Supermicro H8QGi/6 and H8QGL Next Generation OC BIOS

Status
Not open for further replies.
Great work!
Does “Allow unsafe CAS latencies” mean it's possible to use tighter memory timing than in OCNG3? But I find no change after turning it on.
And I find a small bug: clockspeed in ocng-utils-4 seems to not work properly for Opteron 6200s while it states "present refclock and CPU clock on AMD Family 10h and 15h processors".
It reports wrong Refclock and Clockspeed for my 4p 6272:
Code:
Clockspeed (OCNG4)
Determining CPU frequency, please wait...
Refclock: 139.998 MHz
Clockspeed: 2099.968 MHz
(should be 200MHz and 2400MHz)

I'm pleased to announce general availability of OCNG4 ROM.
See OP for details.

Note that this release still does not support IL/PD chips.
 
Last edited:
I've found a trick (remove jumper cap of JPB1) that would force IPMI to work properly with OCNG3:
http://hardforum.com/showthread.php?t=1728050

But when I apply it to OCNG4 seems the result is failed...

update: now IPMI works with OCNG4, don't know why it didn't work just now.
update2: after a power cycling, IPMI was lost again...
update3: IPMI comes back after several minutes. Ok, now I can confirm this trick works for OCNG4 as well.

Cause of this issue is unknown. No investigation has been done to date.
 
Last edited:
Great work!
Does “Allow unsafe CAS latencies” mean it's possible to use tighter memory timing than in OCNG3? But I find no change after turning it on.
I should probably reword it. Basically this option allows setting CL5 and lower _if_ XMP
profile allows it (e.g. DDR3-1600 CL6 memory at 1333). Sfield's research shows that
CL5/1333 doesn't work with MC CPUs even though memory has such capability.

And I find a small bug: clockspeed in ocng-utils-4 seems to not work properly for Opteron 6200s while it states "present refclock and CPU clock on AMD Family 10h and 15h processors".
This is a known issue. If software P-state 0 frequency changes post boot (e.g. multiplier
modification using TPC) the program gets confused. Fixing requires determining P0
frequency from ACPI tables instead using multiplier from the CPU. Fix is planned :)

It doesn't support Turbo either (but that should be easy to add).
 
Thanks for the explanation. And what is the use of the option "Force 1.5V DIMM Vdd (DANGEROUS)"?
Would this option make it possible to use the XMP profile of 1.6V or 1.65V 1600 memory sticks?
 
Last edited:
:) You always make great inquiries!

This has been added with 1.35V DIMMs in mind. I have two systems with mixed 1.35/1.5V DIMMs
but, IIRC, there's only one voltage regulator per DIMM bank (it can do either only 1.35V or only 1.5V).

"Force 1.5V" setting forces all DIMMs to run at 1.5V so all memory can run (instead of only-1.35V-DIMMs
with original BIOS).

To answer your question whether that option could let 1.65V DIMMs operate at XMP timings but with
decreased (1.5V) voltage, the answer is no. "Is DIMM voltage supported" check is performed before
the "force voltage" code.

Back in the day, musky did some experiments with 1.65V memory and tried running it at 1.5V (with
looser timings) -- results weren't really satisfactory.

Using d3sak and with minimal hex edits (and SPD reflash) one can try and run 1.65V modules at 1.5V
but I don't think this is a good idea...
 
Last edited:
The problem with 1.65V memory is that you have no idea what it will run at 1.5V. It is doubtful you will get the XMP values. What I did was guess at a value and use Thaiphoon Burner to modify the .bin files to my guess, then eeprog to flash the memory. I basically went from 1333 C9 to 1333 C8. It did work - someone is still running these modules today at 1333 C8 without issue. The performance gain was minimal, which I think is what tear is getting at. If you really want to do this, i would recommend testing all of your modules in a board that allows you to set memory timings. Find something that works for all modules, then purchase Thaiphoon Burner and modify/flash your modules in the other board. Once they are flashed, OCNG will read the new flashed values and use them.
 
I upgraded the bios to V4 and am getting this after installing the OCing utility
ocng-cu: OCNG configuration utility (OCNG4)

Running in interactive mode.

Detected board: 'H8QGL'
/dev/nvram: No such file or directory
Make sure nvram module has been loaded.

EDIT: I worked it out!
sudo modprobe nvram
my knowledge of ubuntu only goes as far as it allows me to set up folding
 
Last edited:
I've updated ocng-utils (4.1).
The only change is new clockspeed utility -- now more robust than ever!

