ASUS M2A-VM + Phenom 9500 experience report
Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee
ASUS M2A-VM + Phenom 9500 experience report
Hi,
Just to share my experience measuring power consumption with my Volcraft Energy Monitor 3000:
- Seasonic S12 II 430 W (220V AC)
- ASUS M2A-VM
- 4x2GB of transcend RAM PC5300
- Maxtor 500GB SATA STM3500320AS
- AMD Phenom 9500, AMD stock cooler
- no case (on my test bench)
- cost of about 620 euros with 19.6% VAT included (France).
Using Linux (text display no X), BIOS cool & quiet and powernow-k8 with ondemand scaling_governor for all 4 cores when the system is idle I got: 77 W, 84VA cosphi 0.92. During boot it went to 95W (spike).
The system is totally silent (subjective 2 meters but in city environment), the cpu fan is at 1200 RPM according to sensors.
When I write to the disk + 8 fpu+int processes the system consumes 147W (cosphi=1), cpu fan goes up to 2000 RPM but stays quite silent (also subjective, but I can sleep at 3 meters of the bench )
With 3 FPU+int processes system consumes 136W (cosphi=1), with two 110W (cosphi=95), with one 95W (cosphi=0.93) and with zero it goes back to 77W within one minute.
One big warning: under heavy load (compilations) the system will get some process stuck (after 24 hours), this is under discussion on linux kernel mailing list. Same board/memory/disk is stable with an Athlon X2 4400+ EE under the same linux kernel so this is a phenom/kernel specific problem.
What is your experience with quad core processors?
Just to share my experience measuring power consumption with my Volcraft Energy Monitor 3000:
- Seasonic S12 II 430 W (220V AC)
- ASUS M2A-VM
- 4x2GB of transcend RAM PC5300
- Maxtor 500GB SATA STM3500320AS
- AMD Phenom 9500, AMD stock cooler
- no case (on my test bench)
- cost of about 620 euros with 19.6% VAT included (France).
Using Linux (text display no X), BIOS cool & quiet and powernow-k8 with ondemand scaling_governor for all 4 cores when the system is idle I got: 77 W, 84VA cosphi 0.92. During boot it went to 95W (spike).
The system is totally silent (subjective 2 meters but in city environment), the cpu fan is at 1200 RPM according to sensors.
When I write to the disk + 8 fpu+int processes the system consumes 147W (cosphi=1), cpu fan goes up to 2000 RPM but stays quite silent (also subjective, but I can sleep at 3 meters of the bench )
With 3 FPU+int processes system consumes 136W (cosphi=1), with two 110W (cosphi=95), with one 95W (cosphi=0.93) and with zero it goes back to 77W within one minute.
One big warning: under heavy load (compilations) the system will get some process stuck (after 24 hours), this is under discussion on linux kernel mailing list. Same board/memory/disk is stable with an Athlon X2 4400+ EE under the same linux kernel so this is a phenom/kernel specific problem.
What is your experience with quad core processors?
-
- Posts: 247
- Joined: Fri Oct 20, 2006 1:00 am
Merci for posting this, guerby. These are very interesting figures. Unfortunately I haven't tested any quad core configurations. I calculated some time ago that a setup similar to yours with a Q6600 (G0 stepping) and a GA-EP35-DS3P would consume around 80W at idle. But I can't remember how I got to that figure. I think I deduced it from several setups I've seen on this forum.
-
- *Lifetime Patron*
- Posts: 1809
- Joined: Sat Apr 24, 2004 1:45 am
- Location: At Home
I tested a Q6600, Gigabyte G33M-S2, Intel IGP, Samsung 400GB SATA, 2x1GB DDR2-900, Silverstone 300W fanless PSU:
Idle / Prime95 x4
Stock voltage – 66.5 / 141W
RMClock - 65.5 / 122.5W
BIOS 1.05V - 58.5 / 107W
I looked at the reviews of the Phenom X4 B3 yesterday and the power consumption figures were disappointing. Worse than that AMD are going to have to price them low to compete with the Q6600 which is bad news for their finances. Intel’s 45nm quads are looking good but supply is low and prices are high for now.
Idle / Prime95 x4
Stock voltage – 66.5 / 141W
RMClock - 65.5 / 122.5W
BIOS 1.05V - 58.5 / 107W
I looked at the reviews of the Phenom X4 B3 yesterday and the power consumption figures were disappointing. Worse than that AMD are going to have to price them low to compete with the Q6600 which is bad news for their finances. Intel’s 45nm quads are looking good but supply is low and prices are high for now.
