• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

L12 Mod Really Does Raise Fsb Up 20+ Mhz!!![Retired sticky]

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.
[OC]This said:

hitechjb1, the CPU Interface being On will boost your memory bandwidth/throughput. CPU Interface being turned off is giving you lack in your memory performance right now. CPU Interface acts similar to the CPU Fast Decode, found in Via KT266/333/400 chipset based boards. If you could manage to enable that, that would be great.

my old xp2700 aiucb + L12mod(for 133mhz) on a NF7 rev1.2@ 238mhz FSB DC CPU Interface enabled & at 4-2-2-2.5 gives me this:

...

hitechjb1, btw, I really appreciate your work. You have put a lot of effort in your Tbred B 1700+ DLT3C 2.5+ GHz air results thead in AMD CPU section. Nice work. :) Later, when i have the time, I will try the L12 mod wire-drop-socket method for my xp1700 (for 200mhz) as you have suggested in your thread.

Thanks for the info.

Actually I tried the CPU Interface, with that ON, the highest FSB I can get is 223 MHz, after the FSB_Sense (L12) mod. With the CPU Interface OFF, I can get FSB to run 230+ MHz.

I also checked with the CPU Interface ENABLED, it did NOT improve memory bandwidth efficiency at all, it still stays around 87-89% regardless of CPU Interface ON or OFF, also regardless of SC or DC.

From your Sandra memory bandwidth, you are getting 97% efficiency. q149 and among many others are getting 95-97% memory efficiency.

I tried almost everything I can think of to improve the memory bandwidth,
- bios versions 1.6, 1.7, and settings, such as spread spectrum, CPU Interface, ...
- chipset drivers, 2.03 and 2.45
- SC, DC,
- memory ras/cas timing, from 5-2-2-2, 6-3-3-2, to 11-2-2-2, ...
still cannot get the efficiency above 90%. Also the memory itself can get 95% efficiency on an A7N8X-DLX, so memory seems not the problem.

The only remaining things I can think of is
- to reinstall Windows from scratch (but I doubt this would help)
- motherboard itself has some hardware problem
- motherboard and memory module have incompatibility problem
 
Last edited:
I am using 2x256 TwinMOS CH-5. I just realized something though.. i am using the 133 > 200 mod.. i thought i did the 166 one before by connecting AH30 to AG31 but it did nothing at all, so i connected AH30 to ground and now i get 200 fsb. Which one should i do and did i do the 166 one right? Btw, i got all my cooling lapping and Vdd mod done.. i dropped 2 of my hardrives that i was using to the ground along with a psu that were on my desk and now my hdd's are f'd i think, but hopefully i can fix it without reformating, i had 3 windows installs i was using on one of them.
 
q149 said:
I am using 2x256 TwinMOS CH-5. I just realized something though.. i am using the 133 > 200 mod.. i thought i did the 166 one before by connecting AH30 to AG31 but it did nothing at all, so i connected AH30 to ground and now i get 200 fsb. Which one should i do and did i do the 166 one right? Btw, i got all my cooling lapping and Vdd mod done.. i dropped 2 of my hardrives that i was using to the ground along with a psu that were on my desk and now my hdd's are f'd i think, but hopefully i can fix it without reformating, i had 3 windows installs i was using on one of them.

For 133 to 166, only connecting AH30 and AH32

For 133 to 200, requires two U-shape wires
connecting AH30 and AH32
and also connecting AG31 and AH32
(as shown in the picture of the link few posts back).

You connected AH30 and AG31, it would do nothing and should do no harm to CPU, since both of them are logic 1 protected through two 10K ohm pullup resistors. But be very careful though, wrong pins may kill motherboard and CPU.
 
I looked at your pic for 133 to 166 and it looks exactly like what i did.. i get 200 fsb though.

On this image i connected the top green dot that says it's AH30 to the closest blue one directly to the rigiht.
socket462_fsb_sense_pin.gif
 
You are correc that the AH32 Vss is on the right (horizontally in pic), and needs to be connected to AH30 to get 133 to 166 mod.

