X48 Rampage Formula Preview.

[SoLoR]

Tell them that while they are at it they could upgrade to latest "offical" (if they cant just put latest ich10r since its working fine) IMSM for ich9r and other modules, but ofc they wont do that... this is only wishful thinking... for example asus is most SLACK company regarding that, while other x58 mobos are all using 8.5.1030 IMSM option rom, asus is still on 8.0.0.something, even on rampage II extreme. There is even more, lets not count intel that regulary updates all those modules on their top end mobos, but i saw that even DFI upgraded their IMSM option rom via bios update on some P45 mobo. I mean like its that HARD!

[Zucker2k]

We want skews on everything; we want very low voltage options on fsbt, cpupll, low gtls; we need vnb pll too, individual dram skews, etc.

[Alien Grey]

Tell them that while they are at it they could upgrade to latest "offical" (if they cant just put latest ich10r since its working fine) IMSM for ich9r and other modules, but ofc they wont do that... this is only wishful thinking... for example asus is most SLACK company regarding that, while other x58 mobos are all using 8.5.1030 IMSM option rom, asus is still on 8.0.0.something, even on rampage II extreme. There is even more, lets not count intel that regulary updates all those modules on their top end mobos, but i saw that even DFI upgraded their IMSM option rom via bios update on some P45 mobo. I mean like its that HARD!

PM Mikeyakame for that.

[Alien Grey]

We want skews on everything; we want very low voltage options on fsbt, cpupll, low gtls; we need vnb pll too, individual dram skews, etc.

PM mikeyakame for that. Most of it will be in his note but i don't know if everything will be in it.

[Shiranui Gen-An]

If Asus does decide to implement more extensive DRAM skewing options, it'd be nice if it does it DFI style, either showing the current skew level as a "level" (like 1-7 or however high they go), or in the actual ps value.

[mikeyakame]

I actually prefer the real register values, of 0x0T-0xFT. 1-16T. If they convert to roughly 50ps, then that is probabbly the MCH's accuracy of skewing clocks. DRAM Command Skews are 25ps at 1T command rate, but all that means is they work off a known turnaround for issuing queued commands that is half the normal. So if we send commands twice as fast, then we can probably make the assumption that they apply 50ps over 2 clock periods but issue 2 commands over those same 2 clocks. So they cheat to get 25ps!

FSB operates at 4 x clocks on the bus, DRAM operates at 2 x clocks on the bus, if DRAM is 50ps at 2x, then FSB will be 100ps at 4x and then we can use the 0T-FT and cover upto 1500ps since the skews themselves are implemented irrespective of the data clocks per cycle.

[Alien Grey]

This is a copy of my post at ASUS Rampage Formula forum.

This is my answer to the dutch ASUS Tech Support.

This is what we like to see added or fixed in BIOS ASUS Rampage Formula 0701. This will give us the BIOS that we deserve to have on our X48 performance board ASUS Rampage Formula.

CPU GTLREF0 - 3, independant control of each input buffer reference divider. 0.60-0.75x 0/2, 0.58x-0x.73x 1/3. Increment at 0.01%

NB GTL from 0.57x - 0.73x. Increment at 0.01%

CPU CLOCK SKEW / NB CLOCK SKEW - independant control of each skew with usable range, ie 0ps - 1200ps.

DRAM CLOCK SKEW A1/A2/B1/B2 - need individual skew adjustment as AI clock twister unpredictable

CPU / PCIE differential amplitude (drive) 400mv-1200mv

SATA PCIE clock frequency adjustment individual of DMI PCIE/NB PCIE frequency. Interference and jitter at high FSB where close to PCI stepping frequency

AI Clock Twister - Requires manual kickstart to work properly at high FSB. Need to set it to another value than current if changing DRAM skew values otherwise it seems to assume the current delay skew values are base values instead of realizing they are changed and just need minor adjustment ie -1T

tREF - Should be added.

REF[T] = REFI[ns] / CK[ns]
; REFI = Self Refresh Interval of IC in microseconds
REFI = 7.8us or 15.6us and is found in SPD table for dimm
; CK = clock period in nanoseconds
CK[ns] = 1000[ns] / DDR2_IO_FREQ[Mhz]

For DDR2-800 which is Intel default value of tREF you get:

REF[T] = ( 7.8 * 1000 ) / ( 1000 / 400 )
= 3120T

So Automatic Self Refresh Turnaround will be 3120 clocks at 800mhz before the next self refresh cycle can begin.

