G34 overclocking central

tear

[H]ard|DCer of the Year 2011
Joined
Jul 25, 2011
Messages
1,568
[will expand this as time permits]

Course of action depends on your hardware:

1. All production chips (6100/6200/6300 series) in Supermicro H8QGi, H8QG6, H8QG7, H8QGL, H8DGi, H8DG6 or H8SGL board

 Use OCNG5: http://hardforum.com/showthread.php?t=1836265

2. Production 6200/6300 series in other boards

 Use TPC (psmax 1 trick): http://hardforum.com/showthread.php?t=1708870
 Note this does not overclock per se. It makes the CPU operate at all-core turbo at all times (something these CPUs have a problem with).

3. ES 6100 (MC) in any board

 Use TPC.

 Disable PowerNow and CPB/Turbo in the BIOS.

 Linux:
 Sample basic code (3000 MHz, 1.25V):
Code:
FREQ=3000
VCORE=1.2500
sudo tpc -fo 1
sudo tpc -set freq $FREQ vcore $VCORE
sleep 1
sudo tpc -fo 0

 Windows:
 Sample basic code (3000 MHz, 1.25V):
Code:
set FREQ=3000
set VCORE=1.2500
TurionPowerControl -fo 1
TurionPowerControl -set freq %FREQ% vcore %VCORE%
TIMEOUT /T 2
TurionPowerControl -fo 0

4. ES 6200/6300 (IL/AD) in any board

 Use TPC:
  
 Enable PowerNow and CPB/Turbo in the BIOS.

 Linux:
 Use Performance scaling governor (or otherwise disable Ondemand scaling governor);
 if you're using [H] Ubuntu appliance or followed Ubuntu desktop installation guide for
 FAH (with fahinstall utility), you're already set.

 Sample basic code (3000 MHz, 1.175V):
Code:
FREQ=3000
VCORE=1.1750
sudo tpc -psmax 6
sleep 1
sudo tpc -fo 1
sudo tpc -set ps 1 freq $FREQ vcore $VCORE
sudo tpc -set ps 0 freq $FREQ vcore $VCORE
sleep 1
sudo tpc -fo 0
sleep 1
sudo tpc -psmax 1

 TPC installation on Linux is covered here: http://hardforum.com/showthread.php?t=1669698

 Windows:
 Modify power plan settings: change Minimum Processor State to 100%

 Sample basic code (3000 MHz, 1.175V), run from admin cmd (critical: one line at a time!)
 or a batch file (most convenient and preferred):
Code:
set FREQ=3000
set VCORE=1.1750
TurionPowerControl -psmax 6
TIMEOUT /T 2
TurionPowerControl -fo 1
TurionPowerControl -set ps 1 freq %FREQ% vcore %VCORE%
TurionPowerControl -set ps 0 freq %FREQ% vcore %VCORE%
TIMEOUT /T 2
TurionPowerControl -fo 0
TIMEOUT /T 2
TurionPowerControl -psmax 1
 
Last edited:

geok1ng

2[H]4U
Joined
Oct 28, 2007
Messages
2,130
we really need a bump here because 6xxx prices on ebay are down and there are plenty of ES for sale as well.
for some weird reason the hardware discussion of F@H builds is done here on DC subforum and not on the multiprocessing systems subforum, where my thread asking for help in a chess machine build is in serious need of expert advice.
 

Nathan_P

[H]ard DCOTM x3
Joined
Mar 2, 2010
Messages
3,479
Probably on here because we have built more insane high end rigs than anyone else.
 
D

Deleted member 12106

Guest
Gathering more data to come to a verdict....
 

Shadowmeph

Limp Gawd
Joined
Dec 3, 2011
Messages
156
I sure wished I could get TPC to run on my win10 eval OS , no problems at all on win8 or 7 and of course Linux but the win 10 preview is a no go
 
D

Deleted member 12106

Guest
It looks like by default DEP on win10 is set at "essential windows programs and services"

Smart screen is within security center. Beyond that I would look into any prerequisites that many not be in win10.
 

