Sockt 939 power draw tested: Norce4 - Uli - Via - Sis - Ati

jojo4u · Post by **jojo4u** » Fri Oct 07, 2005 3:33 am

The german printed magazine C't tested in issue 20/05 recent Athlon 64 socket 939 motherboards.

contenders:
Nforce4 - Msi K8N Neo 4-F
Nforce4 Ultra - Tyan K8E
Nforce4 SLI - Asus A8N SLI Premium
Nforce4 SLI - Gigabyte GA-K8N Pro-SLI
Ati RX482 - Sapphire Pure Innovation
SIS 756 - Asrock 939S56M
Uli M1695 - Asrock 939Dual-Sata2
Via K8T890 - Asus A8V-E SE

test setup:
Venice 3500+, 1x Sata Hdd, 1024GiB RAM, Geforce 6600GT, WinXP

whole system power draw was tested
results idle:
Tyan Nforce4 77W
Gigabyte Nforce4 78W
Msi Nforce4 79W
Sapphire ATI 79W
Asus Nforce4 81W
Asrock SIS 83W
Asrock Uli 86W
Asus VIA 86W

results full load CPU/CPU+VGA

Tyan Nforce4 110/145W
Msi Nforce4 110/148W
Gigabyte Nforce4 111/148W
Sapphire ATI 111/146W
Asrock SIS 113/149W
Asus Nforce4 115/148W
Asrock Uli 115/152W
Asus VIA 125/163W

Conclusion: Nforce4 comes first, ATI second, SIS third, Uli is fourth. VIA lags behind in idle and especially in load.
Nforce4 equipped motherboards are undoubtly hot but on the whole system consumes less than motherboards equipped with alternative chipsets.
The Gigabyte board draws as much power as the contenders. A fact which hasn't been true for some time.

afrost · Post by **afrost** » Fri Oct 07, 2005 11:46 am

so why can every chipset but nForce 4 be cooled with a passive sink that doesn't require heat pipes?

something seems wrong here....

jojo4u · Post by **jojo4u** » Fri Oct 07, 2005 12:37 pm

afrost wrote:so why can every chipset but nForce 4 be cooled with a passive sink that doesn't require heat pipes?

I'm a bit confused as well...
Here are my suggestions:
The nforce4 is single-chip. The nforce4 boards may have more efficient VRM.

anaqer · Post by **anaqer** » Fri Oct 07, 2005 12:42 pm

afrost wrote:something seems wrong here....

The only thing "wrong" is thinking that motherboard=chipset. Power draw from the chipset is only part of the equation, efficiency of VRM circuitry is at least as significant a factor in total mobo power draw.

Tibors · Post by **Tibors** » Fri Oct 07, 2005 3:28 pm

I think a more valid conclusion of this data is that mobo's designed by Asus/Asrock (which is basically the same company) consume more power than others. There are way to few datapoints to draw conclusions about chipsets in this equation. Except for the NF4 there is only one of each.

elec999 · Post by **elec999** » Fri Oct 07, 2005 3:31 pm

Very nice tests.
Thanks

~El~Jefe~ · Post by **~El~Jefe~** » Fri Oct 07, 2005 8:50 pm

er, the results show a FOUR wattage difference pull.......

im assuming that 1-2 watts are within error, im going to doubt that the results are within less than 1 watt of error.

they all seem the same (at draw at least) besides the Asus Via chipset (which probably is the pci-e chipset, which sux bad, the k8t800 pro agp chipset uses HALF the power of any of these).

Also, what isnt mentioned and is REALLY important is that the nforce4 is a combined north and southbridge on one chip. this means that the chip does get really hot, the ULI board has the functions split up, making both moderately warm but never hot.

at idle things look slightly different. however, if you consider how small 2-10 watts is on a TOTAL system draw, it looks that all of these are fine. i duno whats up with the asus board though. I wish they had dfi on here and epox to compare the overclocker type boards to these.

nice though to see it ! thanx

frostedflakes · Post by **frostedflakes** » Fri Oct 07, 2005 10:03 pm

afrost wrote:so why can every chipset but nForce 4 be cooled with a passive sink that doesn't require heat pipes?

something seems wrong here....

I think chipset placement has a lot to do with this. With Via, ATi, etc., the reference northbridge placement is right next to the CPU cooler. Manufacturers probably feel more comfortable using passive heatsinks, because they know that the CPU fan will help with cooling the chipset. nForce3/4 don't have this luxury.

Also, nForce3/4 are single chip solutions, so all your heat is concentrated in one area. And nForce chipset placement makes it difficult for the use of large passive heatsinks, because of interference with video cards. When low profile coolers are all that's feasable, you usually have to go with a fan-cooled or heatpipe solution.

Just my $0.02. I think nForce3/4's biggest problems are placement, and not necessarily heat output.