BTW, the pic is bottom view instead of top view. At first, I did not even see the light blue pins, so it was confusing for me earlier.

For the 200 mod, you need to connect also AG31 to to Vss.

I think you may just refer to my link for the pics for 133 to 166, or 133 to 200. I think they are clear also.
 
I did do it the way for 166 then but i get 200. I guess i will take it out and try again. I still think my FSB is being limited by cpu, i can do the same fsb DC with fastest timings as i can SC with most relaxed timings. And i have Vdd at 1.94 with very good cooling. I'm getting less FSB than my 8rda. btw one of my hdd's is dead from that fall:(
 
Same for me, after I did a 133 to 166 mod, the default setting for the CPU is a 3200+ with FSB 200. Same for the 133 to 200 mod.

I can get to FSB 238 MHz with Vdd at 1.7 V. I tried Vdd at 1.95 V, but such Vdd did not help FSB to go higher, probably the memory is limiting. I think I need to do Vmem mod to get memory higher, but it is not easy to do using SMD grabbler since the component pin is so small.

Also I begin to think CPU maybe limiting FSB, ...

If you know of a good way to do Vmem mod, pls let me know.

In this respect for higher adjustable Vmem and Vdd, the 8RDA3+ (400 Ultra) would be much better (higher in bios, no need for vmod).

The NF7-S Vcore setting can go way high, but its Vmem is pretty limited.
 
I tried the mod again and now all i can select is user define which i understand is correct. Still though i am getting crap FSB.. i can get same fsb DC 3-2-2 as i can SC 3-3-2.5. It's starting to **** me off now.. maybe a different bios will help, from what i have read they can make a huge difference in FSB. I am using 1.4 right now. I guess i will try all the bioses. I will do the Vdimm but i need to get the proper v-r first. I just use solder which is pretty easy once you get used to it.
 
My board came with the Bios 1.6 on it, tried 1.7 Bios, not any noticeable difference on overall and FSB overclocking.

For Vmem, I think the Hardware Doctor Vcc (2nd row right under Vcore) is Vmem. Mine shows 2.98 V at Bios setting of 2.9 V.

What FSB did you get before and after the 133 to 166/200 mod?
 
Before mod i was unstable at 215 and no boot at 220. I am finally getting into the 230's now but not with cpu interface on and less sandra score at aroiund 232 3-2-2-8 as at 220 3-2-2-8 with cpu interface enabled both in DC. I have tried 3 bioses now and they have made a HUGE difference in my max FSB with or without cpu interface. This board is so touchy. Vdd is at 1.93 and i measured a mosfet near the memory slots that might be Vdimm at 3.1. Still doing same FSBs with any memory settings dc/sc or even async.
 
So your board and my board behave very similar, if you look back my posts:

Before mod 215, 220 no boot.

After 133 to 166/200 mod, highest FSB DC 228, SC 230+.
With CPU Interface ON, highest FSB 223.
With CPU Interace OFF, FSB 230+ SC.
Raising Vdd above 1.7V no help.

Sandra memory bandwidth is still low under 90%, regardless of tightest memory timing, bios versions 1.6, 1.7. Chipset drivers 2.45.

What is your Sandra memory score and efficiency now?
 
I scored about the same with 14 bios 220 3-2-2-8 DC /w CPU interface as with 12 bios 230 3-2-2-8 DC /w out CPU interface. I can still do the same FSB regardless of any memory timings or DC/SC so i don't think it's my memory. I raised Vdd to 2.03. Later I am going to reseat my cpu HSF and put everything in my case then test all the bioses. Is it impossible for me to go to 10 or 11 bios now? it fails everytime i try from any other bioses above it.
 
Did some more testings here are some results:

As OC[This] have pointed out in this thread and other threads, CPU Interface enabled delivers higher memory bandwidth.


