Hyperthreading (CPU split into two)

Has anyone tried forcing Windows to just see the CPU as one processor and disabling the dual processor functions? If so, did they see any performance increases? I've seen a few mentions of it but nothing on how well it works.

 

performance increase ??

 

if you disable HT I would expect a decrease in performances, not an increase.
how is 1 CPU better than 2 (though not physical CPU)??

 

run this command in windows

msconfig

go to boot.ini

advanced options

set /NUMPROC= to 1.

Save and the restart

By setting the value to 1, you'll tell windows to limit the running "processor" to 1 logical processor only.

I wouldn't disable the hyperthreading if I were you as it "improve" your AAO's processor capability. If you don't need that extra 30% speedboost, you'll save a bit of power by turning hyperthreading off.

And no, the hyperthreading function does not split the cpu into 2. That's why you don't get 100% speed improvement.

 

I assumed that Hyper-threading meant that you get two 800mhz processors instead of one 1600mhz processor. Obviously not from the comments!

 

Hyperthreading could be described as 1 fake processor and a real processor.

 

hyperthreading allows the cpu to run two 'threads' (applications) at the same time, it uses the idle time from the first thread to run the second one, if you actually watch the cpu graph and max out the cpu on one thread (using an app like cpu burn in or similar ) you will see that the second 'core' sits at 0% most of the time as it has no processing time to work using. so really your not getting 2 cpus just 1 800/1333/1600mhz cpu that can do 2 things at once.

same test on a dual core system would show the second core working normally even if first is maxxed

 

No...

HT works like dj_steve said operating on "two" CPUs 1.6GHz each...

 

The Intel Atom is a inorder execution processor, that means it has no hardware logic to rearrange the instructions it executes.
This is of great importance for performance as a processor only have a fixed amount of units that can handle specific tasks.
It's also important that you realize that instructions may take several "rounds" (cycles) before completion.

Just imagine a processor that has two addition units, one multiply unit and one branch unit.

If we say an addition takes 3 cycles, a multiply takes 5 cycles and a branch 2 cycles. (and for making this example easy we assume the processor can issue one instruction per cycle)

That means if the instruction feed looks like "multiply, add , branch, add, branch, add" the multiply unit will be busy for the next 5 instructions, so if we encounter another multiply operation within these 5 instructions the processor will get jammed here until the previous multiply is completed- that's not the case in this example so we move on.
the add instruction takes 3 cycles in this example, ie if we encounter another add within 3 instructions we would be jammed... or would we? No as the imaginary processor in this example has two ALUs it can issue two adds simultaneously.
Next instruction is a branch, it has a latency of 2 cycles- and obviously there is a branch within two cycles... well you get the point.

Most processors (the expensive ones) have complex logic that can on the fly "optimize" code in various ways, one of these optimizations are just reordering of instructions to avoid these jams as much as possible,
The Intel Atom does not however... and this is where hyperthreading comes into play.

Hyperthreading is basically a duplicate set of registers, and basically runs this additional register set when the "main thread" is jammed- so say we got jammed on the multiply instructions above, that means the processor can issue additions and branches on the other logical thread during this time.

I don't remember the exact number but I think hyperthreading is just like 5% extra circuitry and theoretically may give up to about 30% higher performance in a multithreaded context. It's basically used on all "lowcost" desktop class processors today,
hyperthreading is Intels trademark though, IBM calls it simultaneously multithreading (SMT) and it's used in both the Xbox360 Xenon chip, and the Playstation 3 Cell chip as well...

This is a very simplified explaination, I hope it sheds some light on the technology though.