4400+ X2 doesnt feel that fast

[Zebo]

I got dibs if you want to sell it.

 

[Markfw]

Originally posted by: Zebo
I got dibs if you want to sell it.

Me too ! add me to the list !

 

[Duvie]

When he can show me some pics that prove the second core is even recognized then I will believe the numbers he is getting....

 

[Zebo]

Originally posted by: Markfw900

Originally posted by: Zebo
I got dibs if you want to sell it.

Me too ! add me to the list !

Don't you have like 3 already?

 

[Zebo]

OH and I use Fedora too.. that should improve my chances right

 

[Markfw]

If he can't figure it out, and doesn;t want to, he should just sell it to us. I don;t think he has it set up right. I know, I had a problem with my abit system. Said it had two cores, but wasn;t sending data to both cores. I played around and re-installed teh multi-processor drivers 5 times, and finally it sees two cores. Verified by F@H times.....

I only have 2 now, but I sort of promised one to Duvie, and I am getting hooked, so I want another one or 2 for myself.

 

[bersl2]

Originally posted by: ethen
guys, GCC 4 shows different result, it's 13 seconds faster with X2, btw i design my application in threaded object, so it should run faster on X2, However, i'm talking about the compiler and has nothing to do with my codes . i checked with intel support, yes it have threaded support.

so i'm not sure why it's slow, maybe it has to do with intel compiler implementation on how thread is queued.

No, really, GCC is a single-threaded program. You will not improve the performance of a build unless you instruct make to compile your program two files at a time. You can do this by typing "make -j2" instead of "make." So unless you're trying to compile a large program written as one file, or unless the program is written in such a way that there is just one long dependency chain, you will benefit from doing the above.

Post your Makefile anyway.

 

[ethen]

Okay, here is what i got

Kernel 2.6.12 (NOTE: in SMP)

setup? what do you mean? i use the same setup as my previous CPU, if you mean in CPU i did it correctly, i enabled SMP in kernel (recompiled from source code)
GCC 4 seems to run a bit faster (GCC3.4 doesnt), but still Intel Compiler doesnt run faster than 4000+

i'm sure it's Intel's compiler problem. pretty annoying swapping cpu around just to test things like this

btw i posted screenie, i tried openSSL benchmark, X2 is faster but i'm just pissed intel compiler just dont show improvement, i guess i got to start using GCC4
http://www.linux.io/blurb/sslbench.png (left single mode right threaded mode)

do you guys think it has something to do with Intel's way of implementing threading process?

 

[Duvie]

Do you have anything that shows cpu usage for each individual core so we can see if it is in fact working each core and if so...how much???

 

[ethen]

runs top -i , it seems to work.

total 19.5% 0.0% 0.0% 0.0% 0.0% 0.2% 73.7%
cpu00 20.0% 0.0% 0.0% 0.0% 0.0% 0.1% 74.0%
cpu01 19.1% 0.0% 0.0% 0.0% 0.0% 0.1% 73.4%

 

[Duvie]

What does all that jibberish mean???

 

[Markfw]

Originally posted by: Duvie
Do you have anything that shows cpu usage for each individual core so we can see if it is in fact working each core and if so...how much???

Yes, that how I figured out my problem... To really verify, run 2 instanes of something like F@H with known run times and compare. Even cpu% was lying to me.

 

[ethen]

the first % is the cpu usage the last one is idle

not sure why it doesnt use 100% of the CPU, while GCC use up to 85%

well i guess it really is Intel compiler's fault, thanks guys, i'll stick with it for a while

 

[Sixtyfour]

Originally posted by: Technomancer Personally, I'm not sure how that could be done, as I'd think it would end up bloating the code. But if they ARE doing that, then all the worse for them.

http://www.xtremesystems.org/forums/showthread.php?p=932561#post932561

That's bloating

 

[BlingBlingArsch]

did u actually know what dual-cores are good for when u bought yours?

 

[ethen]

Originally posted by: Sixtyfour

Originally posted by: Technomancer Personally, I'm not sure how that could be done, as I'd think it would end up bloating the code. But if they ARE doing that, then all the worse for them.

http://www.xtremesystems.org/forums/showthread.php?p=932561#post932561

That's bloating

That means i'm not the only one

 

[Dadofamunky]

Have you considered OCing? I would think it would be a fun exercise for you. Heck, I'd be driving that puppy at 2.7 GHz right now if I had one. I run craploads of apps so a dual-CPU would be great for me. I'm just holding off for awhile.

I'm by no means a Linux developer, but I do know my way around an IP stack. I think you made an excellent purchase. You're basically future-proof for a few years. Hell, that's overkill for Linux. It's exactly two of the exact same CPU as I have, in one socket. It's so much fun to watch this stuff continue to advance. Dude, I would be seriously digging into multiprocessing code if I were you.

 

[ethen]

FINALLY!

http://www.betanews.com/article/Suit_Intel_Sabotaged_Compiler_for_AMD/1121274628

It's finally out in the open

 

[bersl2]

I'm still not convinced that all other possibilities have been ruled out.

OK, we saw from top that it recognizes both cores. You say

i checked with intel support, yes it have threaded support.

but I do not think it means what you think it means there.

Can I at least get an ACK that you understand the parallel features of make, and that they are traditionally how compiles are sped up on SMP machines, and that unless the project has a very linear dependency tree, you should see a speed up vs. make without parallelization, regardless of compiler used?

 

[NokiaDude]

Trust me, it's fast. I remember that feeling when I first got my Winchester 3200+. XP was just as fast as a Barton 2500+, but once I tried some games, I noticed a nice boost in FPS.