At the same time, legacy "freqcheck.sh" has been obsoleted.
Use clockspeed instead and please report any issues :)
 
Building new rig, just have 2x 6176 SE, and load OCNG v4 showing speed at 1.8 instead of 2.3Ghz. (Was 2.3Ghz with the stock BIOS). I assuming, I missed something or doing it wrong?

Using Ubuntu 12.04 with Ballistx 1600 Cas 8 with the H8QGi+-F

EDIT> I thought I would add, I clone this image from my 4x 6166HE that went thru at least Gen 2 and Gen 3 upgrades just in case something legacy could have inter feared with Gen 4.

Ran a ocng-cu with reset and after reboot here are the default settings:
Code:
bowlinra@amd-6176:~$ dmesg | grep -o Detected.*
Detected 1800.103 MHz processor.

bowlinra@amd-6176:~$ sudo tpc -dram
[sudo] password for bowlinra: 
TurionPowerControl export (Turion Power States Optimization and Control - by blackshard)

DRAM Configuration Status

Node 0 ---
DCT0: memory type: DDR3 frequency: 1332 MHz
Tcl=9 Trcd=9 Trp=9 Tras=24 Access Mode:1T Trtp=6 Trc=33 Twr=10 Trrd=4 Tcwl=8 Tfaw=20
TrwtWB=7 TrwtTO=6 Twtr=5 Twrrd=2 Twrwr=4 Trdrd=3 Tref=2 Trfc0=0 Trfc1=2 Trfc2=0 Trfc3=0 MaxRdLatency=54
LDIMM0=EMPTY/EMPTY LDIMM1=OK/OK LDIMM2=EMPTY/EMPTY LDIMM3=EMPTY/EMPTY 

DCT1: memory type: DDR3 frequency: 1332 MHz
Tcl=9 Trcd=9 Trp=9 Tras=24 Access Mode:1T Trtp=6 Trc=33 Twr=10 Trrd=4 Tcwl=8 Tfaw=20
TrwtWB=7 TrwtTO=6 Twtr=5 Twrrd=2 Twrwr=4 Trdrd=3 Tref=2 Trfc0=0 Trfc1=2 Trfc2=0 Trfc3=0 MaxRdLatency=54
LDIMM0=EMPTY/EMPTY LDIMM1=OK/OK LDIMM2=EMPTY/EMPTY LDIMM3=EMPTY/EMPTY 

Node 1 ---
DCT0: memory type: DDR3 frequency: 1332 MHz
Tcl=9 Trcd=9 Trp=9 Tras=24 Access Mode:1T Trtp=6 Trc=33 Twr=10 Trrd=4 Tcwl=8 Tfaw=20
TrwtWB=7 TrwtTO=6 Twtr=5 Twrrd=2 Twrwr=4 Trdrd=3 Tref=2 Trfc0=0 Trfc1=2 Trfc2=0 Trfc3=0 MaxRdLatency=52
LDIMM0=EMPTY/EMPTY LDIMM1=OK/OK LDIMM2=EMPTY/EMPTY LDIMM3=EMPTY/EMPTY 

DCT1: memory type: DDR3 frequency: 1332 MHz
Tcl=9 Trcd=9 Trp=9 Tras=24 Access Mode:1T Trtp=6 Trc=33 Twr=10 Trrd=4 Tcwl=8 Tfaw=20
TrwtWB=7 TrwtTO=6 Twtr=5 Twrrd=2 Twrwr=4 Trdrd=3 Tref=2 Trfc0=0 Trfc1=2 Trfc2=0 Trfc3=0 MaxRdLatency=51
LDIMM0=EMPTY/EMPTY LDIMM1=OK/OK LDIMM2=EMPTY/EMPTY LDIMM3=EMPTY/EMPTY 

Node 2 ---
DCT0: memory type: DDR3 frequency: 1332 MHz
Tcl=9 Trcd=9 Trp=9 Tras=24 Access Mode:1T Trtp=6 Trc=33 Twr=10 Trrd=4 Tcwl=8 Tfaw=20
TrwtWB=8 TrwtTO=7 Twtr=5 Twrrd=2 Twrwr=4 Trdrd=3 Tref=2 Trfc0=0 Trfc1=2 Trfc2=0 Trfc3=0 MaxRdLatency=53
LDIMM0=EMPTY/EMPTY LDIMM1=OK/OK LDIMM2=EMPTY/EMPTY LDIMM3=EMPTY/EMPTY 

