Which A64 cpu to go below 1.1V ?

Masure · Post by **Masure** » Sun Sep 10, 2006 11:38 pm

Hi,

I'm a french guy and I read a lot of SPCR articles and searched the forum in order to have an answer with certitude. I'm aiming to make a low cost/power 24/24 7/7 pc and I wouldn't want to be stucked at 1.1v.

I tried CrystalCPUID and RMClock with my main computer Asrock 939DS2 (vid4 to ground modded) + Opteron 144 but I can't go below 1.1v. Same thing on my girlfriend comp with the same mobo (not modded) and a Venice (E3).

I'd like to have the processors (and rev) which are able to undervolt below that limit. I found the undervoltable motherboards topic.

Thanks

jojo4u · Post by **jojo4u** » Mon Sep 11, 2006 12:26 am

I am interested in this as well, let's try to make a list. As far as I know, the barrier kicked in since the Venice E3. Data taken from german Wikipedia. Get CrystalCPUID or RMClock and try to get lower 1.1V as a low multiplier. RMClock or CPU-Z show you the revision.
I've added the Sempron E6 Palermo.

The list starts with E3 or higher!

Athlon 64
E3 Venice E3: bad
E6 Venice E6: bad
E4 San Diego:
E6 San Diego:
F2 Orleans: good

Sempron with activated Cool'n'Quiet (i.e. >= 1800MHz)
E3 Palermo:
E6 Palermo: bad
F2 Manila: good

Sempron without activated Cool'n'Quiet (i.e. < 1800 MHz)
E3 Palermo:
E6 Palermo: good
F2 Manila:

Athlon 64 X2
E4 Manchester: bad
E6 Toledo:
F2 Windsor: good
G1 Brisbane:

Opteron 1xx
E4 Venus: good
E6 Denmark: good
F2 Santa Ana:

Mobile Athlon 64
E5 Newark:

Mobile Sempron
E6 Albany:
E6 Roma:
F2 Keene:

Turion
E5 Lancaster: good(?)
F2 Richmond:

Turion X2
F2 Taylor:
F2 Trinidad:

Mikael · Post by **Mikael** » Mon Sep 11, 2006 12:50 am

The Athlon64 Orleans (rev. F, Socket AM2) can go lower than 1.1V. I have this system myself and tried 1.0V @ 1.2GHz a few days ago. No problem.

I'm going to try for the lowest stable vcore at 1GHz, just for fun. I could report back here if you're interested.

Masure · Post by **Masure** » Mon Sep 11, 2006 12:59 am

The problem is when someone wants to post his result, he has to be sure that's the processor limit and not the motherboard limit. We have to be sure we're using a "below 1.1v motherboard capable" when testing.

We can also add the 3000+ Winchester D0. It's worth than venice but a 90nm processor to and maybe "maybe 1.1v below capable".

Here is what I got from french wikipedia :

Winchester (90 nm SOI) [modifier]
frÃ©quence : 1800 - 2200 MHz
numÃ©ro de sÃ©rie : D0
Cache L1: 64 + 64 Kio (DonnÃ©es + Instructions)
Cache L2: 512 Kio, fullspeed
Cool'n'Quiet, MMX, 3DNow! Ã©tendu, SSE et SSE2, x86-64 et NX bit
Socket 939, HyperTransport (1000 MHz, HT1000), DDR-SDRAM Double canal
VCore: 1,40 V
TPD : 67 W
Sortie: 2004

scorp · Post by **scorp** » Mon Sep 11, 2006 3:13 am

As far as I know only Rev. >E has the 1.1v limit; on 3 different CPUs (Venice E3, Manchester E4, Venice E6) I can confirm that 1.1v is the lowest possible voltage. The mobos I used are ASUS A8V Deluxe, Abit KN8 Ultra, DFI NF4 SLI-DR, DFI nF3 250. The only CPUs that I've seen with less than 1.1v are an old NewCastle and a Turion64.

QuietOC · Post by **QuietOC** » Mon Sep 11, 2006 4:03 am

All the E revision C'n'Q chips I've tested have locked out the ability to undervolt under the 1.1V setting. My Tforce 6100 motherboard runs the 1.1V setting at 1.06~1.07V.

Chips with the <1.1V setting lock:

Athlon 64 3400+ S754 E3
Athlon 64 3200+ S939 E6
Athlon 64 3200+ S754 E6
Athlon 64 3000+ S754 E6
Sempron 3100+ S754 E6

Chips that will undervolt under 1.1V

Athlon 64 2800+ S754 CG (130nm) 800MHz @ 0.85V
Sempron 3100+ S754 D0 (64-bit enabled) 800MHz @ 0.900V
Sempron 2800+ D0 (no C'n'Q, 64-bit disabled)
Sempron 2500+ E6 (no C'n'Q) 1.4GHz @ 0.88V (stock clockspeed)

Masure · Post by **Masure** » Mon Sep 11, 2006 4:11 am

I would have liked to have a S939 motherboard (GF6150 ideally) but it seems that it's hard to find the right CPU.

Maybe the winchester D0 ?

