tests done by battlenonsense?
reply
Shouldn't ram speed and cpu clock frequency influence this? My cpu latency goes down significantly when i oc my ram from 2666 cl16 (stock) to 4000 cl16 and again a little when i oc then 9900k to 5.1 from 4.7. Or are they at base already much faster then the input from the mouse with a 1000 hz polling rate?
Also what mouse was used and was this confirmed to be set at 1000 hz polling rate? My 1000 polling rate mouse has a default of 200 Hz for some reason which gave me serious grief before i figured that out. Also why are your latencys so much lower then those tests done by battlenonsense?
reply
Axel
when my pc has problems i used a useful little program from www.Thesycon.de , it's called DPC Latency Checker
what helps me to determine something is defective or whether driver problems.
with this i could prevent a manufacturer from refusing the guarantee of my mainboard. I was able to explain what wasn't working properly by doing the tests that are described in the PDF and it was definitely not due to my driver software
in my case it was a broken PCIe controller chip that was responsible for the LAN and USB 3.0
By the way: installation-free software
reply
when my pc has problems i used a useful little program from www.Thesycon.de , it's called DPC Latency Checker
what helps me to determine something is defective or whether driver problems.
with this i could prevent a manufacturer from refusing the guarantee of my mainboard. I was able to explain what wasn't working properly by doing the tests that are described in the PDF and it was definitely not due to my driver software
in my case it was a broken PCIe controller chip that was responsible for the LAN and USB 3.0
By the way: installation-free software
reply
jmich69
I'd be really interested in seeing a video cover and explain how latency and performance scales across RAM with different speeds and timings (primarily CAS). I feel most of us have a general understanding; higher speeds and lower CAS = better. But how much of a difference is there really between sticks of the same speed, but different timings? 3600MHz C18 vs 3600MHz C16 and 3600MHz C14? And then how about 3600MHz C18 vs 3200MHz C14?
reply
I'd be really interested in seeing a video cover and explain how latency and performance scales across RAM with different speeds and timings (primarily CAS). I feel most of us have a general understanding; higher speeds and lower CAS = better. But how much of a difference is there really between sticks of the same speed, but different timings? 3600MHz C18 vs 3600MHz C16 and 3600MHz C14? And then how about 3600MHz C18 vs 3200MHz C14?
reply
_Nism0
I've never liked these tests. Having both an Intel and AMD setup, all overclocked, 4300Mhz RAM (3800Mhz on Ryzen), all C-States off and a debloated Windows install, all for the lowest DPC latency, there is a noticable difference, Intel has lower input and DPC latency. Not that I'm a fanboy, I recommend Ryzen to 90% of people and am looking forward to 4000 series but this from my own testing and my communities general opinion.
reply
I've never liked these tests. Having both an Intel and AMD setup, all overclocked, 4300Mhz RAM (3800Mhz on Ryzen), all C-States off and a debloated Windows install, all for the lowest DPC latency, there is a noticable difference, Intel has lower input and DPC latency. Not that I'm a fanboy, I recommend Ryzen to 90% of people and am looking forward to 4000 series but this from my own testing and my communities general opinion.
reply
Add a review, comment
Other channel videos