DCT1: memory type: DDR3 frequency: 1332 MHz
Tcl=9 Trcd=9 Trp=9 Tras=24 Access Mode:1T Trtp=6 Trc=33 Twr=10 Trrd=4 Tcwl=8 Tfaw=20
TrwtWB=7 TrwtTO=6 Twtr=5 Twrrd=2 Twrwr=4 Trdrd=3 Tref=2 Trfc0=0 Trfc1=2 Trfc2=0 Trfc3=0 MaxRdLatency=54
LDIMM0=EMPTY/EMPTY LDIMM1=OK/OK LDIMM2=EMPTY/EMPTY LDIMM3=EMPTY/EMPTY 

Node 3 ---
DCT0: memory type: DDR3 frequency: 1332 MHz
Tcl=9 Trcd=9 Trp=9 Tras=24 Access Mode:1T Trtp=6 Trc=33 Twr=10 Trrd=4 Tcwl=8 Tfaw=20
TrwtWB=7 TrwtTO=6 Twtr=5 Twrrd=2 Twrwr=4 Trdrd=3 Tref=2 Trfc0=0 Trfc1=2 Trfc2=0 Trfc3=0 MaxRdLatency=52
LDIMM0=EMPTY/EMPTY LDIMM1=OK/OK LDIMM2=EMPTY/EMPTY LDIMM3=EMPTY/EMPTY 

DCT1: memory type: DDR3 frequency: 1332 MHz
Tcl=9 Trcd=9 Trp=9 Tras=24 Access Mode:1T Trtp=6 Trc=33 Twr=10 Trrd=4 Tcwl=8 Tfaw=20
TrwtWB=7 TrwtTO=6 Twtr=5 Twrrd=2 Twrwr=4 Trdrd=3 Tref=2 Trfc0=0 Trfc1=2 Trfc2=0 Trfc3=0 MaxRdLatency=50
LDIMM0=EMPTY/EMPTY LDIMM1=OK/OK LDIMM2=EMPTY/EMPTY LDIMM3=EMPTY/EMPTY 

Done.
 
Last edited:
ocng-utils now include "clockspeed" utility; just go about calling "sudo clockspeed"

It's currently _the_ recommended method of obtaining CPU frequency.
 
I'm apparently missing where are the instructions are for the new utilities..

clockspeed is getting the right speed. It's just not booting that way. Really appreicate the assistance.

Code:
bowlinra@amd-6176:~$ dmesg | grep -o Detected.*
Detected 1799.932 MHz processor.

bowlinra@amd-6176:~$ sudo clockspeed
[sudo] password for bowlinra: 
Clockspeed (OCNG4.1)
Family 10h
Running, please wait...
Refclock: 200.002 MHz
Clockspeed: 2300.026 MHz

bowlinra@amd-6176:~$ dmesg | grep -o Detected.*
Detected 1799.932 MHz processor.
bowlinra@amd-6176:~$
 
It's just bad reporting (due to use NBCLK as TSC source).

Just don't look at dmesg any more :)
 
Last edited:
So this is normal.. Proceed with OC using clockspeed for actual size reference.. Cool.
 
Minor update of ocng-utils (4.2) is now available.

The only change is prettier ht-retries script (when used in 80x24 terminal -- thanks go to musky).
 
Looks like the issue was caused by calling AFUDOS.EXE without SM-mandated parameters which
rendered the board inoperable (beyond emergency restore via SUPER.ROM).

When flashing manually, make sure to always provide SM options, e.g.
Code:
AFUDOS.EXE G60NG4.A11 /P /B /N /C /R
 
Hi tear,
I don't want to push you, but how is overclocking support for PL/IL retail chips going?:)Got a nice offer for 4 6386's:)
 
Thanks guys for the awesome utility. Tried one of the first versions but had issues due to ram timings. Currently trying out 10% with no problems and looking to maybe go to 15% before leaving the rig for a few weeks. 4x6176se H8QGi-+F
 
Hi tear,
I don't want to push you, but how is overclocking support for PL/IL retail chips going?:)Got a nice offer for 4 6386's:)
Perhaps in a few months. Been very busy recently.

Thanks guys for the awesome utility. Tried one of the first versions but had issues due to ram timings. Currently trying out 10% with no problems and looking to maybe go to 15% before leaving the rig for a few weeks. 4x6176se H8QGi-+F
If you run into memory issues you can always try to either relax memory timings (see
advanced options of ocng-cu) or disable use of XMP (more drastic).
 
heh, forgot about the increased power demands. Poped a PSU :(
Back up though, got 10% no problem but 12.5 or 15 were bootable but instalock when folding commenced. Might be a while before I find time to tinker with all those options, did I miss it, is there a decent description of all these new options?
 
