Are core2duos slower than single cores?

Derek · Post by **Derek** » Thu Dec 07, 2006 9:42 am

Exactly, you should get whichever cpu suits your needs and be happy with it. There will always be better ones at doing other things and you have to just take an average and go with it.

Rescator · Post by **Rescator** » Thu Dec 07, 2006 1:06 pm

Derek wrote:Shouldn't the single core be busy with all the little things that make the computer tick while the dual core has a friend to help out and should be belting along.

The "little things" are so insignificant compared to the cpu drain the loop causes, it's like throwing pebbles in the ocean, you can't create waves.
The other cpu/cores may make the system seem more responsive.
Almost as if you had put Delay(0) inside that loop but without the speed loss due to the constant context switching.

To put it another way, increase the run of the loop,
then start it and use a browser. (they usually run multiple threads)
Depending on how the OS does traffic control, if your lucky it will realize one core/cpu is very busy with that loop and put the browser and it's threads on the other cpu's. Hopefully allowing you to surf smoothly while the other cpu is working up a nice sweat.

Trying the same on a single cpu system and it will be like crawling through glue.

Derek · Post by **Derek** » Thu Dec 07, 2006 1:15 pm

If you bring task manager up and watch the cpu usage when you run the first example both cores go from being barely used to about 50%, then one carries on up to about 66% and the other goes down to about 30-40%.

As you say, if I start a browser or other program then the lowest core picks up again and everything is quite smooth.

The only thing that puzzles me is why such a short program isn't cached in one cores cache and just run by that one core.

Still, this is why I don't design CPU's!!

Rescator · Post by **Rescator** » Fri Dec 08, 2006 8:33 pm

No idea, it looks small enough to be but...
Does core2duo use a shared L2 cache?
If so that could be why, if not then I got no clue.

Derek · Post by **Derek** » Fri Dec 08, 2006 11:40 pm

Yeah, it's shared dynamically. Whichever core gets there first gets to use as much as it needs.

At least that's the way I think I read it!

Psychophanta · Post by **Psychophanta** » Sat Dec 09, 2006 10:23 am

Still, parallel processing is not reached as it should be.
Thus, i'd not purchase a core duo nowadays.
The same could be say for 64 bit, even i ignore how is the 64-bit treatment by linux 64bit OSes, however i suspect linux is also lacky in this matter.

dracflamloc · Post by **dracflamloc** » Mon Dec 11, 2006 4:01 am

actually most linux distros have had great 64 bit support for the past couple years... with near-full repositories.

NoahPhense · Post by **NoahPhense** » Sat Dec 30, 2006 6:19 pm

My Dual Core 2.16 runs 50-60% faster than my P4 2.4.

Cores work at half the speed of the over-all number, i.e my 2.16 has
2 cores running at 1.08ghz each.

Yes, true multithreading. Just like having 2 processors.

- np

johnfinch · Post by **johnfinch** » Sun Dec 31, 2006 8:16 pm

For those MAC OSX users, here are the results using a Mac Book Pro 1.83 dual core running Parallels desktop build 1970 (latest stable) and windows XP Home

Code: Select all

OpenConsole() 
start = ElapsedMilliseconds() 
For n = 1 To 50000000 
x.f = x.f+(1.0/Sqr(n)) 
Next 
stop = ElapsedMilliseconds() 
PrintN(Str(stop-start)) 
PrintN(StrF(x.f)) 
Input()

2403 without debugger
7241 with debugger

Not as bad as I thought really.

Psychophanta · Post by **Psychophanta** » Sun Dec 31, 2006 8:40 pm

here:

1625
5313

AMD sempron 3000+ (1800MHz) no dual

Derek · Post by **Derek** » Sun Dec 31, 2006 8:47 pm

I read somewhere that amd's have a lookup table on the chip for SQR's which is why I said somewhere in this thread that amd's are faster, and that goes to prove it!.