For this video we re-used the 7700k and the 1800x from the previous video, configured the same.
We also added a 1080Ti to the Ryzen system to see if we could push the fps up to see if "thats too many fps for gta to handle" was part of the problem we were seeing on the intel system (e.g. so many fps = stutters as we've seen on like 4 core gta v systems).
Interestingly, the intel numbers on the graph look somewhat smoother and more consistent than the Ryzen system EXCEPT for the stutters that last ~100ms. We can see the stutters in both canned benchmarks and real playthroughs at 1080p on GTA V.
We think GTA V is probably just crap to test at high fps, but we're not 100% sure what to make of it.
We looked for correlation between the stutters and many things we could log in the system. We also checked input lag or input latency and built a mouse to help us measure input lag and confirmed it seems identical between systems on the three games we tested.
USB was set to 1000hz
Interestingly, the CPU utilization was not pegged at 100% according to the graph on either system.
This game engine is mysterious. It has the best ocean water effects I've ever seen. The DRAM utilization with 3rd Party Mods and Textures is enormous even using 8GB VRAM GPUs. One user showed over 10GB of system RAM used.
Not to try to add to your work list with your benchmarks, but once you find out which is more efficient or even on all 4 setups it would be interesting. To see some comparisons with this hardware running Linux and games to see what kind of difference there is not only between OS's but if there are any performance surprises between Intel and AMD on Linux systems.
Awesome deep dive, but where was Wendell's Learnatorium?
In all seriousness, it seems to me after this series of videos like the argument between Ryzen and Intel for comparable chips is down to such minutia (for everyday general mix of gaming, workload, etc.) that the only way to make a decision between the two is based on finances. I'll be building a new system in a few months, and it would be hard to justify buying an Intel CPU on a performance-per-dollar basis.
Also, were the frames counted in that hardware input test between LED and animation the game frames or the 180FPS camera frames?
@wendell Thanks for the video. Your testing is very interesting, looking at all the data shows what is really going on in a game. I would like to see what would happen if you go into the AND and Nvidia software and just cap the fps at 60 - 90 fps and see which one plays the game better. A game running at over 60 fps and very smooth does not in my opinion have a benefit. Stutter and screen tear free and 60 fps would be very nice. Smooth game play is what I look for.
is the data for these comparisons available? thinking that violin plots would be a good way to visualise this data, and can easily have both frametime and fps as primary and secondary axis.
I wrote this on YT: (reposting it here since YT/Google sometimes eats my comments) Looks more like a problem with the engine. Maybe some unhandled internal bufferoverflow that triggers an internal crash-handler, that can be better smoothed out with more threads? Does an Intel 6+ core system show the same behaviour? If "no" it is a good hint for that. If the engine has problems with the high FPS try locking it (like to 150) or downclock the i7. I def. think it is not a problem with either of the CPUs it is that freakin' game.
Huh a lot of user names that I've seen in other places... It looks like there is a group of people that are recurrently stirring up hate . These people cannot be helped, they've already made up their mind as to what they want to think without seeing any evidence. They're not worth trying to address even.
Was the i7 maxed out (100%) when these long frame times appeared? I get these kind of stutters too, even in CSGO or BF4, when my i5 is maxed out. I havent noticed it that bad in GTA V yet (I guess the reason for that is simple, I am at a GPU limit in that game). I am playing on a 120Hz display, so I can tell when something is not smooth.
Some of those people need to get a life. They're acting like it's a political scandal involving high ranking government officials having sex with underage orphans.
Anyway, on the subject of frame times. I noticed that pretty much all of my games have abysmal frame times if my mouse polling rate is 1000hz and some games even if it's 500hz. As soon as I drop down to 250hz I get buttery smooth gameplay in all of the games that I currently have installed. The Witcher 3 is the most notorious example. It stutters a bit even on 500hz. Maybe my CPU (i5-4590) just isn't good enough to deal 1ms polling rate, but at least CS GO can still be smooth at 500hz which is all that I care about.
High IO polling always has been a significant waste of CPU cycles in game loops. Ideally you want to apply Fourier Theory here and have your mouse polling set to about twice your framerate to maintain perfect sync and not waste CPU cycles polling for IO. That's why your 250Hz is likely so smooth, assuming you are using a 144Hz monitor and hitting about as many FPS.
Infact if your mouse sensor where to respond too slowly while being polled you could cause the game loop to hitch too while it's waiting for the mouse data. This assumes that you are using raw input and the game does not timeout if it misses a USB frame from the mouse input. At lower polling rates the mouse data is essentially an average vector of the last x timesteps. So you're not loosing much either.
The quality of the videos just keep getting better and better all the time and I think @Wendell has a more confident look when he is doing his videos compared to the old days. Best of luck to Wendell and the rest of the team at LEVELONETECHS in the future.
I have to say I think something funky is going on here because I have played GTAV for quite some hours on both my 6600k and 7700k I have now and noticed no stuttering or high frame times as mentioned. Obviously my personal experience is anecdotal at best but this is confusing to me. My setup isnt far off from the testing rigs with 2 x 128GB SSDs in raid 0 and a GTX1070. Though I'm not running 4k or above 60hz, I am spanning across 3 1080p monitors.
I feel conflicted here. On one hand I have @wendell's undeniable proof its happening, and on the other I have my personal machine that it doesnt seem to happen with. Maybe there's something to running those debloat windows scripts.
