GF100 evaluation thread

[RobertR1]

Dual Monitor users need to read this:

http://www.legitreviews.com/article/1258/15/

 

[rpg.314]

Dual Monitor users need to read this:

http://www.legitreviews.com/article/1258/15/

I think it should be fixable in drivers.

 

[CarstenS]

What about the option "470 yay, 480 nay!"?
IMO 470 is the more well-rounded package compared to the larger thingie.

 

[Squilliam]

Is it me or have a bunch of 5870 cards just sold out on Newegg? There are only 4 cards available and at that only 1 is a reference design whilst the others are higher priced variants.

 

[Silent_Buddha]

This stuff would have been great 6 months ago. Now its pretty average in that it doesn't really change the game and for most people the prices/performance will be no different in a few weeks to how they were two weeks ago with the exception that they can now pick their favourite colour. Red goes faster!

6 months ago this still wouldn't have been great, although it'd be slightly less bad.

We have fan noise somewhere between HD 2900 XT -> GTX 480 -> 5800 FX.

Both idle and load power similar to 2x 5870's. I don't think R600 was even that bad compared to G80.

The one saving grace, with regard to games, is that at least it's generally faster than 5870.

But 13-15% faster (at least in the reviews I've read so far) doesn't quite make all of above into a good thing even had it launched on time.

Out of the two the GTX 470 is at least reasonable-ish.

So it may not be another Nv30, but it's a bit of a mix of R520 (not as good) and R600 (not as bad in that Nvidia doesn't have to price the 480 similar to the competitions salvage part).

Just imagine Nvidia's pain if it had to launch when 5870 was retailing for 379 USD.

Regards,
SB

 

[rpg.314]

What about the option "470 yay, 480 nay!"?
IMO 470 is the more well-rounded package compared to the larger thingie.

You are the second person to ask for it, so I guess I messed up there.

 

[Alexko]

Crazy power draw and thermals, unreasonable noise, 6 months late, huge die, questionable availability... All that and it's 5 to 20% faster than the 5870 depending on your resolution and AA settings?

That's a big "meh" for me, and it's only "meh" because ATI has kept the 5870 above it's launch MSRP. The same relationship pretty much applies to the GTX 470 and HD 5850.

That said, it's nice that AA, even in 8X, is now almost free.

 

[nyt]

470 is so close to 5850 that given its well known shortcomings (power/noise/late) I can't see how it is a better value. Meh, not OK in my book. Even if the 5850 should be cheaper (then I would have bought it instead of a 5770).

 

[CarstenS]

Just take into consideration that not everyone goes by pure length of Fps bars. There's other factors that can decide if a card's interesting for you of if it isn't.

While i can see the power/noise issue (at least 470 has idle power somewhat in check, albeit only at GTX285 level, not the 20ish watts a 5850 uses), i fail to see "lateness" as factor for someone who makes his buying decision right now. Lateness has been punished by all those who've bought a 58x0 in the meantime and rightfully see no reason to switch card at this time.

 

[NathansFortune]

I think both cards are too late and too hot to be anything other than meh. That said, I think Fermi has a lot of legs and Nvidia have got their hard work out of the way and should be set up for easy scalability for a couple of gens. Now they need to work on getting more out of Fermi in the way of computational power while lowering die size and cost.

I think a 28nm version of GF100 will be very competitive and even a 6GPC version would give a linear increase in power for very little added to the core so GTX5x0 series could be very competitive for Nv. We'll see.

 

[liolio]

I think both cards are too late and too hot to be anything other than meh. That said, I think Fermi has a lot of legs and Nvidia have got their hard work out of the way and should be set up for easy scalability for a couple of gens. Now they need to work on getting more out of Fermi in the way of computational power while lowering die size and cost.

I think a 28nm version of GF100 will be very competitive and even a 6GPC version would give a linear increase in power for very little added to the core so GTX5x0 series could be very competitive for Nv. We'll see.

Funnily I think the exact opposite, ATI won't to do much to beat them as far as single chip GPUs are concerned. 28nm process may prove as tricky as 40nm one, Nvidia may not be able to continue to design crazy huge chip, they just hit the power limit, they have no room left, etc.

 

[Jawed]

NVidia's kinda lucky it's so late - if it was on time there'd have been no D3D11 "benchmarks" to put any kind of sheen on it. It would have been laughed off court, instead of being welcomed with a sort of embarrassed silence.

I think the architecture's promising. Texturing and generally getting data into the ALUs looks very healthy. Tessellation and rasterisation look robust. It's forward-looking. Running older games that don't scale well, faster, is over-rated.

The actual product, well, it's pretty poor if you're a gamer. If you're doing CUDA with a workstation you're prolly gonna have a ball - when it arrives, or if you don't need double-precision right now.

If NVidia can get the next chips to look less embarrassing as products and NVidia can start executing again, Fermi architecture might yet prove itself manufacturable.

Jawed

 

[Rebel44]

For me its fail - I wouldnt pay for either more than for 5850 or 5870 because 470 and 480 are louder, more power hungry and hotter.

Fermi = R600 (neither of them is/was as bad as NV30 which was worst GPU ever IMO).

 

[liolio]

By the way I voted meh for both cards as it only makes sense for super enthusiastic audience and because availlability will be a joke for a while.
I also wonder about the merit of the architecture in regard to scaled down derivatives.

 

[Bouncing Zabaglione Bros.]

This is what I wrote in the other thread:

Yeah, it seems 470/480 is at best equal or a little faster than the 5850/5870, whilst using more power, being hotter, and noisier. The power/noise videos at [H] are completely off-putting just because of the noise alone. At the same time, it's not the fastest card you can buy - if frames are what's important, you can still trump the 480 with a 5970.

All in all, Fermi has turned out to be what I expected - a problematic product that was forced out the door late and compromised, and manages to only match the competition, but at the cost of more money, heat, power, and noise. It's a stopgap product to get Nvidia into the DX11 market, but it's costing them in terms of the larger die and poor yields. Nvidia will have to improve things over the next couple of quarters, either with a respin or improved TSMC process.

There is also no sign of mainstream products to compete with AMDs full range, while AMD may be able to bring out it's next generation by Q3/Q4, just as an improved Fermi and it's mainstream versions begin to come through in quantity. With lowered prices, factory overclocks, and two gig cards from AMD, competition will get even harder for 470/480.

I just can't see any reason to buy either of these cards over their AMD counterparts. Performance is very similar, and when you take into account the big increase in noise, power, cost and heat of the Nvdia cards, the AMD cards are more attractive hands down.

The 470/480 are just not competitive cards unless you are a corner case that must have niche capabilities (ie Cuda or Physx). When you look at them as a whole, including performance, noise, heat, cost, power, the Nvidia cards are just not good when compared to the AMD products. I wouldn't have a 470/480 if you gave me one because of the noise alone - I couldn't live with it.

 

[rpg.314]

I think the architecture's promising. Texturing and generally getting data into the ALUs looks very healthy. Tessellation and rasterisation look robust. It's forward-looking. Running older games that don't scale well, faster, is over-rated.

What are the improvements texturing wise? TUs are now part of SMs, clocked higher, but that's that. Are you including unified texture caches in this?

I think all of ati's recent execution prowess/track record is going to be put to a very stern test as they try to climb the distributed setup/raster/tessellation mountain. Not to mention a brand new process due for the end of this year. We'll know if it was just nv fucking up or the new bits are the really hard deal. Having said that, ati has the luxury of spreading out the changes over cypress->hecaton->NI.

 

[seahawk]

I am on the "Meh" side in the current revision of GF100. It consumes too much power and the cooling solution is too loud for my taste. As I do not only use 3D for gaming a quite card is most important to me.

The features and architecture look promising. I can imagine buying a GF104 for my PC at home. (if they get it to around 150-170W load.

 

[trinibwoy]

Why no option for GTX470 yay! and GTX480 no-no?

Yeah I don't know what people think is wrong with the 470. It's faster and more expensive than the 5850 and a wash perf/$ wise. I voted Yaay for both. It might be "Meh" compared to Cypress but the 480 is a decent upgrade feature and performance wise from my 285 and that's where I'm coming from.

Power consumption is the one major ding against it. I want to hear reports from early adopters on noise though, that could kill it for me. Reviews have been inconsistent on that front.

I just can't see any reason to buy either of these cards over their AMD counterparts.

Comfort that you won't be missing out on anything? Like Carsten said there are lots of reasons people spend their money and they won't necessarily share yours.

 

[Bouncing Zabaglione Bros.]

Power consumption is the one major ding against it. I want to hear reports from early adopters on noise though, that could kill it for me. Reviews have been inconsistent on that front.

[H] have got open case videos of all the major cards with their noise going from idle to full load in their review. The noise/power usage is pretty awesome.

 

[ferro]

Meh for both. It's clear they had to push the power beyond reasonable limits to beat the 5850 and 5870 by a few percents. The 5970 remains out of reach, and a dual GF100 card seems out of the question.

I think it relies on brand-loyalty for its sales.