Lets say I change DRAM Frequency to 1066mhz or 1200Mhz.

For DDR2_FREQ = 1066Mhz , IO_FREQ = 533Mhz, CK[ns] = 1000 / 533 = ~1.86ns, REFI = 7.8us/7800ns

REF[T] = 7800 / (1000/533)
= 4157T

For DDR2_FREQ = 1200Mhz, IO_FREQ = 600Mhz
CK[ns] = 1000 / 600 = ~1.667ns, REFI = 7.8us/7800ns

REF[T] = 7800 / (1000/600)
= 4680T

With respect to those calculations above, if that minimum clock turnaround value tREF I've shown is already greater than the Intel default of 3120T then we may have a problem during Self Refresh.

If the value of that register is only 3120T regardless, and if me the "end user" decides to use DDR2-1200 I now am in a position that's not worth smiling about.

If I actually need 4680 clock periods to meet the minimum turnaround at 1200Mhz and I only have 3120 clock periods, then....

Well the answer is simple, we start automatic refresh/precharge 1560 clock periods earlier than expected and as such best case scenario we may just increase traffic on the DRAM bus and externally on the FSB when page faults begin to occur because data expected isn't the data being read.

Worst case scenario may be physical wear and tear from the repaging of memory unnecessarily, to a more localized problem of refreshing incomplete data during burst read/writes leading to continuous refreshing of bad memory data and corrupting devices, files, etc.


CPU Fan Control for 3 pin CPU Fan.

Some say that this is already fixed. I have a 3 pin CPU Fan connected and i don't see any speed increase no matter what profile i use.
If this is already fixed the problem could be the profiles. Some say that the CPU Fan only starts spinning faster when the CPU reaches 60° Celsius.

Silent Mode - The CPU Fan is always at low speed (No problem the CPU Fan is always at low speed)
Optimal Mode - The CPU Fan starts spinning faster when the CPU temperature raises. (I don't know if this works because my CPU is watercooled)
Performance Mode - The CPU Fan starts spinning faster according to the load on the CPU. (If this is how it works i can say that this isn't working. If it's based on temperature the CPU Fan should be at the maximum speed around 40° Celsius. An increase in CPU Fan speed above 60° Celsius is to late.)

Chassis Fan Ratios - Lower Fan speeds should be possible (50% is good if it's really 50% - A Fan with a maximum of 2500 rpm at 50% should spin around 1250-1300 rpm)

Optional Fan Ratios - Lower Fan speeds should be possible (50% is good if it's really 50% - A Fan with a maximum of 2500 rpm at 50% should spin around 1250-1300 rpm)

I think everyone can live with that.

Thanks to you guys helping me out and especially mikeyakame for all the hard work that he has done.

[Roger_D25]

Thanks very much guys (especially A-Gray and mikeyakame) for all your support in this thread! Thanks also for doing your best to get continued BIOS support/upgrades for this motherboard, the changes you have suggested would be awesome and it seems that there is actually a chance they will get implemented! Keep pushing it guys!

[Alien Grey]

Ok guys

I received some good news today.

Soon we will see CPU and NB clock skew in a new beta release that they are going to give me.

tREF is still under testing and won't be available in this beta release.

They didn't find anything very wrong with the CPU Fan Control. Not much will be changed on that part.
I'm still going to ask if they can do something about it for the Performance profile and give them my reason why i ask it.

My contact didn't receive anything about the other features we asked. This may take a longer time before they decide what can be done.

And don't worry i will share this beta BIOS with you guys if i receive it.

[spoof]

This is just greate. Maybe if someone else send message about on-die GTL refs will see them too.

In the GLT's tread i have some posts about what i'm testing now. Since, this is the right tread, i'll cry my tears here - i started from bottom - 333x8.5, all voltages at minimum. All the way to 450 (i tested with 10mhz step) i'll i need was vCPU and two step on the VTT, vNB and one PLL. Stable at 1.35 vCPU, 1.26 VTT, 1.31 vNB, 1.52 PLL, GPU GTL 65x, NB GTL 67x. Just 5mhz up to 455 and i'm in the "lockups world". One step to vNB and VTT doesn't help, now i rised VTT to 1.30. Probably, again i won't be able to reach 471 stable ... ;/

[Alien Grey]

I can give you mine for FSB 465MHz and 8 multiplier. You can find them here on the ASUS Rampage Formula forum.

