Yeah exactly, the way my test code is set up is it bursts 30,000 particles that start on size Vector2f(1.f, 1.f) then slowly transition to the next size if there was another size assigned to it, I noticed that if I go smaller like Vector2f(0.5f, 0.5f) it works alright fine, but the large scaling (5x and above) causes horrible performance.
There is slightly more performance using no texture over using a texture at these smaller sizes but not a whole lot of impact.