AMD has used clustered multithreading (CMT) and while they had decent performance for multithreaded applications they failed largely on single core performance as a result of their architecture design. They have changed now to using SMT or simultaneous multithreading. Following more along the lines now of intel's architecture which should overall be a boon to their single core performance as well as helping multithreaded applications still. Overall Ryzen will be a very good CPU, but amd cpus before are worse unless you are just running things that are multithreaded.
AMD bet on multicore parallel processing but software (namely games, other software transitioned much faster) were way to slow in transition and weve seen quite a bit of progress in last 2 years. The result is that Intels dual core offerings fall behind (lower FPS and stutter in games) similarly priced FX 83xx and that 100$ FX 83xx competes nicely with Intels quad core 250$ i5s
Witcher 3 , which was released on May 19 2015.
(relevant part starts at 3:29, but first part shows dual core pentium G3258 OCed to 4,5 GHz gets completey hammered and gives very poor experience, while i3 falls to ~30 FPS and gives choppy performance)
AMD bet on multicore parallel processing but software (namely games, other software transitioned much faster) were way to slow in transition and weve seen quite a bit of progress in last 2 years. The result is that Intels dual core offerings fall behind (lower FPS and stutter in games) similarly priced FX 83xx and that 100$ FX 83xx competes nicely with Intels quad core 250$ i5s
Witcher 3 , which was released on May 19 2015.
(relevant part starts at 3:29, but first part shows dual core pentium G3258 OCed to 4,5 GHz gets completey hammered and gives very poor experience, while i3 falls to ~30 FPS and gives choppy performance)
There's a few reasons you are seeing that on those benchmarks. The i3 has 2 physical cores that are being split into 4. The i5 has 4 physical cores that are not being split. And that i7 has 4 cores but they are being split. A lot of the i5's are better for gaming for that reason alone. But overall performance of the processors are increasing. They just aren't increasing for gaming. Unless they come out with adapters that fully utilize the i7 it will always run slower in this case.
AMD bet on multicore parallel processing but software (namely games, other software transitioned much faster) were way to slow in transition and weve seen quite a bit of progress in last 2 years. The result is that Intels dual core offerings fall behind (lower FPS and stutter in games) similarly priced FX 83xx and that 100$ FX 83xx competes nicely with Intels quad core 250$ i5s
Witcher 3 , which was released on May 19 2015.
(relevant part starts at 3:29, but first part shows dual core pentium G3258 OCed to 4,5 GHz gets completey hammered and gives very poor experience, while i3 falls to ~30 FPS and gives choppy performance)
There's a few reasons you are seeing that on those benchmarks. The i3 has 2 physical cores that are being split into 4. The i5 has 4 physical cores that are not being split. And that i7 has 4 cores but they are being split. A lot of the i5's are better for gaming for that reason alone. But overall performance of the processors are increasing. They just aren't increasing for gaming. Unless they come out with adapters that fully utilize the i7 it will always run slower in this case.
Hyperthreading won't even be used at all unless Windows thinks you need more cores than you physically have, as it is aware of which logical cores correspond to the same physical core. So it's very rare for disabling hyperthreading to improve performance. This used to not be the case, as if you needed two threads, sometimes Vista would put them on the same core and then they'd interfere with each other. That's one of the things Microsoft fixed in Windows 7.
AMD bet on multicore parallel processing but software (namely games, other software transitioned much faster) were way to slow in transition and weve seen quite a bit of progress in last 2 years. The result is that Intels dual core offerings fall behind (lower FPS and stutter in games) similarly priced FX 83xx and that 100$ FX 83xx competes nicely with Intels quad core 250$ i5s
Witcher 3 , which was released on May 19 2015.
(relevant part starts at 3:29, but first part shows dual core pentium G3258 OCed to 4,5 GHz gets completey hammered and gives very poor experience, while i3 falls to ~30 FPS and gives choppy performance)
There's a few reasons you are seeing that on those benchmarks. The i3 has 2 physical cores that are being split into 4. The i5 has 4 physical cores that are not being split. And that i7 has 4 cores but they are being split. A lot of the i5's are better for gaming for that reason alone. But overall performance of the processors are increasing. They just aren't increasing for gaming. Unless they come out with adapters that fully utilize the i7 it will always run slower in this case.
Hyperthreading won't even be used at all unless Windows thinks you need more cores than you physically have, as it is aware of which logical cores correspond to the same physical core. So it's very rare for disabling hyperthreading to improve performance. This used to not be the case, as if you needed two threads, sometimes Vista would put them on the same core and then they'd interfere with each other. That's one of the things Microsoft fixed in Windows 7.