As posted earlier, with the CPU Interface enabled, highest DC FSB is 223 MHz (with Vdd mod to 1.87 V measured, with a fixed resistor of 3 KOhm and bios Vdd setting at 1.6V). I am able to get a memory bandwidth efficiency of 95-96% in DC (problem posted in previous posts solved), in line with what others have been reported, ... Though the highest FSB I can get now is not as high as up to 240 MHz or more reported by others.

The highest memory efficiency with CPU Interface disabled is 87-89%, FSB can go as high as 228 MHz DC.

With CPU Interface disabled, for SC, FSB can go 230+, efficiency 1-2 % lower on average. So as reported in previous posts, the difference between SC and DC is still only 1-2%. So choosing between SC and DC is a matter of the tradeoff between the FSB gain in SC vs the performance gain in DC. Difference between the two on overall performance is within 1-2% (either way). Also depending on how many memory modules are needed, i.e. if two modules are needed say for 512MB x 2 = 1 GB total, DC would be the choice.


With the CPU Interface enabled, the higher memory bandwidth of 5-7% (compared to CPU Interface disabled), is significant and out benefits the FSB gain (which is less than 5-7% or 10-15 MHz) with CPU Interface disabled.

Also with CPU Interface enabled, the differece between SC and DC efficiency is within 2-3 % (SC being lower). So same argument as before: choosing between SC and DC is a matter of the tradeoff between the FSB gain in SC vs the performance gain in DC. Difference between the two on overall performance is within 1-2% (either way). Also depending on how many memory modules are needed, i.e. if two module are needed say for 512MB x 2 = 1 GB total, DC would be the choice.


PS:
The NF7-S rev 2.0 I have seems to be very sensitive to some bios setting such as CPU Interface, memory timing, even CPU multiplier, ..., and among other things, ... wrong setting means no post/boot. Reasons yet TBD.

Before the FSB 166/200 mod, it did not boot at 220 MHz FSB.
The memory efficiency drops to 80% if CPU is running at low speed around 1.5 GHz.

One more strange thing: at FSB 223 MHz, it is stable at x11, x10.5 and x9.5 multipliers, but it cannot boot in windows at x10 multiplier. I don't know whether this has something to do with CPU also, since it is running with stock HSF.
 
Last edited:
I am about to give up on this board. It's soo damn touchy and wierd. Nothing makes sence.. it's like i have to find the perfect FSB/CPU speed/Multi combination that it loves or else it will be unstable even if anything is at slower speeds or same speeds that are stable at other settings.
 
I have had no luck with the AH30 to ground mod. Never has gone to 166fsb default when reset. I even pulled the mobo and soldered wire to backside. no luck. Is it possible that my vdd and vdimm resistor mods are interfering?? I even tried closing bothe sense 0 and 1 but system wouldn't post (scared me for a while there) And yes I am VERY certain I used the correct holes.
 
q149 said:
I am about to give up on this board. It's soo damn touchy and wierd. Nothing makes sence.. it's like i have to find the perfect FSB/CPU speed/Multi combination that it loves or else it will be unstable even if anything is at slower speeds or same speeds that are stable at other settings.

Yours behaves very similar to mine, and all your description describes what I have been seeing on mine. Seldom have a board behaving like this particular one I have.

It works at certain setting of bios, fsb, multiplier. If things are changed for faster or slower, it would become unstable, or unexpected, ... It seems to me for the board I have, its system behavior is hard to predict and extrapolate smoothly, ... Look like some kind of system timing issues, ...
 
Last edited:
I would also like an easy method on a vmem mod I would love to take this ram to the next level......I am not sure if the pin mod will help me but I think I might give it a go:D
 
climbski said:
Try SMC grabbers but that resistor is very tiny. I just used my awesome soldering skilz...

Unless I can find another point to do the Vmem mod. The known point to that resistor is too small and no hook point for an SMD grabber.

Anyway, the current problem for my board is not memory, but it cannot go higher than 223 MHz DC with CPU Interface enabled.
 
seems as tough it is a no go for me and getting 200fsb.....not to say that 196 is bad ting with this ram.....I just wanted like everyone else more......:D
 
Back