ASUS A8N SLI Deluxe + Athlon64 3000+ Overclocking Woes!

On Wed, 11 May 2005 09:13:16 +0200, David Rasmussen wrote:

I've tried version 1004, 1006 and 1007 (newest). No dice

So if I bought a new system with the same RAM, the same CPU, the same
motherboard, should I expect to be able to overclock more? And if so,
which part of my current system is at fault? And if not, why not? I want
to overclock my CPU, damnit! Everybody else is getting 2.4-2.6 GHz

I'd say the MB is what's holding you back. I don't think you could find a
CPU that wouldn't do at least 9x233. And if you lowered ram to the lowest
speed, it's probably not that.

Here's more info.

http://forums.anandtech.com/messagev...readid=1562941

http://forums.anandtech.com/messagev...readid=1512756

One message from the above link.

Well I used this one
http://forums.anandtech.com/messagev...readid=1497607
"Quick and dirty A64 clocking guide" over the weekend and am running my
A64 3000+ @ 2430 (9x270) with just a little room. (Many thanks to Zebo
for demystifying OC'ing)

Here's what my settings are
BIOS=1002 (my pci lock works as documented)
FSB = 270
CPU Multipier = 9x
HTT = 3x
vCore = 1.50V
DIMM = 2.85V
Memory locked at 333Mhz
CAS = 2.5
RAS to CAS = 3
Row Precharge = 3
Min RAS cycle = 6
So I guess that's 2.5-3-3-6

From this, I'd say you just don't have something right.

--
Abit KT7-Raid (KT133) Tbred B core CPU @2400MHz (24x100FSB)
My server http://wesnewell.no-ip.com/cpu.php
Verizon server http://mysite.verizon.net/res0exft/cpu.htm

Wes Newell wrote:
I'd say the MB is what's holding you back.

Okay.

I don't think you could find a
CPU that wouldn't do at least 9x233.

That's what I gather too, reading around on the net.

And if you lowered ram to the lowest
speed, it's probably not that.


Agreed.

Here's more info.

http://forums.anandtech.com/messagev...readid=1562941

http://forums.anandtech.com/messagev...readid=1512756


I will check those.

One message from the above link.

Well I used this one
http://forums.anandtech.com/messagev...readid=1497607
"Quick and dirty A64 clocking guide" over the weekend and am running my
A64 3000+ @ 2430 (9x270) with just a little room. (Many thanks to Zebo
for demystifying OC'ing)


I know that guide. I followed that guide.

Here's what my settings are
BIOS=1002 (my pci lock works as documented)
FSB = 270
CPU Multipier = 9x
HTT = 3x
vCore = 1.50V
DIMM = 2.85V
Memory locked at 333Mhz
CAS = 2.5
RAS to CAS = 3
Row Precharge = 3
Min RAS cycle = 6
So I guess that's 2.5-3-3-6


Sounds sort of like what everybody else is doing, and what I can't do.

From this, I'd say you just don't have something right.


I promise I have done everything people have told me, and I have
understood it too. Assume that I am right. And that this happened to
you. What would you do? What would you replace? Mobo?

/David

Actually, I am wondering how much software means. I am doing this on a
fresh Windows XP installation, so that shouldn't be the issue in itself.
But when I boot on Knoppix (a live Linux distribution on cd), it seems I
can push LDT/FSB a bit further before getting problems. But still not
much. It stops then at 240 or so. And might still not be totally stable,
I only booted and ran some intensive calculations and a memory stress
test for 2-3 minutes. Still, it didn't croak (before I pushed it further).

Is there a driver issue or something like that here? Is there some piece
of crucial software or driver I have forgotten to install (Nvidia stuff,
nTune, Athlon64 driver (yes it exists), Cool'n'Quiet etc.)?

I also tried disabling all sorts of hardware on the motherboard (sound,
SATA, USB, FireWire etc.). It didn't help.