Why is the i7 running slower then the i5? I believe that is what I was trying to explain. At the basic level the cores are different for that i5. They aren't sharing resources like the cores of the i7 and even that i3. The i3 has 2 cores running 4 threads which might help but you would have to compare it will a regular dual core processor. Meanwhile that i5 is just running straight 4 cores.
Then the i7 also running 4 cores has HT which gives it 8 threads. But somehow its not helping the game at all and is even running slower. But its a lot better overall performance just not for games.
But HT is literally the only reason that i7 is running slower. Its probably only using the 2 cores.
AMD bet on multicore parallel processing but software (namely games, other software transitioned much faster) were way to slow in transition and weve seen quite a bit of progress in last 2 years. The result is that Intels dual core offerings fall behind (lower FPS and stutter in games) similarly priced FX 83xx and that 100$ FX 83xx competes nicely with Intels quad core 250$ i5s
Witcher 3 , which was released on May 19 2015.
(relevant part starts at 3:29, but first part shows dual core pentium G3258 OCed to 4,5 GHz gets completey hammered and gives very poor experience, while i3 falls to ~30 FPS and gives choppy performance)
There's a few reasons you are seeing that on those benchmarks. The i3 has 2 physical cores that are being split into 4. The i5 has 4 physical cores that are not being split. And that i7 has 4 cores but they are being split. A lot of the i5's are better for gaming for that reason alone. But overall performance of the processors are increasing. They just aren't increasing for gaming. Unless they come out with adapters that fully utilize the i7 it will always run slower in this case.
Hyperthreading won't even be used at all unless Windows thinks you need more cores than you physically have, as it is aware of which logical cores correspond to the same physical core. So it's very rare for disabling hyperthreading to improve performance. This used to not be the case, as if you needed two threads, sometimes Vista would put them on the same core and then they'd interfere with each other. That's one of the things Microsoft fixed in Windows 7.
Why is the i7 running slower then the i5? I believe that is what I was trying to explain. At the basic level the cores are different for that i5. They aren't sharing resources like the cores of the i7 and even that i3. The i3 has 2 cores running 4 threads which might help but you would have to compare it will a regular dual core processor. Meanwhile that i5 is just running straight 4 cores.
Then the i7 also running 4 cores has HT which gives it 8 threads. But somehow its not helping the game at all and is even running slower. But its a lot better overall performance just not for games.
But HT is literally the only reason that i7 is running slower. Its probably only using the 2 cores.
A difference of 0.4% is well within experimental error for most benchmarks.
AMD bet on multicore parallel processing but software (namely games, other software transitioned much faster) were way to slow in transition and weve seen quite a bit of progress in last 2 years. The result is that Intels dual core offerings fall behind (lower FPS and stutter in games) similarly priced FX 83xx and that 100$ FX 83xx competes nicely with Intels quad core 250$ i5s
Witcher 3 , which was released on May 19 2015.
(relevant part starts at 3:29, but first part shows dual core pentium G3258 OCed to 4,5 GHz gets completey hammered and gives very poor experience, while i3 falls to ~30 FPS and gives choppy performance)
There's a few reasons you are seeing that on those benchmarks. The i3 has 2 physical cores that are being split into 4. The i5 has 4 physical cores that are not being split. And that i7 has 4 cores but they are being split. A lot of the i5's are better for gaming for that reason alone. But overall performance of the processors are increasing. They just aren't increasing for gaming. Unless they come out with adapters that fully utilize the i7 it will always run slower in this case.
Hyperthreading won't even be used at all unless Windows thinks you need more cores than you physically have, as it is aware of which logical cores correspond to the same physical core. So it's very rare for disabling hyperthreading to improve performance. This used to not be the case, as if you needed two threads, sometimes Vista would put them on the same core and then they'd interfere with each other. That's one of the things Microsoft fixed in Windows 7.
Why is the i7 running slower then the i5? I believe that is what I was trying to explain. At the basic level the cores are different for that i5. They aren't sharing resources like the cores of the i7 and even that i3. The i3 has 2 cores running 4 threads which might help but you would have to compare it will a regular dual core processor. Meanwhile that i5 is just running straight 4 cores.
Then the i7 also running 4 cores has HT which gives it 8 threads. But somehow its not helping the game at all and is even running slower. But its a lot better overall performance just not for games.
