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

[muzz]

This mod seems to work best for most folks when using 10 bios , I had to flash to 10 to enable cpu interface on all multi's.

 

[Hound]

Ok, I set everything back to how it was, but I'm still longing for higher FSB... So, one more quick question: is there any other pin I could use to set the default to 166 other than AH30? On another forum I believe someone used the AH28 pin. I'm a complete noob about this subject.

 

[El<(')>Maxi]

No, AH30 is the only pin to use(aside from AG31) to change FSB sense's. What multi are you using when trying 205 fsb and over?

 

[IGUANAS]

can u go to 200mhz default in any way??

 

[Cúchulainn]

I finally had to breakdown, pull the prommy apart, and give the L12 mod a try this morning before work Piece o' cake to add the c-shaped wire to the socket. I used AH30-32 and my 2100+ booted up just fine as a 2700+.

I didn't have too much time to tweak before leaving for work but did noticed a few things. First, this cpu/mobo absolutely would not go over 223fsb without locking up switching to the second test in prime95. Without fail it did this at all multipliers, memory settings, and vcore I tried. Of course I had to immediately test this out after the mod and it chugged away just fine at 230. I did have to back off the mem timings with my twinmos (from 2-2-2-5 to 2-3-3-6), though I was able to leave the cpu interface enabled. So, I'm hoping I can get at least the same MHz at 0.5 multiplier down from where I was before (223x12 to 233x11.5 or so.)

Anyhow, no question this simple little mod helped in my case.

 

[Hound]

Without the mod I am using 205X12 and it's stable. With the mod, I couldn't use a FSB above 200 no matter what mem timings or multiplier I used. I went as low as 200X9 on the multiplier and as relaxed as 3-6-6-11 on the timings (of course I tried a bunch of other more 'normal' settings). I can't believe that I'm the only one who has had negative results from the mod... It just has me wondering if I did something wrong (not much to screw up tho - C-shaped wire in the appropriate socket).

 

[Mr B]

I don't know....maybe I'm missing something here....

I understand the concept, in which you're basically fooling the CPU and board into thinking it's a 2700+. But how is this going to help you if the chip maxes out at say, 11 x 205, or the DDR you have just doesn't want to go past 210, no matter what voltage you put to it??

How does changing the "default speed" effect the overall overclock and add more headroom to either the CPU or DDR??

Just curious....

 

[Cúchulainn]

Ach, nope, you're not the only one. I left my cpu at 234x11.5 before work crunching seti. This was after testing it in the morning with prime for a few hours. Everything seemed to be going fine as far as achieving a higher fsb.

Throughout the day I would check my work unit times to see if there was any difference. Didn't notice much difference but it's sort of hard to tell with only a few units. When I came home I decided to see how it performed in 3dmark2001 since this machine is primarily used for gaming. 3dmark just crashed to the desktop as did ut2k3. So I backed off a few on the fsb and tried again. Major artifacts and general corruption. I kept going down until I was at my old threshhold of 223 and the artifacts only lessened a bit but never went away entirely. I just decided to let the tests finish at my old speed of 223x12 since the corruption wasn't that bad any more. Score was 16.5k when I normally get 19.1k!!

I also tried several bioses to no avail. So it looks like even though I could no question get a higher fsb it was useless in the games I play. Perhaps vdd was too low?

Anyhow, mods out now and back to normal

 

[q149]

I don't know....maybe I'm missing something here....

I understand the concept, in which you're basically fooling the CPU and board into thinking it's a 2700+. But how is this going to help you if the chip maxes out at say, 11 x 205, or the DDR you have just doesn't want to go past 210, no matter what voltage you put to it??

How does changing the "default speed" effect the overall overclock and add more headroom to either the CPU or DDR??

Just curious....

There is really no good explanation, it's just an issue with nf7 rev 2.0 boards.. You can only get really high FSB with a default 333 fsb cpu. The earlier revision nf7's had the opposite problem, you needed a default 266 fsb cpu to get really high FSB.

 

[Lithan]

Umm I got 230fsb on my 1700+ with bios 11. 228 with bios 17. I'm curious to try this and see if I can get 260 but I'd need faster ram first.

 

[Mr B]

There is really no good explanation, it's just an issue with nf7 rev 2.0 boards.. You can only get really high FSB with a default 333 fsb cpu. The earlier revision nf7's had the opposite problem, you needed a default 266 fsb cpu to get really high FSB.

