Are you seeing this behaviour with all writers or with a specific one? 

It is a new simple model with 2 PostGIS readers and 6 shapefile writers and a few simple testers in between. It took him more than 3 hours to write it out. Normally, this would go very quickly.

I will let a few other existing models run, but I find it not normal.

Try turning on the Log Timestamp Information from options then Translation— maybe it can help you figure out exactly where the issue is happening

Is feature caching enabled? That can have significant impact on all aspects of run time

I always use the feature cache but I also tried without it and it took so long too.  Eventually it does write it away but I noticed it took a long time.  With other models the problem seems to be less. Maybe it is because this was a new model and the others already existed? No idea, I'm going to keep track. 

Either way, 3 hours is long time if usually this is really quick. There must be something else going on. 

I would try and dig into this a bit more. 

If DB is involved it might be that the reading is slow, and if you are trying to write to a network location it could be that there is rather an issue with the network rathe than FME.

The bottom of your log should give you some clues at least as to where performance is going:

e.g., 

FME Session Duration: 16.7 seconds. (CPU: 6.5s user, 7.3s system)

Here the workspace took 16.7 seconds, however, FME was working for 14 (6.7 + 7.3) seconds. In this case FME waited 2.7 seconds. If you see something like 3 hours in run time but also a very small CPU time then this is a good indication that network issues were involved somehow. 
 

I dit it again and this was the result

I dit it again and this was the result

 

OK yeah, so here indeed there is only about 25 mins of CPU time being used - is 25 mins about what you saw pre upgrade? It does suggest that FME is waiting somewhere.

Things I would check:

- Is the workspace running from a network location?
   - This can slow down execution for several reasons.  
- Are you writing to a network location?
   - Does the performance change if you write the data to a local location?
- Memory usage is pretty high - it could be that memory usage is too high, this could certainly be something which may have changed with your upgrade. A slightly higher memory usage could push the process above a limit causing a slow down.
- Are you still running with feature caching on? 
   - This should probably be turned off.
- Where is your fme temp location? Is it local and on an SSD?
  - FME will output quite a few temp files, this can be especially true when the memory of a process gets really high. A temp location on a slow hard drive or worse still a network drive can have a really big impact on performance. 
- Try just reading the data from the database into an inspector (disable the rest of the workflow). Here you can rule out the slow database read.

Logging with timestamps on as ​@mohamedalsobh suggested (Tools > FME Options > Translation > Log Settings > Log Timestamp Information)