The results are pretty surprising, though, I wouldn't have expected SiS, ULi, and Via to be so far up on the list. But there are so many things that could potentially effect motherboard power consumption (efficiency of the CPU Vreg circuitry, additional onboard features such as SATA, firewire, and GbE controllers, etc.) that I'd take those results with a grain of salt.

rpsgc · Post by **rpsgc** » Sat Oct 08, 2005 12:51 am

I guess the new GeForce 6100/6150, nForce 410/430 and nForce 4 SLI X16 will fare better at the power consumption level (at least the first two!).

dukla2000 · Post by **dukla2000** » Sat Oct 08, 2005 12:01 pm

anaqer wrote:
afrost wrote:something seems wrong here....
The only thing "wrong" is thinking that motherboard=chipset. Power draw from the chipset is only part of the equation, efficiency of VRM circuitry is at least as significant a factor in total mobo power draw.

Agreed: there just isn't sufficient data about the efficiency of the several onboard power circuits to draw any conclusions about the chipsets themselves.

Arguably the 2 Asrock boards use the same power circuit design & components and even so there is no significant difference between them.

Similarly the 2 Asus boards may also in which case there may be room to draw a conclusion. But I am concerned that at idle the delta between the 2 (Asus) boards is 5W (around 6%) whereas at peak it is 15W (around 10%). I would want to understand why this is non-linear before I got too sure of deciding the chipset contribution.

(I guess we can assume these results were all with the exact same psu, keyboard, mouse & monitor, let alone the exact same hdd, ram sticks, CPU, hsf & 6800 etc. But what were the ambient temps and input mains voltages? Being C't I hope none of these were variable but ... And how did they account for and verify any BIOS influences such as Spread Spectrum and Disable PCI Clock or even ACPI variations between the boards?)

ps - it may well be that despite ignoring all the niceities, it is true that all NF4 mobos use less power than all KT8 boards. Good reason to buy NF4 even if it may be some implementation choices rather than the chipset itself.

afrost · Post by **afrost** » Sat Oct 08, 2005 1:03 pm

Well if all you guys think this information is correct then this makes my life easier.

This whole time I have been waiting around for a good ATI or ULI board because I wanted something that could be easily cooled passively, and would contribute less heat to overall case temperature.

It seems there isn't a good reason not to go with one of the tried and true heatpipe boards from Asus or Abit.

~El~Jefe~ · Post by **~El~Jefe~** » Sat Oct 08, 2005 3:07 pm

did not anyone read what I wrote????

the northbridge on a nforce has all of the crap plus more you wont use crammed on one chip. this is how it can more easily get performance. I should "performance" in quotes because the real life differences are about 1% at best.

the ULI board from asrock has less functions on the northbridge, more on the southbridge, and even a sata-II chip off of that. it still was shown to be in the top 10% in terms of gaming speed for popular boards.

nforce4 is hot because it crams everything into a smaller space.

If I am wrong, then I guess the info from ocworkbench I got was wrong as well.
I dont think I am though.

He wants to cool passively without getting hot, uli does that without a performance hit. The heat pipe guys still get really hot, it just is stable at that super hot speed. plus, those special heat pipe gaming boards are 2-3x the cost of the uli.

uli 1695 is overlooked in this forum. elsewhere it is known to be a great option for passive cooling with performance and of course, its the most upgradable system out there, upwards and downwards. I dont have stock in the company

but, I just know how good the board is.

it even accepts the new M2 chip. how many boards do that???????????????

It does!
daughter board to run the new chip, cheap addon, has a dedicated slot socket for adding the new m2.

Tried and true??? abit has been horrible. im on one right now. it is the last of the tried and true ones. KV8 Pro. and when it came out for a year + it was horrible with more than 1 bank of ram filled in cool and quiet mode.
now the new ones are rated at slug snail pace compared to very cheap options like epox. I think even foxcon beats it out for pci-e gaming performances.

Go read the asus forums webpage, everyone has the same issues. both the abit and asus bios programs have gone through many patches with still major issues. and for money and less performance, i couldnt see the point in either of them this year.

afrost · Post by **afrost** » Sat Oct 08, 2005 7:32 pm

I was just commenting on what I should purchase based on what I think is important. You constantly go off on rants on this forum.....it's just computer parts dude....chill

~El~Jefe~ · Post by **~El~Jefe~** » Sat Oct 08, 2005 8:07 pm

bah. so. computer parts. better than ranting about politics. that's a load of waste, or about religion, that's worse.

at least computer parts have an ultimate end point to where they do something better or worse or etc.

anyways, I tend to dislike anything that is a just going with the herd and not a scientifically sound thing.

afrost · Post by **afrost** » Sun Oct 09, 2005 10:18 am

So anyway I am considering the Asrock board. There are a both positives and negatives to it in my mind. It is cheap, and it has an AGP slot, and it seems pretty close to nForce in performance. However, it doesn't have the best overclocking options.....and I get the sense that the people doing the Vdimm mods aren't quite as concerned with silence and stability as I am. The expansion card for a future processor won't be free....and what kind of cooling can you really attach to a card sticking out of the board like that?

