Supermicro H8QGi/6 and H8QGL Next Generation OC BIOS

Status
Not open for further replies.
dwdawg said:
Tear, I have a question
Node 0 0000 0000 0000 0000 001c 0000 0000 0000
Node 1 0000 0000 0000 0000 0000 0000 0000 0000
Node 2 0000 0000 0000 0000 0000 0000 0000 0000
Node 3 0000 0000 0000 0000 0000 0000 0000 0000
Node 4 0000 0000 0000 0000 0000 0000 0000 0000
Node 5 0000 51a2 0000 0000 0000 0000 0000 0000
Node 6 0000 0000 3928 0000 0000 0000 0000 0000
Node 7 0000 0000 0000 0000 0000 0000 0000 0000

As you can see, I have three positions with retry numbers. It's always the same three. And this is at a pretty low 10% OC.

Do they correspond to individual memory sticks?
I ordered three new sticks based on the theory that they do.
How would I go about mapping the above output to individual slots?
Click to expand...

That is a lot of retries. Something seems awry.
 
Posidon42 said:
That is a lot of retries. Something seems awry.
Click to expand...
Yeah... Node 5/Link 1/Sublink 0 and Node 6/Link 2/Sublink 0 have one thing in common;
they both connect to CPU1 (just making an observation).

In any way, dwdawg, you should first check how quickly they're rising. Run the script,
wait 5-10 seconds, then re-run it. Post results afterwards.

I've got few ideas but let's see what we're dealing with...
 
tear, I totally agree, it must have been a WU fluke.
But forget about that, I've got good news: successfully booted with 233 16!!! Woohooo! :D

Isn't it something that 232 didn't post but 233 did? I had similar things happening with Intel systems but I wasn't very hopeful to be honest. This is great! It's possible I can push further :)
Running fah now, let's see how it goes till morning.
 
I'd probably ask you whether 233 boots reliably... (power-cycle ten times -- will it POST 10 out of 10?).

Whatever it is, it has to do with HT, it's been the nemesis of this endeavor...

HT issues are on the investigation list so they will (hopefully) be dealt with (tentative ETA is few weeks from now).
 
You may be onto something here, mister. My non-232-booting config does boot at 233 :eek:
I'm sensing productive weekend ahead...

And by the way, F9 indeed loads optimal defaults. Please pardon my distrust :)
 
Last edited:
Just wanted to say thanks again tear for all you are doing, this is some amazing stuff
 
Oh well, extra memory never hurts :)
Think I need to reseat CPU 1?

Node 0 0000 0000 0000 0000 001c 0000 0000 0000
Node 1 0000 0000 0000 0000 0000 0000 0000 0000
Node 2 0000 0000 0000 0000 0000 0000 0000 0000
Node 3 0000 0000 0000 0000 0000 0000 0000 0000
Node 4 0000 0000 0000 0000 0000 0000 0000 0000
Node 5 0000 5a59 0000 0000 0000 0000 0000 0000
Node 6 0000 0000 32ff 0000 0000 0000 0000 0000
Node 7 0000 0000 0000 0000 0000 0000 0000 0000

Node 0 0000 0000 0000 0000 001c 0000 0000 0000
Node 1 0000 0000 0000 0000 0000 0000 0000 0000
Node 2 0000 0000 0000 0000 0000 0000 0000 0000
Node 3 0000 0000 0000 0000 0000 0000 0000 0000
Node 4 0000 0000 0000 0000 0000 0000 0000 0000
Node 5 0000 5a5b 0000 0000 0000 0000 0000 0000
Node 6 0000 0000 338c 0000 0000 0000 0000 0000
Node 7 0000 0000 0000 0000 0000 0000 0000 0000
 
It probably wouldn't be a bad idea to reseat CPU1. Make sure the contact on the bottom of the chip are clean before you reseat it.
 
tear said:
You may be onto something here, mister. My non-232-booting config does boot at 233 :eek:
I'm sensing productive weekend ahead...

And by the way, F9 indeed loads optimal defaults. Please pardon my distrust :)
Click to expand...

Haha, I'm glad I tapped into something new! :D
My 6166s folded just fine through the night with 233 16. Stock voltages.

The thing is the same thing happened with 231 16: the first time it posted (after a power-cycle) and I started the client, the TPF was abnormally high at 16:37. Then, I power-cycled again and the second time around the TPF went down to an expected level at 14:52 (although it is a bit higher than what I saw with 231, I did not test it on the same frames, no direct comparison is not valid).

So, what I'm saying is, with higher refclocks (>230) somehow a second power-cycle is helping. I don't know why, but this has been my experience with 231 and 233.

