- Joined
- Aug 23, 2001
Well folks the proof is in the pudding I guess but it is hard to put numbers on it and these numbers simply do not do the benefits of hyperthreading justice. I don't know about you people but when I am playing a game I do not like it to slow down to 0.5 FPS for a period of 2 seconds which is what happend in my latest round of benchmarks. In both tests HT was turned on and off otherwise everything else was the same and the priority for flaskMPEG was set to high while 3dmark2001 was set to normal. Here are my results.
With HT turned off this is what I got running 3dmark2001 and flaskMPEG together.
Here is what it was with HT on.
Now with HT turned off 3dmark would come to a stop for a period of sometimes around approx 2 seconds but with HT on it never slowed much at all and there was no jumpiness.Remember folks I want to get some work done so flasks priority was set to high.
The reason why I even scored 10000 without HT on was because the game benchmarks would run 80fps then stop to 0.5fps. It would go back and forth like that but the average was enough that it still scored half ways decent so you see people for some things like for sure gaming it is really hard to put a number on how well HT benefits it. If you can see or not you will notice that flask was going to take around 24 hours to do what I wanted it to do and if you have a app that will run that long then it would really suck to have to wait for your encoding to get done so you could say GAME of all things.
I have never used flask but I know what it is supposed to do so my settings may have been off but the stress on the CPU should have been adequate.
This test really shows indeed how great the benefits of hyperthreading are.
I did some more hypthreading benchmarks. I ran two instances of super_pi together. Here they are.
With HT on one instance. 46 sec.
With HT on two instances 1 min 5 sec.
With HT off one instance 47 sec.
With HT off two instances 1 min 35 sec.
So with HT off it took 2 PI's 1 min 35 seconds(95 seconds) or 47.5 seconds a piece which makes sense since it is not multithreading.
With HT on it took only 1 minute 5 seconds (65seconds) or 32.5 seconds a piece to run 1 million places.
To run a single instance it took 46~47 seconds. I think the HT benchmark was better since the background programs won't affect it as much. I think for single apps that is where they get the 5%~10%. With this kinds of benchmarking it is clearly better. Now if you take 46/65 you will see how I get my 75%. Close enough anyway.
Now when you look at this with a benchmark that takes hours instead of minutes or seconds we are talking about alot of time. Seti times with HT are cut in half pretty much if you have not seen those benchmarks.
This is the way it pretty much is for all benchmarks that are done like this which very few do.
If you want to try it open two PI's and set one so you hit the space bar to activate it and the other with the left mouse click and do them simultaneously hitting the space bar .5 second or sooner before the left click. If you left click first then your other PI will not be able to run by space bar activation.
You mathmaticians can use my above data to do some numerb crunching on how affective HT is or is not. Taking 15 sec off a individual PI time is pretty grand.
With HT turned off this is what I got running 3dmark2001 and flaskMPEG together.
Here is what it was with HT on.
Now with HT turned off 3dmark would come to a stop for a period of sometimes around approx 2 seconds but with HT on it never slowed much at all and there was no jumpiness.Remember folks I want to get some work done so flasks priority was set to high.
The reason why I even scored 10000 without HT on was because the game benchmarks would run 80fps then stop to 0.5fps. It would go back and forth like that but the average was enough that it still scored half ways decent so you see people for some things like for sure gaming it is really hard to put a number on how well HT benefits it. If you can see or not you will notice that flask was going to take around 24 hours to do what I wanted it to do and if you have a app that will run that long then it would really suck to have to wait for your encoding to get done so you could say GAME of all things.
I have never used flask but I know what it is supposed to do so my settings may have been off but the stress on the CPU should have been adequate.
This test really shows indeed how great the benefits of hyperthreading are.
I did some more hypthreading benchmarks. I ran two instances of super_pi together. Here they are.
With HT on one instance. 46 sec.
With HT on two instances 1 min 5 sec.
With HT off one instance 47 sec.
With HT off two instances 1 min 35 sec.
So with HT off it took 2 PI's 1 min 35 seconds(95 seconds) or 47.5 seconds a piece which makes sense since it is not multithreading.
With HT on it took only 1 minute 5 seconds (65seconds) or 32.5 seconds a piece to run 1 million places.
To run a single instance it took 46~47 seconds. I think the HT benchmark was better since the background programs won't affect it as much. I think for single apps that is where they get the 5%~10%. With this kinds of benchmarking it is clearly better. Now if you take 46/65 you will see how I get my 75%. Close enough anyway.
Now when you look at this with a benchmark that takes hours instead of minutes or seconds we are talking about alot of time. Seti times with HT are cut in half pretty much if you have not seen those benchmarks.
This is the way it pretty much is for all benchmarks that are done like this which very few do.
If you want to try it open two PI's and set one so you hit the space bar to activate it and the other with the left mouse click and do them simultaneously hitting the space bar .5 second or sooner before the left click. If you left click first then your other PI will not be able to run by space bar activation.
You mathmaticians can use my above data to do some numerb crunching on how affective HT is or is not. Taking 15 sec off a individual PI time is pretty grand.