SemiAccurate wrong about Nvidia 480GTX power use

A

argor

n00b
Joined
Dec 20, 2009
Messages
35
http://www.semiaccurate.com/2010/03/12/semiaccurate-wrong-about-nvidia-480gtx-power-use/
GDC 2010: Officially raised 50W, OEMs livid
IF YOU WANT to learn about Nvidia's Tesla and GTX480 cards at GDC, don't ask Nvidia, they have problems with the truth. The real story is found with the users, and they have interesting things to say about the upcoming card's upward bound TDP.

If you recall, the official story is that the card, in it's cut down and underclocked version, pulls 225W. That number, along with stunningly poor performance, has lead to some notable backpedaling. If that isn't bad enough, some sources at GDC told SemiAccurate that Nvidia jacked the TDP up 50W last week without warning.

We will be the first to admit we were wrong about the TDPs of the cards. At CES we said the GTX480s shown there were pulling 280W, something Nvidia vehemently denied. Engineers beavering away at the things Dear Leader thinks are important, the style of wheels for his Ferrari, have been pulled off to work on cards for some unfathomable reason. Working hard, they have managed to reduce the TDP of the cards 5W to 275W. Yeah, Nvidia finally admitted that the card is the burning pig anyone who has used one knows it is.

There are two problems here, one internal and one external. The internal one is that this is a big flag saying they admit defeat, and have no hope of fixing the problems that plague the chip. Nvidia can't get the power to a reasonable level, and that is the end of it. The only way to get numbers of chips to salable quantities is to jack power through the roof to mask the broken architecture, so that is what they are doing.

More problematic is what about the OEMs? Officially raising the TDP three weeks before launch by a very substantial 50W is massively stupid, you just can't do this to OEMs without causing them lots of pain. For high end desktops with lots of space, that can be worked around, but if the system is a little closer to the edge, 20+% more TDP can have a profound and negative effect on cooling.

Even worse, think about all the companies that make Fermi based Tesla cards. If you put four in a system, and Nvidia jacks TDP 50W per card, that is 200W more you have to dissipate. Three weeks before launch, your cases are built and in a warehouse, your cooling system is finished, and you don't have time to change things, much less test them. 200W is a lot in a 2U server case. 21 of these in a 42U rack is an added 4.2kW that you need to dissipate, roughly 3 hair dryers on full blast.

Then there are the laptops. I feel bad for those guys, first Bumpgate, now this. There is no way you can redesign a laptop cooling system in six months, much less three weeks. Silly ODMs, no cookie, but you will probably be blamed by Nvidia PR for 'screwing up' so badly.

In the end, Fermi is turning into a running bad joke. You have to wonder about how many high margin orders will be shown the door when word of this leaks out. Nvidia may be "Oak Ridged" a few more times yet.S|A
Click to expand...
 
lewlz. so glad i got my 5870 a couple weeks ago.
 
You might want to mention in cliffs that Charlie is only "wrong" by 5w :D

A lot of people don't want to read that entire article just to get the one tidbit.
 
In the words or tim taylor...


Needs more powah!! *grunt grunt grunt*

(idk lol)
 
Of course it's broken, so I wouldn't buy one. That's 100W more than my 5870 and almost equal to a 5970.
 
lol @ nvidia

but then again.. why is charlie talking about laptops? its not like the laptop version of this chip was supposed to be released in 3 weeks and there are no laptops lined up with this design..
 
Where are the guys blasting that site because he makes stuff up only to hurt nVidia's reputation? Normally there are several posts of that nature by now.
 
TomGreen321 said:
Where are the guys blasting that site because he makes stuff up only to hurt nVidia's reputation? Normally there are several posts of that nature by now.
Click to expand...
He's been right so many times in the last eight months that it's hard to blast him anymore.

He called the Fermi delay long before Nvidia admitted anything was wrong, which they denied--still claiming they were on schedule for a December release.

He knew about the high power usage.

He predicted the die size accurately.

He knew they were trimming shaders from the original 512.

Not his doing, but he also gave attention to the fact that the Fermi showed at a technology conference was a pretty laughable mock-up.
 
If I'm not mistaken, he thought the GTX 480 will have 448 SPs and TDP of 225W.
So, now it looks to be 275W. They might have clocked it back to 512?
 
If it has voltage control sounds like you would really have to jack it up for a small return and stability.

But what realistically could they have done?, scrap it 6months ago or hope respins and new revisions would fix or at least downplay the issues - they chose the right decision i would say it just didn't work like they wanted/hoped.

Expect a very aggressive refresh/next gen card based on this architect to make up *if* they ever sort it out, although now might be the time to scrap it rather than hope refreshes will fix it and make the same mistake.
 
If it at least matches 5970 performance I guess this would be acceptable. Pure conjecture but maybe it does and nVidia thinks it's worth the PR hit.
 
DontBeEvil said:
If I'm not mistaken, he thought the GTX 480 will have 448 SPs and TDP of 225W.
So, now it looks to be 275W. They might have clocked it back to 512?
Click to expand...