I'm going to power-cycle it a few more times for ya after this WU finishes ;)

ht-retries has been just zeroes all along. From what I understand ht-retries is giving you the cumulative number of retries, right? So, if I see zeroes now, it means there were no retries over the night, correct?
 
theGryphon said:
ht-retries has been just zeroes all along. From what I understand ht-retries is giving you the cumulative number of retries, right? So, if I see zeroes now, it means there were no retries over the night, correct?
Click to expand...

That is correct. The retries are cumulative since power on. If you reboot the system, the counter will not reset. It is only after a power cycle (power off, power on) that the counters are reset.
 
theGryphon said:
Haha, I'm glad I tapped into something new! :D
My 6166s folded just fine through the night with 233 16. Stock voltages.

The thing is the same thing happened with 231 16: the first time it posted (after a power-cycle) and I started the client, the TPF was abnormally high at 16:37. Then, I power-cycled again and the second time around the TPF went down to an expected level at 14:52 (although it is a bit higher than what I saw with 231, I did not test it on the same frames, no direct comparison is not valid).

So, what I'm saying is, with higher refclocks (>230) somehow a second power-cycle is helping. I don't know why, but this has been my experience with 231 and 233.

I'm going to power-cycle it a few more times for ya after this WU finishes ;)

ht-retries has been just zeroes all along. From what I understand ht-retries is giving you the cumulative number of retries, right? So, if I see zeroes now, it means there were no retries over the night, correct?
Click to expand...

I have observed the same behavior as far as reboots / Power Cycles go. I have noticed that If at higher settings say 237 on 6174’s ,or so If I power cycle the rig and check retries after it comes back up if there are any the TPF will be slower. If I repower cycle until there are none and sooner or latter there will be none then the TPF drops and the rig will run with no problems. Also over time and with each WU the TPF appears to increase until I do a power cycle then it will drop again. (Possible Memory Issue?) These are just observations in no way have they been controlled test and could be purely coincidence but might be worth others watching and seeing if they are seeing the same or similar behavior.
 
Thanks firedfly!
Interesting Grandpa. I don't like the idea of losing TPF over time and I'll sure to be checking on that.

tear, I'm sorry, I won't be able to power-cycle with 233 anymore. How about I give you 235? :p
Yep, just posted 235 16, and folding away :D
 
theGryphon said:
....... with 233 16. Stock voltages.
Click to expand...

About to finish up my 6166HE 4P install and saw this post.
I remember reading something about the second parameter in the command, in this case the "16", what does that represent?
And I previously saw a mention that 6166HEs can mod the voltage. What is the command for controlling that if I may ask?
Thanks.
 
Core32 said:
About to finish up my 6166HE 4P install and saw this post.
I remember reading something about the second parameter in the command, in this case the "16", what does that represent?
And I previously saw a mention that 6166HEs can mod the voltage. What is the command for controlling that if I may ask?
Thanks.
Click to expand...

I'm just a student here really, but I can tell you that 16 controls the memory base timings, and it should be used only if your memory has XMP 1600 profile. Even so, it is not supported (tear's instructions), and it's very much experimental from what I gather. I'm taking the risk :)

The voltage control is done using TPC. There are three commands to issue in a row, but I'm such a noob in this, I'll let tear or another experienced user guide you.
 
musky said:
It probably wouldn't be a bad idea to reseat CPU1. Make sure the contact on the bottom of the chip are clean before you reseat it.
Click to expand...
Yeah... but... but... I meant to do couple more things before resorting to that.
As it's critical that only one thing is changed at a time -- have you reseated it yet, dawg? :D
 
theGryphon said:
The thing is the same thing happened with 231 16: the first time it posted (after a power-cycle) and I started the client, the TPF was abnormally high at 16:37. Then, I power-cycled again and the second time around the TPF went down to an expected level at 14:52 (although it is a bit higher than what I saw with 231, I did not test it on the same frames, no direct comparison is not valid).

So, what I'm saying is, with higher refclocks (>230) somehow a second power-cycle is helping. I don't know why, but this has been my experience with 231 and 233.

I'm going to power-cycle it a few more times for ya after this WU finishes ;)
Click to expand...
I strongly suggest that you back up one WU (along with all client files, ofc) and keep it
for benchmarking (make sure to enable "clock frequently has errors" option so that
passed deadline won't cause the client to delete the WU).

What you should also be doing is checking applied frequency after each boot
(while experimenting) to either confirm of disprove frequency misapplication theory.