The alliterative HT tuning will help when you get failed ht-reties. The other ones are self explanatory.
 
Looks like the issue was caused by calling AFUDOS.EXE without SM-mandated parameters which
rendered the board inoperable (beyond emergency restore via SUPER.ROM).

When flashing manually, make sure to always provide SM options, e.g.
Code:
AFUDOS.EXE G60NG4.A11 /P /B /N /C /R

I tried this (using a bootable USB stick) but all i get in response is:

usage: afudos /i<rom filename> [/o<save rom filename>]

Motherboard: H8QGI-F


Any thoughts?

EDIT: Nevermind - I finally managed to flash it - got the CPU's running at 2.53GHz but Memory dropped from 1333MHz to 1066MHz according to sudo tpc -dram ???
 
Last edited:
Multiply 1066 by your refclock and divide by 200 to get effective DDR3 memory rate.
 
Thanks tear

Got it all up and running @ 220 baseclock

8103 frame times are 9:10 (average) as can be seen here:

http://dl.dropboxusercontent.com/u/25298160/Folding%20Web%20Stats/summary.html

Just look for Cronus - it's currently showing effective frame times (which will be a little out as it's also averaging in the old frame times)

The nice man from Fedex arrived today with four E5-4650's (ES) :D - now all I need is the motherboard (which has been delayed about a week)
 
Last edited:
I'm hoping I could get a little help, I'm very new to 4p systems / folding. I have tried loading the bios and totally bricked my board :-( No beeps without ram, black screen, tried clearing cmos and supermicro BIOS recovery. Also tried the following posted in this thread.

Unplug the PSU/ Let it sit for about 5 to 10 minutes. Clear the CMOS with a paper clip for a good 15 to 20 seconds. Wait another 5 to 10 minutes. Clear the CMOS again for 15 to 20 seconds. Plug the PSU back in and it should power up.

The board is being repaired/replaced but would still like to get this BIOS on it. Here are my hardware specs.

Motherboard: H8QGL-IF-0
CPU: x4 6344
Memory: x16 8GB 1866 Crucial Ballistix Model BLE8G3D1869DE1TX0

I flashed the board using the AMIBIOS Flash Utility from AMI's website and booted on win98 usb dos disk. I used the /P /B /N /C /E /K perimeters that update the whole bios according to AMI manual. Were these right perimeters to use? here was the result.

Code:
C:\>AFUDOS CL1NC4.511 /P /B /N /C /E /K
- Bootblock checksum ... ok
- Module checksums ..... ok
- Warning: BIOS doesn't have NCB, NCB programming canceled
- Erasing flash ........... done
- Writing flash ............ done
- Verifying flash .......... done
- Erasing NVRAM ...... done
- Writing NVRAM ....... done
- Verifying NVRAM ..... done
- Erasing Bootblock .... done
- Writing Bootblock ..... done
- Verifying Bootblock ... done
- Warning: BIOS doesn't have EC, EC programming canceled
- CMOS checksum destroyed
- Program ended normally.

C:\>

I read the BIOS does not support oc on abu dhabi but may in the future. However the BIOS should still work without OC capabilities on abu dhabi correct? Because really just looking to load this bios to get support for my memory "Crucial Ballistix DDR3-1866" Its only operating at 600Mhz dual channel on stock bios. Does anybody know of any 8GB desktop sticks that will operate at quad channel 1866 on this board?
 
This BIOS does not support AD chips at all. It won't even boot with them (which is what you're seeing).

You need a 6100 series chip to recover (cheap 6128 should do) or, alternatively, BIOS backup module ($170-ish) with a pre-flashed chip ($2 + someone who preflashes it).

Your RMA was probably premature. I wish you had touched base with us prior to RMA.

EDIT: Supermicro-mandated options are: /P /B /N /C /R (they were also provided few posts above). Why and where did you get the "AMI" options from?

EDIT: I've reworded the first post a little bit so hopefully no one runs into this issue ever again
 
Last edited:
Wow fast reply. Hmm I see, well I still have the board and can cancel the RMA have not shipped it yet. Will they even RMA it with a modded bios? Seems like getting a 6128 will be cheapest and easiest thing to do. Did I flash the bios correctly? I'm concerned it may be corrupted.
 
From what I can tell, lack of /R and presence of /E /K shouldn't be harmful.