But HT is literally the only reason that i7 is running slower. Its probably only using the 2 cores.
A difference of 0.4% is well within experimental error for most benchmarks.
You know, this is an interesting academic discussion.
But honestly, I don't really care if Windows says I have only one half of one core, or a bajillion cores, because all that ultimately matters is how the system performs in applications and games that I am currently using. If it takes 12,080 cores or 1 core, to get to a performance level that I'm happy with, then other than it being somewhat interesting I don't really care and I'm happy.
AMD bet on multicore parallel processing but software (namely games, other software transitioned much faster) were way to slow in transition and weve seen quite a bit of progress in last 2 years. The result is that Intels dual core offerings fall behind (lower FPS and stutter in games) similarly priced FX 83xx and that 100$ FX 83xx competes nicely with Intels quad core 250$ i5s
Witcher 3 , which was released on May 19 2015.
(relevant part starts at 3:29, but first part shows dual core pentium G3258 OCed to 4,5 GHz gets completey hammered and gives very poor experience, while i3 falls to ~30 FPS and gives choppy performance)
There's a few reasons you are seeing that on those benchmarks. The i3 has 2 physical cores that are being split into 4. The i5 has 4 physical cores that are not being split. And that i7 has 4 cores but they are being split. A lot of the i5's are better for gaming for that reason alone. But overall performance of the processors are increasing. They just aren't increasing for gaming. Unless they come out with adapters that fully utilize the i7 it will always run slower in this case.
Hyperthreading won't even be used at all unless Windows thinks you need more cores than you physically have, as it is aware of which logical cores correspond to the same physical core. So it's very rare for disabling hyperthreading to improve performance. This used to not be the case, as if you needed two threads, sometimes Vista would put them on the same core and then they'd interfere with each other. That's one of the things Microsoft fixed in Windows 7.
Why is the i7 running slower then the i5? I believe that is what I was trying to explain. At the basic level the cores are different for that i5. They aren't sharing resources like the cores of the i7 and even that i3. The i3 has 2 cores running 4 threads which might help but you would have to compare it will a regular dual core processor. Meanwhile that i5 is just running straight 4 cores.
Then the i7 also running 4 cores has HT which gives it 8 threads. But somehow its not helping the game at all and is even running slower. But its a lot better overall performance just not for games.
But HT is literally the only reason that i7 is running slower. Its probably only using the 2 cores.
A difference of 0.4% is well within experimental error for most benchmarks.
True. . Not bad at all really.
It doesn't seem easy to find anything about the inner workings of HT. It appears to always be turned on and makes me wonder if that is simply the case. I mean when looking at task manager and assigning cores they are always present and turned on. So they won't ever get turned off. You would hope they are turned off if the system needs to use all the cpu but at the same time windows might just use all the cores by spreading out the tasks no matter what you do.
AMD bet on multicore parallel processing but software (namely games, other software transitioned much faster) were way to slow in transition and weve seen quite a bit of progress in last 2 years. The result is that Intels dual core offerings fall behind (lower FPS and stutter in games) similarly priced FX 83xx and that 100$ FX 83xx competes nicely with Intels quad core 250$ i5s
Witcher 3 , which was released on May 19 2015.
(relevant part starts at 3:29, but first part shows dual core pentium G3258 OCed to 4,5 GHz gets completey hammered and gives very poor experience, while i3 falls to ~30 FPS and gives choppy performance)
There's a few reasons you are seeing that on those benchmarks. The i3 has 2 physical cores that are being split into 4. The i5 has 4 physical cores that are not being split. And that i7 has 4 cores but they are being split. A lot of the i5's are better for gaming for that reason alone. But overall performance of the processors are increasing. They just aren't increasing for gaming. Unless they come out with adapters that fully utilize the i7 it will always run slower in this case.
Hyperthreading won't even be used at all unless Windows thinks you need more cores than you physically have, as it is aware of which logical cores correspond to the same physical core. So it's very rare for disabling hyperthreading to improve performance. This used to not be the case, as if you needed two threads, sometimes Vista would put them on the same core and then they'd interfere with each other. That's one of the things Microsoft fixed in Windows 7.
Why is the i7 running slower then the i5? I believe that is what I was trying to explain. At the basic level the cores are different for that i5. They aren't sharing resources like the cores of the i7 and even that i3. The i3 has 2 cores running 4 threads which might help but you would have to compare it will a regular dual core processor. Meanwhile that i5 is just running straight 4 cores.
