**Unified** DFI LanParty UT 790FX-M2R-M2RS / Phenom /Data Base Thread**

[vext]

Good news, the board is back.

I tried tony's reverse battery method. I left the totaly unplugged board with the clear cmos jumper set and the batter mounted revers alone fopr a few hours and now it's working again. Glad I had luck this time.

Interesting. My old Epox 939 board used to lock up occasionally. I could shock it back to life by removing the battery, setting the clear cmos jumper for about 10 minutes. Then I would set everything back, and boot up while pressing the insert key. The Epox also had a POST LCD, and would lock up at "FF" rather than "C1".

I recently ran lm-sensors configure while installing Debian Linux on my DFI M2R + 9850BE. It locked everything up and the LED read "C1". I tried my "Epox" technique but it didn't help. I had to RMA both the board and cpu. Since then I've stayed away from lm-sensors and speedfan. Unfortunately those were my two favorite monitoring programs...

If it happens again, I'll try your method. The difference is that you reversed the battery and let it sit for a long time.

[justapost]

I recently ran lm-sensors configure while installing Debian Linux on my DFI M2R + 9850BE. It locked everything up and the LED read "C1". I tried my "Epox" technique but it didn't help. I had to RMA both the board and cpu. Since then I've stayed away from lm-sensors and speedfan. Unfortunately those were my two favorite monitoring programs...

There is no need to run sensors-detect. The only module you need for sensor monitorin is "it87". Seems I have an robus bios chip here.
The second time I ran into the C1 issue, removing one dimm helped. That dimm made problems for a few minutes and seems to work ok now.
The old sensors-detect routine will trigger the chip at 0x2e to appliy an >3V vdimm. That should harm the dimm's and the chip's nb. I wonder why it also results in faulty bioses.
Did you get both RMAed? If so did they say what was defect.

If it happens again, I'll try your method. The difference is that you reversed the battery and let it sit for a long time.

That was tony's tip from a few weeks ago, whom I crossposted here.

[Rammsteiner]

I was just gaming with stock settings. Although I had CnQ and Vcore/Vnb still at auto because of MSR OC attempts. But this shouldnt matter.

But I was gaming and system froze, blackscreen... So I set CnQ off and set the default Vcore and Vnb manually.

If it happens again Ill try to downclock the GPU (maybe the pencil is degrading on it). If not... Im going mad...

Then Ill RMA the M2R and Phenom as well. Im even considering to get Crosshair II instead. This is driving me nuts.

However, not making any conclusions yet. But Im getting a deja vu from my X2 + M2R...

[EDIT]
Didnt help crashed again. Downclocked the GPU this time.

[Brother Esau]

how come new bios don't allow nb fid/did change when enabling cool'n'quiet?

Which new Bios? 4/16/08 Official or 5/5/2008 Beta?

[justapost]

I think 3-07 was the lates version where it was possible to change the nb vid/did if cnq was enabled.

[Brother Esau]

@Achim.....question? Through you're personal Experience with Phenom and 790FX do you find that HCI MemTest works well and is accurate for testing Ram Stability in Windows?

[justapost]

@Achim.....question? Through you're personal Experience with Phenom and 790FX do you find that HCI MemTest works well and is accurate for testing Ram Stability in Windows?

Good question.
I never used HCI memtest, I rely on memtest86+ http://www.memtest.org/.

[AlabamaCajun]

Ramm, sorry if I missed it but have you visited the screens such as Genie, RAM and PWM and hit F7. I had to do this to get stable. Do RAM first and reboot up to the Save DMI, I think Post 96. Then Restart there and reenter BIOS and F7 the other 2 then reboot. This is the only way I've found that completely clears the controller tables.

JP, thanks for the Register info, that's GOLD

I could not resist playing with the clocks again. Still on the 5-X-2008 Beta which for me seems stable.
I went ahead and hit the volts back to 1.28 after a few days running stock at 1.275. Bumped the lock to 209 set CPU multi back to Auto, not 12.5. It booted and ran all night, folding from 8pm to 8am before I shut it down to let the rooms cool. Night temps are on the low 60s requiring little AC so not much cool air goes there, too muggy for cracking the window.

2.61G is not too back but I may still be able to get 2.7G+ solid if I can find the best way to cool this thing.

[seeker010]

I think 3-07 was the lates version where it was possible to change the nb vid/did if cnq was enabled.

yes. you are right. do we know why DFI made this change? Also, what does cpu stepping p0 and p1 mean?

[Rammsteiner]

Ramm, sorry if I missed it but have you visited the screens such as Genie, RAM and PWM and hit F7. I had to do this to get stable. Do RAM first and reboot up to the Save DMI, I think Post 96. Then Restart there and reenter BIOS and F7 the other 2 then reboot. This is the only way I've found that completely clears the controller tables.

Thx for that, didnt know that yet.

How do you mean stable though? As in less/no BSOD's, no strange freezes? At stock or as well with OC'd settings?

