A8N32: SB>NB Overclock "Auto"

Cisco

n00b
Joined
Nov 7, 2005
Messages
7
In the jumperfree page, right below the FSB setting is the SB>NB frequency setting. That together with the SB>NB multiplier in the Chipset page control the SB>NB hypertransport--just like the FSB and K8>NB multi control the standard hypertransport in this, and all other nf4 motherboards.

But just what the heck does "Auto" mean there when overclocking the FSB? As far as I can see it can mean one of two things: 1) it matches the FSB frequency in which case the SB>NB multi should be set the same as the K8>NB multi; or 2) it keeps the SB>NB frequency at 200 regardless of what you do with the FSB. In this 2nd case the SB>NB multi should be kept at 5X... I guess. This is based on the assumption that this 2nd hypertransport also wants to be at 2000 or below for stability... which is just an educated guess.


Is there a program out there that will read both hypertransport frequencies separately so we can figure out just what auto does?

Since this setting is unique to this MB, or rather, to SLI-16X motherboards, there isn't a lot of info about this 2nd hypertransport anywhere. I see a lot of trial and error going on with SB>NB settings--both the frequency and the multi setting--but I don't think any of us know for sure yet what the optimal settings are here, specialy when OC'ing the FSB. I tend to match the SB>NB to whatever I do to the FSB and it seems to work well that way but don't ask me why lol. Others claim that locking it at 200 with the multi at 5 works best.

Neither the chipset developers (nVidia) nor ASUS are saying anything about this that I can find.
 
the simple answer is to NOT use auto.. just set both to (4X) or (800) for FSB of 201-250, (3X) or (600) for FSB of 251-333.
 
Yes. That is simple. So... I gather you also don't know? :)

If anyone knows--or has a link to somewhere where this is being discussed--whether it is best to...

1. leave SB>NB at 200 X 5 or
2. match SB>NB to whatever the K8>NB is set to (e.g 250 X 4, 320 X 3,...)

As I said, this SB>NB secondary HTT is a new thing in SLI-X16 boards. Absent any technical explanations from the developers, I'm still going on the assumption that both HTTs should have identical frequencies (i.e. option #2 above) but I could be totaly wrong.

In case I wasn't clear in my first post, the reason I would like to know what "Auto" is doing is that this could give us a clue as to what ASUS thinks should happen to the SB>NB HTT when overclocking K8>NB HTT.

Anyone?
 
I'd also be interested in a decent answer to this question. At the minute I have mine set at 260 x 4 for the K8 and 201 x 5 for the SB. Changing either of these multipliers stops me booting at all.

I'm in a frustrating place with this OC at the moment. Upping the FSB to 270 (leaving everything else the same) yields a stable 12 hours on prime95. But then the next time I switch my PC off it won't get in to windows until I drop it back to 260.

Can't seem to get POST above 270 under any circumstances, no matter which other settings I change up or down :(

Was kinda looking for the answer to the above question to suggest my next step :D

On a side note, I've got clock gen showing me what the CPU, FSB, RAM, PCI-E and PCI speeds are, and my FSB keeps fluctuating by 0.01 up and down (which in turn fluctuaes the RAM and CPU speed). Is this normal?
 
I was able to get 346MHz FSB out of the ASUS A8N32-SLI Deluxe I tested. Not that they will all do this, but I would think some other component is holding you back.

What is your complete system configuration?
 
Not meaning to hijack the OP's thread, so sorry if you think I am Cisco :) (if you do, lemme know and I'll ask a mod to split the thread ;) )

System = Opteron 144, A8n32-SLI, 1GB crucial Ballistix tracer DDR500, PoV 7800GT

OK, from the top, as of this very minute - stable:
______________________
Jumper free configuration:
-AI overclocking [Manual]
-CPU FSB Frequency [265]
SB to NB OverClock [Manual]
SB to NB Frequency [201]
Adjust PCIE Frequency [101]
PEG Link mode [Auto]
~~~
All overvoltage options enabled
~~~
DDR Vcore [2.80V]
FID/VID change [Manual]
Processor multiplier [9x]
Processor voltage [1.550V]
DDR clock skew [Auto]
______________________
CPU config
GART error reposting [disabled]
MTRR mapping [continuous]
1T/2T memory timing [Auto]
>>Memoy settings:
Memclock Mode [Limit]
Memclock Value [183MHz]
MCT timing mode [Manual]
CL [2.5]
TRAS [10]
TRP [4]
TRCD [4]
~~~~~~~~
Everything else in this section is [Auto]
~~~~~~~~
______________________
Chipset
K8 to NB Frequency [4x]
K8 to NB Linkwidth [Auto]
SB to NB Frequency [5x]
SB to NB Linkwidth [16 - 16]
______________________

Gives me: CPU @ 2385MHz, RAM @ 238.5 (timings in BIOS are the rated timings), PCIE @ 100, PCI @ 33 according to clock gen.

What do you reckon?
 
Forthy said:
Not meaning to hijack the OP's thread, so sorry if you think I am Cisco :) (if you do, lemme know and I'll ask a mod to split the thread ;) )

No, not at all--all of this is interesting stuff. It's also interesting that having the 2 HTTs out of sync with each other works better for you than having them both the same.

You're also going over the conventional HTT by 120 (265 X 4 X2 = 2120) and it seems to be working for you also.

I assume that when you said you tried and failed to cold-boot @ 270 you also tried changing the K8>NB to the safer 3X? If not, it's definitely worth a shot to bring the HTT back down below 1000 (2000) and see if that does the trick.
 
Yeah 270 x 4 works, as does 270 x 3.

Until I cold boot :(

Still playing though, 'cos I can't remember wht I've tried and what I haven't :D

[edit] Can you list your settings so I can compare?
 
Forthy said:
Yeah 270 x 4 works, as does 270 x 3.

Well, I say that, but now it won't boot at 270 x 3 :confused:. 270 x 4 has booted and is currently prime95'ing, but I can't see any stability issues at the minute.....

Pffft, this thing is complicated.

 
OK, so last night it was stable for a few hours at 270 x 4.

This morning, it gets in to windows for about 10 seconds before it locks up. :mad:

Back to 260 we go....

This cycle is geting annoying :(
 
Ok, so here are my test results using 3dmark06

Build:
Opteron 165 CPU Multiplier: 9x
A8N32-SLI Deluxe
2GB RAM
7900GT @ 625 core / 1750 mem

Test 1:
FSB/HTT: 270 x3 (810)
SB to NB: 270x3 (810)


-------------------------------
Test2:
FSB/HTT: 270x3 (810)
SB to NB 250x4 (1000)



-------------------------------
Test3:
FSB/HTT: 270x3 (810)
SB to NB: 200x5 (1000)


So, in conclusion, it looks like it really doesn't matter what the SB to NB speeds are, they're going to be fairly comparable.
 
Back
Top