We can hope..., but I'd wager that the 225W TDP was an optimistic guess at where the TDP would be if they could fix GF100's rather high leakage in time for launch. Like Charlie says, it appears that admitting 275W TDP is essentially admitting that they've run out of time/options and they'll just have to release the product the way it currently is.

If B1 silicon comes back with much lower leakage and better yields, you can bet a new GTX485 will soon follow with the full 512 shaders and possibly higher clocks.
 
fattypants said:
He's been right so many times in the last eight months that it's hard to blast him anymore.

He called the Fermi delay long before Nvidia admitted anything was wrong, which they denied--still claiming they were on schedule for a December release.

He knew about the high power usage.

He predicted the die size accurately.

He knew they were trimming shaders from the original 512.

Not his doing, but he also gave attention to the fact that the Fermi showed at a technology conference was a pretty laughable mock-up.
Click to expand...

Shaders for 480 are still 512, he got it wrong.
Die size is still not known so how the hell did he get this one right?
Power draw based on some guy at GDC. Hell, it could have been an ATI guy telling him that. How about waiting for reviews to make claims to his accuracy about other stuff out side of the time for fermi. And people far more in the know than him are saying die size in the 500mm range, not 550-575 as he is.
 
teenk said:
If it at least matches 5970 performance I guess this would be acceptable. Pure conjecture but maybe it does and nVidia thinks it's worth the PR hit.
Click to expand...

Unfortunately for us and video card prices, most leaked estimates put it at just above the 5870's level, meaning it probably won't be anywhere near the 5970 except in a few heavily nvidia-biased games like FC2.
 
And here I was hoping that Nvidia had somehow managed to show up with a card that was only moderately hotter than the 5870 and had significantly better performance. Not likely, but it'd be nice.

Oh well. I suppose its EXISTENCE will cause some 5000-series price drops, but it's still disappointing.
 
XMAN245 said:
Shaders for 480 are still 512, he got it wrong.
Die size is still not known so how the hell did he get this one right?
Power draw based on some guy at GDC. Hell, it could have been an ATI guy telling him that. How about waiting for reviews to make claims to his accuracy about other stuff out side of the time for fermi. And people far more in the know than him are saying die size in the 500mm range, not 550-575 as he is.
Click to expand...

1) How do you know?
2) Remains to be seen, indeed
3) Remains to be seen, indeed. Though, he was pretty much spot on with the Unigine Benchmark, wasn't he?
4) Didn't he say 530mm^2?


Kaldskryke
Yep, that's another possibility.
 
Raap said:
Unfortunately for us and video card prices, most leaked estimates put it at just above the 5870's level, meaning it probably won't be anywhere near the 5970 except in a few heavily nvidia-biased games like FC2.
Click to expand...

I agree but I'm still treating those leaked benches with a big grain of salt. I think nVidia found performance to be worth it at such high TDP/wattage levels.

Afti said:
And here I was hoping that Nvidia had somehow managed to show up with a card that was only moderately hotter than the 5870 and had significantly better performance. Not likely, but it'd be nice.

Oh well. I suppose its EXISTENCE will cause some 5000-series price drops, but it's still disappointing.
Click to expand...

If all the rumors are true I really doubt ATI cards go down in price. Sad really, I've been jonesing to buy another 5870 when the price drops. I might have to wait for a while more.
 
XMAN245 said:
Shaders for 480 are still 512, he got it wrong. Actually he's stated that the 480 has 512 cores several times
Die size is still not known so how the hell did he get this one right?We'll see, but right now it looks like the lower limit is 525mm^2. He predicted a lower limit of 530mm^2 back in July
Power draw based on some guy at GDC. Hell, it could have been an ATI guy telling him that. How about waiting for reviews to make claims to his accuracy about other stuff out side of the time for fermi. And people far more in the know than him are saying die size in the 500mm range, not 550-575 as he is. Who are these people and where are they hiding. 3B transistors on 40nm will not be 500mm. 525 is probably the smallest possible die size. As for power consumption, let's just take a wild stab at this and say that if the Tesla part at 448SP's and low clocks had a TDP of 225W that a 480SP part at high clocks will pull a heck of a lot more power
Click to expand...

So yeah, considering how tightly held the information about this chip is, he has been surprisingly accurate with his stories. If he's wavered between "530mm^2 ++" and "550mm^2", excuse him, he clearly wants the sensationalist story and has enough lee way not to be perfectly accurate on the figures. But even if it hits 525mm^2, he's still "right" since the question is whether it is 23mm*23mm or 24*23 or 24*24, and up until at least October he said it is 530mm^2 or larger. Picking nits about whether he was off by 1-5% on die size is just asinine.

Power consumption he's had down. Performance he's had down. Of course you have a bone to pick, but that's your modus operanda. I wonder what you'll be saying in 1.5 weeks?
 
Last edited by a moderator:
275 watts for a single gpu video card? Yeesh, thats only 25W less than the dual GPU 5970. Considering how much ATI had to work to get the 5970 into a 300W envelope, I really doubt that Nvidia will have a dual GPU card to fight with the 5970 any time soon. 275x2 = 550W! :eek: And the magic mark for maximum power consumption of a single computer part is apparently 300W. Ouch.

