GF100 evaluation thread

You're probably right, but if I were to judge by my card, which does 1GHz on default voltage, then they do overclock pretty nicely.

Speaking of oc'ing - check it out - http://www.fudzilla.com/content/view/18265/1/

That's some serious stuff . Almost P30k in Vantage - single GTX480.

 

Yea, power consumption especially is the problem. Once the leakage is fixed and yields are up, I would love to see what a fully enabled chip, with decent clocks, can do

 

I've just checked the prices again, and it looks like there is a slightly cheaper 470 now available for pre-order that puts the price a few pounds under the 5870

Cheapest products in each class at OCUK. The 470/480 don't look that attractive pricewise compared to their AMD counterparts.

5850 - £220.99
GTX470 - £304.99
5870 - £314.00
GTX480 - £445.99
5890 - £499.99

 

I don't doubt there are a decent number of Cypress chips out there that can hit 1GHz on stock cooling and with sufficiently low voltage as to keep power and thermals in check. The question is: is it enough to create a whole new SKU for?

 

About GTX480's diminishing advantage once it reached 2500x1600 resolution : what are you guys suggesting ? is it an architectural problem ? or a driver problem ?

I read somewhere (can't remember exactly) , that it could be a combination of both ..

If true , then GF100 architecture is definitely worse than GT200 in that regard.

 

I voted Meh for both. Architecture looks good, Fermi2 could potentially be a winner (much like R600 -> R770), as long as nVidia manages to work out the kinks and start executing on timely basis.

For now I'm not impressed, 10-15% extra speed for 30% extra price, along with much more power, heat and noise? No thank you, even if Fermi would be priced on the 5850/5870 level, I still would think twice before picking them instead of 5800 series, but at least they would be competitive, they are not at the moment. Small initial batch is meant for the fanboys-only IMO.

Picture this: according to Techpowerup:
http://www.techpowerup.com/reviews/N..._Fermi/32.html

Difference at 2560x1600 is 7% only, 5870 with 2GB will shave few more %, and AMD could easily boost clocks of picked chips 10-15% more, and launch 5890 2GB tomorrow at 499$, and in bigger quantities than Fermi.

And thats not even talking about a real refresh, AMD had almost a year for it (Evergreen chips were ready way before launch, AMD waited for TSMC to ramp up).

The question is - will AMD launch 5890 2GB with higher clocks to finish Nvidia off for this generation, or they wont bother, and just milk 5800 series till NI shows up in the 2H?

 

At those prices the 480 looks bad yes. The 470 is fine though - it regularly bumps up next to the 5870 and is most often faster than the 5850.

HardwareCanucks review. Nvidia claims there's some driver issue causing an unexpected fall-off at 2560x1600. Should be easy enough to verify when new drivers arrive.

 

Like I said, any numbers I provide would be irrelevant because you've already made up your mind. Premature accusations, whining etc. like I suggested the thing you need most is that group hug. I have to make some adjustments to my Ignore list now.

 

The 5870 is cooler, quieter, uses less power, and has generally better performance than the 470 for £10 more, and you get that £10 difference back in your first power bill! After that, you're making money!

 

Any data to back it up? I've calculated average difference from all games tested at TPU (since they tested most games - old and new ones) at 1920x1200. Average Fermi's advantage is 14.6%, while average advantage for games, which get less than 120 FPS (on either board) is 15,3%. Pretty consistent, I'd say.

 

My bold. There was only ever one company that could provide a satisfactory upgrade for trinibwoy's requirements. I'll be kind and call it "brand loyalty".

 

You got all that from me linking to a computerbase article to support a point which you have yet to address? Amazing! I haven't made up my mind about anything - the numbers are right there in plain sight.

And here I thought silly season was over.

Of course. Why in the world would I go from a 285 and lose features by choosing an AMD product? What are they offering me besides a quieter fan? You make it sound like a hard decision

 

Thats what Charlie thinks, but IMO Fermi2 will take at least another 6+ months and I doubt nVidia will be happy with current A3 till then, they have to keep delivering now, in whatever quantities they can. B1 respin should be ready about now, it would make sense to start shipping higher-yielding lower-power parts ASAP, instead of waiting for Fermi2 miracle in undefined future.

Thats because GTX480 competition is not 5870, but 5970 (if its single or multiple-gpu's cards are not relevant, its companies strategy to go with huge monolithic die or not). Since nVidia high-end cant beat AMD's high-end, they have to lower the prices, much like AMD did with R600 and derivatives.

 

This early in the video cards' life, smart money is on driver problem.

 

Yeah I figured. I thought he had some objectivity left.

 

Performance, DX11? Although I agree that for most people it might not be worthwile, as it's not worth it to upgrade from Cypress to Fermi.

Out of curiosity - what features do you use (not have) that made your 285 irreplaceable thus far? For me, PhysX is dead in the water (with laughable 10-15 titles that support it), so I had no problem switching from 285 to 5870... And since I don't code for CUDA, it's not useful either.

 

I'd be lost without my shuttered glasses too.

What if those were titles you actually wanted to play?

 

Not this again...Why do some feel the need to try and justify that guy's wrongs, by twisting everything into a "reality" that doesn't exist ?

The GTX 480 was never meant to compete with the HD 5970. Never did NVIDIA counter a dual GPU with a single GPU and this time, it's not different. Even more so, because of the delays.

 

You would be right if AMD would launch 5890(?) basing on the same not-preselected chips, but since even they OC very nicely, its not hard to imagine specially binned GPUs would do just fine with higher clocks 24x7. What to speak of respin (or two) AMD could have pulled in the meantime?