[AlabamaCajun]

Stable is when I can carry out normal tasks, surfing, folding, gaming and not see any glitches. Solid is what I call one that runs weeks without any problems though I will through a game error if it cleanly quits to desktop and reports an error that happens after doing something wild that may have just pushed the game into a buggy area of code. Some of the simulator games can be pushed over the edge. Even though I get suspect of a hardware problem and will be on guard.
After all the flakiness I saw with the GBT 790FX-DS5 that sits in a box, this is stable.

[KTE]

If they removed the option of FID+CnQ, it may be because of a problem with it though, someone should ask them
They rarely remove it for no reason.

[aGeoM]

If they removed the option of FID+CnQ, it may be because of a problem with it though, someone should ask them
They rarely remove it for no reason.

The reason is, before you could set it the way you wanted, but after you reboot/boot to windows, it reseted to default, now you can't have it enabled just because it wouldn't work. Makes sense to me. Don't know if my English makes for you.

[Brother Esau]

Update:.........5/5/2008 Beta Run at the same settings a my Bios Tempate Overclock on Front page with Ram @ 1066 OCZ Tony's Modified SPD @ 1.87Vdimm CPU @ 1.23750vcore set OCCT Crashed 4hrs 46 minutes so technically that equals 8 1/2 hrs of Prime95 I am quite positive its due to heat nmore then likely as it is 90F in my Room Currently

Overall Bios seems OK but needs some refinement is all

Once again there are USB Keyboard Issues Present in this Bios when in Pre Boot Prior to Post the USB Keyboard does not Respond and you have to Restart Manually to get the Delete Key to Function to get into Bios!

Also some NIC Latency Issues "Not Much' but it is there.

Also another Minor Annoyance is that after saving settings and Restarting the Splash Screen / Post Screen does not appear and just is Black until Raid Controller Bios Post appears!

Can any of you gentlemen Running 5/5/2008 "Beta" with the Board set up on the Bench get a accurate DMM Reading on CPU , NB Voltage please and post offset variants?

[KTE]

Thanks aGeoM. It was a problem on many AMD boards, even when they disable it.

When they leave CnQ and FID changing on together, the clockspeeds through the TSC are read wrongly many times [downclock to read 2500 at all times].

[justapost]

The odd thing is that different nb vid's and did's worked fine here with the 9600BE and the 4-07 bios. The problem aGeom described did not occure with all phenoms. !

[justapost]

A small update about the phenoms smbus thermal interface.

On the DFI/Sapphire 790FX boards this interface is accessible like a separate chip at the address 0x4c. It can be read out undel linux with

i2cdump 0 0x4c

     0  1  2  3  4  5  6  7  8  9  a  b  c  d  e  f    0123456789abcdef
00: 00 16 00 00 08 00 00 46 00 00 00 00 00 00 00 00    .?..?..F........
10: 80 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00    ?...............
20: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00    ................
30: 47 80 00 00 00 00 00 00 00 00 00 00 00 00 00 00    G?..............

This is a typical output. I found the value at address 0x01 holds the core temp in celsius. 0x30 is always that temp + 49°C-50°C.
The value at 0x0a seems to be the max temp the chip is rated for. I only have an 9600BE here and that cpu shows 0x46 means 70°C here.

However there is another way to read out this interface the smbus way.
There are two registers on the Miscanselous Control Register (F3).
Register 0x1E8 is used as an address pointer and 0x1EC can be used to read and write data to that address.
So if you startup Baredit and select "B.0 D.24 F.3" you can write the desired address to the register 0x1E8 and after you select "B.0 D.24 F.3" again you get the value in 0x1EC


This screenshot shows the data of register 0x01 of the thermal sideband smbus interface means it's the current core temp (0x18 =>24°C).

Note: This way of access also works on the GBT780G board with the 9600BE. This board does not offer an i2c/smbus access at 0x4c and on no other address but via the registers 0x1E8 and 0x1EC it works. On both boards registers 0x1E4 has bit 3 set to 0b, which means the thermal side band smbus interface is not deactivated.

May I ask that someone with an 9850BE or an other phenom different from an 9600BE posts the values of the registers 0x06 and 0x46 here.

1. Start BarEdit
2. Select "B.0 D.24 F.3"
3. Write 0x06 into register 0x1E8
4. Select "B.0 D.24 F.3"
5. Write down the value of register 0x1EC

Do the same with 0x46 instead of 0x06 at step three.

This data will help to identify the chip and skip some probing routines.

[aGeoM]

The odd thing is that different nb vid's and did's worked fine here with the 9600BE and the 4-07 bios. The problem aGeom described did not occure with all phenoms. !

NB-VID works if I set +xxx% over default.
NB-FID if I set it to >06~what ever APM shows always 2000MHz, then after reboot and enter BIOS NB-FID=06, so the only way to set ie. FID=09(2400) is with Baredit tool after a reboot to windows.

This was tested with 9850BE + Bios 307.

Edit

Bios 505:

BIOS CPU-VID=+103.xx%
BIOS CPU-NB-VID=103.xx%



Well...VIDs are working as BIOS 307, CPU-FID can be changed but resets to default (2500), they should ghost this option to.

