NVIDIA GF100 & Friends speculation

Well, idle power consumption is a bit lower than the 5770 (or the 5750). Load power consumption is a bit higher, but not by much (comparable to the idle power consumption advantage). Noise is also lower.

So, the question is, which situation is likely to be the more common one? Idle or load? I'd be willing to bet that for most people, the GTS 450 will actually end up consuming less power in total than the 5770, plus it's cooler and quieter. However, it also doesn't perform as well, so its probably really seems to be price/performance, not power/performance.

Obviously SLI GTS 450's are just a bad option all around.

 

Are you sure about those "idle" numbers?
I mean, is the GF106 just like GF100, that is has an "idle" clock and a 2D clock, which is comparable to AMD's "idle" numbers.

 

Well, I'm going by Anandtech here:
http://www.anandtech.com/show/3909/nvidias-geforce-gts-450-pushing-fermi-in-to-the-mainstream/16

Shows the reference GTS 450 (and three of the four retail ones) three watts below the 5750, six watts below the 5770 at idle. A quick look at Tom's Hardware shows similar numbers.

 

what's weird is that the TDP of the 450 is mentioned everywhere (108W) but Annand says he didn't get it..

So yeah, no conclusion on the idle mode, I think this one clocks up just like the other Geforces once for 2D mode, but i'm not sure.

http://www.techpowerup.com/reviews/MSI/GeForce_GTS_450_Cyclone_OC/26.html TPU has the the idle mode just in between the 5750 and 5770.

 

Well, he was specifically talking about Idle TDP, which should obviously be much lower.

At any rate, power consumption is not a problem with this card.

 

At least it has double precision.

 

slightly overclocked 4-years old G80 offers similar performance to default GTS450 - despite GF106 has almost twice as many transistors... What's wrong with them?

 

Pretty sure the slightly overclocked 4-years old G80 also has much, much higher power consumption.

 

I agree, but that's caused by ancient manufacturing process. My point was architecture and (almost half) efficiency.

 

Why not compare a X1950 XTX to a modern day entry-card? WAY faster, but only half or less transistors …

 

It also has much more BW.

 

Hmm, that is an interesting point. I mean, this sort of increase in transistor count without an increase in game performance is normal with time, as architectures do tend to add features with time. But it does seem like a rather excessive increase in transistors for very similar clock speeds and performance, doesn't it?

 

G80 is vastly deficient in features - the rumour being that its deficit lead to D3D10 having features cut late in the game.

 

So, how much extra performance will GF106 deliver with 192-bit interface along with the extra ROPs? 20%?

Seems to me that NVidia might enable all the ROPs and increase clocks to fight Barts. Such a part would come pretty close to GTX460 768MB in performance.

 

1950XTX:380M transistors
HD5670: 627M transistors

HD5670 is about 40% faster than 1950XTX.

 

My bet is 5%.

Wider memory bus won't help by itself (128bit GTS450 is slower than 128bit HD5770, so available bandwidth is unlikely its limiting factor). And ROPs? GF104 with 32 ROPs and 33% more BW is about 10% faster than GF104 with 24 ROPs/192bit. And the ROPs and bus are fed by 7 SMs. GF106 with its 4 SMs won't be able to utilize this massive back-end.

 

Yes, I think nvidia deserves some credit for providing a decent reference cooler even with this class of cards (its only problem seems to be with SLI if two cards are packed densely, but that's not really a good idea anyway). Doesn't look though that too many manufacturers will use it.

I think the idle power advantage is due to downvolting memory - no matter how it is achieved, it's definitely nice to see things are improving there.
I think nvidia chose wisely not to try to beat HD5770 with the reference card by just upping clocks and voltages - everybody would have complained about a card with performance similar to HD5770 but power draw of HD5850 (or GTX460, for that matter). Still, the higher clocks (and voltage!) compared to GTX460 already show up, but the compromise isn't too bad.
Overall this looks all solid, beating GTS250 (with lower power draw) on average easily (OMG nvidia can finally retire g92b!!!), sometimes even close to GTX260, so this definitely is an improvement.
Looks like AMD is willing to fight back with price drops of HD5770, about time (well for consumers...), which could make life difficult for the GTS450 - though HD6xxx will probably be a much bigger problem.
One advantage GF106 (like GF104 too) has seems to be it's a bit less dependent on memory bandwidth compared to AMD's chips, hence memory chips should be a bit cheaper. It is also DP capable, which some people might miss from Juniper.
The die size still has me puzzled though - GF104 vs. Cypress is (depending on which of the two numbers you believe) either same die size or 10% more, yet GF106 vs. Juniper is nearly 50% more, despite both chips being mostly half of their respective brethren. In theory GF106 should downscale more effectively (due to less shared logic which doesn't downscale with Juniper) - granted Juniper also dropped DP which should save some transistors but still the difference is so big that even the (not even used) additional ROP partition imho isn't enough to explain it. At least the larger the chip the more easy it is to cool but that's certainly not the reason for it to be so big...

 

That's the problem though, at least in the US. Reference GTS 450 is priced the same as reference 5770 but is generally slower. OC'd GTS 450 has one card at 10 USD higher and others even higher than that. At that point you'd be better off just going for a GTX 460 at 170 USD. Taking prices from the Anandtech article since I don't see GTS 450 for sale yet.

And as Anandtech mentioned, Nvidia is the one that put the GTS 450 in that position. By aggressively dropping price on the GTX 470, they made AMD lower the price of the 5770.

So basically it has Idle power when computer isn't in use, 2D power when computer is in use at the desktop or 2D program, and 3D power use?

As well ouch at the overclock power useage (passing reference GTS 460 768 MB and 5850) and overclock temps (same thing).

Certainly not horribly priced, but definitely puts zero pressue on 57xx line. As well it's interesting that Nvidia's OGL perf at this level (in Wolfenstein at least) doesn't even allow the OC cards to surpass the 5770 and puts the reference 450 at approximately the same speed as the 5750.

Regards,
SB

 

An entry card would be more in the range of 5450 or 4350... Anyway, the point is, that it's worthless to compare older architectures with less features with modern cards - on any side.

How's this evil plan:
Nvidia is deliberately making their chips big in order to occupy more die space to prevent AMD from getting more wafers out of TSMC. Just stockpiling more of the same small chips won't cut it, because at some point you'd need an inventory writeoff which doesn't please shareholders. So: Big chips, big market share!

*SCNR*

 

1950XTX 380M transistors
HD4350 242M transistors
HD5670 627M transistors