Because it's so cheap I feel like picking it up just to give it a chance.

I do have a bit of experience with the AN8 ultra. In only one night of playing with it when building for a friend, I was able to get a 3000+ venice up to 2.5 GHz and rock solid prime95 stable for 12 hours. I felt that I could easily go further but I didn't have time and was mostly concerned with stability. I have read about the problems people have with it....it went together really easy for me.

I haven't really researched the Asus stuff very closely.....but it seems like since they ship so many boards, the boards should be pretty decent.

Anyway back on topic.......

JanW · Post by **JanW** » Mon Oct 17, 2005 1:52 pm

I think, as dukla2000 pointed out, the fact that fractional differences increase with load is suspicious. I expected them to decrease, as at higher load the CPU+GPU fraction of the total power consumption presumably increases. As a result, I don't think that this data can be interpreted as representing chipset power dissipation. Differences in chipset power consumption could be easily masked by a difference in, say, actual Vcore (e.g. a M/B delivering 1.38 instead of 1.35V, or whatever the stock Vcore of that CPU is).

I know c't has a really good reputation (I don't know about todays quality, stopped subscribing 7 years ago), but unless they measured Vcore (and voltage supplied to the graphics card) with a multimeter in idle and under load and showed it to be identical between boards, this data is probably completely useless.

jojo4u · Post by **jojo4u** » Mon Oct 17, 2005 2:20 pm

This test coveres quite a bit of the available boards with alternative chipsets. So it might not be good in showing how energy-efficient one chipset is, but it shows that the nforce4 is not as bad as everybody thinks (including me).

About the reputation: Yes, it's still the best general IT magazine available in germany. But sometimes their test do not go into detail like online reviews. Good example: SPCR PSU reviews show much more efficiency data. On the other hand they have superior equipment for the majority of all tests than online reviewers

Mats · Post by **Mats** » Mon Oct 17, 2005 3:38 pm

afrost: All overclockers with S939 in mind should get an Opteron, but I guess you all know about that already. There's simply no reason to get an A64 for overclocking, except for availability.

=assassin= · Post by **=assassin=** » Mon Oct 24, 2005 7:13 am

afrost wrote:so why can every chipset but nForce 4 be cooled with a passive sink that doesn't require heat pipes?

something seems wrong here....

My Dad's Athlon 64 machine with K8NF-9 motherboard (Nforce 4x chipset) has a passive chipset cooler - in EVEREST it reads 69C, so quite hot.

I'm hoping Gigabyte still do the passive version only as I intend to get the same board for my own build.

afrost · Post by **afrost** » Mon Oct 24, 2005 7:27 am

Mats wrote:afrost: All overclockers with S939 in mind should get an Opteron, but I guess you all know about that already. There's simply no reason to get an A64 for overclocking, except for availability.

Yep, I have an Opty 146 arriving from Ewiz tomorrow as a matter of fact

I decided to try out the Asrock board since they are cheap, and when some ATI chipset boards come out that I like I will pick up one of those. I have high hopes for the Abit ATI board with heatpipe.

Mats · Post by **Mats** » Mon Oct 24, 2005 7:29 am

=assassin= wrote:
afrost wrote:so why can every chipset but nForce 4 be cooled with a passive sink that doesn't require heat pipes?

something seems wrong here....
My Dad's Athlon 64 machine with K8NF-9 motherboard (Nforce 4x chipset) has a passive chipset cooler - in EVEREST it reads 69C, so quite hot.

I'm hoping Gigabyte still do the passive version only as I intend to get the same board for my own build.

Yes, there are some nF4 mobos with no fan out there. Mostly those not intended for overclocking.

69 degrees, that's quite hot. I'd apply new thermal paste on it. Many manufactureres does that badly.

hravn · Post by **hravn** » Mon Oct 24, 2005 11:24 pm

Mats wrote:afrost: All overclockers with S939 in mind should get an Opteron, but I guess you all know about that already. There's simply no reason to get an A64 for overclocking, except for availability.

Unless you are going dualcore. Isn't the dualcore 939 opterons alot more expensive than the 3800?

Mats · Post by **Mats** » Tue Oct 25, 2005 12:01 am

hravn wrote:
Mats wrote:afrost: All overclockers with S939 in mind should get an Opteron, but I guess you all know about that already. There's simply no reason to get an A64 for overclocking, except for availability.
Unless you are going dualcore. Isn't the dualcore 939 opterons alot more expensive than the 3800?

Yes they are, and not really comparable since all dual core Opterons have 2x1 MB cache.

hravn · Post by **hravn** » Tue Oct 25, 2005 4:57 am

Yeah, just meant that if you're on a budget and want dualcore, Athlon64 is still the way to go...