A friend is going to check if his opteron has the lock this evening (he owns a non modded Asrock 939DS2)

QuietOC · Post by **QuietOC** » Mon Sep 11, 2006 4:24 am

Masure wrote:I would have liked to have a S939 motherboard (GF6150 ideally) but it seems that it's hard to find the right CPU.

Maybe the winchester D0 ?

I'd suggest going AM2. People have reported that their normal Athlon 64 F2 chips can undervolt below 1.1V. Check out vg30et's posts on this thread.

You can save idle power with the D0 revision chips, but they have to be clocked all the way down to 800Mhz to undervolt down to ~0.9V. AMD's newer revisions are able to run much faster using the same amount of voltage (using very little extra power.)

Masure · Post by **Masure** » Mon Sep 11, 2006 4:45 am

QuietOC wrote:I'd suggest going AM2. People have reported that their normal Athlon 64 F2 chips can undervolt below 1.1V.

I focussed on the 939 socket cause I had some ddr sticks to reuse but since the MSI with 6150 chipset exists, I can think about it

QuietOC wrote:You can save idle power with the D0 revision chips, but they have to be clocked all the way down to 800Mhz to undervolt down to ~0.9V. AMD's newer revisions are able to run much faster using the same amount of voltage (using very little extra power.)

Are you meaning thats it will be a bios underclocking and using soft CnQ is impossible ? That's crap.

Concerning newer revisions quality they're better, it's a fact you're right.

widman · Post by **widman** » Mon Sep 11, 2006 4:54 am

Tried myself, these cpu cannot go below 1.1V

E3 Venice 3000+
E4 Manchester X2 3800

and on the same motherboad, Clawhammer CG possible to set 0.85V.

I read here, many winchester cpu possible to set ~0.8V

QuietOC · Post by **QuietOC** » Mon Sep 11, 2006 4:58 am

Masure wrote:Are you meaning thats it will be a bios underclocking and using soft CnQ is impossible ? That's crap.

No, the D0 Sempron 3100+ mentioned above is now running in an ASUS barebone (Terminator T2-AE1) that has no BIOS CPU controls. The C'n'Q mechanism allows selecting multipliers down to 4X for the S754 C'n'Q chips. All you need is a motherboard with C'n'Q support and software (CrystalCPUID or RMClock for example) to control clockspeed (via multipliers) and voltage.

Mats · Post by **Mats** » Mon Sep 11, 2006 9:39 am

E4 Venus: goes under 1.1 V.

jojo4u · Post by **jojo4u** » Mon Sep 11, 2006 10:19 am

All your input has been added to my post.

Masure · Post by **Masure** » Mon Sep 11, 2006 10:47 am

I did post a reply with my friend's venus screenshoot. Were I dreaming ?

frostedflakes · Post by **frostedflakes** » Mon Sep 11, 2006 1:12 pm

My AM2 Athlon64 X2 3800+ also allows <1.1V in RMClock 2.1 (confirmed with Biostar hardware monitor), but system crashes a few seconds. Think it's a software issue, been meaning to try the newer version of RMClock (2.15) and CrystalCPUID.

It would appear, though, that all CnQ enabled AM2 chips support <1.1V. Nice of AMD to listen to all of us quiet PC enthusiasts and allow this adjustment.

Masure · Post by **Masure** » Mon Sep 11, 2006 2:29 pm

Here is my friend's opteron screenshoot with an unmodded Asrock DualSATA2 :

Nil Einne · Post by **Nil Einne** » Thu Sep 14, 2006 9:00 am

The A64 3200+ AM2 I have is able to do 1.050. I'm only testing 1000mhz however. I'm assembling this for a friend, primarily trying to make sure it should be stable with CnQ (which is 1.1V at 1000mhz so I undervolted it more just to help pick up errors, no problems so far). This is done with Crystal, the mobo doesn't support bios Vcore adjustments.

dragmor · Post by **dragmor** » Thu Sep 14, 2006 3:51 pm

I think this is more complicated. Starting with the E revision I believe that AMD put microcode in the CPU that tells the software that the minimum voltage is 1.1v. However this is an artificial limit (can some dev for RMClock or Crystal confirm this?).

My E3 3500+
- IS limited to 1.1v when using one of these applications.
- IS NOT limited when using the bios (800mhz 0.8v works fine).
- IS NOT limited to 1.1v when using CnQ (2.2ghz 1.2v, 1ghz 0.9v) as the shuttle just drops the multiplier to 5 and the voltage by 0.3v.

In effect the shuttle seems to ignore the voltage limit placed by AMD. I think this is why you get weird results whenever someone looks into this.

I'm sure one of the software devs could tell us for certain.

Masure · Post by **Masure** » Fri Sep 15, 2006 12:03 am

i don't understand when you tell :

- IS limited to 1.1v when using one of these applications.

and

- IS NOT limited to 1.1v when using CnQ (2.2ghz 1.2v, 1ghz 0.9v) as the shuttle just drops the multiplier to 5 and the voltage by 0.3v.

