jay_567 said:
I'm trying to find the ideal cpu voltage for the A64 3200+(939)/venice
core so I can manually set it my A8N-E BIOS. I haven't bought the cpu
yet, but heard the voltage isn't printed on venice core processors.
AMD's website says the voltage is "variable" for that cpu. While the
voltage settings are Winchester (1.40) and Newcastle (1.50). I'm a
newbie at building up systems. All I'm looking for is overall stability.
Can anyone tell me what voltage setting I should use for this cpu?
My recommendation would be to set it to "Auto", boot the system,
and use Windows utilities, to see what it is doing. Either observe
the Vcore with Asus Probe (with Cool N' Quiet not installed, or
perhaps disabled in the BIOS) or perhaps use utilities like CPUZ,
clockgen, and the like, to see what those utilities feel the VID is
set to.
Looking at the AMD web site for tech docs is a real eye opener. It
looks like they haven't updated anything for quite a while, as the
tech docs section is full of 2004 dated documents. It would appear
AMD is determined to leave Venice/SanDiego undocumented. I did find
this pseudo-documentation, which you've probably already found:
http://www.amd.com/us-en/assets/content_type/DownloadableAssets/K8_Electrical_Spec_Rev_ENG.pdf
Looking in the 31411 S939 functional spec, it says VID voltage is
determined by two things. There is a "metal mask" value used before
PWR_OK. A "value programmed during device manufacturing" comes out
on the VID pins, once PWR_OK is sent to the processor. In the above
document, where it says a device has a "variable" VID, this is more
or less the same practice as at Intel - devices are individually tested
on the production line, a "good operating value" for the device is
established, and that is programmed into the device. The ones with
extremely low voltage values might become DTR or mobile chips or the
like. (The chips are sorted, but in this case, are not tracked as
closely as you, the customer, would like.) The "losers" become desktop
chips. It could be that you would find 1.4 or 1.45 or 1.5V programmed
into the device, meaning the value was determined on the production
line, but they could not be bothered sorting and labelling boxes for
that express purpose. The same thing that Intel does for their
"unstated VID" devices. That allows the fab to send barrels of chips
to a packaging area or plant, without having to keep individual
statistics on the chips. (I.e. ADA3000DAA4BP is printed on the lid
of each 3000+ Venice, but nobody keeps track of the 1.4, 1.45, 1.5V
part of the deal. One of the letters in the part number is coded
to indicate a "variable" voltage, meaning they aren't keeping
track.)
I think going the "Auto" route, you'll figure out some way of
reading back the "programmed" value, as long as Cool N' Quiet is
disabled. When CNQ is running, the value seen will depend on the
P-state at that instant in time, so if CNQ is running, run a
copy of Prime95 to get the VID to go to the max P-state value.
If you do find out the value, report back with what you find.
According to this site, the VID is in the range of 1.35V to 1.4V.
http://fab51.com/cpu/guide/opn-64.html
The five bit VID code goes from 0.800 to 1.550 in steps of 0.025.
See page 267 of this doc. AFAIK this table is the right one.
http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/26094.PDF
Paul
