Supermicro H8QGi/6 and H8QGL Next Generation OC BIOS

Status
Not open for further replies.
Refclock of 240 on 6174s is exceptional. I doubt you'll be able to squeeze more beyond that
but you sure can try if you want to...

I'd also recommend keeping fans upright to avoid fluid leak and premature bearing wear.
 
I've added voltcheck to ocng-utils (now at version 4.4).

To install/upgrade follow steps 10 through 14 from the first post.

First post now also includes complete list of OCNG utilities.
 
This might sound stupid for all you Linux-Guru's in this forum, but I finally got my H8QG6-F board w/4x6380 (EngSample) running, but I can't figure how I can change the CPU Multiplier!
Can anybody help me (I have 0.01% experience with Linux)?

BTW: Thanks a 1,000,000 for making the "Ubuntu Desktop Installation Guide for F@H" (http://hardforum.com/showthread.php?t=1601608) - It got me going in less than 15min!!
 
TPC wiki is pretty much stale and you're unlikely to get good results by using wiki examples.

I have updated this: http://hardforum.com/showthread.php?t=1765747

Item #4 contains most reliable/safe OC sequence for ES IL/AD chips.

As musky noted, this thread is meant for OCNG BIOS [retail, MC chips]; if you have
further questions, please create new thread.
 
Finally got around to flashing my BIOS with the [H] mod. Gotta say folks, tear, others who made this magic I'm impressed. I posted my stuff over at the other place regarding numbers.

I'm currently pushing a refclock of 240 with 0 ht-retries. I'm gonna pop it over to 245 in a day to see if it can maintain that refclock. Temps are mild and averaging 35C with 212's.

Thanks folks for everything you do. If I knew about this forum before the 'other' one I'd probably be here. Alas, loyalty and all. :)
 
I'm currently pushing a refclock of 240 with 0 ht-retries. I'm gonna pop it over to 245 in a day to see if it can maintain that refclock. Temps are mild and averaging 35C with 212's.

What chips? Do note that those temps are not real temps, it is an arbitrary scale that gives you a measurement to measure by, but doesn't mean what it says (like my wife).

Do watch it though, at first it may look good and 1/2 or more through the WU it crashes out on you...

Alas, loyalty and all. :)

Here, drink this kool-aid :)
 
I'm watching closely sc0tty8... trust me... So far I'm at 250 and stable as hell. Seeing a minor incremental increase in temps and TPF... so I may just stay there for a bit. See how things go.

I'll let ya know...

I'll settle for a beer or bourbon... can't stand Kool-Aid. :)

Oh... 6168's.
 
Okay... settled in at 250 and completed the 8101 without complications. Working on an 8104 now and watching the system. HT-retries is still rock solid with no boogey men and I'm sitting pretty hoping for the best and that this works for awhile.
 
I just upgraded one of my 4P folders to Ubuntu 12.04 LTS, clean install, from the link in this thread. The system:
12.04 LTS
Board: H8QGi+-F
BIOS: G60NG4.A11
4P, G34 6172
16 X 2GB
DDR3 1600/8/8/8/24, 1.5v

This system was previously running 10.10 with the previous OC utilities, BIOS G60NG3, and Turion Power Control utiilties.Issue: I installed ocng-utils.4.5 and did not receive any installation errors, at least that I could recognize. I followed the instructions in this thread to the letter. At the end of the installation, without closing terminal, I ran ocng-cu and set my preferred refclock. After a power-down, wait, and reboot, only one of the utils functions would work - ht-retries.

sudo ocng-cu resulted in "
[FONT=&quot]“/dev/nvram: No such file or directory Make sure nvram module has been loaded”
sudo clockspeed resulted in "invalid refclock"

I reinstalled ocng 4.5 and opened ocng-cu in the same terminal session. It showed the refclock that I had set previously, so ocng has some functionality.

What's wrong? What should I do?

