Gigabyte GA-MA74GM-S2: AMD 690G, Take 2

Want to talk about one of the articles in SPCR? Here's the forum for you.
MikeC
Site Admin
Posts: 12285
Joined: Sun Aug 11, 2002 3:26 pm
Location: Vancouver, BC, Canada
Contact:

Gigabyte GA-MA74GM-S2: AMD 690G, Take 2

Post by MikeC » Thu Aug 28, 2008 9:10 am


BillyBuerger
Patron of SPCR
Posts: 857
Joined: Fri Dec 27, 2002 1:49 pm
Location: Somerset, WI - USA
Contact:

Post by BillyBuerger » Thu Aug 28, 2008 9:44 am

Thanks for the review. There's been very little about the 740G and I think it came out around the same time at the 780G. I contemplated the HDMI version (GA-MA74GM-S2H) for my HTPC build. But decided on the 780G for the UVD stuff. That is once I finally found some info about what the 740G was and what was different from the 780G. But I always thought the 740G would make a great work PC. I have a 690G for my own work PC. Haven't gotten the chance to build any new ones since that though. We've got too many old P4s laying around still. =(

QuietOC
Posts: 1407
Joined: Tue Dec 13, 2005 1:08 pm
Location: Michigan
Contact:

Post by QuietOC » Thu Aug 28, 2008 10:40 am

Has anyone actually taken the heatsinks off a 740G and 690G and verified they are different chips (i.e., the 740G being smaller)?

From AMDs own specifications the 690G idles slightly higher than 1W and the 780G just below 1W. The 9W idle power difference highlighted in the article can't be due to the different chipsets.

I think the idle difference is mostly due to VRM design--especially in light of how poor this 740G motherboard becomes at higher loads. This is just a motherboard designed for low wattage CPUs.

MikeC
Site Admin
Posts: 12285
Joined: Sun Aug 11, 2002 3:26 pm
Location: Vancouver, BC, Canada
Contact:

Post by MikeC » Thu Aug 28, 2008 11:01 am

QuietOC wrote:Has anyone actually taken the heatsinks off a 740G and 690G and verified they are different chips (i.e., the 740G being smaller)?

From AMDs own specifications the 690G idles slightly higher than 1W and the 780G just below 1W. The 9W idle power difference highlighted in the article can't be due to the different chipsets.

I think the idle difference is mostly due to VRM design--especially in light of how poor this 740G motherboard becomes at higher loads. This is just a motherboard designed for low wattage CPUs.
This is possible. We've only had the one 740G board to examine, and our generalization about the chipset may be premature. Regardless, we'd still highly recommend this Gigabyte for a low power home server or desktop.

Greg F.
Posts: 372
Joined: Wed Nov 23, 2005 6:51 am
Location: Seattle

Post by Greg F. » Thu Aug 28, 2008 11:31 am

Poorly placed power connectors can also airflow and make the system more thermally challenging. page 2
Just tryin' to help. :roll:

lm
Friend of SPCR
Posts: 1251
Joined: Wed Dec 17, 2003 6:14 am
Location: Finland

Post by lm » Thu Aug 28, 2008 12:20 pm

Some questions:

1) Some Gigabyte boards have had DPC latency spikes. That means sometimes the system becomes unresponsive for a couple milliseconds, which might for example make music playback skip a beat. Easily tested with the tool here:

http://www.thesycon.de/deu/latency_check.shtml

Maybe you could add this into your benchmarks, as it is very easy to run it and just add a screenshot of it in action, nothing else is needed. No other site seems to test it, and it's important for real time tasks.

2) What is the maximum resolution that the IGPU supports? Can't go smaller after I got used to 2560x1600.

3) Can the integrated audio give bit-perfect digital output, or does it force resampling?

