Current Generation Games Analysis Technical Discussion [2020-2021] [XBSX|S, PS5, PC]

Tessellating does not mean adding more polygons tho in general?

We're those Cross-Platform performance tests done in Like-for-Like with no AI on? Because AI placement and density is random and decisive to performance.

 

only way to compare framerate like for like would be to remove all traffic and all pedestrians. Parked cars are in the same positions in both versions.
Then fly through the same streets at max speed, but without these two elements i think the framerate is pretty much locked to 30fps.

 

in all open world games you have some randomines aspect, you can't compare visualty screen to screen in easy way but still no problem for benchmarking, just test few minutes walking in city, few minutes drving car and few minutes camera flying full speed and you will have some stats

 

it's not an openworld game, it's a big environment demo and you can remove the randomness of traffic and pedestrians, keeping only static assets.

 

but without trafic and pedestrians would be rather artifical benchmark imo

 

i agree with this on the overall discussion of benchmarking a game, there's little one could do. But if you're going to cherry pick a particular sample and find a very tiny variance, that is what Alex is trying to get at.
It's okay if you're looking the overall benchmark of a lot of different areas and noticing a continuous trend of +X, (ie comparing the mean frame time over the course of the whole car scene and finding a 5% variance in FPS) but to zoom in and cherry pick a small section of a noisy line, at least in my line of work, is not a solid defensible position.

 

Watch Dogs and GTA have fixed and controlled benchmarks that you can run though. nxdude didnt even match the footage for his own video, nor the same time of today, nor the same driving dirrection. Even if we cant account for randomness, he didnt do a single thing that he had control over.

 

yea lol. haha. tell me about it lol.

I'm only talking methodology. I can run some analysis on resolution if people really want, it will take some time to render the results back to the movie, but I can't make the demo run the same path twice.

 

From what I've seen, this isn't a debate on him, just specific issues with the benchmarking method taken.
Which I would think should be able to be discussed in isolation without getting into "those debates"

 

He actually doesn't really make any real commentary on it, though he says there's up to a 19% fps difference, he doesn't really go into specifics. I assume he's looking at min/max for what he recorded. The easiest way to settle that is to run a specific course several times over and take the min/max/mean of each run and see what the variances are like before making a statement, but he says not to take the numbers for anything.

But if we ourselves take his video and cherry pick a particular area to prove his point, then you're running into methodology issues.

 

Oh yeah i did not say it was ideal, but that in the case of that demo, that would be the only way to have similar runs between xsx and ps5.

 

I've not watched the video.
Guess I'm just highlighting that I think what @Shortbread said in regards to @Phantom88 point can be discussed without it getting bogged down.
As I think his point was general enough about how he saw the comparison being made for the discussion not to go down the drain.

 

The Matrix demo performs better on the ps5 and this is not news. I have both the Series X and the PS5 and have tested both rigorously. The Series X also has more bugs but again, it's a tech demo. We can't really draw any conclusions regarding the hardware from it but, his assertion that the PS5 version runs slightly better than the Series X version is not even remotely controversial. It's objectively and measurably true.

 

That may be the case, but the data provided doesn't prove that, hence the discussion on correct process necessary to determine this where the supporting evidence provided fails. Simply, ten minutes of similar gameplay processed to get average framerates should show any constant, notable delta between platforms.

 

disagree thats the only way to have similar run but for sure easiest and imo useless

 

Not sure if you agree or disagree here. Tessellation is a covering of a surface with tiles. I think one can use it in the context of videogame by meaning "tessellating = adding more polygons". I believe based on one of his posts NXGamer actually uses that word in that way.