I'll give it a run tonight when I get off work and see what it takes to get those kinds of numbers.
How regularly does the frame hitching occur actually? Is it within about the same space time intervals? Or near randomly distributed?
I remember seeing in the video that for most of the examples you where turning a corner or walking down the alley something that would lead to resources of previously occluded objects being loaded and processed.
Maybe try exacerbating the problem with a slow HDD or so and see if they line up somehow.
Yes, and this is the key. Ingroup/outgroup is strong in that thread, and particularly stands out in some of the wording and attempts to redefine the topic being spoken of to fit the op's ad-hoc-triggered narrative pandering to the ingroup (an interesting topic in and of itself: https://en.wikipedia.org/wiki/Ingroups_and_outgroups )
The only way to reach the reddit ingroup in question is to break the group identity for a moment, and they are having too much of a ball taking turns to speak to begin taking turns to hear a word. It was nice that @wendell actually got a word in for anyone outside the group who would read through the response in future, though. Addressing the ingroup directly for any other purpose than indirectly reaching out to their less rabid audience is a complete waste of time. Now, @wendell, this is probably going to be successful other than perhaps being TLDR to most people who aren't (at least potentially) L1 ingroup. A succinct statement of Socratic method as it applies to the Ryzen review could also have been just as effective (of course, posted as reply to a few of the top comments).
While there are ways to break up in/out group think*, there needs to exist some common ground for it to happen, and some individual motivation to reach it. I have seen no individual motivation to reach common ground that could be stronger than the ingroup member's individual motivation for bias-confirmation, and the common ground of academic reasoning (have proof, will tell) offered by Wendell may not be the game to match the arena.
*and it also takes time, patience, and stomach.
Internet. Makes 40 year olds 12 again since September 1993.
Stuttering rears its ugly head. I also tried 1080 which did stutter once or twice, but not as much as 720.
This is the resolution I normally play at which feels smooth enough to me.
Seems stupid targetting Wendell then as his coverage of amd has always been good.
People just want a witch hunt / school yard fight
Cool, you are seeing some stutters, too, some not super high but some kinda high
well at 720 and 1920 x 1080 it definitely looks like your graphs. 5760 x 1080 feels normal to me even though it shows theres some high frame times at the beginning of each scene. The 2 biggest culprits during the 1920 test were the pier and when the camera is chasing the jet. They were also when I happened to be getting the best FPS.
Edit: My wife has a 6600k and GTX1070 on windows 8.1. I could test GTAV there as well if you see reason to.
Gamers Nexus dropping GTA for testing (trying to find suitable trollface image)
The way people are describing these stutters, and the way they're shown in the video, I'm thinking software problem too. More specifically an unexpected scheduling of threads causing a deadlock / livelock between threads, and a watchdog process spotting this after X amount of time between frames. It then kicks the engine on.
Would fit the symtoms of:
- High frame rates (maybe the draw thread is finishing earlier than expected)
- Differences between CPU architectures. One architecture might just make a certain thread timing more likely.
- Video seems to stutter on areas of quite low complexity, so doesn't suggest a resource paging issue or something like that, or just that the system is overloaded.
Honestly I think you're mining a vein that has little to offer. You're looking at an issue Rockstar QA should have found (Possibly they did find it, and that's why there's a watchdog process - a glitch is better than a hang when you're pushing for a release deadline). It's not anything which sheds any light on the suitability of a particular CPU for gaming IMO.
I think we're finding the limits of using GTA V as a benchmark - for anything beyond telling you how well GTA V runs.
Thing is I get the stutters on an i7 with smt enabled.
I know I am just feeding the trolls who want to make something big out of this but.. I HAD TO!!
I just couldnt resist..
Edit: Updated pic
Edit: switched back.. preferred the original..
3 Likes
@wendell
Could those frametime drops in intel be caused by background Windows activity? (talking about Fallout 4)
Could you please try using Windows game mode to test if they still occur?
As for GTA V, I think this might not be the case as it occurs in specific parts. Does this really make the experience unpleasant? Is there a reason why one would choose an i5?
After another day or so of testing, confirmed it is 98% likely it is the game engine. using the nvidia profile utility to limit the game to 160 fps, problem prettymuch disappears.
How did you get the number 98%?
to confirm the last 2% we'd need to do something to push the fps up to the same on ryzen to confirm that ryzen has the issue at high fps, thus completely eliminating the possibility of differences between platforms.
Instead a 1080 and a 1080ti on ryzen have similar frame rates (1080ti is a bit higher) suggesting some type of bottleneck outside both the gpu and the cpu.
Since we couldn't push up ryzen fps, we pushed down intel fps, which did actually make things better.
3 Likes