Shadowmeph

Limp Gawd
Joined
Dec 3, 2011
Messages
156
Tose are off, what happens when I use TPC is that my screen flashes then after a few seconds my system logs out when I try to log back in it repeats the same thing flashing away then logs out, on the very rare occasion that it does let me log back in my system is not OCed at all.
At first I had thought that maybe the OC way to high so I just ran TPC at the stock clocks for my CPU and the same thing happens.
this erks me because tpc works great on all of my other OS's just this one is giving me problems
 

tear

[H]ard|DCer of the Year 2011
Joined
Jul 25, 2011
Messages
1,568
Please file an issue at tpc's project page (google code).

TPC uses WinRing0 (library+kernel driver) for MSR and PCI config space accesses which
is where compatibility issues arise most frequently.

The problem with WinRing0 is that it's no longer maintained, source is gone and there are
some scenarios it doesn't handle very well (vide issue #28).

Proper way of dealing with this and other issues is switching to a different/smarter approach,
except I haven't been able to find a similar tool.

That said, Windows 10 support isn't something that will happen soon.
 

Shadowmeph

Limp Gawd
Joined
Dec 3, 2011
Messages
156
Please file an issue at tpc's project page (google code).

TPC uses WinRing0 (library+kernel driver) for MSR and PCI config space accesses which
is where compatibility issues arise most frequently.

The problem with WinRing0 is that it's no longer maintained, source is gone and there are
some scenarios it doesn't handle very well (vide issue #28).

Proper way of dealing with this and other issues is switching to a different/smarter approach,
except I haven't been able to find a similar tool.

That said, Windows 10 support isn't something that will happen soon.

I never noticed that I could submit to the google code page thak you for the info
 

Shadowmeph

Limp Gawd
Joined
Dec 3, 2011
Messages
156
Finally I am able to OC in windows 10 preview I upgraded the OS to build 9879 and then reinstalled my ATI vidcard drivers everything is all good now :)
 

terrabyte89

Weaksauce
Joined
Nov 6, 2004
Messages
120
Hi guys! I know G34 is a bit last decade but I was hoping some people might be able to help me out with some overclocking. I've had great success with ocng and ocing retail opterons :). But this time I've got some ES chips and I was trying to get them running a bit quicker first for some benchies and then more boinc!

Running H8QGi+-f running ocng5 bios
4x Spicy 2.1ghz IL 16 core (62xx) chips.
16x 1gb sticks of DDR3 1333.
Windows server 2008 R2.

Runs fine at stock, pumps through a few WU's in boinc and ran a few quick benchies, seems ok for an IL @ 2.1ghz.

Powernow is enabled in BIOS and windows minimum processor time is set to 100% and I've loaded TurionPowercontrol and setup a bat file to run the OC using the sample code here. I first tried the sample code directly but then ran it a bit slower to see if it made a difference.

Code:
set FREQ=2400
set VCORE=1.1000
TurionPowerControl -psmax 6
TIMEOUT /T 2
TurionPowerControl -psmax 1
TurionPowerControl -set ps 1 freq %FREQ% vcore %VCORE%
TurionPowerControl -set ps 0 freq %FREQ% vcore %VCORE%
TIMEOUT /T 2
TurionPowerControl -fo 0
TIMEOUT /T 2
TurionPowerControl -psmax 1

Basically what's happening is the script runs, says done and all good but then the PC runs like a dog. Wprime 32m takes like 5 mins (instead of 11seconds). Windows lag moving them around, loading anything off the hdd takes forever. The PC doesn't crash just runs horribly slowly.

I ran turionpowercontrol -l and it shows all the nodes with their pstates. First Pstate lists 2400mhz and 1.1v then each pstate after is a step down in clock speed and volts just as you would expect. But CPU-Z still shows 2100mhz. Reboot and it goes back to defaults?