Here is the message in terminal following sudo ocng-cu: "ocng-cu: OCNG configuration utility (OCNG4.5), Running in interactive mode. Detected board: 'H8QG6', /dev/nvram: No such file or directory, Make sure nvram module has been loaded"

[/FONT]
[FONT=&quot]Update: After reinstalling (yet again), the following commands invoke the proper function - sudo ht-retries, sudo version, and sudo voltcheck. Also, like before, right after a fresh installation, leaving the terminal open, I was able to check/change the refclock, but not after power down and restart.

[/FONT]
 
Last edited:
Thanks for the feedback :)

I'm seeing you followed Ubuntu Desktop Installation Guide. That said, you need to load
appropriate modules for these utilities to work.

Specifically, need to load cpuid and msr for clockspeed and nvram for ocng-cu (per the
message it produced).

To have system load said modules at boot, do the following:
Code:
echo msr | sudo tee -a /etc/modules
echo cpuid | sudo tee -a /etc/modules
echo nvram | sudo tee -a /etc/modules

You can load them manually right now as well (to avoid unnecessary reboot):
Code:
sudo modprobe msr
sudo modprobe cpuid
sudo modprobe nvram

Once the modules are loaded, clockspeed and ocng-cu should work without additional intervention.

Let me revise setup instructions a little bit so we don't run into this problem again...

UPDATE: instructions are now revised, sorry about the hassle
 
Last edited:
Smooth! ocng-utils-4.5 seemingly is working perfectly now. All the commands bring up the correct functions now.

This was the 'proof of concept' 4P upgrade to Ubuntu 12.04 and OCNG 4.5. Fixing the OCNG utilities was the last main issue.

Thanks, Tear, thanks Musky.
 
Exciting news! Happy to have donated a wee bit to its development as a way of saying 'thanks'!
 
Donated! What a great idea. My 4Ps sent me a message in unison over my home network, "Area 51 rocks!"
 
Donated as well and just wanted to say "thank-you" for all the great Folding Tools that you have created for Folders!! :cool:
 
Hi,

My board won't POST anymore.
I followed the reset CMOS procedure. Doesn't fix it.
What should I do ?

Thanks.

PS : I did not panic.
 
Can you walk us through what you did, and when it stopped working?

I was doing some overclock. I first tested the max OC value [262] (I think), it worked for a time and turned off. I tried [250], it worked for a time and then the CPU was acting weird. I tried 240 and after that my board stopped working.

My board is a H8QGL.
 
Last edited:
What chips and psu? How are you clearing the cmos?

I use 4 x 6164HE. Watercooled.
PSU is ENERMAX REVOLUTION 85+ 1020W.
Before the overclocking, it's been working well for 2 weeks.

I'm not sure about the process of clearing the CMOS, on this thread it is said to not remove the battery but wait over 15 sec. On the board's manual it is said to remove the battery.
I tried both.

Any advice ?

EDIT : I waited 12 hours with board unplugged from psu and it finally turned on. I can resume my OC.

EDIT : The system crashed after few minutes. Can't turn it on again.
 
Last edited:
When you say it won't power on, do the fans spin at all or not?

Your PSU is a multi-rail design. It's possible that there's not enough current in your EPS12V
rail(s) and overcurrent protection is kicking in (you need to remove AC power for a while
to "reset" the OCP) thus causing your symptoms.

In any event, I highly recommend switching to a single-rail PSU (I figure 650W should be
enough for your application). If you have a spare 550-600W single-rail PSU you could try
testing with it, too.
 
Last edited:
bios_pads.jpg


Look for these 2 points and ground them out with a screw driver, hold it there for 15-30 seconds. Power up and see what happens.

For those chips your PSU is huge like xbox, a smaller one as tear has eluded to will also work.
 
Also, given you're using HE chips, you can bump the CPU voltage slightly which should
stabilize the chips at higher refclocks (but you should generally take your time with OCing).

The process is currently slightly complex but there are plans to make it more straightforward.

You should install tpc (http://hardforum.com/showthread.php?t=1669698 -- though note
that you only need to load Optimal Defaults with OCNG, no need for additional tweaks).

