Testing out the GeForce 8200 chipset...(ASUS M3N78-EMH HDMI)
Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee
Testing out the GeForce 8200 chipset...(ASUS M3N78-EMH HDMI)
Okay, as I said a few days ago, I have bought an ASUS M3N78-EMH HDMI which is equipped with the massively delayed GeForce 8200 chipset. I don't know what promted ASUS to release their boards ahead of most others, but they've been available for some time and I decided to have a go at it.
I will not test HD playback capabilites and I will not go over the board's features. HD playback has been tested by Anandtech and the features are available on ASUS' site. What I will do is measure the power consumption of the system, compare it to the Gigabyte GA-G33M-S2H (Intel G33) and also provide 3DMark06 scores for both setups. Let's roll...
Setups
Both setups are identical, except for the motherboards and CPUs.
AMD
Corsair HX520W
ASUS M3N78-EMH HDMI (beta BIOS 222)
Athlon64 X2 4400+ Brisbane rev G2 (default VID 1.300V)
2GB DDR2-800 5-5-5-15-2T
Samsung SpinPoint HD321KJ 320GB
Netgear WG311v3 WLAN PCI card
Scythe S-Flex 120mm 1200RPM
Intel
Corsair HX520W
Gigabyte GA-G33M-S2H (BIOS F3)
Pentium Dual Core E2160
2GB DDR2-800 5-5-5-15-2T
Samsung SpinPoint HD321KJ 320GB
Netgear WG311v3 WLAN PCI card
Scythe S-Flex 120mm 1200RPM
NOTE: SpeedStep and C'n'Q disabled for all tests.
Power consumption tests
The best result in each test will be marked green.
Both setups stock
AMD: 2.3GHz (11.5*200) (1.312V idle, 1.344V load)
Intel: 1.8GHz (9*200) (1.280V idle, 1.264V load)
Idle:
AMD: 67W
Intel: 61W
Load (Orthos):
AMD: 109W
Intel: 91W
Load (3DMark06):
AMD: 107W
Intel: 83W
Load (3DMark06+Orthos):
AMD: 115W
Intel: 93W
Both setups on stock clocks and max undervolt
AMD: 2.3GHz (11.5*200) (0.992V idle, 1.008V load)
Intel: 1.8GHz (9*200) (1.072V idle, 1.056V load)
Idle:
AMD: 56W
Intel: 59W
Load (Orthos):
AMD: 79W
Intel: 79W
Load (3DMark06):
AMD: 79W
Intel: 77W
Load (3DMark06+Orthos):
AMD: 83W
Intel: 83W
Both setups overclocked to ~2.7GHz
AMD: 2.66GHz (11.5*230) (1.10V idle, 1.10V load)
Intel: 2.7GHz (9*300) (1.312V idle, 1.280V load)
Idle:
AMD: 60W
Intel: 67W
Load (Orthos):
AMD: 92W
Intel: 109W
Load (3DMark06):
AMD: 88W
Intel: 99W
Load (3DMark06+Orthos):
AMD: 99W
Intel: 109W
Minimum power settings
Note: The Intel couldn't run on lower voltage, but the AMD possibly could. The CPU voltage used for the 4400+ was the lowest available in Crystal CPUID.
AMD: 1.0GHz (5*200), DDR2-400, (0.816V idle, 0.832V load)
Intel: 1.2GHz (6*200) (0.960V idle, 0.960V load)
Idle:
AMD: 52W
Intel: 55W
Load (Orthos):
AMD: 60W
Intel: 69W
Power consumption tests conclusions
Intel wins easily at stock settings, it's a tie with stock clocks and undervolting and AMD wins easily at the overclocked settings. Now, the overclocked results aren't entirely fair, since the Intel CPU is considerably faster, but they're still interesting. It's also interesting to see how much the power consumption of the AMD system is affected by the undervolt. This 4400+ seems to be an unusually good undervolter, though.
Although hard to tell from these figures, the GeForce 8200 chipset doesn't seem all that power hungry under graphics load. The full load figures with the undervolted (stock clock) setup are pretty impressive, I think. It should also be mentioned that the heatsink gets hot, but not extremely so. At the end of the 3DMark test, I could still keep my fingers on the heatsink indefinitely, although it was a bit uncomfortable. The test system was not assembled in a case, though, so actual temps in a closed system may be a bit different.
