Amazing power characteristics for A64 "Venice"!

All about them.

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

Post Reply
Mikael
Posts: 206
Joined: Mon Dec 06, 2004 3:12 am
Location: Gothenburg, Sweden

Amazing power characteristics for A64 "Venice"!

Post by Mikael » Mon May 02, 2005 10:35 am

Look at this very comprehensive test done by Lost Circuits:

http://www.lostcircuits.com/cpu/amd_venice/

They measure the 2.4GHz "Venice" to 30W power consumption during load (without taking into account motherboard power circuitry efficiency)! That's bloody amazing! The test also shows the 2.4GHz Venice to consume the same amount of power as a 2.2GHz Winchester. In other words, Venice is actually even more efficient than Winchester!

Looking at these tests, one can draw the conclusion that a 2.2GHz Venice core A64 undervolted to about 1.2V probably doesn't output more than approximately 20W. Now, that's just great! :D

Anyway, let's hear your opinions on this!

Green Shoes
*Lifetime Patron*
Posts: 477
Joined: Thu Feb 17, 2005 6:41 am
Location: Nashville, TN

Post by Green Shoes » Mon May 02, 2005 10:51 am


StarfishChris
Posts: 968
Joined: Fri Jan 07, 2005 7:13 pm
Location: Bristol, UK
Contact:

Post by StarfishChris » Mon May 02, 2005 10:52 am

You were beaten to it - though whether this forum is more appropriate or not, I'm not sure.

Edit: looks like I was too! :O

Mikael
Posts: 206
Joined: Mon Dec 06, 2004 3:12 am
Location: Gothenburg, Sweden

Post by Mikael » Mon May 02, 2005 11:20 am

Ohh, well, might as well have one in the right forum. :) As a "CPU guy" I would probably have missed that other thread. Well, apparently I did. Duh! :P

NeilBlanchard
Moderator
Posts: 7681
Joined: Mon Dec 09, 2002 7:11 pm
Location: Maynard, MA, Eaarth
Contact:

Post by NeilBlanchard » Mon May 02, 2005 3:28 pm

Hello:

I linked my thread over here -- it is in a better place...

The Venice is looking mighty tempting -- especially if they address some of the memory controller issues.

BTW, I tend to trust Lost Circuits more than I do XBit, FWIW. Since all the CPU's were tested in the same way, then their relative power consumption would seem to be accurate, at least.

Green Shoes
*Lifetime Patron*
Posts: 477
Joined: Thu Feb 17, 2005 6:41 am
Location: Nashville, TN

Post by Green Shoes » Mon May 02, 2005 3:32 pm

NeilBlanchard wrote:BTW, I tend to trust Lost Circuits more than I do XBit, FWIW.
Judging from the response your post on this got, you might be the only one who does. They did mention their outdated equipment in the article, though, so I'm not sure why anyone is acting like that's news.

Not that it matters to me. I'm building at the end of the month and it's gonna be a San Diego all the way :mrgreen:

Wraith
Posts: 139
Joined: Sat Dec 27, 2003 1:57 pm
Location: Minnesota

Post by Wraith » Mon May 02, 2005 6:29 pm

Windows XP has power savings features built in, similar to the CPU-IDLE and CPU-COOL features known from the Windows 9X environment as part of the ACPI feature set. Cool'N Quiet was disabled in the BIOS to disallow change in core frequency.
I wasn't quite sure what power-saving features of XP they're referring to. Do they mean it's just running at 0% CPU usage? They're claiming the Venice and Winchester are 8-10W on idle, without using Cool & Quiet, so at stock voltage and clock? :|

yeha
Posts: 292
Joined: Thu Jan 13, 2005 7:54 pm

Post by yeha » Mon May 02, 2005 11:02 pm

well cool'n'quiet consists of powernow transitions to lower multipliers/voltages, whereas typical operating system idle loops consist of issuing the hlt instruction to instruct the cpu that nothing important is happening. cool'n'quiet transitions and hlt instructions do different things, and while both will reduce power consumption by some degree, using both in concert will provide the best power savings.

jojo4u
Posts: 806
Joined: Sat Dec 14, 2002 7:00 am
Location: Germany

Post by jojo4u » Tue May 03, 2005 12:28 am

lostcircuits wrote:Windows XP has power savings features built in, similar to the CPU-IDLE and CPU-COOL features known from the Windows 9X environment as part of the ACPI feature set. Cool'N Quiet was disabled in the BIOS to disallow change in core frequency.
A quote from my "software cooling" thread:
jojo4u wrote: When there is nothing to to, the operating system has the possibility to send three different signals to the CPU:
no operation, HLT, STPGNT.
Windows 9x versions without ACPI installed use "no operation".
Windows 2000 and XP without ACPI installed use HLT.
All Windows' with ACPI installed use ether STPGNT or HLT.
Linux without ACPI kernel support use HLT, with ACPI kernel support they use STPGNT.
There is a Linux-Howto available.

Once the chipset recieves a HLT or STPGNT is disconnects the FSB from the CPU. The CPU now uses a divider (8...512) to clock itself down significantly.

There is no powersaving difference between the HLT and the STPGNT command on an Athlon (XP) platform.
Enabling this feature doesn't bring any performance hit, as xbitlabs.com showed in this arcticle under "Heat Dissipation".
And from my Cool'n'Quiet thread
jojo4u wrote: The Athlon 64 puts itself to sleep when the HLT/STPGNT instruction is sent.

MattHelm
Posts: 41
Joined: Fri Aug 27, 2004 5:38 pm
Location: Chicago, IL

Post by MattHelm » Fri May 06, 2005 7:00 am

I was wondering how well the rev E's will undervolt, underclock. I'm building a video file server that I want to have great transfer speeds (gbit ethernet) but use as little power as I can. This server will be on 24/7, and I'm looking for ways to lower the electric usage. Sounds like the rev E's were made for my server. :D


BTW, the reason for a socket 939, is I also wanted some future expantions if I decide to use it for more than just files. (I can slap a full 4000+ in if I need it)


Also, anyone know if you can shut down a RAID array, then spin it back up? (???) :?: I know, wrong area.

Post Reply