Assume I am an intelligent user, and that I have read up on most of what
there is to say about A64 OCing, and that I have followed it all to the
letter and even tried some other stuff and experimented, and I still
don't get above 225 MHz. Imagine I am you And you experience this.
What do you do?

/David

In article , David Rasmussen
wrote:


Now, again, assuming that I have done anything right and there is no way
to get above 225 MHz, which part of my system should I replace? If it
was the CPU clock that couldn't go higher than 2000 MHz, I'd say:
replace the CPU. If it was the RAM that couldn't go higher than 200 MHz,
I'd say: replace the RAM. Now, when it's the LDT/FSB that can't go very
far above it's stock 200 MHz, what do I replace? What part of the system
is responsible for the LDT/FSB? The Mobo? The CPU? The RAM?

/David

Looks like it happens to other people also:
http://www.xtremesystems.org/forums/...ad.php?t=52145

The FSB clock is a "processor internal" issue. Which says,
if you didn't know anything else, you would conclude this is
a processor issue.

FSB * LDT multiplier = Hypertransport bus.

Dropping LDT multiplier should make a crappy Hypertransport bus
or a crappy motherboard chip work.

Changing memory divider should correct for a processor memory
controller that is not working well. Even some crappy RAM I had
here, would eventually work if the frequency was dropped low
enough. Memory parameters that might defy adjustment, would be
things like setup and hold time, but then failing to meet those
parameters would result in failure at any frequency (even 1 Hz).

Logic dictates to replace the processor. Does the processor have
a return policy ?

You simply do not have enough evidence to make a call as to
which _single_ component to replace. You can replace a whole
bunch of stuff (like a guy doing work on a contract hardware
support would do - change keyboard and mouse too :-)), but
you'd lose our respect if you did that :-)

Your decision is a time sensitive one, as if the retailer has
a returns period or a returns policy, that may force the issue
for you. You could return the motherboard before the return
period has expired, for example. I doubt the processor has as
generous a return policy as the other components (as people
would abuse a generous return policy, to just search for
good overclocking processors etc).

I cannot help but feel you are missing something, like Vcore,
Vdimm, etc adjustment. There are some motherboards that have
obvious design flaws, and reproducible overclock limits caused
by some architectural decisions. But there are some people out
there who manage to get higher internal clock than you have
managed.

Any chance of testing the processor on another S939 motherboard ?

Paul

"David Rasmussen" wrote in message
. ..
Pete M Williams wrote:

NOW, DAVID!

A few questions that need answering.

Can I ask what ratio's you have tried running your memory at?


All ratios below the stock speed. And the stock speed of course.


I will ask again. What ratio's did you try?
Sorry but I do not think you have grasped this yet.



You do know what the different DDR settings in the BIOS of the A8N
represent with respect to dividers don't you?


I'm not sure what you mean here. I understand the divider. I don't fully
understand what "CAS Latency" is but I know where it and other similar
settings should be set according to the manufacturer of my memory. I also
tried a lot of different settings. Very conservative and slow settings.
Experimenting with 1T and 2T etc.
Didn't help.


RAM setting (DDR266, DDR333, DDR400, DDR466, etc.) uses a ratio to take the
FSB/HTT speed and multiply it by a fraction to set the RAM speed.
Those settings are as follows:
DDR 400 (PC3200 RAM):1/1 ratio
DDR 333 (PC2700 RAM): 5/6 ratio
DDR 266 (PC 2100 RAM): 2/3 ratio
So if you are setting the BIOS to DDR 400 or and maybe Auto too you are
running your RAM at 1:1 meaning that your "crappy" RAM is not able to keep
up!
With your RAM you will have to lower the setting to DDR 333 or lower or you
will be overclocking the RAM by to much.

Paul wrote:

I cannot help but feel you are missing something, like Vcore,

Nope. Tried from 1.30V-1.55V and even more I think.

Vdimm,

Tried from 2.4V-2.9V. And again, I am not trying to overclock the RAM.

etc adjustment. There are some motherboards that have
obvious design flaws, and reproducible overclock limits caused
by some architectural decisions. But there are some people out
there who manage to get higher internal clock than you have
managed.