Performance in 3DMark06
Please note that the Intel G33 doesn't support Shader Model 3.0, so the final 3DMark score isn't directly comparable to the GeForce 8200.
Both setups stock
AMD : 2.3GHz (11.5*200) (1.312V idle, 1.344V load)
Intel: 1.8GHz (9*200) (1.280V idle, 1.264V load)
AMD:
3DMarks: 840
3DMark06 SM2.0: 294
3DMark06 SM3.0: 294
3DMark06 CPU: 1761
Intel:
3DMarks: 304
3DMark06 SM2.0: 140
3DMark06 SM3.0: N/A
3DMark06 CPU: 1507
Both setups overclocked to ~2.7GHz
AMD : 2.66GHz (11.5*230) (1.10V idle, 1.10V load)
Intel: 2.7GHz (9*300) (1.312V idle, 1.280V load)
AMD:
3DMarks: 972
3DMark06 SM2.0: 339
3DMark06 SM3.0: 342
3DMark06 CPU: 2030
Intel:
3DMarks: 400
3DMark06 SM2.0: 183
3DMark06 SM3.0: N/A
3DMark06 CPU: 2249
Final conclusions
Neither drivers nor BIOS are finalized for the GeForce 8200 yet, so performance may change down the line. All-in-all it looks good and the board works quite well. The BIOS is lacking a little at the moment, having some quite cryptic features and no CPU vcore setting. The voltage can however be changed using Crystal CPUID, which is what I used during my tests.
Finally, I should mention that my board exhibits a slight buzzing during certain CPU loads. This buzzing seems to increase in volume as CPU power consumption increases. Due to this, I will ask for a replacement or possibly a refund. We'll see how things work out.
I hope you enjoyed this little sneak peek of the GeForce 8200.
I will not test HD playback capabilites and I will not go over the board's features. HD playback has been tested by Anandtech and the features are available on ASUS' site. What I will do is measure the power consumption of the system, compare it to the Gigabyte GA-G33M-S2H (Intel G33) and also provide 3DMark06 scores for both setups. Let's roll...
Setups
Both setups are identical, except for the motherboards and CPUs.
AMD
Corsair HX520W
ASUS M3N78-EMH HDMI (beta BIOS 222)
Athlon64 X2 4400+ Brisbane rev G2 (default VID 1.300V)
2GB DDR2-800 5-5-5-15-2T
Samsung SpinPoint HD321KJ 320GB
Netgear WG311v3 WLAN PCI card
Scythe S-Flex 120mm 1200RPM
Intel
Corsair HX520W
Gigabyte GA-G33M-S2H (BIOS F3)
Pentium Dual Core E2160
2GB DDR2-800 5-5-5-15-2T
Samsung SpinPoint HD321KJ 320GB
Netgear WG311v3 WLAN PCI card
Scythe S-Flex 120mm 1200RPM
NOTE: SpeedStep and C'n'Q disabled for all tests.
Power consumption tests
The best result in each test will be marked green.
Both setups stock
AMD: 2.3GHz (11.5*200) (1.312V idle, 1.344V load)
Intel: 1.8GHz (9*200) (1.280V idle, 1.264V load)
Idle:
AMD: 67W
Intel: 61W
Load (Orthos):
AMD: 109W
Intel: 91W
Load (3DMark06):
AMD: 107W
Intel: 83W
Load (3DMark06+Orthos):
AMD: 115W
Intel: 93W
Both setups on stock clocks and max undervolt
AMD: 2.3GHz (11.5*200) (0.992V idle, 1.008V load)
Intel: 1.8GHz (9*200) (1.072V idle, 1.056V load)
Idle:
AMD: 56W
Intel: 59W
Load (Orthos):
AMD: 79W
Intel: 79W
Load (3DMark06):
AMD: 79W
Intel: 77W
Load (3DMark06+Orthos):
AMD: 83W
Intel: 83W
Both setups overclocked to ~2.7GHz
AMD: 2.66GHz (11.5*230) (1.10V idle, 1.10V load)
Intel: 2.7GHz (9*300) (1.312V idle, 1.280V load)
Idle:
AMD: 60W
Intel: 67W
Load (Orthos):
AMD: 92W
Intel: 109W
Load (3DMark06):
AMD: 88W
Intel: 99W
Load (3DMark06+Orthos):
AMD: 99W
Intel: 109W
Minimum power settings
Note: The Intel couldn't run on lower voltage, but the AMD possibly could. The CPU voltage used for the 4400+ was the lowest available in Crystal CPUID.