Do mention the windows driven CnQ in the last case ? A shuttle provided application ?

jaganath · Post by **jaganath** » Fri Sep 15, 2006 1:47 am

- IS NOT limited to 1.1v when using CnQ (2.2ghz 1.2v, 1ghz 0.9v)

Are these the stock CnQ P-states for the 3500? The spec sheet indicates that the stock volts for stock speed should be 1.35/1.40V?

http://www.amdcompare.com/us-en/desktop ... 3500DAA4BP

dragmor · Post by **dragmor** » Fri Sep 15, 2006 3:55 pm

jaganath wrote:
- IS NOT limited to 1.1v when using CnQ (2.2ghz 1.2v, 1ghz 0.9v)
Are these the stock CnQ P-states for the 3500? The spec sheet indicates that the stock volts for stock speed should be 1.35/1.40V?

http://www.amdcompare.com/us-en/desktop ... 3500DAA4BP

To answer both of you. I am using the stock windows CnQ, with the P states set by windows. No software is controlling this.

However jaganath is right, the voltages are not the standard set by AMD. In the bios of my machine I have the 3500+ set to 1.2v instead of default. Because of the way the Shuttle's bios handles CnQ (drops the multipler to 5x and the volatage by 0.3v) it results in a lower voltage. If I was overvolting then both P States would be overvolted.

QuietOC · Post by **QuietOC** » Fri Sep 15, 2006 6:32 pm

dragmor wrote:However jaganath is right, the voltages are not the standard set by AMD. In the bios of my machine I have the 3500+ set to 1.2v instead of default. Because of the way the Shuttle's bios handles CnQ (drops the multipler to 5x and the volatage by 0.3v) it results in a lower voltage. If I was overvolting then both P States would be overvolted.

Your Shuttle seems to be the exception. There no reason why any motherboard could do the same and just ignore these limits, but I am guessing that you still are unable enable the voltage states under the "1.1V" setting for your CPU.

StrongSilentType · Post by **StrongSilentType** » Sat Sep 23, 2006 11:39 am

dragmor wrote:I think this is more complicated. Starting with the E revision I believe that AMD put microcode in the CPU that tells the software that the minimum voltage is 1.1v. However this is an artificial limit (can some dev for RMClock or Crystal confirm this?).

My E3 3500+
- IS limited to 1.1v when using one of these applications.
- IS NOT limited when using the bios (800mhz 0.8v works fine).
- IS NOT limited to 1.1v when using CnQ (2.2ghz 1.2v, 1ghz 0.9v) as the shuttle just drops the multiplier to 5 and the voltage by 0.3v.

In effect the shuttle seems to ignore the voltage limit placed by AMD. I think this is why you get weird results whenever someone looks into this.

I'm sure one of the software devs could tell us for certain.

The author of RMClock say something about running in less than 1.1v here:
http://forum.rightmark.org/topic.cgi?id=6:717

And some more about not being able to change it below 1.1v here:
http://forum.rightmark.org/topic.cgi?id=6:261

I think you can go below 1.1v from the BIOS, but can't change it in Windows as the CPU will ignore the request to change.

burebista · Post by **burebista** » Sat Sep 23, 2006 8:42 pm

MB ABIT KN9
CPU X2 3800+
And RMClock.

jaganath · Post by **jaganath** » Sun Sep 24, 2006 1:09 am

Seems like AM2 hasn't been lumbered with the Vcore lock. Good stuff. I am very tempted by the AM2 X2 3800, do you know what is the lowest voltage where it will run stable at stock speed? thanks.

burebista · Post by **burebista** » Sun Sep 24, 2006 1:39 am

I haven't tried with stock speed and lower voltage (usually I run my computer 90% of time like in above picture) but a friend managed to run stable at stock speed and 1.1 Vcore.

Erssa · Post by **Erssa** » Sun Sep 24, 2006 7:35 am

widman wrote:Tried myself, these cpu cannot go below 1.1V

E3 Venice 3000+
E4 Manchester X2 3800

and on the same motherboad, Clawhammer CG possible to set 0.85V.

I read here, many winchester cpu possible to set ~0.8V

You made me test my E4 Manchester x2 3800+ to confirm this. You are correct it doesn't go below 1.1V. I also found out during my testing that my cpu is stable at default clock speed (2000Mhz), even at the lowest possible voltage.

jojo4u · Post by **jojo4u** » Tue Oct 17, 2006 10:44 am

*push* please report for the missing cores, especially socket F Semprons are missing!

ryboto · Post by **ryboto** » Sat Nov 11, 2006 4:07 pm

My opteron 170 will do close to 2000mhz at 1.075v. The lowest I can get it to at it's lowest multiplier, 4x, is 1.025v, but keep in mind my fsb is at 260, so, I might even be able to go lower at 200fsb.

edit: oh, and it's a Denmark. Oddly, I'm limited in the bios, I can't go below the stock, 1.35v, but with crystal cpuid, I can go below 1.1.

Tobias · Post by **Tobias** » Sun Nov 19, 2006 3:22 am

Assembled my new system yesterday. I think I hve a Windsor (AMD X2 4200 Socket AM2) on which I've managed the following p-states with CrystalCPUID

200MHz*[email protected]
200MHz*[email protected]
200MHz*[email protected]

So this chip definately goes below 1.1V.