Bakker et al (2015)
show evidence that Dimetrodon (Fig. 1) fed on aquatic prey as there were too few terrestrial reptilian herbivores to sustain their numbers.
From the abstract
“In restorations, Dimetrodon often appear feeding upon large land herbivores, e.g., Diadectes and Edaphosaurus. Base Theory (AFBT) recognizes non-terrestrial prey as key for dimetrodont food webs. Over 45% of the bones are severely tooth-marked; ubiquitous shed Dimetrodon teeth are mingled with tooth-marked bones in every depositional unit. The CBB lacks any structures that indicate high current energy, so the hydraulic forces probably did not wash in bones from beyond the trough, though bloated whole carcasses could have floated in. There are 39 Dimetrodon, one each of the large herbivores Edaphosaurus and Diadectes, three of the large non-herbivore, non-apex carnivore Secodontosaurus, and three of the semi-terrestrial amphibian Eryops calculated form postcrania. Did benthic amphibians and fish fill the gap in prey? The benthic amphibian Diplocaulus is abundant in every bone-rich unit. Xenacanth sharks are very common in several layers; each shark carried a large, well ossified head spine. AFBT is corroborated: dimetrodonts fed intensively on aquatic prey at the CBB.”
There is also a good Dimetrodon video (52 min.)
on YouTube featuring Dr. Bakker as he describes how the vast majority of Dimetrodon tails are missing, neatly cut and probably carried away for their meat (because that’s where the most of it is!) by other Dimetrodons.
Bakker RT et al. 2015. Dimetrodon and the earliest apex predators: The Craddock bone bed and George Ranch Facies show that aquatic prey, not herbivores, were key food sources. Journal of Vertebrate Paleontology abstracts.