[spoof]

Thanks, but the didn't help ... i mean, my setting are with more voltages. VTT is what help stoping the lockups, and 1.36 VTT i just for 455, my goal is 470. I'm now 100%, that it's CPU problem. 6x472 - no problem, 8.5x473 - hard lockups.
A question - after flashing P5E to RF, and 04** RF to 0701 RF i haven't clear CMOS(never cleared it). Could this help or clearing is just after flashing to reset BIOS?
I'm just hopeless now, for 450 i need really low voltages, for 455 i'm rising them and it just keep on locking up ...

[mikeyakame]

Thanks, but the didn't help ... i mean, my setting are with more voltages. VTT is what help stoping the lockups, and 1.36 VTT i just for 455, my goal is 470. I'm now 100%, that it's CPU problem. 6x472 - no problem, 8.5x473 - hard lockups.
A question - after flashing P5E to RF, and 04** RF to 0701 RF i haven't clear CMOS(never cleared it). Could this help or clearing is just after flashing to reset BIOS?
I'm just hopeless now, for 450 i need really low voltages, for 455 i'm rising them and it just keep on locking up ...

Ok try this, forget all the settings you are using right now and start fresh.

Start at 466mhz FSB with 7x multi and PL at 9 on 12:10 divider. Leave CPU GTL on auto, set NB gtl to 0.63x. Start with 1.24v VTT, 1.41-1.45v Vnb, and whatever Vcore you need to be stable and guaranteed error free. 1.2-1.25v I assume would be enough? DRAM Clock Skews to Normal, and advanced/delay if you need to. DRAM Static Read to Enabled, Vsb to 1.075v, Vsb_pll to 1.50v, Vcpu_pll to 1.54v, Vdimm to whatever you need for ~ 1120mhz, DRAM Controller REF to DDR2_REF or -10Mv, others are ok on auto, PCIE frequency to 102-104mhz. AI Clock Twister if you have it set it to moderate.

And start from here. If you don't have all those options don't worry too much, but if I missed some let me know what they are. Also put a 40mm fan on the NB heatsink if you haven't got one already.

And with Vtt, for me sometimes changing a few Mhz FSB above whatever I'm using requires me to add 0.04v - 0.06v Vtt, then I may go up 2-3mhz more and have to drop Vtt back down 0.08-0.1v to get stable and adjust GTLs all over again. Everytime you change Vcore, Vnb or Vtt you have to adjust CPU_GTL, NB_GTL or both respectively.

[Alien Grey]

A question - after flashing P5E to RF, and 04** RF to 0701 RF i haven't clear CMOS(never cleared it). Could this help or clearing is just after flashing to reset BIOS?
I'm just hopeless now, for 450 i need really low voltages, for 455 i'm rising them and it just keep on locking up ...

You can try that. At first everything looked stable for me with BIOS 0701 and FSB465MHz. After a reboot everything was a mess again, so I tried to clear the CMOS. It really helped. Everything was stable again after a Clear CMOS.

[mikeyakame]

I'm with Alien on that. Clear CMOS seems to correct bugs which set in sometimes from combinations of settings or bug in bios assembly code resulting from the former. What ever the reason, clearing the cmos fixes most bios code bug problems.

[DEFLORATOR]

A-Grey
These are tons of issues and features to be corrected/added in future bios release for RF. But i will not be able to enjoy it at all if ASUS doesn't correct that problem when my Adaptec 3405 Raid controller cannot work with bioses newer than 0403. For God's sake pls. tell the ASUS to fix this stupid problem!

[spoof]

I really, really highly appreciate you help, both of you mikeyakame and A-Grey, i'm really thankful. I'll try first with clearing CMOS and than with mikeyakames settings and later i'll post what are the results.

One more time thank you so much, both! ((:

[Alien Grey]

A-Grey
These are tons of issues and features to be corrected/added in future bios release for RF. But i will not be able to enjoy it at all if ASUS doesn't correct that problem when my Adaptec 3405 Raid controller cannot work with bioses newer than 0403. For God's sake pls. tell the ASUS to fix this stupid problem!

Strange that you say that. I thought they fixed the RAID Controller issue with BIOS 0601.
Did you try a Clear CMOS after you've flashed your board with a new BIOS?