MikeC
Site Admin
Posts: 12285
Joined: Sun Aug 11, 2002 3:26 pm
Location: Vancouver, BC, Canada
Contact:

Post by MikeC » Thu Aug 28, 2008 1:34 pm

lm wrote:Some questions:

1) Some Gigabyte boards have had DPC latency spikes. That means sometimes the system becomes unresponsive for a couple milliseconds, which might for example make music playback skip a beat. Easily tested with the tool here:

http://www.thesycon.de/deu/latency_check.shtml

Maybe you could add this into your benchmarks, as it is very easy to run it and just add a screenshot of it in action, nothing else is needed. No other site seems to test it, and it's important for real time tasks.

2) What is the maximum resolution that the IGPU supports? Can't go smaller after I got used to 2560x1600.

3) Can the integrated audio give bit-perfect digital output, or does it force resampling?
1) interesting. We'll try it.

2) 2560x1600 seems feasible. Directly tied to video ram -- 2560x1600x 3 BYTES is what is needed -- 12mb is no problem.

3) that's more an OS issue than the audio card, afaik.

Greg F.
Posts: 372
Joined: Wed Nov 23, 2005 6:51 am
Location: Seattle

Post by Greg F. » Thu Aug 28, 2008 7:05 pm

The Biostar version of this board boasts a FSB twice that of the Gigs. 2000 versus 1000. In practical terms what would that mean for Blueray and HTPC?
If you have an OS hard drive that was used in a Gigabyte 690G could you plug it into a 740G and then just update some drivers and be good to go? :oops:

http://www.newegg.com/Product/ProductLi ... rchInDesc=

lm
Friend of SPCR
Posts: 1251
Joined: Wed Dec 17, 2003 6:14 am
Location: Finland

Post by lm » Thu Aug 28, 2008 11:45 pm

MikeC wrote: 2) 2560x1600 seems feasible. Directly tied to video ram -- 2560x1600x 3 BYTES is what is needed -- 12mb is no problem.

I wish it was so easy. DVI is bandwidth-limited, and it's up to the connector what is supported. There's single link and dual link DVI connectors. They are the same size, single link just does not connect all pins.

Single Link DVI connection is limited to 1920×1200 @ 60 Hz and Dual Link DVI is limited to 2560 × 1600 @ 60 Hz.

More info in http://en.wikipedia.org/wiki/Digital_Visual_Interface

Sadly, not all GPUs have Dual Link DVI connectors, even though ANY card nowadays would have enough onboard RAM. With discrete GPUs, it's usually mentioned which kind of connector they have.

I tried reading Gigabyte's webpages and the product manual, but there is no mention anywhere about the Dual Link capability. So I just have to assume they only have single link, unless sending them email.
MikeC wrote: 3) that's more an OS issue than the audio card, afaik.
I'm under the impression that some chips are only capable of f.ex. 48kHz output, which means forced resampling for 44.1kHz data. It would be better if not, but maybe it's cheaper to make that kind of audio chips. This is yet another minor annoyance of the computer world, since if there were no forced resampling anywhere, any mobo could be used for bit-perfect audio source for an amp that takes digital input.

prodeous
Posts: 90
Joined: Sun Nov 30, 2003 8:01 pm
Location: Poland
Contact:

Post by prodeous » Fri Aug 29, 2008 12:31 am

Very nice review.

I was waiting for this before I decided what MB should i use for my cluster system. This seems to be one of the most logical choices.

Question is how well would it work with a 65W phenom x4. I doubt it would be as simple as adding 20w to the figures.

QuietOC
Posts: 1407
Joined: Tue Dec 13, 2005 1:08 pm
Location: Michigan
Contact:

Post by QuietOC » Fri Aug 29, 2008 5:11 am

lm wrote:Sadly, not all GPUs have Dual Link DVI connectors, even though ANY card nowadays would have enough onboard RAM. With discrete GPUs, it's usually mentioned which kind of connector they have.
All the HD series Radeons support dual dual-link ports, and all the X1xxx series supported at least one dual-link port. The 690G/740G is an X700 based, though. I will see if I can find out anything.

According to wikipedia the 690G supports dual-link. Not the greatest source, I know. AMD's own website says it "Supports a maximum resolution of 2048x1536 @ 32bpp" but that is probably the analog output.

croddie
Posts: 541
Joined: Wed Mar 03, 2004 8:52 pm

Post by croddie » Fri Aug 29, 2008 8:55 pm

Good article, thanks. Impressive idle system power.
Combined with a 45nm processor when they come out (Q1 2009?) this should make for an easy to cool office system.

bjorn
Posts: 5
Joined: Sat Aug 30, 2008 2:20 am
Location: Sweden

Post by bjorn » Sat Aug 30, 2008 2:32 am

I have been looking forward to this review, thanks.

How would the power consumption be affected if you would downclock the IGP? For a low power server, this would be very interesting to know.

The bios of this board allows you to lower the VGA core clock from default 400 MHz to 200MHz. On the Gigabyte 780G board, you can go as low as 150MHz. If you also consider that the 780G is manufactured at a lower 55nm process, 780G might actually be the better choice for a low power server.

amenokaji
Posts: 1
Joined: Sat Aug 30, 2008 3:24 am
Location: ohio

Post by amenokaji » Sat Aug 30, 2008 3:27 am

Great review, one thing i've been wondering about this board, will it fit a 8800gts? i notice one of the chips is extreamly close to the PCI-e slot, and that alone is holding me back from purchasing it. in my current system there's about a paper thin inch between my card and the ram due to placement, and i'm not entirely sure if the card would go over the chip in that system or not fit at all. Any info on that would be great:)

prodeous
Posts: 90
Joined: Sun Nov 30, 2003 8:01 pm
Location: Poland
Contact:

Post by prodeous » Sat Aug 30, 2008 9:32 am

I'm wondering one thing.

Is the 740 chipset supported by AMD Overdrive Tool?

prodeous
Posts: 90
Joined: Sun Nov 30, 2003 8:01 pm
Location: Poland
Contact:

Post by prodeous » Sun Aug 31, 2008 12:47 am

And another question.

With the 690 chipset, when a non Phenom was placed, voltage in BIOS was available. The instance you placed a Phenom (in my case 9500) the ability to adjust the voltage was not available.

Granted, if the AMD overdrive tool is working, then in a windows environment this is not to important. But in case of a non Windows environment, bios option would be simpler.

npp
Posts: 35
Joined: Wed Sep 20, 2006 8:43 am

Post by npp » Sun Aug 31, 2008 4:05 am

lm wrote:Some questions:

3) Can the integrated audio give bit-perfect digital output, or does it force resampling?
I think Mike is right, it has much to do with the OS (especially if you intend using Windows XP/Vista). I suggest you try this:

http://www.asio4all.com

This driver is one of my finest discoveries on the software front lately :) It provides ASIO support (i.e. bit-identical transport of the audio data to the soundcard, bypassing the Windows kernel, so at least any resampling there can be avoided) for virtually all modern sound devices, so it should work fine with this board's codec, too. Of course, the application you're using to play back your cds/dvds/etc. should be capable of dealing with the ASIO interface.

As far as I can see, the specifications of the ALC888 codec are pretty impressive, Realtek claims support for 16/20/24 bit depths and 44,1/48/96/192Khz sapmling rates throught S/PDIF-out on their web site - if this holds true in reality, you may be able to get some decent output combined with the mentioned driver.

loimlo
Posts: 762
Joined: Thu May 03, 2007 3:58 am
Location: Formosa

Post by loimlo » Mon Sep 01, 2008 9:24 am

Well, all intergrated chipsets on the market including Intel, AMD, nVIDIA didn't support dual-link DVI regardless of G45, 790GX, GF8300, 945G, 690G, and GF6100. I think it's their policy to protect their discrete VGA business. You just can't get every traits of discrete cousins like simultaneous DVI/HDMI digitals and have to put up with DSUB/DVI or DSUB/HDMI if you want dual monitors.

prodeous
Posts: 90
Joined: Sun Nov 30, 2003 8:01 pm
Location: Poland
Contact:

Post by prodeous » Wed Sep 03, 2008 1:47 am

I was wondering if anyone was able to find the answers to my quesitons about the 740G chipset and Phenom?