Most people, if not all, that have OCed with an ASUS A8N SLI Deluxe seem
to get much higher LDT than me.

Any chance of testing the processor on another S939 motherboard ?


Hmm. Not easily. But of course it is a good suggestion to test every
single component in another system.

/David

Pete M Williams wrote:

I will ask again. What ratio's did you try?
Sorry but I do not think you have grasped this yet.


Why do you think that?


RAM setting (DDR266, DDR333, DDR400, DDR466, etc.) uses a ratio to take the
FSB/HTT speed and multiply it by a fraction to set the RAM speed.

I know. Have I said otherwise?

Those settings are as follows:
DDR 400 (PC3200 RAM):1/1 ratio
DDR 333 (PC2700 RAM): 5/6 ratio
DDR 266 (PC 2100 RAM): 2/3 ratio

I know.

So if you are setting the BIOS to DDR 400 or and maybe Auto too you are
running your RAM at 1:1 meaning that your "crappy" RAM is not able to keep
up!

I know.

With your RAM you will have to lower the setting to DDR 333 or lower or you
will be overclocking the RAM by to much.


I know.

As stated: I have tried all possible ratios/dividers below stock speed.
In other words, I tried DDR333 (5/6), DDR266 (2/3) and DDR200(1/2) and
all other options below DDR400 (1/1), if any. Just to show you that I
understand it (which you seem to doubt for some reason), I tried the
following:

LDT at, say, 240 (yeah, I tried that, but it is just an example) instead
of 200 MHz.
HT multiplier at 4 (HTT speed 4*240 MHz = 960 MHz instead of 1000 Mhz)
CPU multiplier at 7 (speed: 7*240 MHz = 1680 MHz instead of 1800 MHz)
RAM at DDR266 (2/3), so instead of running DDR400 (200 MHz), it runs
240*2/3 = 160 MHz.

And, as I said, I verified all these numbers in CPUZ, ClockGen NF4 and
similar programs.

At least I am getting verified that my case is really an odd one.
Otherwise, people wouldn't be so insistant in their "Are you sure you
understand this? Are you sure you did it right?" etc.

/David

On Wed, 11 May 2005 19:28:14 +0200, David Rasmussen wrote:

don't get above 225 MHz. Imagine I am you And you experience this.
What do you do?

Well, I don't run windows for one thing.:-) But when I'm overclocking I
always boot with a memtest cd and give it a good test before booting from
the HD. Since you say you can get 240 running Knoppix, then maybe it's a
driver problem in WXP.

--
Abit KT7-Raid (KT133) Tbred B core CPU @2400MHz (24x100FSB)
My server http://wesnewell.no-ip.com/cpu.php
Verizon server http://mysite.verizon.net/res0exft/cpu.htm

Wes Newell wrote:

Well, I don't run windows for one thing.:-)

Heheh... Neither do I sometimes. But I want Windows to work on this
particular machine.

But when I'm overclocking I
always boot with a memtest cd and give it a good test before booting from
the HD.

I did that too.

Since you say you can get 240 running Knoppix, then maybe it's a
driver problem in WXP.


Maybe, but 240 isn't high either. Not high enough 9*240 MHz = 2160
MHz. Hardly impressive when everybody else is getting 2400 Mhz and
similar. Also, the 240 Mhz wasn't entirely stable in Linux.

What would you do in my case? Replace the CPU?

/David

On Wed, 11 May 2005 14:52:33 -0400, Paul wrote:

FSB * LDT multiplier = Hypertransport bus.

Please refrain form using the defunct term LDT. That name/terminology
(Lightining Data Transport) died with the new name of HyperTransport and
I'm sure it's going to confuse the hell out of some people. There is no
LDT bus in the A64 system.

--
Abit KT7-Raid (KT133) Tbred B core CPU @2400MHz (24x100FSB)
My server http://wesnewell.no-ip.com/cpu.php
Verizon server http://mysite.verizon.net/res0exft/cpu.htm