G80FTW
Active Member
I work at FedEx in the warehouse where we unload sort and load packages. Yea. Anyway. Iv recently discovered that our inbound operates very much like a processor in many ways.
So we have 2 main inbound doors where we can originate packages and process them much faster than the other 3 doors we have. So lets say these 2 doors are 2 processing cores. Each one having one unloader (which would be where the information for the processor is coming from) a belt that the packages go on that moves at a rather slow speed and is very thin (thinik of this as total bandwidth for the processor core) and 4 scanners to actually process the information. So that would be a 4 pipeline processor core right? So 8 altogether I suppose.
So, the unload rate is about 3k packages per hour. Roughly it takes us an hour to finish a 3k load trailer on those 2 doors. So if we were to use both of these "cores" the information would be processed theoretically at 6k pph. Impressive and interesting I know. But 6k pieces of volume would be completed in one hour (or 6k instructions per hour, whatever).
Now lets say for whatever reason my boss wants us to use only one door and double-chute the volume to both scanning stations. This would be where I would consider this hyper-threading. He is taking one door, or processor core, and running the volume, or information, down 2 separate processing paths (parallel threads right?).
BUT, he does not understand there is a problem with doing this. He thinks that by adding one person to unload that he can double the speed of a single trailer and get it down quicker. Well, the belt, or total bandwidth, we talked about. That remains the same. And that belt can only handle about 3k pph at its best. Maybe a little more. So he thinks that by putting another person in there he can double the speed it comes out and use 8 scanners rather than 4 to process them. But the belt cannot handle 6k pph at all ever, so it doesnt exactly work the way he thinks it does.
Now, I watched and all 3 trailers each finished in about 50 minutes exactly. Which is a whopping 10 minutes faster than they would have if we did not "hyper-thread" but instead used the 2 cores. While this may sound good, there is another bigger problem here. While it took 50 mins to complete one, there was only one trailer being done. But had we used both "cores" individually, not only would we save man power but we would also increase production. Since 2 trailers would be done within 1 hour. Which means 6k done in one hour vs 3k done in 50 mins.
Another more important problem here, is that not only does it take more people to do it this way, they cant seem to keep BOTH stations busy. Which means you will always have at least 1-3 scanners not working. Which is effectively 1-3 pipelines not being used. A waste.
So maybe not EXACTLY hyper-threading, but if you seen it like I did I think that would be the first thing that would come to mind.
Hope you enjoyed and hope maybe it was informative to someone out there. An easy way to explain Hyper-Threading maybe?
So we have 2 main inbound doors where we can originate packages and process them much faster than the other 3 doors we have. So lets say these 2 doors are 2 processing cores. Each one having one unloader (which would be where the information for the processor is coming from) a belt that the packages go on that moves at a rather slow speed and is very thin (thinik of this as total bandwidth for the processor core) and 4 scanners to actually process the information. So that would be a 4 pipeline processor core right? So 8 altogether I suppose.
So, the unload rate is about 3k packages per hour. Roughly it takes us an hour to finish a 3k load trailer on those 2 doors. So if we were to use both of these "cores" the information would be processed theoretically at 6k pph. Impressive and interesting I know. But 6k pieces of volume would be completed in one hour (or 6k instructions per hour, whatever).
Now lets say for whatever reason my boss wants us to use only one door and double-chute the volume to both scanning stations. This would be where I would consider this hyper-threading. He is taking one door, or processor core, and running the volume, or information, down 2 separate processing paths (parallel threads right?).
BUT, he does not understand there is a problem with doing this. He thinks that by adding one person to unload that he can double the speed of a single trailer and get it down quicker. Well, the belt, or total bandwidth, we talked about. That remains the same. And that belt can only handle about 3k pph at its best. Maybe a little more. So he thinks that by putting another person in there he can double the speed it comes out and use 8 scanners rather than 4 to process them. But the belt cannot handle 6k pph at all ever, so it doesnt exactly work the way he thinks it does.
Now, I watched and all 3 trailers each finished in about 50 minutes exactly. Which is a whopping 10 minutes faster than they would have if we did not "hyper-thread" but instead used the 2 cores. While this may sound good, there is another bigger problem here. While it took 50 mins to complete one, there was only one trailer being done. But had we used both "cores" individually, not only would we save man power but we would also increase production. Since 2 trailers would be done within 1 hour. Which means 6k done in one hour vs 3k done in 50 mins.
Another more important problem here, is that not only does it take more people to do it this way, they cant seem to keep BOTH stations busy. Which means you will always have at least 1-3 scanners not working. Which is effectively 1-3 pipelines not being used. A waste.
So maybe not EXACTLY hyper-threading, but if you seen it like I did I think that would be the first thing that would come to mind.
Hope you enjoyed and hope maybe it was informative to someone out there. An easy way to explain Hyper-Threading maybe?
Last edited:
