Page 1 of 3

Gigabyte 780G GA-MA78GM-S2H

Posted: Tue Mar 11, 2008 2:18 pm
by NX3
I've posted most of this as a follow up to the SPCR review of this board in this forum : viewtopic.php?t=46828

For other people to get in on the thread and start a discussion the key points are here.

The board is a direct replacement for my Gigabyte 690v which I've previously posted about in this forum. I bought the 780g from Scan.co.uk as a drop in replacement. XP didn't like that so I did a repair install which worked fine. Later that week it was announced officially at CeBIT and hybrid crossfire to be Vista only so I switch to Vista.

Power readings, the setup :

Athlon64 x2 4000 (Brisbane), Gigabyte 780G, Pioneer DVDR-110, Seasonic S12 380, WD 160gb hard drive, AMD stock cpu fan, 12v 80cm case fan on low speed. Using the integrated 780g video card on DVI.

Vista Desktop (Aero OFF) idle is 40/41 watts CNQ ON
Vista Desktop (Aero OFF) idle is 50/52 watts CNQ OFF
Vista Desktop (Aero ON) idle is 41/42 watts CNQ ON
Vista Desktop (Aero ON) idle is 50/51 watts CNQ OFF

I quote two figures for the watts as it flickers between then. Interesting Aero doesn't really add any overhead.

As mentioned Vista on power saving mode keeps setting the minimum cpu idle to 100% from 5% which is annoying.

I used Prime95 to try to max out to get a full cpu hammering power usage but I only ever hit about 70% on either cores. So I don't have a full load reading, suggestion on how to archive this are welcome.

With the Sapphire HD 3450

Vista Desktop (Aero OFF) idle is 53 watts CNQ ON

That heatsink on the 3450 is very hot even at desktop, not just in 3d stuff. 3dmark06 scored 1692.

With Aero off the system has run all evening with a hitch. With Aero on I suffered driver crashes but the desktop recovered it so I didn't have to reboot. With hyrbrid crossfire enable I suffered desktop crashes and a few instant reboots.

Some 3dmark06 scores for comparison :

690v mobo, ATI HD 2400, Cats 8.2, XP : 1373
780G mobo, IGP, Cats 8.3, XP : 1132
780G mobo, IGP + ATI HD 2400, Cats 8.3 Hybrid crossfire enabled, Vista : 2000'ish
780G mobo, ATI HD 3450, Cats 8.3, Vista : 1692
780G mobo, IGP + ATI HD 3450, Cats 8.3 Hybrid crossfire enabled, Vista : <coming soon>

Overall a great board, good 3d performance but cats 8.3 aren't stable in 3d mode.

Posted: Wed Mar 12, 2008 4:52 pm
by austinbike
Put in a 120mm high capacity fan and a notebook hard drive and then you'll really take the power down ;)

Posted: Wed Mar 12, 2008 6:33 pm
by Jubei
I'm considering this board, but the high NB temps are abit scary since I live in the tropics . . . ambient temps in summer is already around 27-30C so if the NB runs hot in Europe and N. America in winter, it is going to really burn in summer.

Otherwise, this boards fits my needs very well.

Posted: Thu Mar 13, 2008 10:32 pm
by xen
Tom's Hardware measurements for the same board:

CPU: Athlon X2 4850E (2.5ghz, underclocked to 1.0ghz)
RAM: 2 GB A-DATA DDR2-800 (not clear whether 2 dimms or single)
CPU cooler: Box
Graphics: IGP
HDD: WD Caviar SE (WD3200AAJS, 3.5" 7200RPM)
PSU: most likely Coolermaster RS850-EMBA ATX 2.2, 850 W

Idle power draw @1.0ghz: 41.7W

690G board (Gigabyte GA-MA69GM-S2H) with the same hardware: 44.1W
(but note that the GA-MA69GM-S2H is a power hog (it uses 8-16W more than the BioStar TA690G-AM2 according to Frank2003's comparison)).

Perhaps the GA-MA78GM-S2H is as power hungry as the GA-MA69GM-S2H, which means that 780G boards by other manufacturors may prove to be less power hungry than their 690G counter parts. Comparing a Gigabyte 780G to a Asus 690G (as in the sprc article) may not be a representative comparison as regards the chipsets.

Posted: Sun Mar 23, 2008 7:09 am
by Maverickâ„¢
X NX3

What about power consumption with previous 690V board?

Posted: Mon Mar 24, 2008 8:06 am
by NX3

Posted: Mon Mar 24, 2008 9:14 am
by Palindroman
How do you think the Gigabyte 780G motherboard compares to the Gigabyte 690V motherboard when it comes to power consumption, NX3? Which one is more efficient, you think?

Posted: Mon Mar 24, 2008 9:55 am
by xen
Do you think the 690V is that much different from the 690G?

Posted: Mon Mar 24, 2008 12:41 pm
by thejamppa
xen wrote:Do you think the 690V is that much different from the 690G?
V is "light" version of the G-model with lower frequencies and some things disabled. in 780V doesn't have UVD but 780G has. I am not sure with 690G/V relations though. So in theory it should use less energy aswell.

Posted: Mon Mar 24, 2008 2:34 pm
by xen
Aight. I've only seen a couple of 690V's in the store listings, they're supposed to be targeted at businesses right?

Posted: Tue Mar 25, 2008 6:46 am
by Palindroman
I've posted this somewhere before but this is what I measured last year:

Image

Unfortunately I haven't got the time or money to test the Gigabyte 780G board. I was hoping a potential client would want this board but the project was canceled.

Does anyone know when the nVidia 8200 chipset is due?

Posted: Tue Mar 25, 2008 8:08 am
by xen
There is a store in the Netherlands that lists the 8200 as being available:

Azerty, Asus M3N78-EMH HDMI,Retail

product page by Asus

Posted: Tue Mar 25, 2008 8:12 am
by Mikael
I have one of those ASUS GeForce 8200 boards coming. Should be here tomorrow. I'll probably make a new thread about it when I've put it through its paces.

Posted: Tue Mar 25, 2008 2:46 pm
by NX3
Palindroman, swapping the 690v to the 780g everything else the same I'm using about 38-40 watts as I was before when idle. So it compares well imo.

Xen, the 690v used a 3-4 watts less than the 690g according to wikipedia.

thejamppa, 690v does not have the digital interface "lacks support for TMDS output and no HDMI output" otherwise it was the same. The frequencies etc are bios features and they vary from manufacture. The gigabyte had most options exposed on the last bios but not CPU voltage controls.

The differences to the 780G and 780v are bested explained in this table on wikipedia :
http://en.wikipedia.org/wiki/AMD_700_ch ... d_graphics

Posted: Wed Mar 26, 2008 12:54 am
by amp
I have bought 2 of these motherboards.

Setup 1 was:
BE-2350 2.1ghz
Drive: 160GB WD AAJS Single Platter
PSU: Earthwatts SA-380 In Case OEM/Version, I wonder.
CD: Liteon DVD
Windows XP
No add in cards.
FAN: Low rpm 120mm antec tricool
CPU: Fan Low RPM cooler master


41W with power management, so 1ghz clocked.
50W Idle.
85W Othos/Prime95
87W with othos in background and Flatout 2 Basic Gameplay
No voltage adjustments.


Machine 2, No Fan case fan, Same CPU stock HSF, 36GB Raptor, EA-430 PSU
43.5ish with cool and quiet
50.5ish windows default.
When hard drive is seeking it spikes about 5-7W ,


I would like use the 780g as a Linux server although it nullifies the point of hdmi and graphics power. I haven't seen any information comparing the 780g and the 770 chipset/65nm both at idle with no graphics card. What i have inferred about the 780g chip is that it is power smart on the hardware level, with its graphical rendering, so at a console 24/7 it is very efficient to the point of almost no GPU use? Comments appreciated.

Posted: Wed Mar 26, 2008 2:50 am
by Palindroman
NX3 wrote:Palindroman, swapping the 690v to the 780g everything else the same I'm using about 38-40 watts as I was before when idle. So it compares well imo.

Xen, the 690v used a 3-4 watts less than the 690g according to wikipedia.

thejamppa, 690v does not have the digital interface "lacks support for TMDS output and no HDMI output" otherwise it was the same. The frequencies etc are bios features and they vary from manufacture. The gigabyte had most options exposed on the last bios but not CPU voltage controls.

The differences to the 780G and 780v are bested explained in this table on wikipedia :
http://en.wikipedia.org/wiki/AMD_700_ch ... d_graphics
Thanks a lot, NX3!

So, it looks like the nVidia 7050 chipset is more power efficient than the 780G chipset (although other boards probably consume less power than this Gigabyte). Let's see what the nVidia 8200 has to offer.

Posted: Wed Mar 26, 2008 2:12 pm
by NX3
From what I read on here the 7050 wasn't as good a the 690/780 but I maybe wrong. Have you got some figures for it ?

The 770 is a interesting chipset, it uses very little power and if paired with a suitable low spec / power card I guess it will use less power than the 780 for your linux console. Say a old PCI video card, they can't draw much power, I've got a 2mb Matrox knocking about. With a 770 thats going to be a nice server console setup and use very little power I'd have thought.

Posted: Fri Mar 28, 2008 1:39 am
by Maverickâ„¢
From this tests it seems that it's very difficult to go below 30 watts, even using a pico PSU and a 2.5" drive maybe power can drop about 10 watts (reaching 30/31)

a guy from Silenthardware.de reached an amazing consumption of 16.5 watts as you can see here --> http://www.silenthardware.de/forum/inde ... opic=23348

So, the difference is too big, maybe the AliveNF7 is a little miracle? :?:

Posted: Fri Mar 28, 2008 2:07 am
by Palindroman
NX3 wrote:From what I read on here the 7050 wasn't as good a the 690/780 but I maybe wrong. Have you got some figures for it ?
After testing the 690V (which was a disappointment) I haven't tested any 690G boards. I now see that the higher power consumption was mainly due to Gigabyte, but I still think nVidia chipsets are more efficient.

The Asrock is 7050, the Gigabyte in the middle is 6150, the one on the right is 7025:

Image

Posted: Fri Mar 28, 2008 2:11 am
by nutball
Maverickâ„¢ wrote: So, the difference is too big, maybe the AliveNF7 is a little miracle? :?:
... or something. My build of an NF7 + X2 3600 @ 1.0GHz/0.8V + 2GB RAM + 2.5" hard-drive + picoPSU doesn't go below 30W. Take from that what you will. It doesn't with a Sempron 3200+ either.

Posted: Fri Mar 28, 2008 2:43 am
by Maverickâ„¢
According to this Tom's test --> http://www.tomshardware.com/2008/03/04/ ... age14.html
..the difference in idle between a Sempron and an Athlon X2 can't be so hight..
in this test even the X2 consumes less power than the 90nm Sempron (I suppose Manila core)

what do you think?

Posted: Fri Mar 28, 2008 3:00 am
by nutball
Maverickâ„¢ wrote:what do you think?
I think that one or more of the results are inaccurate.

If I had to point a finger at the source of the inaccuracy it would be pointed at the power meter(s) being used in the various tests.

I'd also tend to lean toward the lower power consumption figures being more likely to be inaccurate (not because my personal readings are high, just that the low figures seem too good to be true).

I also am quite likely wrong :)

Posted: Fri Mar 28, 2008 1:19 pm
by xen
Whatever the power consumption, it is by a long distance inadequate to heat my home :P. A single electrical radiator uses about 2500W. So that must mean that my gas usage in these months is at least 3000W for many hours a day. Whether my computer adds 16W or 30W (ex monitor), is not of *that* great interest to me :P. I would like to see some real heating usage statistics. Perhaps we need to put things a bit more into perspective. I think we just need more trees.

But do go on ;).

Posted: Sat Mar 29, 2008 9:56 am
by xen
Does anyone know how the power consumption of such boards are really built up?

