I guess most people don't remember Glide was simply a harshly abbreviated version of OpenGL that simply made it easier to implement and add 3dfx-specific extensions on the fly. Being much more compact, if offered a good performance advantage as well.
As soon as NVidia released their first Alpha OpenGL "game drivers", GLquake and the likes could immediately be played on a Riva128 card as well, using their OpenGL components as fully complatible Glide substitute.
The Kronos group was notoriously slow (if at all) implementing new rendering techniques and back in those days, alot of 3dfx "quick&dirty" routines and Drivers naturally didn't meet the very high OpenGL standards to even enter the slow certification & validation procedures.
Took them many years to realize games were valid OpenGL applications too, and leave their AutoCAD mindset behind.
IMHO Mantle may bring a general speedup in the rendering API that has the potential to improve speed across all supported devices.
If they manage to yield a sufficient advantage over OpenGL/DirectX and make it cross-platform... They might find enough momentum to establish it.
Unfortunately, as witnessed more and more often, rendering speed or quality doesn't make better games by itself
To me it seems almost all big publishers (even previously good and reliable ones) are migrating to the "cash cow" syndrome or simply abandon their game once enough cash is made...
UbiSoft's FarCry3 comes readily to mind, one of the more interesting yet buggiest titles ever released with a literally catastrophic multiplayer stability. Unpatched since almost a year now and v1.05 Multiplayer still plays like a v0.92alpha (internal testing only).
Whatever AMD does with it, it
better had much more qualified and solid support than their Catalyst Drivers sometimes had in the past.