More information is certainly needed and I think @No_One_Is_Innocent got the important points correct. Reiterating just the questions:
-Will switching to EtHash reduce TPS?
-Will this affect the DAG size?
Short Answer: I am confident Conflux is ready and able to take on the risk of switching to ETHash for the perceived benefit of breaking down barriers and getting as many miners on the network as possible.
Long Answer: Assuming the hashing algorithm change has NO NEGATIVE SIDE EFFECTS on the Tree Graph consensus or TPS.
I think the hope is more miners come to Conflux, which ultimately drives overall adoption and exposure in the mainstream crypto community. All positives in my eyes.
As previously mentioned by others Octopus is unique to Conflux. I love how Conflux can and will build bespoke for their needs and abandoning something unique for something seemingly losing its main purpose does feel like a step back technologically, although this view may be flawed, ETHash is proven to be viable at massive-scale.
@joseconflu and @Ghost also so made important points about other ways to onboard AMD cards and fight DAG size v. average GPU memory capacity. These approaches seem like a more conservative option to achieve some of the same goals.
In the end I would be happy with either Octopus or ETHash, but making the switch to ETHASH feels like a higher-risk, higher-reward option. The team has invested significantly in and seems confident about the PoS layer as a protective measure against PoW based attack vectors.
There could be a massive network effect if hashrate grows orders of magnitude from current levels and personally I would be fine if my small mining rig becomes less impactful if the network gains adoption and presumably CFX gains value. There is a set output rate after all, so price should respond to a large change in hashrate.