[DEFLORATOR]

Strange that you say that. I thought they fixed the RAID Controller issue with BIOS 0601.
Did you try a Clear CMOS after you've flashed your board with a new BIOS?

Yes, of course I cleared bios. I tried all bioses, any newer than 0403 doesn't work when Adaptec 3405 is installed. The difference was in the behavior of PC: with 0406 and 0407 PC will not boot at all, only removing of raid card helped; starting from bios 0408 PC starts, but Adaptec isn't seen for system. Raid controller has the newest firmware (I tried other versions too, but that didn't helped).

[Alien Grey]

Did you contact the ASUS Tech Support for your problem?
You can send them a Technical Inquiry about your problem with their new BIOSes and your Raid controller.
That's the first thing you should do. Now that they are working on the BIOS they could try to solve the problem for you.

I'll do my part and I'll tell them about the problem when they contact me.

[spoof]

I think, i maybe found a sweet spot at 460FSB for the CPU - this means no lockups(for 1 hour 30 min). BUT( there is always but) i got after 1 hour 30 minutes "Final result was *******, expected *******" on core #3 in Prime95 Small-FFT. Ran Memtest, as i expected - errors (not mass errors, but ofter 3% 1,2 errors). Last time i had this problem(there was mass error, no just from time to time) 67x NB GTL fix it, but now it's the same on 67x and 63x. DRAM Controller to DDR2_REF -10mV doesn't help, high vNB too. My 8800PC are to 1105 with 1.90* real, so i don't think they are the problem. PL is 7, which i think is not the problem too. Any ideas?

I forgot - the problem causes BSOD too with 0x0000101 or 0x0000124.

[DEFLORATOR]

Did you contact the ASUS Tech Support for your problem?
You can send them a Technical Inquiry about your problem with their new BIOSes and your Raid controller.
That's the first thing you should do. Now that they are working on the BIOS they could try to solve the problem for you.

I'll do my part and I'll tell them about the problem when they contact me.

Thanks, I sent inquiry today. It's Chinese New Year these days, till Feb 1st they have days off, hopefully tech support will revert to me after holidays.

[Alien Grey]

I've send a mail to my contact for your issue with your RAID card.
All you can do now is hope that they will/can fix it.

[EOD]

I dont know how A-Grey and Mikeyakame are able to run such a low NB voltage for their FSB OCs. I recently upgraded to a Q9650 and am running 467 FSB x9 at 4.2GHz 16+ hrs stable Prime95 with the following voltages in bios:

vcore: 1.33v
PLL: 1.54v
NB: 1.69v (as tRD of 9 I can lower it to 1.65v)
DRAM: 1.98v (They are rated for 2.1v, but runs table as 2x2Gb at 1.8v and 4x2Gb at 1.94v)
FSB Term Volt: 1.38v (but Hardware Monitor in bios and Everest report it as 1.26v)

LLC: disabled
CPU GTL: x.63
NB GTL: x.67

Spread spectrums disabled


Running DDR2 934 5-4-4-12 45
Ai Clock Twister at moderate
Transaction Booster at 8


But as soon as I drop the NB voltage to 1.67, one core fails. My NB temps never go above 42°C, but I hate running the voltage so high.

[mikeyakame]

Dont know for certain, but it might be to do with your divider? And FSB Term Voltage, i set 1.40v, my bios/everest report 1.46-147v, i set 1.47v Vnb reports about the same.

I'm about to start considering that my board might be a female, I could swear it's that time of the month

I gotta stop pulling my comp apart because every time I do I end up spending 1/2-1 day fixing problems that I dont know how to fix by moving cables around and playing musical chairs with pci-e power and fan headers till eventually after a while it stops acting like a child that didn't get its toy it wanted and works normally again. Dont know if its my board or not. I would only imagine it is, because the problem it self appears as many fluctuating voltages in hardware monitor, by fluctuating i mean 0.05-0.1v, bouncing around like a pogo stick. Then during the exorcism I perform it does all kinds of evil things like corrupting and wiping my cmos!

Argh I can fix most things but this one has me baffled, it's been on going for 2 - 3 months and as many times as Ive fixed it up I still don't know how exactly I fixed it.

The only thing positive to come out of it is my CPU temps have plummeted from high 50s to high 40s and all I did was replace the thermal paste I replaced recently, and shimmed down my TRUE so it clamps better. Even Linpack temps are amazing now, 65c full load!