-
- Posts: 247
- Joined: Fri Oct 20, 2006 1:00 am
How about this from Anand:
Configuration:
CPU:
AMD Phenom 9850 (2.5GHz)
AMD Phenom 9750 (2.4GHz)
AMD Phenom 9550 (2.2GHz)
AMD Phenom 9600 (2.3GHz)
AMD Phenom 9500 (2.2GHz)
Intel Core 2 Quad Q9300 (2.50GHz/1333MHz)
Intel Core 2 Quad Q6600 (2.40GHz/1066MHz)
Intel Core 2 Duo E8200 (2.66GHz/1333MHz)
Intel Core 2 Duo E6750 (2.66GHz/1333MHz)
Motherboard:
ASUS P5E3 Deluxe (X38)
MSI K9A2 Platinum (790FX)
Chipset:
Intel X38
AMD 790FX
Chipset Drivers:
Intel 8.1.1.1010 (Intel)
AMD Catalyst 8.3
Hard Disk:
Western Digital Raptor 150GB
Memory:
Corsair XMS2 DDR2-800 4-4-4-12 (1GB x 2)
Corsair XMS3 DDR3-1066 7-7-7-20 (1GB x 2)
Video Card:
eVGA GeForce 8800 GT SSC
Video Drivers:
NVIDIA ForceWare 169.25
Desktop Resolution:
1920 x 1200
OS:
Windows Vista Ultimate 32-bit
Configuration:
CPU:
AMD Phenom 9850 (2.5GHz)
AMD Phenom 9750 (2.4GHz)
AMD Phenom 9550 (2.2GHz)
AMD Phenom 9600 (2.3GHz)
AMD Phenom 9500 (2.2GHz)
Intel Core 2 Quad Q9300 (2.50GHz/1333MHz)
Intel Core 2 Quad Q6600 (2.40GHz/1066MHz)
Intel Core 2 Duo E8200 (2.66GHz/1333MHz)
Intel Core 2 Duo E6750 (2.66GHz/1333MHz)
Motherboard:
ASUS P5E3 Deluxe (X38)
MSI K9A2 Platinum (790FX)
Chipset:
Intel X38
AMD 790FX
Chipset Drivers:
Intel 8.1.1.1010 (Intel)
AMD Catalyst 8.3
Hard Disk:
Western Digital Raptor 150GB
Memory:
Corsair XMS2 DDR2-800 4-4-4-12 (1GB x 2)
Corsair XMS3 DDR3-1066 7-7-7-20 (1GB x 2)
Video Card:
eVGA GeForce 8800 GT SSC
Video Drivers:
NVIDIA ForceWare 169.25
Desktop Resolution:
1920 x 1200
OS:
Windows Vista Ultimate 32-bit
Thanks for posting! Does the consumption goes up core per core? Ie test with 1xprime95, 2xprimer95 ...smilingcrow wrote:I tested a Q6600, Gigabyte G33M-S2, Intel IGP, Samsung 400GB SATA, 2x1GB DDR2-900, Silverstone 300W fanless PSU:
Idle / Prime95 x4
Stock voltage – 66.5 / 141W
RMClock - 65.5 / 122.5W
BIOS 1.05V - 58.5 / 107W
-
- Posts: 247
- Joined: Fri Oct 20, 2006 1:00 am
-
- *Lifetime Patron*
- Posts: 1809
- Joined: Sat Apr 24, 2004 1:45 am
- Location: At Home
It’s not a linear relationship. Below is a series of tests using a Q6600 running 1 to 4 instances of Prime95. The base line is higher here than for the other test as it had a VGA card installed:guerby wrote:Does the consumption goes up core per core? Ie test with 1xprime95, 2xprimer95 ...
Idle / x1/x2/x3/x4
Bios 1.05V - 70.5 / 91/102/114/120
Stock voltage – 77 / 116/138/151/159.5
Another way of looking at is the extra power consumed per instance of Prime95. So the 1st reading is x1 – idle, the 2nd is x2 – x1 etc
x1/x2/x3/x4
Bios 1.05V - 20.5/11/12/6
Stock voltage – 39/22/13/8.5
The cores themselves on the chip varied also. I tested 4 combinations of running 2 instances of P95; not sure why I didn’t try all 6 combinations!
Cores / Power
0+1 – 140
0+2 – 140.5
0+3 – 136
1+2 – 133.5
Intel’s current quads have a crude architecture compared to AMD’s native quads with their individual clock speeds per core. It would be interesting to compare the two in terms of scaling.
-
- *Lifetime Patron*
- Posts: 1809
- Joined: Sat Apr 24, 2004 1:45 am
- Location: At Home
It’s hard to get a picture of how these two platforms compare as the various reviews use different motherboards, high power GPUs etc.Palindroman wrote:How about this from Anand
After looking at the data this is my estimate for what is achievable using integrated graphics on systems using mainstream desktop components:
CPU / Voltage / Idle / Load
Q6600 / Stock / 66.5 / 141
Q9300 / Stock / 56 / 101
9850 / Stock / 70 / 185
Q6600 / BIOS / 58.5 / 107
It’ll be interesting to see how the Q9300 and 9850 under-volt; has anyone seen data for under-volting them?
I have a suspicion that Intel’s 45nm process won’t have the same gains when under-volting as their 65nm process.
-
- *Lifetime Patron*
- Posts: 1809
- Joined: Sat Apr 24, 2004 1:45 am
- Location: At Home
Windows doesn’t do that very much from what I’ve read and seen. I set the affinity manually to lock the processes to a particular core so it’s certainly not the case in these tests.guerby wrote:May be Microsoft Windows is moving the process across cores, I'm using Linux which is known not to move process around needlessly.
