Need help with K7N2 Delta-ILSR

Status
Not open for further replies.

TheBoss68

Gawd
Joined
Jul 9, 2004
Messages
632
In my ignorance I cut the 4 pin plug off my psu and discarded it quit sometime ago. My ? is, With this mobo and an XP 2500+ Barton would the lack of that 4 pin plug make a significant performance loss for my cpu, seeing that the plug is to supply power to the cpu?
If so it would surely explain my low performance numbers in sisoft and 3dmark on this mobo...I mean...my old KT4 Ultra with a 2500+ oc'd 2.04 gets better numbers than my K7N2 and 2800+ oc'd to 2.27.
both rigs have Kingston PC3200 512mb only diff is the K7N2 has Dual channel and they are matched sticks.
HELPPPPPPPPPPPPPPP!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
 
The lack of the plug shouldn't affect performance, only stability.

The plug is there to provide a cheaper way to construct the board's PCB, as running tons of power accross a mobo can increase production costs. Having a closer source to the CPU for such power makes for a cheaper produced mobo.

The performance drop is likely due to a different issue, but remember that Sandra and 3dMark do not equal real world performance.

One reason why you may see raw bandwidth lower for memory is that the Athlon's bus can be saturated rather quickly with only a single channel of ram. The second channel of ram will prevent processing slowdowns when heavy DMA action is happening- basically provides for more seamless multitasking as opposed to bumping your scores... There's a good article at lostcircuits.com about this... tacked on to the A7N8X-X review, if i recall correctly.
 
But its not just the memory bandwidth scores....my 2800+ at stock speed gets lower numbers than my 2500+ at stock speeds...and they are both Barton cores....The 2500+ is on my older MSI KT4 Ultra, and the 2800+ is on the newer MSI K7N2 Delta...from my reading around...I was under the impression that the nforce 2 platform was the best for the AMD chips....in my case it seems that the VIA KT400 is the best. I can agree on the stability issue....it is extremely hard to get a g00d overclock on my 2800+ anything over 182 fsb and the voltage at 1.70 and it locks at POST. Any other ideas i can try?????
 
Are you running yoj rmemory in sync w/ the fsb, and at what?

What timings are you using?
 
I am running Kingston PC3200 Dual channel and it is running async. its at 400mHz and the cpu is at 166 mHz at the moment....I reattached the 4 pin plug and i saw a very slight incresse in performance....but nothing near what i expected. The thing is all I have heard about the nforce 2 chipset is that its the bomb for the amd's...but my via run better with my 2500+ than the nforce with my 2800+
 
some one tell me how to get images uploaded and i can give some shots of cpu-z and sisoft to show what is going on
 
I think there's a free image hosting thread stickied to the FS/FT Forums.

First of all, running memory in sync w/ the proccessor is one of themost important things that you can do for nF2 performance. VIA chipsets tend to run better async.

Another big difference is one of RAM timings. 2-2-2 @ 166 beats the snot out of 2.5-3-3 @ 200 (or even 220) every day of the week. Droping the memory to 166 will yeild you a real world performance gain, but not neccessarily a synthetic gain (likely a loss).

Please recall that SiSoft is a synthetic benchmark that does not translate into real world performance. I've never been able to hit some of their preloaded numbers- with any of the systems that i've assembled and tested (hundreds). You will see the difference in the platform when you crank up a few of the following things- a few dozen gig network transfer, while loading up some mp3's to play while fragging people in UT2K4. You should see a huge difference in this arena given the same specs except mobo.

As for your OCability, there's quite a good chance that you got a chip that won't do much better than what its been stamped at.... nothing to do with the mobo or chipset.
As you should note from the [H]'s "Overclocking the 2500+" article, some Bartons are the limiting factor in FSB oc'ing.
 
Damn...a GA Boy...not far from my home town of Griffin.
Anyway...just got outta the bios again....i have it running stable with FSB@ 195 and Ram 1:1. timings @3/3/4/8, its kingston Value Ram Dual Channel so it won't do much more than that. And its not the preloaded scores in sisoft im am refering to, My 2500+ on a KT4 is getting higher all around than my 2800+ on the K7N2 let me get the pics working and i can post and show.
 
Make sure you use the same exact version of Sandra (down to the build #)...... But i'm sure you've already considered that....

Here's my memory bandwidth score, i'm at 11.5x166 right now to give my a/c a break for the summer months. Timings are 2-3-3-7.

sandra.jpg
 
CPU-z CPU-z Sisoft Sisoft This is my 2800+
I think I bout got it figured out. But still not what was hoping for!
And my 2500+ @ 12.5x166 on my KT4 Ultra is getting mem scores of Int buff 2513 and Float buff 2349
Let me oc it to where you have yours and see what it does.
 
Your images don't work.

3dMark01 is a better motherboard/memory bandwidth/latency/cpu tester.... 03 is generally graphics card restricted (a 700mhz overclock won't change the score more than 10 points, in most cases).
 
You are lucky that your system posts without the 4 pin power connector, and OC ...
 
k...just ran 3DMark01 and got a score of 16947.
Not bad for a Radeon 9700pro clocked at 370/624

Waiting for my hosting site to let me post pics!!!
 
Status
Not open for further replies.
Back
Top