X48 Rampage Formula Preview.

[Zucker2k]

So you tried the 1N command rate aye Hows your head doing?

I found I needed to advance CMD Skew about half the amount I had to advance CLK skew on mine.

We did ask for individual clock skews and GTLs, Asus guys asked why do we need them LOL ouch I say.

We can keep bugging though. We'll wait and see when the next beta bios comes out as they are supposed to be adding tREF timing. Perhaps one of my other many requested options will be added too who knows. I'm pretty sure their bios programmers have their hands full at the moment working on all the I7 bios problems they've got heh!

I had tried 1N clocking with Maximus Formula/Rampage Formula in the past with dismal success; same story with this bios.

We do need individual gtl control; also lower gtl than 62x would be very helpful. Asus' tactics is equal to neutering in order to promote sales of their newer offerings. This board is very capable as it is, but it can get better easily - all that is needed is for the decision-makers to issue the order.

[Alien Grey]

We do need individual gtl control; also lower gtl than 62x would be very helpful. Asus' tactics is equal to neutering in order to promote sales of their newer offerings. This board is very capable as it is, but it can get better easily - all that is needed is for the decision-makers to issue the order.

I totally agree with that. That's why we should try to convince them that a BIOS on our Rampage Formula needs these settings. It's a X48 board. This chipset was made by Intel for performance and needs a BIOS with all these settings available. Anything less isn't acceptable.
Although the CPU and NB Clocks Skew, that they added in the BIOS gives us better stability and higher clocks, it doesn't fix all the problems.
I still have to inform them about what we think of this BIOS but I'll wait for that till I tested this BIOS with my G.Skill memory at 1200MHz.
When I'm going to inform them I'll need an explanation with what we need in the BIOS and Why we need it. The "Why" is very important. That's the only way to convince them.
And don't worry the person at ASUS that i'm going to send it to is going to understand everything that we're talking about. That's why it's so hard to convince them.

[neo_rtr]

With respect i have to admit my ignorance for the GTL signaling, But i have to ask in order to understand. I been reading alot, still not making sence, though i dont ask to have everything server on a silver plater.
So I have a QX6700. I assume this Cpu that costed 1000$ back then is able to do at leat 4Ghz, Let me try to get to 3.8Ghz 1st.
The QX6700 is a 65nm processor
According to an other thread Maximus Formula II Club i got the following (i have seen then before, just forgot the page just now)

It says
Vtt x 0.635+ Item = Vtt x 0.667

What is Vtt: It’s front-side bus (FSB) termination voltage (thats is the bios setting: FSB Termination Voltage In my case 1,63V )
Question what the "ITEM" ? IS it a give nr?

AnandTecg had back in the day the following reference for GTL
BIOS Set VTT Measured VTT BIOS set GTL Measured GTL 67% of VTT
1.50....... 1.53 .............0.63x............1.022............... ...1.025
1.52........1.55............. 0.63x............1.036..................1.038
1.54........1.57..............0.63x............1.0 51..................1.051
1.56........1.60..............0.63x............1.0 65..................1.072
1.58........1.62..............0.63x............1.0 80..................1.085
1.60........1.64..............0.63x............1.0 90..................1.098
How do i use the above data in this formula?
Or does this formula here doing the same thing?

Example:
65nm CPU
FSB Termination: 1,63V


Formula:
Vtt x 0.635 + Item = Vtt x 0.667
1,63V x 0.635 + Item = 1,63V x 0.667

0.85725v + Item = 0.90045v
Item = 0.90045 - 0.85725
Item = 0.043v
Item= 40mv

Now round up to the nearest selectable value in the bios is 40mv and set the 0.635x Item value
to 40mv (positive), usually the (0 &2) core.
You have now made the 0.635x GTL voltage as close as possible to the 0.667x GTL voltage.


Bios setting Example:
CPU GTL REFERENCE (0) 40mv
CPU GTL REFERENCE (1) Auto
CPU GTL REFERENCE (2) 40mv
CPU GTL REFERENCE (3) Auto


Also there are only two values
CPU GTL Voltage Reference [] .067x
NB GTL Voltage Reference [] .63x

And one more question , is the above calculation to get the CPU Voltage i need or what the CPU GTL V will be?
(hope making sence, writing from 4" screen..hehe

[Alien Grey]

Forget about all those calculations. With the Rampage Formula BIOS you can only use static values for GTL's. As we don't know what the real voltage is it's all a bit trial and error.
What I did was following the example from Anandtech. They are saying to use CPU GTL 0.63X and NB GTL 0.67X for 45Nm CPU's. That's what I did and it looks like it's right and that ASUS follows this to. That doesn't mean that you won't need a different GTL but it's a good starting point.
What you should do is follow the example that Anandtech gives for 65Nm CPU's and start with CPU GTL 0.65X and NB GTL 0.67X.
Also try lower FSB Termination Voltage. You don't need that high FSB Termination Voltage for FSB <475MHz. On the other hand perhaps you do need it since it's a Maximus board flashed with a Rampage BIOS.

