GF100 evaluation thread
- Thread starter rpg.314
- Start date
lol @ 3:15
I still think HD 2900 XT was more disappointing. It was only sorta on par with 8800GTS 640. It was hotter. AA imploded performance, and AF was more demanding for it than for G80. The FP16 focus and its insane bandwidth proved incredibly off target because today a lowly 4670 with the same 320 shaders but ~1/4 the bandwidth is generally its equal in modern games!
And 3870 too showed some of R600's ugliness when it dropped 100W off load consumption!! I expect this to happen in NV's refresh too. Fermi is a mess on the manufacturing side and there must be huge gains to be had.
And 3870 too showed some of R600's ugliness when it dropped 100W off load consumption!! I expect this to happen in NV's refresh too. Fermi is a mess on the manufacturing side and there must be huge gains to be had.
I'm still convinced looking at thier line up that the 5850 and 5870 were meant to be $200 and $300 . The 5830 is just a salvage part that was done later on because prices were kept very high. I think before launch amd knew nvidia was delayed and by a while and decided to take advantage o fthe price. Low 40nm yields pushed prices up even more because of demand.
I hope at some point these cards drop down in price otherwise i think dx 11 uptake will be really harmed
Why? Juniper/Redwood are very good value.
Bouncing Zabaglione Bros.
Legend
Problem is that if it takes B1 to fix some of these things, it's quite possible the market will have changed around 470/480. If NI makes it out in Q3/Q4, then by the time Nvidia have got Fermi fixed, it will be up against the next generation.
Huge gains are not enough, it's got to be taken in context with the state of the market at the time, and I'm not sure Nvidia has much time left given that they've already blown half a year getting it out the door even in it's current overvolted and cut-down state.
Sure Nvidia can hang around and wait for TSMC to improve their 40nm process, but then AMD will be eating their lunch (again) by the end of the year.
FrameBuffer
Banned
hmm 5.21% voted YAH for Both !!
Let me guess who.. Razor, Silus, Degustor, Xman and .. ?
Let me guess who.. Razor, Silus, Degustor, Xman and .. ?
Yoga master, did you swim all the way from the other side of the island?
In any case considering your recent annoyance at ATI about their power usage with multiple monitors, you'd be positively furious with Nvidia had you got a 480 at launch.
The issue here is that theres no way that they are going to be able to make a laptop derivative even with a cut down version as they'd simply burn too much power. Does that give AMD full rein over the laptop space?
I suspect that their one saving grace is that they are both supply constrained at this point. Our favourite Dave probably couldn't cut the price even if he wanted to. His margins are probably excellent but the overall profit is probably lower than it could have been had he twice the number of wafers to work with. I wonder if the 5830 is more due to the fact that they want to increase overall volume of their parts as they are still struggling to balance supply/demand than any desire to fill a hole in their pricing structure.
Don't worry, after rain comes sunshine, after NV30 comes NV35.
Sontin.
Hmm..., nice smell. Is that a gf104, fresh from the oven?
I actually voted meh for both. They are good cards, just the power consumption is too high.
Bouncing Zabaglione Bros.
Legend
Of course that is the alternative. Junk Fermi in it's current form, do a redesign and respin, and launch 570/580 by the end of the year. Then make a video where your engineers make fun of the customers who bought 470/480.
If Nvidia had any sense, they would have been working on the fixed/redesigned product as a contingency from six months ago when it became obvious their design was not going to cut the mustard on TSMC's 40nm. Depends if anyone at Nvidia had the guts to stand up and say Fermi was not going to fly in it's intended form and they'd better start thinking of something else.
Maybe they were too busy thinking of how they could sidestep into the HPC market to avoid competition in the graphics market, but with these temps and power requirements, that's not looking likely either.
I disagree , the only advantage they offer over last gen is dx 11 and lower heat/ power. Its not really anythign to write home about if you own a half decent last gen card
Hmm..., nice smell. Is that a gf104, fresh from the oven?
GF10x is inherently the same design and as long as they share the same parts (i.e. the crappy GDDR5 controller carried over from the same ill-fated GT21x parts) they will look just as flawed compared to competitive parts.
We'd have to wait for Fermi2; leaner design, more fun! (It's like this really obese girlfriend, Sure it's fun she'll help you tick your "carnal fun" checklists, but it would be better if she was 100 pounds (or 100mm2) lighter.
What? Ljonberg hasn't voted yet?
Indeed, there's no real upgrade from RV770 to RV840.
Rootax
Veteran
The architecture seems great to me, keeping in mind that nVidia did a lot of thing for computing efficiency too, with >2x performances over gtx285 when the application is a little optimized.
Yeah they're late & hot, but at least they have a solid base now, and with better fabrication process, thing will get better...i guess.
Yeah they're late & hot, but at least they have a solid base now, and with better fabrication process, thing will get better...i guess.
I'd say, junk the B1. Shoot for gf10x and 28 nm.