ATI still has the fastest card on the market. 5970. The success of Fermi will depend on how much the 480 can beat the 5870 and how cost effective it is in relation. Nvidia obviously can't price the 480 at $600 when it won't even beat the 5970 that sells at that price point. The 5870 costs $400, so IF the 480 can beat that by about 20% performance, I'd expect the pricing of Nvidia's new flagship card to come in around $500 per video card.
 
I'm not too knowledgeable on the subject, but assuming it does use the 275w and they are throwing more power at the card just to make it work, what do you think this will do to the failure rate and overclockabilitity of the gtx480?
If it does only outperform a 5870 by 10-15% but can barely oc above stock, an average 5870 oc should just about close the gap.
With all the rumors of cutting shaders to make it run and how much power it's using, could we see a large percentage of them needing rma? That would surely suck if it costs $500+.
 
With the rumored limited numbers at release, getting one needing a RMA would not be pretty. :eek:
 
Kowan said:
With the rumored limited numbers at release, getting one needing a RMA would not be pretty. :eek:
Click to expand...

i wonder if any of the video card companies are going to offer lifetime warranty with fermi...
 
TomGreen321 said:
Where are the guys blasting that site because he makes stuff up only to hurt nVidia's reputation? Normally there are several posts of that nature by now.
Click to expand...

lol thats how i feel. those people sure are quiet.
 
tvdang7 said:
lol thats how i feel. those people sure are quiet.
Click to expand...

At this point, I think people are tired talking about Chuck. His whole being 'controversial' spiel has gotten boring.

If he reported his rumors without so much self flattery and bias he would be much more credible and have some semblance of being a professional.

The content of his ridiculously lengthy articles would amount to a sentence or two at a decent site like Anandtech.
 
My main rig F@H 24/7 pulls up to 800watts from the wall. My 1200watt PSU isn't worked as hard as a lower wattage one would be.
 
time for AMD to buy nvidias pc graphics cards business. nvidia should focus on smaller things like mobile graphics and ion.
 
LawGiver said:
time for AMD to buy nvidias pc graphics cards business. nvidia should focus on smaller things like mobile graphics and ion.
Click to expand...

Nvidia still has more than majority share of the discreet card market. Even if Fermi flops, they are in no real danger. Isn't AMD still bleeding money anyway? What are they going to buy Nvidia's PC graphics business with? Hopes and dreams? A 5970?
 
What do the investors think about this little "oopsie"? Do they just go by the big Iraqi information minister quality PR and not what's really going on?
 
I'm still going to wait for the proper reviews instead of relying on more speculation.
 
Wow 275W for 1 freakin card?

Man Might as well go with a 5970.....

Ugh I was thinking maybe ill go Fermi...Hopefully the honest reviews will show differently, but With what Kyle posted. NO WAY im going to upgrade my PSU just to go Multi-monitor gaming. I freakin Paid $280 bucks for this damn thing, No way in hell im going to get a 1k watt PSU For 2 nvidia cards. I wont do it.
 
kllrnohj said:
Nvidia still has more than majority share of the discreet card market. Even if Fermi flops, they are in no real danger. Isn't AMD still bleeding money anyway? What are they going to buy Nvidia's PC graphics business with? Hopes and dreams? A 5970?
Click to expand...

Here's another worrying about financials. All we want are GPUs, not financial forecasts.
Toyota is a top car maker but they are taking a PR beating now. Hyundai is doing pretty good but they're not even a blip on the radar.
People will always want quality, reliable and affordable products from any manufacture no matter what their financials look like. Fermi is not that.
 
CoW]8(0);1035446698 said:
At this point, I think people are tired talking about Chuck. His whole being 'controversial' spiel has gotten boring.
Click to expand...

this. and I think he or what he writes doesn’t deserve the privilege of being mentioned and discussed a lot around here.

even if he was accurate about everything he said (and he’s sure not), he says it in a way that doesn’t respect the readers minds and intelligence.
 
Still not official and personally I don't like the chance that this is wrong as it will give the trolls a chance to hijack threads when it is released about how the card is so great because the rumours were not true.
 
akotlar said:
So yeah, considering how tightly held the information about this chip is, he has been surprisingly accurate with his stories. If he's wavered between "530mm^2 ++" and "550mm^2", excuse him, he clearly wants the sensationalist story and has enough lee way not to be perfectly accurate on the figures. But even if it hits 525mm^2, he's still "right" since the question is whether it is 23mm*23mm or 24*23 or 24*24, and up until at least October he said it is 530mm^2 or larger. Picking nits about whether he was off by 1-5% on die size is just asinine.

Power consumption he's had down. Performance he's had down. Of course you have a bone to pick, but that's your modus operanda. I wonder what you'll be saying in 1.5 weeks?
Click to expand...

That funny, I haven't seen one review of the cards yet, have you? Noone knows what the TPD or power draw is for the card cept for the reviewers who aint leakin it as I'm sure their is atleast 1 site out there that would love to spill/leak that kinda of info. Least of all Charlie and his GDC someone. GTX480 is 512SPs. Again, the ONLY thing he has been accurate about is the time, and none of that was why he said it was.
 
You must log in or register to reply here.
Back
Top