Do this while the system is idling (try copying and pasting as typing it manually exposes you
to a risk of failure):
Code:
sudo tpc -l | grep -A5 Node.[0246] | grep Maximum.allowed | tr -d '()V' | cut -f 11 -d ' ' | (N=0 ; while read V ; do echo $V | grep '1\.55' && continue ; echo Node $N/$(($N+1)): $V ; sudo tpc -set node $N vcore $V node $(($N+1)) vcore $V ; N=$(($N+2)) ; done)
sudo tpc -fo 1
sleep 1
sudo tpc -fo 0
 
Last edited:
It worked after grounding pads? With those chips idk use tpc to Max volts and try for 238-244.
 
oops - sorry, posted original content in wrong thread
 
Last edited:
@tear: glad to hear you are continuing the great work on the OCNG overclocking bios for AMD G34 opterons!

<snip>

I'm not sure its really "worth it", it seems easier, and perhaps more "safe" to buy some GTX 780 TIs ...but the point is i just think its MUCH MORE FUN building a 4p machine. I we do it for the fun, right? :)
Indeed :)

pps. Is is generally better (presently) to fold using 6100 series Opterons than 6200 or 6300 series? I have seen people mention that, and i would like to know as im in for purchasing G34 chips soon :)

A bit of background on AMD chips:

All ES AMD G34 chips can be tuned/overclocked today.
OC of all production AMD G34 chips will be possible within a month's time (6100 series
can be overclocked today, OC of 6200/6300 will be possible the moment 6200/6300 OCNG
is released -- http://area51dev.blogspot.com/).

I'd say go for 6300 (either ES -- part numbers ending with G54 -- or production), if you can.

The downside of 6300 chips is they're more expensive to procure than Xeon E5 chips...

That, combined with lower performance/Watt, doesn't make AMD 6300 chips very
desirable (if you care about power usage). At the same time, 6300 series are the most
efficient Opterons one can get and you said you wanted a 4P G34, so... :)

Wrt 6200 chips, you could get ES chips (avoid part numbers ending in *G43, ideally
get *G47 e.g. http://www.ebay.com/itm/271366567099, *G44/*G45 should be fine
in vast majority of configurations).
6200 retail are fundamentally fine, too but are usually more expensive than their ES
counterparts.
6200 when tuned, can be visibly more efficient than 6100 series, too.

6100 series, while not super efficient are easy to obtain so.. there you have it.


HTH!
 
As an eBay Associate, HardForum may earn from qualifying purchases.
Btw are there any good reasons to go with one over the other - of the 3 motherboards you mention? I think i can find the H8QGL-IF+ board somewhat cheaper than the other two.
Any recommendations regarding best 4p G34 motherboard (that works with this OC BIOS of course!) ?
 
Btw are there any good reasons to go with one over the other - of the 3 motherboards you mention? I think i can find the H8QGL-IF+ board somewhat cheaper than the other two.
Any recommendations regarding best 4p G34 motherboard (that works with this OC BIOS of course!) ?
H8QGi/H8QG6 (but not -LN4F variant) and H8QGL are supported by OCNG.
H8QG7 is currently not supported (but may become supported in future).

You may want to compare features of the boards on SM website, the differences are in the
following areas:
- socket layout
- number of DIMM slots
- configuration of PCI-E slots
- presence of HTX slot
- presence of 4-port GigE

H8QGL offers better voltage regulation (than Gi/G6) so if you may end up with
production chips, I'd recommend it == allows production chips to be OC'd higher than
Gi/G6. It's also the chepest board out there.

OTOH, voltage regulation quality argument is pretty much out when you know you'll be
using ES chips as one can always manually compensate for it...

Two things you should know about GL:
(a) socket layout limits the use of enthusiast (high performance) HSFs to musky modded
&#8194;&#8194;&#8194;212+/EVO (at least I can't think of anything else)
(b) few individuals have burnt their GL boards at very high loads (mind they were using ES
&#8194;&#8194;&#8194;chips); I haven't had a problem with my GL (went to 900W on it, which is high but not
&#8194;&#8194;&#8194;extreme)
 
Status
Not open for further replies.
Back
Top