I'm hoping I'm missing something simple here? Any help is much appreciated.

edit: Also something else very odd. I can't get accurate CPU temps? Using either coretemp or tpc -temp. They all just show the board temp of like 30c. Running 100% load with 212 coolers I get 33c tops, which is definitely not correct. This happened on both my h8qgi+f boards and my retail 6170 chips and these spicy ones.
 

terrabyte89

Weaksauce
Joined
Nov 6, 2004
Messages
120
Worked it out (sort of). Used the script from Andrew Hazelden's blog .

Now just have the issue with this shitty board giving my opteron temperatures! I'm 99% sure my opterons don't run at 27c at full load. I've tried, coretemp, tpc -temp, hwm blackbox. Everything just gives what I think is the board sensor rather than the actual CPU.
 

tear

[H]ard|DCer of the Year 2011
Joined
Jul 25, 2011
Messages
1,568
The script should have worked. I'll double check in spare while.

Temps in TPC come from on-die sensors.

Also, the units are not Celsius degrees; from AMD documentation:
Tctl is a non-physical temperature on an arbitrary scale measured in degrees.
It does not represent an actual physical temperature like die or case temperature.
Instead, it specifies the processor temperature relative to the point at which the system
must supply the maximum cooling for the processor’s specified maximum case
temperature and maximum thermal power dissipation.

Virtually all AMD CPUs bear max Tctl of 70 deg regardless of Tcase or Tjmax or whatnot.
IOW, you could look at your 27 deg Tctl as: "43 deg Tctl until CPU starts throttling".
OTOH, OC'd processors stop working long before you reach max Tctl so manipulating max Tctl doesn't change much in the picture.

It is possible that your sensors are damaged or miscalibrated.

If you need readings in human temp scales you'll need a thermocouple or somesuch :)
 

tear

[H]ard|DCer of the Year 2011
Joined
Jul 25, 2011
Messages
1,568
I'm hoping I'm missing something simple here? Any help is much appreciated.
Actually, you weren't missing anything.

A bug sneaked into Windows version of the script (no idea how); fixed it -- should be fine now.

Thanks for the report :)

BTW, I'm thinking of integrating multiplier changes with OCNG to ultimately remove the need to run TPC at boot (w/ES chips).
 
Last edited:

terrabyte89

Weaksauce
Joined
Nov 6, 2004
Messages
120
Hi tear, thank you for getting back to me :).

Wow, that temperature stuff is complicated haha! So I guess I'll find out it's too hot when it just shuts down?
 

tear

[H]ard|DCer of the Year 2011
Joined
Jul 25, 2011
Messages
1,568
CPU will typically malfunction (crash the app/OS or freeze) before
any thermal conditions are met.

Side note: we have seen CPUs with broken thermals that would signal
overheat (and cause the board to shut the system down) even at stock
speeds and medium load.
 

rvborgh

Weaksauce
Joined
Jul 1, 2014
Messages
108
hi Folks,

I'm running OCNG 5.3 on my H8QGi-F running 4 Magny-Cours 61xx Extra Spicys at 3.0 GHz.

OCNG allowed my existing G.Skill memory to run at its rated XMP profile at DDR3-1600 (CL9 at 1T) - its set to manual in the bios at 800 MHz.

i was wondering if anyone has had success running the 61xx processors with higher performance DDR3-1866 (CL9) memory? i was thinking of purchasing 4 sticks and giving it a go for CPU 0. The thing about it is that my CPU/Northbridges are stuck at 1800 MHz. i have read that this gives each die an upper memory performance limit of 28.8 GB/s. Two channels worth of DDR3-1866 (2x14.9 GB/s) would put me over that limit.

Has anyone had any success using these higher performance DDR3-1866 sticks with the H8QGi-f series of boards and OCNG?

Also... how to boot with the Northbridge volts set higher? With OCNG... i can slightly up the CPU/NB by raising the reference clock... but anything more than 205... results in an unreliable system. i suspect it probably is because the stockish 1.08v CPU/NB voltage is not enough to run at the higher speeds.

Any info appreciated.

This system runs fantastic as a home PC by the way. Kind of overkill but i love it.
 
Top