Then the i7 also running 4 cores has HT which gives it 8 threads. But somehow its not helping the game at all and is even running slower. But its a lot better overall performance just not for games.
But HT is literally the only reason that i7 is running slower. Its probably only using the 2 cores.
A difference of 0.4% is well within experimental error for most benchmarks.
True. . Not bad at all really.
It doesn't seem easy to find anything about the inner workings of HT. It appears to always be turned on and makes me wonder if that is simply the case. I mean when looking at task manager and assigning cores they are always present and turned on. So they won't ever get turned off. You would hope they are turned off if the system needs to use all the cpu but at the same time windows might just use all the cores by spreading out the tasks no matter what you do.
Windows 7 or later knows which pairs of logical cores correspond to the same physical core. If one of the logical cores on a physical core is already in use, Windows won't assign anything to the other unless all of the physical cores are already in use. If you have a processor with hyperthreading, you can see how this works at a high level by opening up Task Manager and looking at the activity on each core.
Quizzical said: Windows 7 or later knows which pairs of logical cores correspond to the same physical core. If one of the logical cores on a physical core is already in use, Windows won't assign anything to the other unless all of the physical cores are already in use. If you have a processor with hyperthreading, you can see how this works at a high level by opening up Task Manager and looking at the activity on each core.
Quizzical said: Windows 7 or later knows which pairs of logical cores correspond to the same physical core. If one of the logical cores on a physical core is already in use, Windows won't assign anything to the other unless all of the physical cores are already in use. If you have a processor with hyperthreading, you can see how this works at a high level by opening up Task Manager and looking at the activity on each core.
AMD bet on multicore parallel processing but software (namely games, other software transitioned much faster) were way to slow in transition and weve seen quite a bit of progress in last 2 years. The result is that Intels dual core offerings fall behind (lower FPS and stutter in games) similarly priced FX 83xx and that 100$ FX 83xx competes nicely with Intels quad core 250$ i5s
Witcher 3 , which was released on May 19 2015.
(relevant part starts at 3:29, but first part shows dual core pentium G3258 OCed to 4,5 GHz gets completey hammered and gives very poor experience, while i3 falls to ~30 FPS and gives choppy performance)
There's a few reasons you are seeing that on those benchmarks. The i3 has 2 physical cores that are being split into 4. The i5 has 4 physical cores that are not being split. And that i7 has 4 cores but they are being split. A lot of the i5's are better for gaming for that reason alone. But overall performance of the processors are increasing. They just aren't increasing for gaming. Unless they come out with adapters that fully utilize the i7 it will always run slower in this case.
Hyperthreading won't even be used at all unless Windows thinks you need more cores than you physically have, as it is aware of which logical cores correspond to the same physical core. So it's very rare for disabling hyperthreading to improve performance. This used to not be the case, as if you needed two threads, sometimes Vista would put them on the same core and then they'd interfere with each other. That's one of the things Microsoft fixed in Windows 7.
Why is the i7 running slower then the i5? I believe that is what I was trying to explain. At the basic level the cores are different for that i5. They aren't sharing resources like the cores of the i7 and even that i3. The i3 has 2 cores running 4 threads which might help but you would have to compare it will a regular dual core processor. Meanwhile that i5 is just running straight 4 cores.
Then the i7 also running 4 cores has HT which gives it 8 threads. But somehow its not helping the game at all and is even running slower. But its a lot better overall performance just not for games.
But HT is literally the only reason that i7 is running slower. Its probably only using the 2 cores.
A difference of 0.4% is well within experimental error for most benchmarks.
True. . Not bad at all really.
It doesn't seem easy to find anything about the inner workings of HT. It appears to always be turned on and makes me wonder if that is simply the case. I mean when looking at task manager and assigning cores they are always present and turned on. So they won't ever get turned off. You would hope they are turned off if the system needs to use all the cpu but at the same time windows might just use all the cores by spreading out the tasks no matter what you do.
Windows 7 or later knows which pairs of logical cores correspond to the same physical core. If one of the logical cores on a physical core is already in use, Windows won't assign anything to the other unless all of the physical cores are already in use. If you have a processor with hyperthreading, you can see how this works at a high level by opening up Task Manager and looking at the activity on each core.
No - sorry. Windows does lock CPU cores, it is a known problem. In fact, if you have 2 or more cores, than more than likely only one core is used and the others are locked. I am using Park Control to deal with this and it helps free the CPU cores up.