Hmmmm...

*bookmarks thread*

It's a bit of a bugger to get at mine, the way the H20 is routed, but the next time the block comes off, I'll give it a shot...nothing ventured, nothing gained.

Thanks for the explaination....

B.

 

[q149]

Np, good luck. I have an NF7 coming that i should get tomorrow or moday.. hopefully i can go well above and beyond my 235 limit on my 8rda. I will post my results before/after L12 mod and before/after vdd/vdimm mods. There is 2 ways to do the L12 mod btw if it hasn't been mentioned. You can connect 2 pins on the cpu or 2 holes in the cpu socket, or you can connect the 2nd bridge to the left of where it says L12 on the cpu(3rd bridge).

 

[Mr B]

Np, good luck. I have an NF7 coming that i should get tomorrow or moday.. hopefully i can go well above and beyond my 235 limit on my 8rda. I will post my results before/after L12 mod and before/after vdd/vdimm mods. There is 2 ways to do the L12 mod btw if it hasn't been mentioned. You can connect 2 pins on the cpu or 2 holes in the cpu socket, or you can connect the 2nd bridge to the left of where it says L12 on the cpu(3rd bridge).

I'd go with taking a strand of wire and putting it in the socket, I think. I had to do that on my Iwill XP333-R to get it to run my 2100+ "B". It wanted to run at 20 x 100FSB, and putting a jumper in the socket was the only way to get a realistic multipier/FSB ratio.... It took me forever to figure out the foolish thing was trying to boot at 20/100FSB because it wouldn't recognize a multi higher than 12.5x.

B.

 

[hitechjb1]

Got a NF7-S rev 2.0, it could only do 215 MHz FSB stable, no boot above 220 MHz.

This is the mod I did to make it to run higher FSB, 230+.

It looks like the NF7-S rev 2.0 behave quite differently for various owners as reported regarded to CPU FSB_Sense setting 133/166/200, memory, some bios setting (such as CPU Interface), ... Reasons are yet unkown.

FSB_Sense mod to set Tbred B default FSB from 133 to 166, or to 200

NF7-S rev 2.0 FSB_Sense 166 and 200 mod (from 133) test results

NF7-S rev 2.0 FSB_Sense 166 and 200 mod (from 133) test results

Did the FSB_Sense mod for setting default FSB from 133 to 200 for a Tbred B 1700+ DLT3c.

Reset CMOS, the NF7-S rev 2.0 system booted up. The default bios showed the CPU as a 3200+ 200 MHz FSB, very interesting.

The FSB_Sense mod for 133 to 166 helped stable FSB from 215 to 232+. Chipset Vdd 1.7V. PC3500 C2 6-3-3-2.

These are the highest FSB attained from the 200 mod, and no noticeable improvement (on max FSB) from the 166 mod.
FSB is running 3DMark01/03 stable at

3DMark 01/03 w/ 8500 128 MB, 275/275

Single channel:

238 MHz x 9, PC3500 C2 512 MB x 1 SC, 6-3-3-3, 2.9 V,
Sandra BW efficiency 86-87% (e.g. 3290/2975) <---- highest
3DMark 01 10457
3DMark 03 1544

234 MHz x 9, PC3500 C2 512 MB x 1 SC, 6-3-3-2, 2.9 V,
Sandra BW efficency 87-88% (e.g. 3278/2963)
3DMark 01 10521 <---- highest
3DMark 03 1551 <---- highest

228 MHz x 9, PC3500 C2 512 MB x 1 SC, 6-3-3-2, 2.9 V,
Sandra BW efficency 87-88% (e.g. 3218/2900)
3DMark 01 10335
3DMark 03 1540

Dual channel:

228 MHz x 9, PC3500 C2 512 MB x 2 DC, 6-3-3-2, 2.9 V,
Sandra BW efficiency 88-89% (e.g. 3222/2961)
3DMark 01 10451
3DMark 03 1548

Still testing, ...

- DC loses about 5-10 MHz from SC, as expected.
- But the Sandra BW efficiency 87-89% seems to be low compared to 95% seen mostly reported ??
(For the same memory module, same timing, I can get 95% memory efficiency from A7N8X-DLX SC).
- Also the performance from DC just ties or slightly worse than the 5-10 MHz higher FSB from the SC.
- Using NF7-S rev 2.0 BIOS 1.7, Nvidia chipset driver 2.45

As shown above, have not seen a difference in highest FSB attained between the 166 mod and the 200 mod. But a big difference from default 133 FSB.