And, for symmetry, you should also start the bench WU several times (from scratch)
"within" single boot ("inverse" of what you've tried so far).

theGryphon said:
ht-retries has been just zeroes all along. From what I understand ht-retries is giving you the cumulative number of retries, right? So, if I see zeroes now, it means there were no retries over the night, correct?
Click to expand...
That is correct.
 
theGryphon said:
I'm just a student here really, but I can tell you that 16 controls the memory base timings, and it should be used only if your memory has XMP 1600 profile. Even so, it is not supported (tear's instructions), and it's very much experimental from what I gather. I'm taking the risk :)

The voltage control is done using TPC. There are three commands to issue in a row, but I'm such a noob in this, I'll let tear or another experienced user guide you.
Click to expand...
Yes, well. I don't mind documenting some of the flags but I'm not sure whether
this is the time or the place, to be honest. I don't want to contribute to existing confusion;
will have to consult the elders on this one.

Neither I want this thread to become TPC support forum.
Check this post at AMDzone for instructions that should work
with most recent TPC version. For further inquiries please use AMDzone forum
(TPC has its thread there somewhere) or post in a new thread :D
 
That frequency misapplication thing was not a theory, or a claim, was just a thought and I don't think that's the case actually. Frequency has been reported correctly after every power-cycle, so I don't think there's any problem with the script or the BIOS. I think it has to do with how stressing the higher refclocks are for the system. Is there any other way to diagnose this other than the ht retries?

By the way, interestingly, I did not have any problem with 235. It's now folding at 14:32, which is around what I would expect (not a benchmark to be compared to previous numbers but it's not abnormal like 16:37, etc.).

I'm very happy with how it's going. ht-retries is reporting full zeroes. Stock voltages. Temps 43-47C. :D

I'm starting to think that my board is liking odd refclocks. It denied 232 but worked fine with 231, 233 and 235.
 
theGryphon said:
Frequency has been reported correctly after every power-cycle (...)
Click to expand...
Good. Then that's off the table. Per explanation in my post you should exclude GROMACS fluke
as well...

theGryphon said:
I'm starting to think that my board is liking odd refclocks. It denied 232 but worked fine with 231, 233 and 235.
Click to expand...
Nah. Anything 233+ would probably work. I'm looking into this.
 
Nope. I was waiting on the word from master Yoda :D
I figure the low overclock and frequency of retries might give you a good testbed to help figure out the other problems.

Your humble (my wife disagrees with this) servant awaits.


tear said:
Yeah... but... but... I meant to do couple more things before resorting to that.
As it's critical that only one thing is changed at a time -- have you reseated it yet, dawg? :D
Click to expand...
 
theGryphon said:
Isn't it something that 232 didn't post but 233 did? I had similar things happening with Intel systems but I wasn't very hopeful to be honest. This is great! It's possible I can push further :)
Click to expand...
The spirits hath spoken! They're telling me... they're telling me 209 will not boot for you
either:eek:
 
dwdawg said:
Nope. I was waiting on the word from master Yoda :D
I figure the low overclock and frequency of retries might give you a good testbed to help figure out the other problems.
Click to expand...
Most excellent! Can you hop on IRC one evening? That would expedite whole process.

dwdawg said:
(...) (my wife disagrees with this) (...)
Click to expand...
Certain disputes are meant not to ever take place... :D
 
241 is a go. 242 is a no-go.

List of frequencies to avoid (until new version comes out): 209, 232, 242, 262.
 
tear said:
241 is a go. 242 is a no-go.

List of frequencies to avoid (until new version comes out): 209, 232, 242, 262.
Click to expand...

Wow, so you weren't kidding??! I really thought you were :eek:
How did you find out? I mean, how does it work?
 
I can. Tonight happens to be a good night.
If I miss you, I'll keep trying.

tear said:
Most excellent! Can you hop on IRC one evening? That would expedite whole process.


Certain disputes are meant not to ever take place... :D
Click to expand...

Yeah? tell HER that! :p
 
theGryphon said:
Wow, so you weren't kidding??! I really thought you were :eek:
How did you find out? I mean, how does it work?
Click to expand...

Tear has an endless supply of lab rats for testing BCLK. PETA should be knocking on the door anytime soon.
 
musky said:
FAQ

Q: How can I view my current CPU frequency?
A: NOTE: this works in Ubuntu 10.10 - it may not work in newer versions of Ubuntu - please let me know your results. If you have set your clocksource=hpet (see previous post), you can see the current CPU frequency by parsing it out of dmesg:
Code: 
dmesg | grep -o Detected.*
Detected 2749.890 MHz processor.
Click to expand...

Well I'm using 11.04 and I hope this is returning the nb speed, 6176 at 200 this says 1800mhz, all the way up to 261 it reported 2349mhz (multiples of 9x ref clock) I'm still just playing around, had a no boot at 262.... Seems my cheap cl7 1066 that did 1333 on stock bios only did 1066 at 200 and then down clocked to 800 for anything above that. Went back to my gskill eco's and they do 1333 at 200 and 1066 everything above that.
 
First, thank you tear for everything!