AMD: 1.0GHz (5*200), DDR2-400, (0.816V idle, 0.832V load)
Intel: 1.2GHz (6*200) (0.960V idle, 0.960V load)
Idle:
AMD: 52W
Intel: 55W
Load (Orthos):
AMD: 60W
Intel: 69W
Power consumption tests conclusions
Intel wins easily at stock settings, it's a tie with stock clocks and undervolting and AMD wins easily at the overclocked settings. Now, the overclocked results aren't entirely fair, since the Intel CPU is considerably faster, but they're still interesting. It's also interesting to see how much the power consumption of the AMD system is affected by the undervolt. This 4400+ seems to be an unusually good undervolter, though.
Although hard to tell from these figures, the GeForce 8200 chipset doesn't seem all that power hungry under graphics load. The full load figures with the undervolted (stock clock) setup are pretty impressive, I think. It should also be mentioned that the heatsink gets hot, but not extremely so. At the end of the 3DMark test, I could still keep my fingers on the heatsink indefinitely, although it was a bit uncomfortable. The test system was not assembled in a case, though, so actual temps in a closed system may be a bit different.
Performance in 3DMark06
Please note that the Intel G33 doesn't support Shader Model 3.0, so the final 3DMark score isn't directly comparable to the GeForce 8200.
Both setups stock
AMD : 2.3GHz (11.5*200) (1.312V idle, 1.344V load)
Intel: 1.8GHz (9*200) (1.280V idle, 1.264V load)
AMD:
3DMarks: 840
3DMark06 SM2.0: 294
3DMark06 SM3.0: 294
3DMark06 CPU: 1761
Intel:
3DMarks: 304
3DMark06 SM2.0: 140
3DMark06 SM3.0: N/A
3DMark06 CPU: 1507
Both setups overclocked to ~2.7GHz
AMD : 2.66GHz (11.5*230) (1.10V idle, 1.10V load)
Intel: 2.7GHz (9*300) (1.312V idle, 1.280V load)
AMD:
3DMarks: 972
3DMark06 SM2.0: 339
3DMark06 SM3.0: 342
3DMark06 CPU: 2030
Intel:
3DMarks: 400
3DMark06 SM2.0: 183
3DMark06 SM3.0: N/A
3DMark06 CPU: 2249
Final conclusions
Neither drivers nor BIOS are finalized for the GeForce 8200 yet, so performance may change down the line. All-in-all it looks good and the board works quite well. The BIOS is lacking a little at the moment, having some quite cryptic features and no CPU vcore setting. The voltage can however be changed using Crystal CPUID, which is what I used during my tests.
Finally, I should mention that my board exhibits a slight buzzing during certain CPU loads. This buzzing seems to increase in volume as CPU power consumption increases. Due to this, I will ask for a replacement or possibly a refund. We'll see how things work out.
I hope you enjoyed this little sneak peek of the GeForce 8200.
Last edited by Mikael on Fri Mar 28, 2008 4:16 am, edited 5 times in total.
-
- Posts: 247
- Joined: Fri Oct 20, 2006 1:00 am
Thanks!Palindroman wrote:Fantastic, Mikael!
WinXP SP2Palindroman wrote:-What OS did you use?
Both disabled. Should have mentioned that.Palindroman wrote:-How about Speedstep and Cool 'n quiet?
The 4400+ is faster at stock, which actually means the undervolted stock AMD setup has the best performance/watt. Generally speaking, the E2160 is 15% faster than the Athlon64 X2 at the same frequency.Palindroman wrote:-How do the CPUs compare (you say overclocked E2160 is faster, what about stock speed)?
