This issue is quite simple. Compared to the latest experimental snapshot, the terrain generation is a less extreme and hilly throughout certain areas in 21w37a. This is especially noticeable with the new jagged peak generation, which experienced drastic decreases in size and altitude as well as less cohesive placement. To an extent this also applies to the caves, which somewhat changed in orientation and no longer form any size-able entrances or extreme interactions with the surface. Note that this issue is valid as @unknown stated that no changes to world generation were made intentionally, but are rather outcomes of changes to the underlying tech to allow for the porting of these features to regular snapshots. He also tweeted out asking the community to report such mismatches and downgrades in the bug tracker in order to help them get the terrain to match Experimental Snapshot 7 as closely as possible.
How to reproduce: Open up a world on Experimental Snapshot 7. Copy the seed and load it up on 21w37a, then explore and compare the differences.
Expected Results: Terrain in 21w37a should fundamentally match Experimental Snapshot 7 terrain, with few less noticeable exceptions.
Actual results: Terrain is highly nerfed in terms of elevation and jaggedness throughout various locations.
DISCLAIMER: None of the images of terrain generation present here have been found or created by me. They are from Twitter accounts who tested the snapshots and publicly shared what they found. ALL CREDIT GOES TO THEM. Please also feel free to test for yourself and share your findings on this issue. Images can really help.
Attachments
Comments 2
Note that Groves currently don't generate trees (MC-236698) which will be fixed for next snapshot. This may contribute to the feeling of less interesting terrain.
Cannot reproduce this, but could use some help testing.
This is probably just a matter of perception. The seeds have a changed a bit. The general terrain elevation is the same, so plateaus and rivers and such should be in the same general locations. But the detailed terrain shape is different (the 3d noise), so jaggedness and overhangs and peaks will be different. Some areas will be more jagged & extreme in experimental snapshot 7, other areas will be more jagged and extreme in 21w37a.
I think what happened here is that people are revisiting known locations from experimental snapshot 7 with dramatic & extreme terrain, and finding that they just look less dramatic on 21w37a, which is statistically likely. That should go both ways. If you find some dramatic terrain in 21w37a and visit those in experimental snapshot 7, the same location in experimental snapshot 7 is likely to be less dramatic.
To test this properly, you need to visit a number of new random locations on both versions and compare. On average the extremeness of the terrain should be similar - sometimes more dramatic on one version, sometimes more dramatic on the other. If not, let us know!