I'd definitely give 6100 series chip a shot (boot it exclusively in CPU1 socket...).
 
I looked in the supermirco 3.0a bios update it did not have a "AFUDOS.EXE" like you said it has a "AFUDOS.SMC" So downloaded the utility from AMI website. Did not used /P /B /N /C /R options because saw wasn't working for someone. The manual that came with the utility said to use /P /B /N /C /E /K

I'm guessing I'm sopose to rename AFUDOS.SMC to AFUDOS.EXE? Sorry for being such noob here.
 
Ok will give 6100 series a shot. Is there anyway supermicro will rma the board with a modded chip?
 
Yeah, 3.0a BIOS package confuses the hell out of people. We'll be providing direct links to
older BIOS packages (they're easier to work with) soon.

Did not used /P /B /N /C /R options because saw wasn't working for someone.
Can you direct me to a specific post, please? I didn't include these options in the first post
for my entertainment, y'know...

I'm guessing I'm sopose to rename AFUDOS.SMC to AFUDOS.EXE? Sorry for being such noob here.
Yes, rename was the right course of action. 'Tis okay. A 6128 chip should get you going :)

To answer your question, we don't have any information on vendor's practice w/custom-flashed boards.
Though if I was the vendor I would reserve the right not to fix such boards.
Also, remember that SM's RMA process may take long weeks (4-6). Your chances are better with a 6128.
 
Last edited:
No its just confusing for people that don't know what there doing totally. Was in middle of night when did this. And you said "ie" thought you meet "in example" and that is not what you were sopose to put.

Darn yeah they said 24hour turn around knew that was bull shit. Yeah this rig needs to be done before 4-6 weeks. Its going to be for hyper-v server.

Um have another question for you what 1866 8GB desktop ram will work on this board with stock bios? Because I have 128GB of ram I can't use. Says 1866 8GB micron sticks tested with it so got cuical. It has very crappy compatibility with the board.

Here is what read that made me use ami options

loudbob n00bie, 25 Days

Quote:
Originally Posted by tear View Post
Looks like the issue was caused by calling AFUDOS.EXE without SM-mandated parameters which
rendered the board inoperable (beyond emergency restore via SUPER.ROM).

When flashing manually, make sure to always provide SM options, e.g.
Code:

AFUDOS.EXE G60NG4.A11 /P /B /N /C /R

I tried this (using a bootable USB stick) but all i get in response is:

usage: afudos /i<rom filename> [/o<save rom filename>]

Motherboard: H8QGI-F


Any thoughts?

EDIT: Nevermind - I finally managed to flash it - got the CPU's running at 2.53GHz but Memory dropped from 1333MHz to 1066MHz according to sudo tpc -dram ???
 
And you said "ie" thought you meet "in example" and that is not what you were sopose to put.
i.e. actually stands for "id est" == "that is" == what you _are_ supposed to put ;)

Crucial DDR3-1866 RAM (assuming you're talking about BLT2G3D1869DT1TX0 model) should function but it won't go beyond DDR3-1333 -- at least not w/o tricks.

Here is what read that made me use ami options
I think you jumped to conclusion here... to me it looked like a case of PEBKAC... :-o
 
I see well I'm learning, Um I have 8GB sticks Model BLE8G3D1869DE1TX0 OS was saying 600MHz dual channel :-( Supermicro's site says DDR3 1866 Quad channel support. Thought could just pop it in and be good to go. I tried some tweaks and tricks in 3.0a bios nothing worked. What are these tricks?
 
600 or 667 MHz? 667 MHz would fit with DDR3-1333 [667 MHz + DDR]. Post BIOS pic if in doubt.

Re tricks
Principles are laid out at http://www.amdzone.com/phpbb3/viewtopic.php?f=521&t=138490&start=225#p218762 and couple following posts.

Note that d3sak, as released, does not support G34-compatible modifications of Crucial DDR3-1866 SPDs so some updates will have to be made
before you can follow the process. For now, think of these posts as an overview.

I would also hold off any memory mods until machine is operating stably at nominal JEDEC speed (DDR3-1333).
 
Last edited:
Was fluctuating but don't think hit 667. Bios was showing it as DDR3-1333 anyways is it difficult to get 1866 speed on AD cores with ram I got or should I return it and get something else? I need every ounce performance out of this thing.
 
Aw hmm darn so pain in but to get working at 1866. I had server2012 and cpuz on it looking at ram. Server is for remotefx hyperv.
 
Status
Not open for further replies.
Back
Top