This is probably NOT normal (I'm disagreeing here with the answer provided by user grg). I own the same iMac configuration you are asking about (late 2014 iMac 5k R9 M295x graphics card, 4GB vRAM). In addition to running the built in display at a higher 'scaled' resolution (2880x1620)—which Apple warns "may affect performance"—, I also have attached a 2560×1440 external display and a 3840x2160 external display. So it seems on that count I'm taxing my GPU harder than you are. But my GPU die temp (measured via iStat Menus) is generally around 75 degrees C, with my fan usually around a silent 1200 rpm (which seems to be its base rate).
My usual computer use involves web browsing, email, Microsoft Office, Slack, Zoom video conferencing, note-taking apps, etc. So not anything too intense from a GPU perspective (other than the fact that I have LOTS of windows open spread across 12 desktops in Mission Control, which does tax the vRAM). It will occasionally get above 90C when I'm doing something graphics intensive (such as zooming around a lot in Google Maps satellite view), but quickly drops down in temperature once that activity is over.
If you are doing something graphically intensive on a regular, ongoing basis (such as watching videos non-stop), that could cause your GPU die temperature to stay around 90C, but it would have to be pretty sustained for the temperature to remain there, as macOS seems to want to keep the GPU temperature below that point, and aggressively ramps up the fan speed when those temperatures are encountered, to try and get the GPU a bit cooler.
Also, if you are in an unusually hot room (and/or have sunlight directly on the iMAC or a heater located near it), that of course can affect the thermal situation.
I'd look to see if certain apps or webpages you might have open could be somehow 'secretly' driving GPU usage and temperatures. Try closing them one by one and seeing if that makes a difference. For instance, it could be that you have webpages open on sites with bitcoin miners, which will tax your GPU. There are browser plug-ins which will block miners, too. When doing this kind of troubleshooting, I find that looking at the "GPU package" power consumption (in Watts) using an app like iStat Menus (also mentioned by grg) is most useful, as it response most quickly to a change in GPU load, before that change is reflected in die temperature. For reference, my GPU package power use hovers around 40W–44W under typical usage (with spikes to around 60W when the GPU is being severely taxed).