Nvidia lowering power supply requirements?

[Evildeus]

You are right, but when i look to Fodder exemple, i'm a bit sceptical if a R420 should not require a 480W PSU also.

Um....Why? R420 consumes about the same or less than the 9800XT. Did the 9800XT require a 480W power supply?

I'm beginning to wonder if you'd ever looked at that chart :/ Where did you get this ludicrious thinking?

 

[Joe DeFuria]

I'm beginning to wonder if you'd ever looked at that chart :/ Where did you get this ludicrious thinking?

I'm beginning to wonder if you ever looked at any charts that attempted to actually stress the GPUs, so that peak relative power consumption of the video cards can actually be measured...in other words, I'm wondering if you just ignored my last post due to some ludicrous thinking on your part.

 

[Bjorn]

I'm beginning to wonder if you ever looked at any charts that attempted to actually stress the GPUs, so that peak relative power consumption of the video cards can actually be measured...in other words, I'm wondering if you just ignored my last post due to some ludicrous thinking on your part.

I agree that we need more tests with heady shader usage before we get any clue on the maximum power usage for these new cards.

But afa the X800 XT vs 9800 XT goes, shouldn't the power usage go up more on the X800 XT then on the 9800 XT in that case ? (since it has twice the amount of shader units, that should of course also be true for the 6800 vs 5950). And this chart already indicates that it draws more power then the 9800 XT.

 

[Evildeus]

I'm beginning to wonder if you ever looked at any charts that attempted to actually stress the GPUs, so that peak relative power consumption of the video cards can actually be measured...in other words, I'm wondering if you just ignored my last post due to some ludicrous thinking on your part.

Yes i do, but i'm still waiting for the proof of the power consumption of the 9800XT > X800XT, at leastn back up your claims.. All the charts i've seen show the contrary.

Perhaps, the target of this chart is to stress the GPU (3dmark) while doing something else in the background (prime).

 

[Joe DeFuria]

But afa the X800 XT vs 9800 XT goes, shouldn't the power usage go up more on the X800 XT then on the 9800 XT in that case ?

Possibly sure. But there's too much architecturally that can impact power consumption, including memory. It may not actually be shader apps that cause worst case power conusumption, but lots of AA, which persumably maxes out memory power.

 

[Joe DeFuria]

Yes i do, but i'm still waiting for the proof of the power consumption of the 9800XT > X800XT, at leastn back up your claims.. All the charts i've seen show the contrary.

http://techreport.com/reviews/2004q2/radeon-x800/index.x?pg=27

Sorry, it shows the X800 Pro is less than the 9800XT peak...and the X800XT is a whopping 6 watts higher than the 9800 XT at peak.

Though it does indicate that on average, and the x800XT is lower in power draw, given the 28 W lower power consumption at idle.

Perhaps, the target of this chart is to stress the GPU (3dmark) while doing something else in the background (prime).

So what's the point of that test? That doesn't show peak GPU power consumption.

 

[Maintank]

I wouldnt expect much proof on the power consumption.

Personally I think Nvidias reqs were a little over the top. The system I own runs on a 300 Watt PS without issue. And from most reports of measuring power draw the 6900 Ultra isnt much more than 15-20% higher than the 5950. That doesnt equal 180 watts IMO.

I am thinki most people can get by with a very good 380 watt PS or a good 430 Watt.

 

[Bjorn]

Sorry, it shows the X800 Pro is less than the 9800XT peak...and the X800XT is a whopping 6 watts higher than the 9800 XT at peak.

Though it does indicate that on average, and the x800XT is lower in power draw, given the 28 W lower power consumption at idle.

The average isn't that interesting afa recommended PSU though.

And while the difference between the 9800 XT and X800 XT is small, there's "only" a 23 W difference between the X800 XT and 6800 Ultra. Hardly any reason for a 130W increase in recommended PSU.

 

[Joe DeFuria]

The average isn't that interesting afa recommended PSU though.

Correct.

And while the difference between the 9800 XT and X800 XT is small, there's "only" a 23 W difference between the X800 XT and 6800 Ultra. Hardly any reason for a 130W increase in recommended PSU.

That all depends on the power draw on the 12V line, and what a significnat number of 350 W power supplies give on that line, vs. 480W power supplies.

 

[Pete]

I don't see the problem with running Prime concurrently with 3DM03. 3DM03 has been shown to be largely GPU-limited, so adding Prime just loads up an otherwise loafing CPU to show a potentially more accurate total system draw (worst case scenario). I guess benching a shader- and CPU-heavy game like Far Cry at 12x10 AA+AF would be a good test, though one with an emphasis on stressing the video card.

 

[Evildeus]

Yes i do, but i'm still waiting for the proof of the power consumption of the 9800XT > X800XT, at leastn back up your claims.. All the charts i've seen show the contrary.

http://techreport.com/reviews/2004q2/radeon-x800/index.x?pg=27

Great, so the X800XT does indeed draw more power than the 9800XT.

As for your other question, perhaps, because you don't have just a GPU in your PC? Otherwise why don't we make the power requirement @150W?

 

[Joe DeFuria]

I don't see the problem with rnuning Prime concurrently with 3DM03. 3DM03 has been shown to be largely GPU-limited, so adding Prime just loads up an otherwise loafing CPU to show a potentially more accurate total system draw (worst case scenario).