May I ask that someone with an 9850BE or an other phenom different from an 9600BE posts the values of the registers 0x06 and 0x46 here.

1. Start BarEdit
2. Select "B.0 D.24 F.3"
3. Write 0x06 into register 0x1E8
4. Select "B.0 D.24 F.3"
5. Write down the value of register 0x1EC

Do the same with 0x46 instead of 0x06 at step three.

This data will help to identify the chip and skip some probing routines.

I wrote 0x06 and 0x46 @ 0x1E8 nothing changed @ 0x1EC, as your instructions.

[justapost]

NB-VID works if I set +xxx% over default.
NB-FID if I set it to >06~what ever APM shows always 2000MHz, then after reboot and enter BIOS NB-FID=06, so the only way to set ie. FID=09(2400) is with Baredit tool after a reboot to windows.

This was tested with 9850BE + Bios 307.

Hmm as usual I mixed up NB-VID and NB-FID in my previous post. I have in mind that i used different nb-fid's with CnQ enabled turing my p-state testings and it worked fine.

I wrote 0x06 and 0x46 @ 0x1E8 nothing changed @ 0x1EC, as your instructions.

Thank you for testing.
Even if you select that "B0 D24 F3" pci register in the dropdown again?
If i write someting into 0x1E8 it first shows 0x00 as it's new value. After I loaded the pci register again by selecting it in the dropdown the value i wrote in shows up and the data from that register shows up in 0x1EC.
Is bit 3 ox 0x1E4 0b on your system?

[aGeoM]

Hmm as usual I mixed up NB-VID and NB-FID in my previous post. I have in mind that i used different nb-fid's with CnQ enabled turing my p-state testings and it worked fine.

I whish it worked for me, this 9850BE, is stable if I set the MSR registers to 2700@default VID, but from BIOS I need to feed some more volts, NB it's ok @2400, so KTE help me to write the ".ini" to use with Sam2008 tool to set CPU speed at 2700, but for NB@2400 I have to manualy change it in Baredit and reboot, thats a PITA.

Thank you for testing.
Even if you select that "B0 D24 F3" pci register in the dropdown again?
If i write someting into 0x1E8 it first shows 0x00 as it's new value. After I loaded the pci register again by selecting it in the dropdown the value i wrote in shows up and the data from that register shows up in 0x1EC.
Is bit 3 ox 0x1E4 0b on your system?

Well after 0x1E8 write first shows 0x00, but after I load it again it reads the value entered at the same adress 0x1E8. ie 0x1E8=0x06

Here's a SS defaul system/0x06@0x1E8:

[vext]

There is no need to run sensors-detect. The only module you need for sensor monitorin is "it87". Seems I have an robus bios chip here.
The second time I ran into the C1 issue, removing one dimm helped. That dimm made problems for a few minutes and seems to work ok now.
The old sensors-detect routine will trigger the chip at 0x2e to appliy an >3V vdimm. That should harm the dimm's and the chip's nb. I wonder why it also results in faulty bioses.
Did you get both RMAed? If so did they say what was defect.

That was tony's tip from a few weeks ago, whom I crossposted here.

Whooee, 3.2v, I can see why things came to a halt. The dimms are ok but I could never get the cpu to work after that. I tried it in two other boards, including a new DFI M2RS. I didn't try a second cpu in the original M2R board, so it might have been ok.

Yes, I RMAed both cpu and board. Newegg didn't get back with me about what the defect might be.

So, sensors-configure is not necessary? I would really like to get lm-sensors working in Linux.

[justapost]

Whooee, 3.2v, I can see why things came to a halt. The dimms are ok but I could never get the cpu to work after that. I tried it in two other boards, including a new DFI M2RS. I didn't try a second cpu in the original M2R board, so it might have been ok.

Yes, I RMAed both cpu and board. Newegg didn't get back with me about what the defect might be.

So, sensors-configure is not necessary? I would really like to get lm-sensors working in Linux.

Hi vext,

I was wrong with the 3,2V the probing used random values so the applied value could have been anything up to 3.2V.
Jean from lm-sensors made a few patches in the last days so the next version 3.0.2 will run just fine on the Sapphire/DFI boards.
Everything lm-sensors dtects is the smbus controller (i2c-piix4) and the sensor-chip (it87). So if you add those modules manual you don't need to run sensors-detect.

[Final8ty]

My Phenom 9850 Black Edition should be here tomorrow
Some people have said in another forum that i will need to re-install Vista64
going from Dual to Quad ?

[Brother Esau]

That would be a wise idea

[vext]

Hi vext,

I was wrong with the 3,2V the probing used random values so the applied value could have been anything up to 3.2V.
Jean from lm-sensors made a few patches in the last days so the next version 3.0.2 will run just fine on the Sapphire/DFI boards.
Everything lm-sensors dtects is the smbus controller (i2c-piix4) and the sensor-chip (it87). So if you add those modules manual you don't need to run sensors-detect.

Thanks for the info. I just checked the lm-sensors site and they have a bulletin which explains the problem and their solution.

I'll try version 3.0.2 as soon as it's posted.

vext