[neo_rtr]

I have some new results

Extreme Tweaker 3,680 mhz
Ai Overclock Tuner : Manual manual
CPU Ratio Control : Manual 8
- Ratio CMOS Setting :
FSB Frequency : 450
FSB Strap to North Bridge : 333


By neo_rtr, shot with DiMAGE X1 at 2009-02-11


By neo_rtr at 2009-02-11

[mikeyakame]

Good stuff mate

I've seen quite a few MFs needing a lot of Vtt so I'm not surprised. Thats good news anyhow, all you can do is find what works for your board right.

Other thing is Asus reads Analog Vtt, which isn't all that much to go by as you increase Vtt voltage. Vtt voltage at the pins is a decoupled voltage, Analog Vtt is decoupled with Digital Vtt to give a final Actual Vtt voltage. I've heard of upto 0.1v difference at the pins with high Vtt voltages. So if you find you need that much Vtt, it's possible the actual Vtt voltage at the CPU and NB land pins may be much lower than what you are setting.

I got my DFI X48 board up and running last night, and there is two things I miss already with the RF.
1) LCD Poster, hell makes life so much easier than looking at a diagnostic display on a board mounted inside a closed case!
2) Phoenix bios is nowhere near as overclock recovery friendly as AMI bios on Asus boards. Rebooting during a post or a locked POST on the DFI board doesn't trigger default settings on next POST. The "home" key is supposed to do this, but since I use a bluetooth keyboard primarily this makes it impossible to hit "home" to boot with defaults. It hardly even gives me enough time between USB Finalize when USB devices are active and INIT Option ROM stage to hit the "del" key to get into setup! I had to turn quick boot off so it runs a long POST memory test to get the time to enter setup!

Other than that it's a nice board so far, definitely no where near as friendly as Asus boards for overclocking with a closed case that I can say for certain!

Edit:

Hopefully this time I write this post prime95 wont lock my system uP!

It's a fun board so far, bios is extremely flexible and learning curve is steep, I'm liking it. Though it took me 2-3hours just to get my ram stable at 1066mhz ! heh. Figured out the set Vdimm is 0.08v higher than the read Vdimm.

Giving up for tonight cause I need to catch up on work sleep! Still haven't got 333Mhz FSB stable yet! By stable I mean to my own acceptable performance and not locking up while doint it

Kinda makes me realize how much we take for granted with Asus boards. The bios does half the work for you, just you dial in numbers and presto very little work for stability! I can dial in 400mhz FSB on my RF without even hardly touching a thing. This DFI board is lucky to get past POST if I set 333mhz FSB and don't play with voltages, timings, and all that stuff.

Speedstep works well though so far. The Special VID Add option is nice, lets you run VID voltage + percentage on top. I believe it's so speedstep doesn't break like on every other board when you change Vcore.

Fan Control works perfect too. You set minimum temperature for fan disabled, which is < 25c, then you set max temperature where fan is 100% duty, which I have 35c. So between 25 and 35c the fan duty is increased linearly on a gradient line. Hell my 2700rpm 51dBa 135-140cfm 120x38mm Delta runs at 8-900rpm on POST and whenever the CPU is idle It's almost quiet now heh

Hell I'm almost tempted to throw my 65dBa 235cfm 120x38mm Delta onto my TRUE if fan control works this well.

I tried running it once on the RF and I removed it after 30seconds because the bios wouldn't run it any less 100% duty! It sounds like a wind tunnel at 100% duty. I once had it sitting on my desk for open case benchmarking and forgot to secure it to something. Turned the system on and it launched itself a good 0.75m off the desk! Ripped the connector out of the power cable toO!

Oh one last thing. The DFI bios sets tRFC correctly! First board I have seen to actually do this.

Memory Bus 532.3 MHz
Refresh Period (tREF) 4171T

tREF is 3120T for DDR2-800.

