Does dual core mean?

Mr Green · 2006-12-11 16:54:39

Ok so you have a dual core system does it mean its twice as fast ?

Does ram have an effect on system performance ?

Can the kernel be configured to run better under dual core ?

and the big one ... is 64 bit better than 32 ?

Not meant as a troll or flame war... I am interested thats all

chrismortimore · 2006-12-11 16:59:26

I remember I asked a friend about this a few months ago (he had just got a fancy X2 machine).

Apparently, if a program is threaded properly, the kernel can migrate threads to each core and give you twice as fast performance.

If the program isn't threaded, it'll only run on one core and be the same as unicore. But, if you run multiple programs (lets face it, when have you only had one process running), it can load balance them, so you experience less slowdown.

But, if you're not doing anything CPU intensive, it really isn't worth it. And most games don't use both cores (true at time of conversation), so don't experience much benefit. Other than theoretically having a core to itself...

But that isn't my experience, so I can't say whether any of that is true.

As for 64 bit and 32 bit, this I can comment on. And I am yet to see a difference.

Mr Green · 2006-12-11 18:01:48

My system is 64 bit but I still run 32 Arch ... had so many problems running 64 Arch [early days!] I gave up

So you cannot say dual core s twice as fast ......

chrismortimore · 2006-12-11 18:06:59

Mr Green wrote:

So you cannot say dual core s twice as fast ......

It's only potentially twice as fast. Apparently...

Hopefully someone here will know first hand...

skottish · 2006-12-11 18:11:03

I run batch processing using multiple instances of Gimp, and I can pin both cores of my processor at 100% for hours at a time. This also works for rendering in Blender if your have the threads set up properly.

Mr Green · 2006-12-11 18:58:35

guess I need to read up on threads ;-)

alger · 2006-12-11 19:01:32

chrismortimore wrote:

Mr Green wrote:
So you cannot say dual core s twice as fast ......
It's only potentially twice as fast. Apparently...
Hopefully someone here will know first hand...

IIRC It's no even potencially twice as fast, due to memory and cache access.

phrakture · 2006-12-11 19:08:51

As a quick rundown, most multicore chips work like so:
Memory access and L2 cache is still the same - a single unit. The bus from L2 to L1 caches are split, so each core has it's own L1 cache, though they're probably smaller than a single core, due to die size limitations.
Each "core" is responsible for all the non-resource work of a typical processor - that is, acting on each opcode - adding, subtracting, reading memory, storing memory, etc.
It does have the potential to be "twice as fast", however, this is typically not the case. Special coding practices need to be taken into account to properly work in parallel on both processors. The usual way of working is to use each core for a separate action, not both cores together, but I'm digressing.
One big fault with multicore processors, is that, due to the shared pipeline, a cache miss, or wrongly predicted branch will, in theory, have more of an impact than on a single core chip. I haven't studied the caching strategies and branch predictors of the newer mutlicore chips, so I cannot claim this as fact, but it is a risk.

Mr Green · 2006-12-11 19:42:27

I run a dual core system and to me its not twice as fast maybe a little quicker than a single core but not but much...

would I notice any speed increase say in running beyond kernel or building kernel for Intel etc... ?

but I'm not sure I want to go down that road

iBertus · 2006-12-11 20:45:50

Mr Green wrote:

I run a dual core system and to me its not twice as fast maybe a little quicker than a single core but not but much...
would I notice any speed increase say in running beyond kernel or building kernel for Intel etc... ?
but I'm not sure I want to go down that road

Probably not. Most claims of greater performance from custom kernel builds can be attributed to the placebo effect. With a dual core system you already have so much processing power that you shouldn't notice a difference with the kernel. Now, if you want to go and recompile your media apps with processor specific flags and SSE2/3 you could pick up a margin of performance.

Mr Green · 2006-12-11 20:58:45

yeah but then you are going down the 'G*****' route ...

performance issues well I only use the web mail ..etc ...

Media mostly music [speakers for Xmas that reminds me!]

would be interesting to compare times against a single core box ....

pauldonnelly · 2006-12-12 06:52:19

What you are likely to see are fewer lags. I don't have a dual core system, but my dual Pentium Pro box, while slow, remains responsive even when doing something demanding on one core.

chrismortimore · 2006-12-12 09:15:25

To really feel dual core, you'd need to run something big and intensive. A price example would be "make" for something like... glibc or something like that. Compile it on a unicore system (with a core speed similar to one of the cores of the dual core), then compile it on the dual core (with an appropriate -j setting), and feel the speeeeeddd!

AndyRTR · 2006-12-12 09:35:59

chrismortimore wrote:

...Compile it on a unicore system (with a core speed similar to one of the cores of the dual core), then compile it on the dual core (with an appropriate -j setting), and feel the speeeeeddd!

...and run into one of those ugly "go back to -j1" errors

glibc was a bad example. often parallel build is broken. sometimes the build will break. sometimes it goes through but one or more files are missing in the pkg. you have to take care :!:

Mr Green · 2006-12-12 09:42:22

do not like messing with flags .... I'll time next kernel build

chrismortimore · 2006-12-12 09:52:12

AndyRTR wrote:

...and run into one of those ugly "go back to -j1" errors
glibc was a bad example. often parallel build is broken. sometimes the build will break. sometimes it goes through but one or more files are missing in the pkg. you have to take care :!:

Really? I've never had issues building glibc using distcc across 3 machines...

How about gcc then? That a better example?