One thing must point out is that:
Currently, CPU is running at slow speed around 2 GHz 1.55V to 2.2 GHz 1.7 V (because using stock HSF).

Have to see whether the FSB_Sense mod affects CPU's full overclocking speed.

 

[hitechjb1]

I wish the NF7-S rev 2.0 Vmem and chipset Vdd have higher bios voltage setting like the 8RDA3+.

The chipset Vdd mode is not hard to do (I tried that).

But the Vmem mod is not easy due to the very small size of the components involved. Anyone has done a Vmem mod.

I found that there are quite a few inconsistent settings reported on the NF7-S rev 2.0, maybe related to CPU, ...

1. Some find the FSB_Sense (or L12) 166/200 mod works, some don't.
For those that seems to work (10-15+ FSB improvement, like my case), are those who get low FSB (under 215 MHz) to begin with. There is no difference between the 166 and the 200 mod on helping max FSB in my case.

After the 166/200 mod, I can now do 228 MHz DC, 230+ MHz SC, PC3500 C2 512MB x 2 memory at 6-3-3-2 @2.9V, Vdd at 1.7 V.

Those who already do 220+ FSB, no much help from the mod.

2. Some say the CPU Interface has to be ENABLE, some say NO. In my case, if CPU Interface is ENABLE, system cannot even boot.

Some say disabling CPU Interface would help max FSB a lot, there is no help on FSB at all in my situation.

3. My Sandra memory bandwidth efficiency is only 87-89%, low compared to many reported 95%. Not much difference before and after the FSB_Sense 166/200 mod.

It looks like the NF7-S rev 2.0 behaves very differently in many cases with CPU, maybe memory, ... I don't yet have any idea whether it is BIOS, CPU, motherboard, ... related.

 

[q149]

Well, i got my NF7 today.. It wasn't even stable at 215 and no boot at 220. I did the L12 mod and now i am stable up to around 225 dc. Everything is still stock though basically, all i have done so far is lap the NB and put the stock cooler back on. I will put mosfet sinks on, my modifed sk-6+ on the NB and do the vdd mod and see what i can do. I don't have an extra 1k v-r for Vdimm right now but anyone know where i can check the vdimm at?

 

[hitechjb1]

It looks like your NF7-S rev 2.0 behaves like mine, before FSB_sense (L12) mod, I got 215 MHz stable, 220 no boot. After mod, 228 DC stable, with Vdd at 1.7V.

Have you run Sandra memory bandwidth, I'd like to see what efficiency you can get. I got only 87-89%. I saw some reported 95%.

In the bios, do you have the CPU Interface enable or disable?

I was not able to do Vmem mod since component involved is so small. Did Vdd mod to 1.95V, but did not help much, probably due to memory.

 

[q149]

I just got done doing Vdd mod and it's at 1.94 right now.. i am too scared to mess in the bios though because it is storming really hard and the power flicks off&on real quick sometimes, i don't want to corrupt my bios.. i will later tonight though. I did run sandra once at 220 8-3-2-2 i think and i scored 3400/3200. I'm doing Vdimm at 2.9 but i'm not sure where to measure it to see what it's really at. I have cpu interface enabled and i hav on the 14 bios but once i get everything straight with my hardware i will test different bios'. Hopefully tonight i will finish all the cooling so i can stick it in my case where i have awesome airflow and mess with the hardware.. the board and memory is brand new though so i'll burn tonight and maybe get some better oc tomorrow.

 

[hitechjb1]

Yes, don't flash bios in stormy weather.

So you are getting 96-97% Sandra memory bandwidth efficiency, which is really good. Also CPU Interface Enable. What memory are you using, and what size?

I cannot boot with CPU Interface Enable.

I tried many things, my memory efficiency is low, only 87-89%. I tried bios 1.6 and 1.7 same things. Also different chipset drivers, 2.03 and 2.45, same thing.

The only thing I can think of is to install Windows XP from scratch. Now the current OS is picked up from a A7N8X-DLX installation, and just reinstalled drivers and latest chipset drivers for the NF7-S rev 2.0.

 

[[OC]This]

I cannot boot with CPU Interface Enable.

I tried many things, my memory efficiency is low, only 87-89%. I tried bios 1.6 and 1.7 same things. Also different chipset drivers, 2.03 and 2.45, same thing.

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.