[Alien Grey]

Good stuff mate

I've seen quite a few MFs needing a lot of Vtt so I'm not surprised. Thats good news anyhow, all you can do is find what works for your board right.

Other thing is Asus reads Analog Vtt, which isn't all that much to go by as you increase Vtt voltage. Vtt voltage at the pins is a decoupled voltage, Analog Vtt is decoupled with Digital Vtt to give a final Actual Vtt voltage. I've heard of upto 0.1v difference at the pins with high Vtt voltages. So if you find you need that much Vtt, it's possible the actual Vtt voltage at the CPU and NB land pins may be much lower than what you are setting.

I got my DFI X48 board up and running last night, and there is two things I miss already with the RF.
1) LCD Poster, hell makes life so much easier than looking at a diagnostic display on a board mounted inside a closed case!
2) Phoenix bios is nowhere near as overclock recovery friendly as AMI bios on Asus boards. Rebooting during a post or a locked POST on the DFI board doesn't trigger default settings on next POST. The "home" key is supposed to do this, but since I use a bluetooth keyboard primarily this makes it impossible to hit "home" to boot with defaults. It hardly even gives me enough time between USB Finalize when USB devices are active and INIT Option ROM stage to hit the "del" key to get into setup! I had to turn quick boot off so it runs a long POST memory test to get the time to enter setup!

Other than that it's a nice board so far, definitely no where near as friendly as Asus boards for overclocking with a closed case that I can say for certain!

Edit:

Hopefully this time I write this post prime95 wont lock my system uP!

It's a fun board so far, bios is extremely flexible and learning curve is steep, I'm liking it. Though it took me 2-3hours just to get my ram stable at 1066mhz ! heh. Figured out the set Vdimm is 0.08v higher than the read Vdimm.

Giving up for tonight cause I need to catch up on work sleep! Still haven't got 333Mhz FSB stable yet! By stable I mean to my own acceptable performance and not locking up while doint it

Kinda makes me realize how much we take for granted with Asus boards. The bios does half the work for you, just you dial in numbers and presto very little work for stability! I can dial in 400mhz FSB on my RF without even hardly touching a thing. This DFI board is lucky to get past POST if I set 333mhz FSB and don't play with voltages, timings, and all that stuff.

Speedstep works well though so far. The Special VID Add option is nice, lets you run VID voltage + percentage on top. I believe it's so speedstep doesn't break like on every other board when you change Vcore.

Fan Control works perfect too. You set minimum temperature for fan disabled, which is < 25c, then you set max temperature where fan is 100% duty, which I have 35c. So between 25 and 35c the fan duty is increased linearly on a gradient line. Hell my 2700rpm 51dBa 135-140cfm 120x38mm Delta runs at 8-900rpm on POST and whenever the CPU is idle It's almost quiet now heh

Hell I'm almost tempted to throw my 65dBa 235cfm 120x38mm Delta onto my TRUE if fan control works this well.

I tried running it once on the RF and I removed it after 30seconds because the bios wouldn't run it any less 100% duty! It sounds like a wind tunnel at 100% duty. I once had it sitting on my desk for open case benchmarking and forgot to secure it to something. Turned the system on and it launched itself a good 0.75m off the desk! Ripped the connector out of the power cable toO!

Oh one last thing. The DFI bios sets tRFC correctly! First board I have seen to actually do this.

Memory Bus 532.3 MHz
Refresh Period (tREF) 4171T

tREF is 3120T for DDR2-800.

That's the kind of BIOS I would like to have on my ASUS Rampage Formula. Overclocking might be easier on the Rampage Formula but the lack of useful settings in the BIOS makes it harder to push it to the limit.

[mikeyakame]

That's the kind of BIOS I would like to have on my ASUS Rampage Formula. Overclocking might be easier on the Rampage Formula but the lack of useful settings in the BIOS makes it harder to push it to the limit.

Oh by the way that pdf is only the icing on the cake There are much more options it doesn't cover too heh

It also displays all cross clocking skews in PS values as autodetected, and are fully adjustable. You can also set the fine delay PS value, it is 5ps by default

It has what is similar to AI Clock Twister, CH1/2 Cross Clocking Setting which you can set the equivalent of Lighter -> Stronger, but not just that it displays all current cross clocking timing values, and each is fully adjustable from the base Cross Clocking strength setting, ie you set More Aggressive (Asus Stronger) to get the base values and you can fine tune it from there

Another crazy option is the Host Slew Rate, which is the Host FSB clock slew rate, rate at which the V/ns rises and falls, used to match clock edges. You can set the Resistance parameters for Slew Rate, MCH ROn, MCH Rtt and a couple of others.