AMD bet on multicore parallel processing but software (namely games, other software transitioned much faster) were way to slow in transition and weve seen quite a bit of progress in last 2 years. The result is that Intels dual core offerings fall behind (lower FPS and stutter in games) similarly priced FX 83xx and that 100$ FX 83xx competes nicely with Intels quad core 250$ i5s
Witcher 3 , which was released on May 19 2015.
(relevant part starts at 3:29, but first part shows dual core pentium G3258 OCed to 4,5 GHz gets completey hammered and gives very poor experience, while i3 falls to ~30 FPS and gives choppy performance)
There's a few reasons you are seeing that on those benchmarks. The i3 has 2 physical cores that are being split into 4. The i5 has 4 physical cores that are not being split. And that i7 has 4 cores but they are being split. A lot of the i5's are better for gaming for that reason alone. But overall performance of the processors are increasing. They just aren't increasing for gaming. Unless they come out with adapters that fully utilize the i7 it will always run slower in this case.
Hyperthreading won't even be used at all unless Windows thinks you need more cores than you physically have, as it is aware of which logical cores correspond to the same physical core. So it's very rare for disabling hyperthreading to improve performance. This used to not be the case, as if you needed two threads, sometimes Vista would put them on the same core and then they'd interfere with each other. That's one of the things Microsoft fixed in Windows 7.
Why is the i7 running slower then the i5? I believe that is what I was trying to explain. At the basic level the cores are different for that i5. They aren't sharing resources like the cores of the i7 and even that i3. The i3 has 2 cores running 4 threads which might help but you would have to compare it will a regular dual core processor. Meanwhile that i5 is just running straight 4 cores.
Then the i7 also running 4 cores has HT which gives it 8 threads. But somehow its not helping the game at all and is even running slower. But its a lot better overall performance just not for games.
But HT is literally the only reason that i7 is running slower. Its probably only using the 2 cores.
A difference of 0.4% is well within experimental error for most benchmarks.
True. . Not bad at all really.
It doesn't seem easy to find anything about the inner workings of HT. It appears to always be turned on and makes me wonder if that is simply the case. I mean when looking at task manager and assigning cores they are always present and turned on. So they won't ever get turned off. You would hope they are turned off if the system needs to use all the cpu but at the same time windows might just use all the cores by spreading out the tasks no matter what you do.
Windows 7 or later knows which pairs of logical cores correspond to the same physical core. If one of the logical cores on a physical core is already in use, Windows won't assign anything to the other unless all of the physical cores are already in use. If you have a processor with hyperthreading, you can see how this works at a high level by opening up Task Manager and looking at the activity on each core.
No - sorry. Windows does lock CPU cores, it is a known problem. In fact, if you have 2 or more cores, than more than likely only one core is used and the others are locked. I am using Park Control to deal with this and it helps free the CPU cores up.
This is the way Windows defaults in my experience - to load up HT cores first, and leave CPUs parked if possible. That allows Turbo frequencies to kick in higher, which if you don't actually need the additional cores, will give you better performance overall.
It talks about how HT works at the hardware level, which is neat.
It doesn't say squat about how the OS scheduler uses HT, because the decision to use a logical core vs a physical core is partly up to the OS system scheduler (partially via CPU affinity, and partially via C6/Core Parking), the hardware just makes it's assets available and does whatever it's told.
It talks about how HT works at the hardware level, which is neat.
It doesn't say squat about how the OS scheduler uses HT, because the decision to use a logical core vs a physical core is partly up to the OS system scheduler (partially via CPU affinity, and partially via C6/Core Parking), the hardware just makes it's assets available and does whatever it's told.
Glad I didn't waste my time reading something that had nothing to do with our topic. Yea I did try to find something and its not easy to come up with exactly how windows handles the cores and when it decides to use them or not use them. I've even found that some single core applications can be spread over a few of the threads even though its something as simple as notepad.
It talks about how HT works at the hardware level, which is neat.
It doesn't say squat about how the OS scheduler uses HT, because the decision to use a logical core vs a physical core is partly up to the OS system scheduler (partially via CPU affinity, and partially via C6/Core Parking), the hardware just makes it's assets available and does whatever it's told.
Glad I didn't waste my time reading something that had nothing to do with our topic. Yea I did try to find something and its not easy to come up with exactly how windows handles the cores and when it decides to use them or not use them. I've even found that some single core applications can be spread over a few of the threads even though its something as simple as notepad.
There is a good reason that there isn't a lot of concrete information available about the Windows Scheduler. Even digging into Linux, which is totally open source, it gets confusing because different distros handle it differently.