I'll get back to you on that one.Palindroman wrote:-What was AMD power consumption when undervolted and underclocked all the way?
-
- Posts: 22
- Joined: Thu Jul 19, 2007 4:38 am
- Location: Italy
Yeah, I'd like to know too. From what I've seen, it seems as if the GF8200 is close to the 780G and maybe slightly more efficient. We'll hopefully know once Anandtech posts their µATX roundup (which should be pretty soon).Maverick™ wrote:Great job, i whish so much to know the power consumption difference between the 8200 and the 780g, and also in comparison with previous chip 7050 and 690g
if someone could...
Excuse me, would you mind sharing fan controlling with us? It's also an important role in silent computer. I hope it's a typical ASUS mid-range board which has 2 controllable fans with three silent/performance/optimal settings. By the way, did M3N78-EMH HDMI support 3pin fan or it was limited to 4pin fan? As for Anandtech, the long overdue 690G/GF7050 round-up was really let me down. I'll just wait and see this time.
Sorry to disturb you so eagerly.
Sorry to disturb you so eagerly.
I returned the board today for a refund, so I can't answer your questions, sorry. The board has a 4-pin connector for the CPU fan, but I don't know how the fan control function is affected by the use of a 3-pin vs a 4-pin fan.
I probably should have tested the fan control function, but since I never use such functions it slipped my mind. I thought about getting the board replaced for one that hopefully wouldn't buzz, but I decided to wait for the more feature packed Gigabyte boards.
I probably should have tested the fan control function, but since I never use such functions it slipped my mind. I thought about getting the board replaced for one that hopefully wouldn't buzz, but I decided to wait for the more feature packed Gigabyte boards.
Fan control works pretty well, on lowest settings my Freezer 64 Pro runs at 1000RPM on idle, 1300 under load and 1500 with CPUburn (cooling a BE 2400).
It does run fans with 3 pin just fine, can't remember whether fan control worked (I believe it did) with those though. I just hooked the two case fans to 5V and left it at that.
All in all the board is decent but the BIOS is really sparse. No undervolting, no nothing.
And it won't boot with my Seasonic S12 II 330.
Using the Etasis EFN300, 1GB RAM and a 1GB Kingston stick, it draws about 35W AC (230V) in idle and 93W under cpuburn. That is in Ubuntu 8.04, using cool and quiet.
It does run fans with 3 pin just fine, can't remember whether fan control worked (I believe it did) with those though. I just hooked the two case fans to 5V and left it at that.
All in all the board is decent but the BIOS is really sparse. No undervolting, no nothing.
And it won't boot with my Seasonic S12 II 330.
Using the Etasis EFN300, 1GB RAM and a 1GB Kingston stick, it draws about 35W AC (230V) in idle and 93W under cpuburn. That is in Ubuntu 8.04, using cool and quiet.
-
- *Lifetime Patron*
- Posts: 1809
- Joined: Sat Apr 24, 2004 1:45 am
- Location: At Home
Personal interest/habit, really. I always have them disabled, for the simple reason that they make so little difference compared to having the system run unervolted at stock clocks. For example, there's a 4W difference in idle power consumption between the 4400+ @ 2.3GHz and 1.0GHz.smilingcrow wrote:Interesting review.
Why did you choose to test with Speedstep & CnQ disabled? It seems an odd choice for a test looking at power consumption.
I guess I should have included it for completeness, but with four tested setups you hopefully get a pretty clear picture of it anyway.
-
- *Lifetime Patron*
- Posts: 1809
- Joined: Sat Apr 24, 2004 1:45 am
- Location: At Home
I tend to see the baseline as being with Speedstep/CnQ enabled, the next level is with software undervolting and the final level is with BIOS under-volting.Mikael wrote:Personal interest/habit, really. I always have them disabled, for the simple reason that they make so little difference compared to having the system run unervolted at stock clocks.
That wouldn’t add another stage to your testing either but bring it more into line with real world scenarios as commercial systems tend to ship with Speedstep/CnQ enabled.