If I look at:
http://techreport.com/articles.x/14424/15
They report a Phenom 9500 system idle at ... 193W !
46W more than the max wattage I was able to draw from my phenom 9500 setup.
Any idea on how they get those numbers? Is it Windows Vista + graphic card which is sucking 116W at idle?
http://techreport.com/articles.x/14424/15
They report a Phenom 9500 system idle at ... 193W !
46W more than the max wattage I was able to draw from my phenom 9500 setup.
Any idea on how they get those numbers? Is it Windows Vista + graphic card which is sucking 116W at idle?
-
- *Lifetime Patron*
- Posts: 1809
- Joined: Sat Apr 24, 2004 1:45 am
- Location: At Home
Scott from Tech Report confirmed by email his measures, so my guess is that 8800 GTX under idle Vista with Aero consumes about 116W...smilingcrow wrote:I don’t know how they manage to get it that high with a single VGA card; that’s quite an achievement.guerby wrote:Any idea on how they get those numbers? Is it Windows Vista + graphic card which is sucking 116W at idle?
I'd love to get an independant confirmation though.
-
- *Lifetime Patron*
- Posts: 1809
- Joined: Sat Apr 24, 2004 1:45 am
- Location: At Home
Take a look at this review at Anandtech which shows an 8800 GTX system consuming 77W more than a HD 3870 at idle. The ATI card consumes roughly 15 to 25W at idle which places the 8800 GTX close to 100W.guerby wrote:Scott from Tech Report confirmed by email his measures, so my guess is that 8800 GTX under idle Vista with Aero consumes about 116W...
I'd love to get an independant confirmation though.
If I compare my own data for a Q6600/IGP against the Techreport Q6600/8800 GTX system the difference is 176 – 66.5 = 109.5W. There are some minor differences between their system and mine but probably no more than 10 – 15W which again puts the 8800 GTX at something like 95W. Staggering.
The Tech Report’s use of an 8800 GTX in the Phenom test skews their power efficiency data as it adds nearly 100W to the baseline power consumption of the system. Surely that makes Phenom look worse than it is as since it is slower than C2Q it will take longer to run the task and will be penalised for longer with the extra 100W.
Exactly the point I made in my answer to Scott answersmilingcrow wrote:Take a look at this review at Anandtech which shows an 8800 GTX system consuming 77W more than a HD 3870 at idle. The ATI card consumes roughly 15 to 25W at idle which places the 8800 GTX close to 100W.
If I compare my own data for a Q6600/IGP against the Techreport Q6600/8800 GTX system the difference is 176 – 66.5 = 109.5W. There are some minor differences between their system and mine but probably no more than 10 – 15W which again puts the 8800 GTX at something like 95W. Staggering.
The Tech Report’s use of an 8800 GTX in the Phenom test skews their power efficiency data as it adds nearly 100W to the baseline power consumption of the system. Surely that makes Phenom look worse than it is as since it is slower than C2Q it will take longer to run the task and will be penalised for longer with the extra 100W.
When no monitor is plugged or if plugged and screen blanked, GPU consumption should be near zero, we're far from it ...Interesting to see that an idle high end card consumes alone nearly more
electricity than a fully cpu-loaded high end quad core complete machine.
Talk about wasting energy...
This base consumption is making it slightly harder comparing your Joule
numbers for cinebench rendering (presented as a cpu test) since 2x the
joules won't mean 2x less energy efficient processor because there's the
idle GPU term which is skewing the numbers (but not the ranking). May be
something to think about for your next article .
Note: BIOS 1705 for the ASUS M2A-VM released on 20080331 fixes all stability problems I had with the previous BIOS version, I got more than 60 hours of stress testing without problem:
So in the end a very nice quad core system with 8 GB of RAM for a cost of about 620 euros with 19.6% VAT included, 80W-140W power draw depending on effective CPU usage.
Code: Select all
guerby@gcc04:~$ uptime
23:46:09 up 2 days, 14:37, 5 users, load average: 17.24, 17.54, 17.62
Try to undervolt via RMClock if you can.guerby wrote:I've never undervolted or overvolted/clocked something. The BIOS has some options for this IIRC.Tomb wrote:Are you able to undervolt Phenom from the bios?
FWIW - here's my Q9450 power consumption numbers:
Asus P5K-VM G33 Mobo
Intel Yorkfield Q9450 45nm Core 2 Quad CPU
Asus 8600GT silent
Corsair 4GB DDR2-800
Corsair 450W PSU
3X WD 1TB GP
Lite On Blu Ray
2.8GHz (1.08V) -> 86W Idle
2.8GHz (1.08V) -> 150W (full OCCT)
2.8GHz (1.08V) -> 175W (full OCCT + RTHDRIBL)
2.8GHz (1.08V) -> 329W (full OCCT + RTHDRIBL w/ 47" 1080p LCD on)