(It may be helpful in terms of understanding the situation if you were clearer what the "different scales" was about, but it probably won't substantially change the answer.)
Neither plot shows any clear indications of heteroskedasticity, or even much of a hint of it.
With so many points it would be useful to have transparency on the points so that depth of shading gave better indication of where most of the mass of points was. Sometimes it can be difficult to see the pattern with large solid points (especially if they might completely overlap).
Aside from the issues just mentioned (that it's hard to see where the heaviest concentration of points is), the first plot doesn't show much of anything out of the ordinary.
The down-sloping lines in the second plot suggests that the response was discrete - probably integer-valued. Outside of that there's nothing especially surprising about it; the tendency to clumping in small concentrated groups along each line suggests at least one of the predictors is also discrete (again probably integers), but there may be another continuous variable*, with a smaller effect.
*(or possibly a discrete variable on a larger range)
If you fit a linear model to a discrete response, you would tend to expect the bands / stripes you see here.
Similar plots can be found in numerous posts on site (searches for terms like residual stripes or diagonal bands turn some of them up, as can search terms like weird residual pattern). Here are two such:
Analysis of the Residuals vs Fitted
What do my residuals graphs say about my data?