Maybe so, but I just didn't think of it at all. If it had entered my mind I probably would have tried it.smilingcrow wrote:I tend to see the baseline as being with Speedstep/CnQ enabled, the next level is with software undervolting and the final level is with BIOS under-volting.
That wouldn’t add another stage to your testing either but bring it more into line with real world scenarios as commercial systems tend to ship with Speedstep/CnQ enabled.
Anyway, with the already tight idle ranges, it's easy to see where the numbers would be for Cool'n'Quiet/SpeedStep. Not very scientific, but it works:
AMD: 56-58W
Intel: 57-59W
I'll try to include the numbers next time.
Well there are no adjustments for undervolting I can find in the BIOS of this board. Asus has really skimped on BIOS features. As for software undervolting, I can't seem to find the patches against Linux enabling it to undervolt AMD K8I tend to see the baseline as being with Speedstep/CnQ enabled, the next level is with software undervolting and the final level is with BIOS under-volting.
Interestingly, I probably do not really need it with the BE 2400 when considering that I idle at 35W vs Mikael's 56W... And I presume his PSU is actually more efficient than mine.
It's probably a combination of a RAM stick less, no WLAN card and more efficient PSU (yes, your PSU is actually more efficient at low loads). I don't think the CPU is the main source of the difference. At 1.0GHz and 0.8V, neither of these CPUs should pull more than 5W.oberbimbo wrote:Interestingly, I probably do not really need it with the BE 2400 when considering that I idle at 35W vs Mikael's 56W... And I presume his PSU is actually more efficient than mine.
You been able to get HD hardware acceleration working well? How about heat? I heard the 8300 was being delayed due to heat. Curious if the 8200 suffers heat problems but at a lesser degree. How about multichannel LPCM via HDMI? Have you been able to test that?
Also what Bios version are you running?
Also what Bios version are you running?
Some guy in a German forum had it running HD DVD on a X2 4400+ (Windows) without problems. Not sure what audio codec he was using, though.
There's NO way you can do full bitrate 1080p in software on Linux right now. Not even on a 3.0 GHz Wolfdale, for all I'm told. I can barely do low bitrate 1080p and even then, pans are stuttering.
There's NO way you can do full bitrate 1080p in software on Linux right now. Not even on a 3.0 GHz Wolfdale, for all I'm told. I can barely do low bitrate 1080p and even then, pans are stuttering.
oberbimbo,
Do you mind doing me a favour if you have some time? I'm looking for the PCI:ID for the onboard ethernet on that mobo. Can you run lspci and find the ethernet card. Then lspci -n -s <###> where ### is the first few digital from the beginning of the line like 00:0a.0 or something. Trying to see if I can modify the bios and integrate gPXE into the mobo.
Thanks!
Do you mind doing me a favour if you have some time? I'm looking for the PCI:ID for the onboard ethernet on that mobo. Can you run lspci and find the ethernet card. Then lspci -n -s <###> where ### is the first few digital from the beginning of the line like 00:0a.0 or something. Trying to see if I can modify the bios and integrate gPXE into the mobo.
Thanks!
There's a new bios dated April 10th version 0404 you may want to try.oberbimbo wrote: Well there are no adjustments for undervolting I can find in the BIOS of this board. Asus has really skimped on BIOS features. As for software undervolting, I can't seem to find the patches against Linux enabling it to undervolt AMD K8
-
- *Lifetime Patron*
- Posts: 1809
- Joined: Sat Apr 24, 2004 1:45 am
- Location: At Home
There’s a review of power consumption for the 780G, GF8200 and G35 chipsets at idle and whilst playing HD video formats at Anandtech.
oberbimbo,badkarma wrote:oberbimbo,
Do you mind doing me a favour if you have some time? I'm looking for the PCI:ID for the onboard ethernet on that mobo. Can you run lspci and find the ethernet card. Then lspci -n -s <###> where ### is the first few digital from the beginning of the line like 00:0a.0 or something. Trying to see if I can modify the bios and integrate gPXE into the mobo.
Thanks!
What distro are you running? I tried my backtrack CD and the ubuntu 8.04 cd installer and both won't boot my Giga-byte Nvidia 8200 mobo properly. Trying to find the PCI:ID for the onboard ethernet.