There is also DRAM Drive Strength, and a host of other DRAM/MCH timings, drives and other settings. It's overwhelming even for me I feel like a child in a fun park

If there is any board that can overclock DDR2 high with 1N command rate it is this board. You can manually set command rate for READ, WRITE and COMMAND to CS (chip select) for BOTH CHANNELS! 6 command rate options in bios! Awesome

Edit:

http://img.lanparty.tw/Upload/Catalo...48T3R_BIOS.pdf

This pdf lists all the options in latest bios. Only thing I have missing is CPU Clock0 / 1 Skew ( CPU / NB ). I think I'll hassle DFI about adding that to next beta bios

I've also got a ton of cool things now to put forward to Asus to add haha Just look at what the DFI X48 bios contains and you'll get a good idea of what we can ask for. Not everything is necessary, actually most will just confuse 90% of RF users, but there are definitely some good options I'd like myself.

I'm seriously intrigued about the possibilities for this option under Host Slew Rate compensation.

MCH Rtt
RTT is the on-die termination resistance measured at VTT/3 of the GTL+ output driver.

Source: http://www.intel.com/design/PentiumI...s/24508601.pdf

3.1.2. AGTL+ SENSITIVITIES

Pull-up resistance

As noted above, the pull-up resistors provide a path for current flow when the signal on the network is
less than the termination voltage. As a result, the value of RTT affects the rise and fall times of the
signals. The impact of this effect is that the system flight time is dependent on the pull-up resistance,
as illustrated by Figure 2. Higher values for RTT result in slower rising edges and faster falling edges.
Lower values for RTT cause faster rising edges and slower falling edges.

The processor substrate contains a 56W resistor at the end of the network. Intel recommends that an
additional 56W resistor be placed on the system board at the other end of the network for a single
processor design. This value represents the optimum balance between rising and falling edges.
In addition to flight times, the pull-up value affects the ringback characteristics of a AGTL+ network
(see section 3.2 for a description of ringback). Lower values for RTT generally result in decreased
ringback on the rising edge. Ringback is not typically a problem on the falling edge.

4.4.2. PRACTICAL CONSIDERATIONS

...

· Higher RTT values tend to increase the amount of ringback on the rising edge, while smaller
values tend to increase the amount of ringback on the falling edge. It is not necessary to budget
for RTT variation if your simulations comprehend the expected manufacturing variation.

...

Still trying to figure out what MCH Ron does exactly, I have a basic idea though.

R(on) is Resistance On Impedance of the Output Driver

R(on) = Vcc(out) / I(oh) ,Vcc(out) / I(ol) at Vcc/2

Nominal R(on) value from what I understand is used to transition between the pull-up and the pull-down drivers. I'm still not completely clear on what it affects, but it's something along the lines of setting up:

Clock (High) - delay from rising clock edge peak (end) to falling clock edge peak (start)
Clock# (low) - delay from falling clock edge trough (end) to rising clock edge trough (start)

What I think this means is if you picture a differential Clock Waveform like this kind of:

\_ /-\ _ /-\_/ (high clock)
/ \_/ \_/ \ (low clock#)

_ part of \_/, or - part of /-\ is determined by MCH Ron resistance value.

I need to ask a friend who designs electronic circuits for a living if I have it correct just in case! heh.

[Michowski]

Lets see, my cpu's highest core usually sat at 36-38c while on desktop. Now all my cores went up about 6-10c just spur of the moment. Highest core is now 47c idle. My room is the same temperature, fans all same speed as before. Nothing else is running compared to when temps were under 40. Why or what could cause the temps to just be 10c higher? Before I used to sit at 450fsb and I went lower to 430 and lower volts for everything. All my temps went down from nb to sytem temps, except for my cpu. That went up 10c. I checked my fluid and it hasnt moved at all. The block is still tight and well secured to the chip. Im lost.

[Zucker2k]

Lets see, my cpu's highest core usually sat at 36-38c while on desktop. Now all my cores went up about 6-10c just spur of the moment. Highest core is now 47c idle. My room is the same temperature, fans all same speed as before. Nothing else is running compared to when temps were under 40. Why or what could cause the temps to just be 10c higher? Before I used to sit at 450fsb and I went lower to 430 and lower volts for everything. All my temps went down from nb to sytem temps, except for my cpu. That went up 10c. I checked my fluid and it hasnt moved at all. The block is still tight and well secured to the chip. Im lost.