1. Is 740G supported by AMD Overdrive tool?
2. Does Bios still show voltage option when Phenom is pluged in?

I'm really tempted by this chipset and a nice 9150e or 9550 undervolted for cluster nodes.

lm
Friend of SPCR
Posts: 1251
Joined: Wed Dec 17, 2003 6:14 am
Location: Finland

Post by lm » Wed Sep 03, 2008 6:01 am

npp wrote:
http://www.asio4all.com

This driver is one of my finest discoveries on the software front lately :) It provides ASIO support (i.e. bit-identical transport of the audio data to the soundcard, bypassing the Windows kernel, so at least any resampling there can be avoided) for virtually all modern sound devices, so it should work fine with this board's codec, too. Of course, the application you're using to play back your cds/dvds/etc. should be capable of dealing with the ASIO interface.
That is interesting. However from what I read on that site, I understood that it does not bypass windows kernel, just uses it in a different way. On this site seems to be a very good explanation of how it works inside:

http://www.head-fi.org/forums/f46/asio4 ... on-221237/

And btw, I'd probably use linux in it, since my main box takes care of my occasional windows usage needs.

npp
Posts: 35
Joined: Wed Sep 20, 2006 8:43 am

Post by npp » Wed Sep 03, 2008 11:49 am

Sure, don't take my explanation for a scientific fact, I'm a total noob :) What I know is that ASIO output works for me even when the Windows Audio Service isn't started at all... And I guess it's a shortcut after all. I'll read the link thoroughly, thanks for replying.

UPDATE> Yes, the picture seems pretty bright in vista/xp, close to what I have thought... However, no idea of how Linux handles audio. I remeber reading somewhere that it's a better way.. So ASIO wouldn't be needed at all. Good luck ;)

bean1975
Posts: 90
Joined: Thu Jan 12, 2006 3:46 am
Location: Vancouver

Post by bean1975 » Sun Sep 07, 2008 4:17 am

I bought this motherboard (based almost solely on this report), added a 4850e with a Minja in an NSK3480 and the machine is fantastic quiet. As it was said the MOSFETs grow hot, I bought MOSFET coolers but after unboxing the motherboard I saw that there are so many little things that look MOSFETs to my eyes. Can we get a pic of which ones get hot?

Monkeh16
Posts: 507
Joined: Sun May 04, 2008 2:57 pm
Location: England

Post by Monkeh16 » Sun Sep 07, 2008 6:48 am

The ones above the CPU socket, by the back panel, are probably the ones you're after.

Lawrence Lee
SPCR Reviewer
Posts: 1115
Joined: Fri Mar 04, 2005 9:07 pm
Location: Vancouver

Post by Lawrence Lee » Sun Sep 07, 2008 9:31 am

Mainlythe MOSFETs in the upper-left hand corner, between the CPU socket and back panel. Also a couple of the box-like components IRC. you can touch them momentarily to determine which ones are the hottest.

PieterPan
Posts: 1
Joined: Tue Sep 09, 2008 6:28 am
Location: Netherlands

How about the PicoPSU

Post by PieterPan » Tue Sep 09, 2008 6:42 am

Mike, Lawrence,
Very interesting review and board. Got my attention for a nice home server.
Is it possible to add a little test with the PicoPSU? I'm very curious to see what this does to the power usage, idle and loaded. Maybe you have one laying around.
Regards,
Pieter

bean1975
Posts: 90
Joined: Thu Jan 12, 2006 3:46 am
Location: Vancouver

Post by bean1975 » Thu Sep 11, 2008 9:12 am

picoPSU, yes! I am very interested. My setup is very similar to the artilce's and the article showed less than 100W power consumption always.

prodeous
Posts: 90
Joined: Sun Nov 30, 2003 8:01 pm
Location: Poland
Contact:

Post by prodeous » Thu Sep 11, 2008 1:30 pm

if there was some answer to the AMD Overdrive tool, BIOS voltage setting and Phenom, I would definately get it and test it. have 4 picos for cluster nodes.

rdibley
Posts: 8
Joined: Sun Sep 28, 2008 7:55 pm
Location: Lancaster, CA

GA-MA74GM-S2 Idle Power Levels

Post by rdibley » Sun Sep 28, 2008 8:56 pm

I am putting together a machine to act as a NAS, a web server, and a general number cruncher. After reading the article on the GA-MA74GM-S2, I put together a similar machine:

AMD Athlon X2 4850e
GA-MA74GM-S2H
Mushkin 2GB (2 x 1GB) DDR2 800
2 Western Digital 1TB drives (WD10EACS)
PicoPSU-120 with 12V, 8.5A Power Adapter
Zalman CNPS 8700 NT
Lian Li PC-V350B
3 PWM Arctic Cooling Case Fans
Debian Linux (Etch stable)

My hope was to put together a machine that had some power to it, but that would idle at a very low power state. My plan is to have it on 24/7, and it will spend most of its time idle. I was also hoping to keep it pretty quiet. I have installed Debian Linux, and configured the two drives in a RAID 1 configuration.

I used a Kill-A-Watt meter to measure the power consumption. After installing the OS and figuring out how to turn on the AMD cool n quiet feature in Linux, I was disappointed to see that it idled at 42W. The review configuration was able to achieve 31W, and I was hoping to get even less than that, assuming that the PicoPSU would be more efficient at the lower power levels than the 400W Seasonic power supply used in the review.

I then started playing around with the configuration to try to determine where my additional power consumption was coming from:

Idle, Complete System: 42W
Idle, 1GB RAM: 41W
Idle, 1GB RAM, 1 Drive: 35W
Idle, 1GB RAM, 1 Drive, CPU Fan only: 31-32W

I should note that I haven’t yet figured out how to throttle back the three case fans. They’re running full speed currently. Also, I had originally chained the three fans together (the Arctic Cooling fans have cables designed for this) and had them powered off the motherboard, but I later realized that the motherboard was putting out 5V instead of 12V. I unplugged that pin and connected them to the 12V off of the floppy connector. I hooked up a scopemeter to the PWM line and it wasn’t pulsing but remained constant at 2.5V (I am assuming this is the high state). I have the system Smart Fan Control set to “Enabledâ€

MikeC
Site Admin
Posts: 12285
Joined: Sun Aug 11, 2002 3:26 pm
Location: Vancouver, BC, Canada
Contact:

Post by MikeC » Sun Sep 28, 2008 10:28 pm

rdibley --

Your WD GP drives idle at ~3W each, compared to about 1W or less for the 2.5" drive of our review. The optical drive pulls just about nothing when it's not spinning.

PWM is more energy efficient than voltage control. The latter does not change power draw compared to straight 12V to the fan. The fan slows down because the VR controller absorbs a higher proportion of the electrical energy. In fact, a VR controller + fan may pull a touch more power than just the fan alone. But each fan probably draws only a watt or so.

Keep in mind that the Killawatt's precision is not that high, at least + or -1W.

a_punker
Posts: 33
Joined: Thu Oct 05, 2006 2:14 am
Location: United Kingdom

High power CPU warning

Post by a_punker » Mon Sep 29, 2008 4:04 am

Hello,

I notice that one or two of you are considering overclocking and putting high TDP CPUs in this board.

There appears to be a problem with the low 7xx chipsets supporting the high power (quad core 9850 etc) CPUs. It is not a fault of the chipset itself, but a design flaw with the board manufacturers that employ them - the on-board power supplies can't handle it.

This board DOES explicitly state that Phenom FX and Phenom x4 are compatible, but here is an article about the situation. . .

http://www.anandtech.com/mb/showdoc.aspx?i=3279&p=1

AL

Post Reply