According to that Tom's hardware article, the 780G northbridge only uses about 1W. If your processor idles at say 5W, and your harddisk idles at 3W, and your RAM uses 2W (I'm just guessing), then how come such systems still usually take 30W? Where does all the other power go? At 80%, 30W should amount to 24W real use, but all components combined should take no more than 11W. Is it the voltage regulators and other circuitry?

Posted: Sat Mar 29, 2008 10:19 am
by nutball
OK, I spent the morning doing some tests on the motherboards which have passed through my HTPC.

Common components:

Athlon 64 X2 3600+
2 x 1GB DDR PC6400 RAM
WD Scorpio 250GB 2.5" drive
120W PicoPSU

I'm using RMClock rather than CnQ, in general use I have it configured to clock the processor at 1.0GHz/0.9V idle up to 2.0GHz/1.1V under load. However it's stable enough at 1.0GHz/0.8V that I can get a power measurement, so that's what I've done. RMClock was configured to adjust clock speed to keep the average CPU load at 75%.

Movie playback is the Coral Reef trailer used in SPCR reviews.

Measurements made with a Brennestuhl PM230 power meter.

All tests under the same install of Windows XP.

Asrock ALiveNF6G-DVI (GeForce 6100 + nForce 430)
Idle: 22-24W
Orthos: 54-56W
Movie playback: 32-34W

Asrock ALiveNF7G (GeForce 7050PV + nForce 630A)
Idle: 22-24W
Orthos: 56-57W
Movie playback: 34-37W

Gigabyte GA-MA78GM-S2H (AMD 780G)
Idle: 29W
Orthos: 66-67W
Movie playback: 39-41W

Posted: Sat Mar 29, 2008 11:06 am
by Palindroman
Looks like your Asrock board does go beneath 30W after all. nutball. 8)

Posted: Sat Mar 29, 2008 2:00 pm
by NyteOwl
Xen: yes, tere is power loss in ther regulators, the fans use power, the LED's use power the power supply circuitry itself uses power - even in sleep mode.

Posted: Sat Mar 29, 2008 11:32 pm
by nutball
Palindroman wrote:Looks like your Asrock board does go beneath 30W after all. nutball. 8)
Yeah, but it didn't when I tested it initially. The memory sticks are different but that shouldn't explain the difference.

It reinforces my view about power meter measurements maybe not really being that reliable... especially when comparing small differences (<~10W).

The "it won't go below 30W" was related to the rebuild in which I replaced the NF6 + stock PSU with the NF7 + PicoPSU. I was disappointed because I'd expected it to break the 30W barrier, but it didn't no matter what I did to it (which is why it stuck in my memory). Maybe some component "breaks in" once it's been run for a day or week or so. But really that all sounds like voodoo to me.

Posted: Sun Mar 30, 2008 12:15 pm
by amp
nutball, can you test your picopsu power supply itself by doing the good old paper clip trick to turn it on. It would be nice to know how much power it uses not running anything.

My earthwatts EA-430 at 118.5V 0.06 Amp 5 watts was the readout. I tested a redundant server power supply/2 module. With 3 high speed fans it used 65 watts by itself. It gives an idea on how much it takes to power the power supply, but once load is applied some of it can disappear since its getting used, and more efficiency kicks in. I am assuming in theory a 100% efficient power supply would have a 0 watt load running itself, without a fan.

Looking at one of the reviews in the brownout scenario for certain power supples the difference between 114v from the utility vs 124v could make a 1-2W difference cause of the efficiency curve.

The Kill O Watt specs says a 0.2% Accuracy, i am fairly sure that they read things under 30 watts just fine, cause i have measured many devices that are in the 5-10 watt area accurately.

Posted: Sun Mar 30, 2008 11:52 pm
by nutball
amp wrote:nutball, can you test your picopsu power supply itself by doing the good old paper clip trick to turn it on. It would be nice to know how much power it uses not running anything.
Just tried this an the power meter reads 0.0W, even with a 120mm fan running. Sounds dodgy to me.