Second, I noticed you guys are all running MC's, lots of 6174's etc. I am building a new 4p rig. I just want to double check before I buy the chips, that even with this bios the IL 6274's should be avoided and won't overclock like the MC 6174's?

Mike
 
I was just thinking.... has anyone been keeping track of what BCLK values work for which CPU/memory combos? I think this would be very useful info.

Keith
 
ctrlbrk said:
First, thank you tear for everything!

Second, I noticed you guys are all running MC's, lots of 6174's etc. I am building a new 4p rig. I just want to double check before I buy the chips, that even with this bios the IL 6274's should be avoided and won't overclock like the MC 6174's?

Mike
Click to expand...

From the FAQ in post 3:

Q: When will this have IL support?
A: "When AMD makes a Family 15h CPU that doesn't suck" - tear

So yes, avoid IL chips.
 
ctrlbrk said:
First, thank you tear for everything!

Second, I noticed you guys are all running MC's, lots of 6174's etc. I am building a new 4p rig. I just want to double check before I buy the chips, that even with this bios the IL 6274's should be avoided and won't overclock like the MC 6174's?

Mike
Click to expand...

The fact that almost no one on this team is running 62xx on this team, even before OC was an option should be a strong answer. Now with OCing as an option it isn't even realy an option.

61xx chips use less power, are better at folding, are cheaper when purchased used, and now can be OC-ed.

62xx chips don't have much going for them in the folding world.
 
musky said:
From the FAQ in post 3:

Q: When will this have IL support?
A: "When AMD makes a Family 15h CPU that doesn't suck" - tear

So yes, avoid IL chips.
Click to expand...

Thanks, sorry I missed that -- I thought I read the entire thread twice :)

Kendrak said:
The fact that almost no one on this team is running 62xx on this team, even before OC was an option should be a strong answer. Now with OCing as an option it isn't even realy an option.

61xx chips use less power, are better at folding, are cheaper when purchased used, and now can be OC-ed.

62xx chips don't have much going for them in the folding world.
Click to expand...

Yes, I figured. But wanted to make sure. The 6274's are actually cheaper than the 6174's on ebay right now!

Appreciate the quick responses guys.
 
402blownstroker said:
I was just thinking.... has anyone been keeping track of what BCLK values work for which CPU/memory combos? I think this would be very useful info.

Keith
Click to expand...


+1

And is it useful to buy ddr3 1600 mhz instead of ddr3 1333 mhz ?

Can we hope better performance ?
 
Jeanjean said:
+1

And is it useful to buy ddr3 1600 mhz instead of ddr3 1333 mhz ?

Can we hope better performance ?
Click to expand...

I'll repost this here also:

The first thing you need to make sure of is that the memory you buy is 1.5V or less. Much of the performance memory on the market is 1.65V, which you do not want.

The CAS ratings are in reference to the rated memory speed. While not quite true, a good rule of thumb is 1333 C7 = 1600 C8 = 1866 C9. You usually end up paying more of a premium for higher frequency memory, which is why so many are running the G.Skill 1333 C7 Ripjaws at $60 for 4 x 2GB. For this application, 1333 C7 is faster than 1600 C9 (which is a common frequency/CAS combination for 1.5V memory.) If money is no object, you can get the Eco series at 1600 C7 (1.35V) or the Ripjaw series at 1600 C6 (1.5V.) Either are faster than the 1333 C7 stuff, and both will cost you $100 or more for 4 x 2GB.

References:
Ripjaw 1333 C7 - http://www.newegg.com/Product/Product.aspx?Item=N82E16820231492
Eco 1600 C7 - http://www.newegg.com/Product/Product.aspx?Item=N82E16820231330
Ripjaw 1600 C6 - http://www.newegg.com/Product/Product.aspx?Item=N82E16820231444
Click to expand...

To answer what i think you are actually asking though, you aren't going to be able to run much over 1333 for memory speed with MC chips. The chips don't support it.
 
402blownstroker said:
I was just thinking.... has anyone been keeping track of what BCLK values work for which CPU/memory combos? I think this would be very useful info.

Keith
Click to expand...

I'm not sure if RAM plays any role in "what BCLK values work" but I'm running these Crucial sticks (http://www.newegg.com/Product/Product.aspx?Item=N82E16820148488) successfully so far. They're 1600 8-8-8, but have a 1333 7-7-7 XMP profile too. They're low profile and that's why I got them over the GSkills (I needed something to slide under my Noctua 92mm's).

My rig folded nicely through the night at 241. No ht retries, temps are 43-47C. Average TPF is 14:10 on 6903. That's almost 500K on 6166HEs :D :knocks on wood:

I'm not done searching for the upper limit yet, but this is just awesome already. Thank you again tear and all with a hand in this!
 
Status
Not open for further replies.
Back
Top