AndyRTR · 2006-12-12 10:17:56

gcc will miss some headers. kernel builds fine with smp and with distcc. maybe distcc and local smp build with -j>1 are somewhat different.

btw: using distcc together with 2 dualcore system really rockz 8)

in most cases where i run into smp issues i added -j1 into our PKGBUILDs. abs will tell you ;-)

chrismortimore · 2006-12-12 10:37:19

AndyRTR wrote:

gcc will miss some headers. kernel builds fine with smp and with distcc. maybe distcc and local smp build with -j>1 are somewhat different.

Maybe. But I'm poor and can't afford a real smp system

brazzmonkey · 2006-12-12 17:51:16

well, i have arch on a desktop with an athlonxp 2200+, 1gb ddr ram and 2x 7200rpm HD.
i have kubuntu on a laptop with with core2duo 1833MHz (mostly running @ 1000 MHz), 2gb ram and 2x 5400 HD.

both are running a 32-bit linux.

Kubuntu is much faster, though i reckon my 2 computers have very differents specs. but i think dual core really kicks ass. i don't think quad-cores will be such an improvement for daily use, though (read : no games, no compilation, etc.).

pauldonnelly · 2006-12-13 07:34:27

brazzmonkey wrote:

Kubuntu is much faster, though i reckon my 2 computers have very differents specs. but i think dual core really kicks ass. i don't think quad-cores will be such an improvement for daily use, though (read : no games, no compilation, etc.).

Yeah, I can't see quad cores being especially useful unless your trying to recompile your kernel while you play games, render 3d artwork, and do something else complex. Even a modern single-core is pretty tricky to tax, since usually when you're waiting it's because a single application is running slower than you'd like.

chrismortimore · 2006-12-13 08:58:26

Out of curiousity, does anyone know if mencoder takes advantage of smp? Cos the only real cpu intensive things I do these days are compiling and mencoder..

johnisevil · 2006-12-14 23:16:02

I've been using dual core now for almost a year and I wouldn't say it's faster like everyone would like to think, it's just that my system never hits any slow downs. Currently I'm using a socket AM2 Athlon 64 X2 4600 at stock speed (2.4ghz), ASUS M2N-SLI motherboard, 2gb of dual channel memory, etc. The biggest factor of my system's speed is the clock speed. A 2.6ghz single core processor will most likely out perform a 2.4ghz dual core in terms of brute speed, but with the dual core you'll be able to stress the system more by being able to throw more at it at once.

iphitus · 2006-12-15 05:41:25

johnisevil wrote:

A 2.6ghz single core processor will most likely out perform a 2.4ghz dual core in terms of brute speed, but with the dual core you'll be able to stress the system more by being able to throw more at it at once.

pretty blanket comment. depends on the task, if it's something that can take advantage of the dual core, for example compiling, then the dual core would outpace the 2.6 easily.

When I got this computer, I did a compile-off against Con Kolivas' 3.06ghz HT P4 with my Pentium D 2.8ghz dual core. Mine smoked it, compiling a kernel of the same config in nearly half the time.

James

johnisevil · 2006-12-15 07:31:52

iphitus wrote:

johnisevil wrote:
A 2.6ghz single core processor will most likely out perform a 2.4ghz dual core in terms of brute speed, but with the dual core you'll be able to stress the system more by being able to throw more at it at once.
pretty blanket comment. depends on the task, if it's something that can take advantage of the dual core, for example compiling, then the dual core would outpace the 2.6 easily.
When I got this computer, I did a compile-off against Con Kolivas' 3.06ghz HT P4 with my Pentium D 2.8ghz dual core. Mine smoked it, compiling a kernel of the same config in nearly half the time.
James

Well, of course. This applies to anything that can utilize both cores simultaneously other than multi-tasking. Compiling is a perfect example of that. What I meant was in situations where both cores can't or won't be used at full speed at the same time. If you were to recompile that kernel with your Pentium D 2.8ghz but only using a 1 core, I'm pretty sure the 3.06ghz processor would out perform it. Now using both cores, your system obviously smoke his.

Mr Green · 2006-12-15 07:44:56

@iphitus do you use any special flags to get better system performance ?

Arch Linux

#1 2006-12-11 16:54:39

Does dual core mean?

#2 2006-12-11 16:59:26

Re: Does dual core mean?

#3 2006-12-11 18:01:48

Re: Does dual core mean?

#4 2006-12-11 18:06:59

Re: Does dual core mean?

#5 2006-12-11 18:11:03

Re: Does dual core mean?

#6 2006-12-11 18:58:35

Re: Does dual core mean?

#7 2006-12-11 19:01:32

Re: Does dual core mean?

#8 2006-12-11 19:08:51

Re: Does dual core mean?

#9 2006-12-11 19:42:27

Re: Does dual core mean?

#10 2006-12-11 20:45:50

Re: Does dual core mean?

#11 2006-12-11 20:58:45

Re: Does dual core mean?

#12 2006-12-12 06:52:19

Re: Does dual core mean?

#13 2006-12-12 09:15:25

Re: Does dual core mean?

#14 2006-12-12 09:35:59

Re: Does dual core mean?

#15 2006-12-12 09:42:22

Re: Does dual core mean?

#16 2006-12-12 09:52:12

Re: Does dual core mean?

#17 2006-12-12 10:17:56

Re: Does dual core mean?

#18 2006-12-12 10:37:19

Re: Does dual core mean?

#19 2006-12-12 17:51:16

Re: Does dual core mean?

#20 2006-12-13 07:34:27

Re: Does dual core mean?

#21 2006-12-13 08:58:26

Re: Does dual core mean?

#22 2006-12-14 23:16:02

Re: Does dual core mean?

#23 2006-12-15 05:41:25

Re: Does dual core mean?

#24 2006-12-15 07:31:52

Re: Does dual core mean?

#25 2006-12-15 07:44:56

Re: Does dual core mean?

Board footer