First, it's pretty complicated. I won't pretend to understand it all, but it has a lot to do with power management, with maintaining data integrity, and with raw performance - so it juggles all of that.
Second, it's proprietary. Windows doesn't publish is publicly because it's their own custom algorithm that they believe gives Windows an advantage over other operating systems.
Third, even just looking at anecdotal evidence, such as your Notepad test, it may vary from system to system, and even on the same system vary between different times. It all has to do with how your power management is configured, the state of various caches and registers, what else is running in the background, and probably even the phase of the moon.
Ridelynn said: It doesn't say squat about how the OS scheduler uses HT
...you need to understand how HT and CPU works first, half of the questions would be moot by then.
Besides, I was replying to filomret's "It doesn't seem easy to find anything about the inner workings of HT."
Guess you haven't been reading anything just sitting there laughing then acting like you know whats going on. We were actually having an intelligent conversation wondering how and when HT works. Not how they build the computer chip and the idea of what it does. We all know what it does. And I'm not about to read through that entire paper to find an answer that google does not have. You don't even know the question and somehow you have given us an answer.
When does HT become active? Is it always being used or just sometimes? If there is a heavy load will HT be disabled and the basic 4 cores used instead? Got a source or just another wall of text to answer those simple basic questions?
You know, this is an interesting academic discussion.
But honestly, I don't really care if Windows says I have only one half of one core, or a bajillion cores, because all that ultimately matters is how the system performs in applications and games that I am currently using. If it takes 12,080 cores or 1 core, to get to a performance level that I'm happy with, then other than it being somewhat interesting I don't really care and I'm happy.
Kinda one of the reasons I'm still running that over-aged 1090T in my system. Though I do start to feel the performance drop with more modern (MMO) games that I play - certainly with the GTX970 alongside causing the CPU to become a bottleneck...
Comments
Witcher 3 , which was released on May 19 2015.
(relevant part starts at 3:29, but first part shows dual core pentium G3258 OCed to 4,5 GHz gets completey hammered and gives very poor experience, while i3 falls to ~30 FPS and gives choppy performance)
Then the i7 also running 4 cores has HT which gives it 8 threads. But somehow its not helping the game at all and is even running slower. But its a lot better overall performance just not for games.
But HT is literally the only reason that i7 is running slower. Its probably only using the 2 cores.
But honestly, I don't really care if Windows says I have only one half of one core, or a bajillion cores, because all that ultimately matters is how the system performs in applications and games that I am currently using. If it takes 12,080 cores or 1 core, to get to a performance level that I'm happy with, then other than it being somewhat interesting I don't really care and I'm happy.
No - sorry. Windows does lock CPU cores, it is a known problem. In fact, if you have 2 or more cores, than more than likely only one core is used and the others are locked. I am using Park Control to deal with this and it helps free the CPU cores up.
http://www.intel.com/content/www/us/en/support/processors/desktop-processors/000006865.html#1c
http://web.archive.org/web/20121019025809/http://www.intel.com/technology/itj/2002/volume06issue01/vol6iss1_hyper_threading_technology.pdf
It talks about how HT works at the hardware level, which is neat.
It doesn't say squat about how the OS scheduler uses HT, because the decision to use a logical core vs a physical core is partly up to the OS system scheduler (partially via CPU affinity, and partially via C6/Core Parking), the hardware just makes it's assets available and does whatever it's told.
https://msdn.microsoft.com/en-us/library/windows/desktop/ms685096(v=vs.85).aspx
https://msdn.microsoft.com/en-us/library/aa394373(v=vs.85).aspx
First, it's pretty complicated. I won't pretend to understand it all, but it has a lot to do with power management, with maintaining data integrity, and with raw performance - so it juggles all of that.
Second, it's proprietary. Windows doesn't publish is publicly because it's their own custom algorithm that they believe gives Windows an advantage over other operating systems.
Third, even just looking at anecdotal evidence, such as your Notepad test, it may vary from system to system, and even on the same system vary between different times. It all has to do with how your power management is configured, the state of various caches and registers, what else is running in the background, and probably even the phase of the moon.
Besides, I was replying to filomret's "It doesn't seem easy to find anything about the inner workings of HT."
When does HT become active? Is it always being used or just sometimes? If there is a heavy load will HT be disabled and the basic 4 cores used instead? Got a source or just another wall of text to answer those simple basic questions?
"going into arguments with idiots is a lost cause, it requires you to stoop down to their level and you can't win"