Added a new video card?

[Michowski]

Nothing. Not one thing has changed for my pc. Same programs and games I play everyday. Just one day it was warmer. Out of curiousity, whats the offset for tj-max supposed to be for the 9550?

[Alien Grey]

Lets see, my cpu's highest core usually sat at 36-38c while on desktop. Now all my cores went up about 6-10c just spur of the moment. Highest core is now 47c idle. My room is the same temperature, fans all same speed as before. Nothing else is running compared to when temps were under 40. Why or what could cause the temps to just be 10c higher? Before I used to sit at 450fsb and I went lower to 430 and lower volts for everything. All my temps went down from nb to sytem temps, except for my cpu. That went up 10c. I checked my fluid and it hasnt moved at all. The block is still tight and well secured to the chip. Im lost.

Dust in the radiator can cause higher temps.

[neo_rtr]

Perfect example that going up is not always the direction one should choose.

But for some time now been trying to get it to work.
My goal was 3.8ghz. At some point i did a cross flash as i call it to the RAMPAGE bios. 8202

It proven to be one of the best things ever. No i have not reach my goal but now days what ever i put on it it will boot.

Any way. been surfing alot to get some more info. This QX6700 is not so commons and by now almost no one using it.
In search of the sunrise has to redo bios settings and from what you can see from previous posts they where high.
So far i have managed to take them down a bit.

One thing i can tell you for sure is that 8x450 fsb needs that 1.50V FSB Termination Voltage. (have tryied almost anything and i found countless ways NOT to do it that way).
So 1.50V FSB and 1.53V NorthBridge works for me.


By neo_rtr, shot with DiMAGE X1 at 2009-02-13


By neo_rtr, shot with DiMAGE X1 at 2009-02-13


By neo_rtr at 2009-02-13

Also notice that the CPU Voltage is now down to 1.46875V / 1.54V CPU PLL

Iam quite sure that there is room from improvment.
One thing i did notice is that taking the CPU voltage 1 click down, I had to take the PLL 1 click down as well. If not, Prime failed on around the 7 or 8th loop.


Also when prime fails going up one click in voltage Do Not always work.
Will not argue that , just an observation. Be flexible! and take your time. It does take time this OC business.

[Alien Grey]

Also when prime fails going up one click in voltage Do Not always work.
Will not argue that , just an observation. Be flexible! and take your time. It does take time this OC business.

Overclocking a Quad on high FSB is a bit of an art. It takes time and practice to see good results.

I did some more tests today with BIOS 0802. With CPU Clock Skew -200 and NB Clock Skew -100 my system runs very smooth on FSB 465MHz. I could lower the NB Voltage and Memory Voltage a bit.

I tried to get it Prime95 stable at FSB 475MHz. It's a no go. My OCZ Reapers HPC PC2-8500 doesn't like to be pushed over 1120MHz with 5-5-5-15... timings.

[greg.m]

Well here i am again. It took a while but....
I rmaed my qx9650 c0 with which i couldnt run 4ghz prime stable with lower than 1.55 real volts(bios 1.6) and yesterday i recieved my new direct from intel... With the old one i could run 3.6 prmstbl with1.4 bios settings and
NOW here is my bios settings


9X400
1066
pl6
vcore 1.3125
nb 1.4
fsb 1.36
cpupll 1.5
vram 2.1
cpu 0.63x
nb 0.67x

And of course for one more time i wanna thank A-Grey so much for his help

here is the results

[Alien Grey]

Well here i am again. It took a while but....
I rmaed my qx9650 c0 with which i couldnt run 4ghz prime stable with lower than 1.55 real volts(bios 1.6) and yesterday i recieved my new direct from intel... With the old one i could run 3.6 prmstbl with1.4 bios settings and
NOW here is my bios settings


9X400
1066
pl6
vcore 1.3125
nb 1.4
fsb 1.36
cpupll 1.5
vram 2.1
cpu 0.63x
nb 0.67x

And of course for one more time i wanna thank A-Grey so much for his help

here is the results

It looks like an RMA for you CPU wasn't a bad idea. A stable 3.6GHz with 1.24V on full load isn't bad at all. I need that to to run 3.6GHz linpack stable.