It can also be introducing stalls to the GPU, lowering GPU power draw.

I would like to see perhaps 3 different tests:

1) Prime95 only (Idle GPU, stressed CPU)
2) 3DMark 03 only ("Idle" CPU (not really though), Stressed GPU)
3) Both simultaneously. (Some stress on both.)

Then you can maybe start to make draw some meaningful results.

 

[AlphaWolf]

The average isn't that interesting afa recommended PSU though.

Correct.

And while the difference between the 9800 XT and X800 XT is small, there's "only" a 23 W difference between the X800 XT and 6800 Ultra. Hardly any reason for a 130W increase in recommended PSU.

That all depends on the power draw on the 12V line, and what a significnat number of 350 W power supplies give on that line, vs. 480W power supplies.

I saw a test where it was taking 5A off the 12V line. I will find the link.

Link Added

 

[Joe DeFuria]

Great, so the X800XT does indeed draw more power than the 9800XT.

Yup, in that that test, it does.

As for your other question, perhaps, because you don't have just a GPU in your PC? Otherwise why don't we make the power requirement @150W?

Bah...I didn't see him burning a DVD whlie running Prime 95, and at the same time run a raid File Server stress test, all while capturing DV video through the FireWire port....I mean, I just don't have a CPU and a GPU in my machine...

 

[Pete]

It can also be introducing stalls to the GPU, lowering GPU power draw.

True, but HW.fr's combo seems to be inducing greater power draw under load than TR's RTHDRIBL-only stress-test, while showing roughly the same idle power draw.

I would like to see perhaps 3 different tests:

1) Prime95 only (Idle GPU, stressed CPU)
2) 3DMark 03 only ("Idle" CPU (not really though), Stressed GPU)
3) Both simultaneously. (Some stress on both.)

Then you can maybe start to make draw some meaningful results.

Agreed, but I'm assuming HW.fr tried all three combos and found the third to show the greatest total power draw. If they did, it's their fault for not communicating their methodology. If they didn't, well, it's time to start cracking the whip, eh?

 

[Bjorn]

Yup, in that that test, it does.

Yes, it's just one test but you must admit that it's not very likely that the X800 XT will draw less power (maximum) then a 9800 XT.

 

[Stryyder]

The average isn't that interesting afa recommended PSU though.

Correct.

And while the difference between the 9800 XT and X800 XT is small, there's "only" a 23 W difference between the X800 XT and 6800 Ultra. Hardly any reason for a 130W increase in recommended PSU.

That all depends on the power draw on the 12V line, and what a significnat number of 350 W power supplies give on that line, vs. 480W power supplies.

I saw a test where it was taking 5A off the 12V line. I will find the link.

Link Added

Shouldn't you have 18 AWG to safely conduct 5 Amps???

Some Common PS outputs on the 12+V rail

Enlight 300 Watt 13Amp 21.00
Sparkle 300 13 Amp 22.00
Antec 300 15 Amp 30.00
Zalman 300 18 Amp 54.00

Sparkle 350 16 AMps
ENlight 350 25 AMps 40.00


Now My Antec True Power 480 gives me 22 Amp ons the 12+ line the difference is that good power supplies do not bleed power from one rail to handle a large drop on another. Secondly good power supplies respond faster to large power draw changes for example the ANTEC 480 will be withing 5% of the output within 1ms for over a 20% load change.

Also just because that Power supply is say rated for 18 Amps doesn't mean that it is constructed to carry 5, 6 or 7 AMPs over one connector better power supplies have better rated wiring or Lower AWG than shitty ones. Also Shitty power supplies don't behave very well under load. I.E when the loads get high the % variance on the rails increase.

So My Guess is that the NVIDIA guys know this, know that every 480 watt power supply out there will work perfectly with the 6800 but know that a certain percentage of the 350 Watt power supplies will choke if under full load or a heavy load on the 12+V rail and that to cover their asses they should recommned the 480 to the reviewers.

What I think you will see is required = Good 350 Watt powersupply Minimum 15 Amps on the 12+V rail Recommend any 480 Watt power supply.

 

[Maintank]

I think you are making a big deal out of nothing.

 

[Stryyder]

I think you are making a big deal out of nothing.

I just bought a Dell PC for my father in law last week How Much you want to Bet that the Power Supply in his new P4 has a rating of about less than 15 Amps on that 12+V rail??

 

[karlotta]

Shouldn't you have 18 AWG to safely conduct 5 Amps???

What I think you will see is required = Good 350 Watt powersupply Minimum 15 Amps on the 12+V rail Recommend any 480 Watt power supply.

No a 18 will carry MUCH more than that, there is a post in this thread that has the AWG rateings. And yes 15amp/12v rail is standard for quality 300/350 PSUs. But then the guys who have 6 IDE/SATA devices , 1-2 gigs of RAM ,Fans all over and a serious vidcard..... 15amps will let your system reboot every so often. ( and then some will think it was the drivers..lol)
good site to chk specs:dont mess around

As for Dell they tune there PSUs to there MBs and total system. Shuttle does the same. The MB is a key part in the power equation. A poor MB will not help a 15amp, you will have to make up for it with some 20amp+ PSU.. To me poor MBs are more of a issue than PSUs for the Do it your self crowd.