[neo_rtr]

Originally Posted by greg.m
Well here i am again. It took a while but....


Greg Did your M/B died and took one last night from the shop, ? Is this result with the new one?

[greg.m]

Originally Posted by greg.m
Well here i am again. It took a while but....


Greg Did your M/B died and took one last night from the shop, ? Is this result with the new one?

Thats true!!
Thursday I received my qx9650 and after i installed everything i couldnt boot I was getting only DET RAM. After i tried everything and i mean everyf@.k..gthing i gave up. Yesterday i bought a new one while im waiting to rma the faulty one. When ill receive the new one ill sell it direct.

[spoof]

I'm facing greate trouble getting LinX working. 475 i have no problem at all, no lockups, no BSOD, Blend Stable 700% Memtest, but i can't pass even 3 loops on LinX. After a lot of tweaking, i think, that i may be RAM issue. So, i put more vNB, PL to 8 and still problem, so i bump vDIMM from 1.88 to 1.90 = 5 loop passed. Than i bump it a little more = 9 loops passet (this is from 25 loop test). Bump bit more = fail after 2 loop. I steped back, again fail. So, bumping vDIMM doesn't help, but i still think, that is just RAM issue, not NB. So, what will be yours suggestions, any help would be good (;

[Ghostleader]

I'm facing greate trouble getting LinX working. 475 i have no problem at all, no lockups, no BSOD, Blend Stable 700% Memtest, but i can't pass even 3 loops on LinX. After a lot of tweaking, i think, that i may be RAM issue. So, i put more vNB, PL to 8 and still problem, so i bump vDIMM from 1.88 to 1.90 = 5 loop passed. Than i bump it a little more = 9 loops passet (this is from 25 loop test). Bump bit more = fail after 2 loop. I steped back, again fail. So, bumping vDIMM doesn't help, but i still think, that is just RAM issue, not NB. So, what will be yours suggestions, any help would be good (;

My guess is that you need more vcore, if you run LinX with max program size and max memory use you probably need 0,10v more vcore then P95 small FFT to be stable.

What about the rest of your voltage and transaction setup?

Post your bios setup, it´s easier to help then.

[Alien Grey]

I'm facing greate trouble getting LinX working. 475 i have no problem at all, no lockups, no BSOD, Blend Stable 700% Memtest, but i can't pass even 3 loops on LinX. After a lot of tweaking, i think, that i may be RAM issue. So, i put more vNB, PL to 8 and still problem, so i bump vDIMM from 1.88 to 1.90 = 5 loop passed. Than i bump it a little more = 9 loops passet (this is from 25 loop test). Bump bit more = fail after 2 loop. I steped back, again fail. So, bumping vDIMM doesn't help, but i still think, that is just RAM issue, not NB. So, what will be yours suggestions, any help would be good (;

Like Ghosleader says. You need a little more vcore for linpack stability. Look how low the vcore goes in Prime on small FFT's. Look how low it goes in Linx and raise the vcore a bit to compensate the higher vdrop.

[spoof]

Later, when i come back hope i'll test, but as far i as measured vcore with DMM,while runing small FFT's i have no vdrop.
Here are my BIOS settings:

[Alien Grey]

I rmaed my qx9650 c0 with which i couldnt run 4ghz prime stable with lower than 1.55 real volts(bios 1.6)

The question is are our first CPU's bad or is it something else?

I think that the CPU's aren't bad because they can run stable at higher FSB/Speed. I blame the BIOS for the problem with to high vcore and vtt.
The problem might be the GTL handling in the BIOS.
I think that two cores always fails because of the poor CPU GTL setup in the BIOS. That make sense since we could only get it stable with to high vcore and vtt. Like that we forced the cpu to be somewhat stable and pass Prime95 without errors.
This might be the best prove that we need independant CPU GTL control.

[Alien Grey]

Later, when i come back hope i'll test, but as far i as measured vcore with DMM,while runing small FFT's i have no vdrop.

Loadline Calibration is Disabled. Like that you always have vdrop on load. The vdrop is higher with linpack and you have to compensate that.

Edit:

Also try with DRAM Controller Voltage REF and DRAM Channel A/B Voltage REF on auto. I have more luck leaving that on auto.

[Ghostleader]

@spoof

You may also consider to upgrade to 0802 so that you can try CPU and NB clock skew.

Zucker2k bios setup might help some, look at post 1656