Cookie Monster
Newcomer
Refresh on the same node? I guess that could work, since they can start cutting some of the "fat". Maybe they've aimed too much this time around.
GF100 evaluation thread
- Thread starter rpg.314
- Start date
Refresh on the same node? I guess that could work, since they can start cutting some of the "fat". Maybe they've aimed too much this time around.
A refresh sooner than this fall wouldn't make much of any sense. Though the rest of the family is likely to come by then.
Only had the chance to comment on the results of the reviews now.
I was surprised with the GTX 480, but underwhelmed with the GTX 470. Although priced correctly the GTX 470 doesn't provide (on most occasions) a good upgrade over a GTX 285, except for the new features of course.
Now the GTX 480, despite all the doom and gloom, is performing more or less as I was expecting and it's not even the full blown chip, since it has some units disabled. Its theoreticals are so well below the competition, yet it manages to beat it almost every time. It does have a drawback: power consumption. I'm guessing they are suffering from leakage problems and that's causing the excessive power draw.
This architecture sure has some legs in it. I can't wait to see what a full blown chip can do (especially if the rumor of it having 128 TMUs is indeed true), with proper clocks (at least 750/1500/4200 Mhz) and the power usage improved.
I was surprised with the GTX 480, but underwhelmed with the GTX 470. Although priced correctly the GTX 470 doesn't provide (on most occasions) a good upgrade over a GTX 285, except for the new features of course.
Now the GTX 480, despite all the doom and gloom, is performing more or less as I was expecting and it's not even the full blown chip, since it has some units disabled. Its theoreticals are so well below the competition, yet it manages to beat it almost every time. It does have a drawback: power consumption. I'm guessing they are suffering from leakage problems and that's causing the excessive power draw.
This architecture sure has some legs in it. I can't wait to see what a full blown chip can do (especially if the rumor of it having 128 TMUs is indeed true), with proper clocks (at least 750/1500/4200 Mhz) and the power usage improved.
That's exactly what I'm thinking. They are probably working on reducing that "fat" (heh) ever since they knew they were going to delay the launch of the cards based on GF100.
Yeah. It would be nice to see GF102 at the same time GF104 is rumored to come out (in 3 months)!
A word of warning. NV's newest campaign involves getting everyone shouting GF104! GF104! and how good it will be compared to GF100.. I can only say.. don't hold your hopes up. Specs put it past GT200, performance and (power!)characteristics don't.
(i'll even make a GF104 swear jar myself, since it's going to be ridiculous the coming months.)
Bouncing Zabaglione Bros.
Legend
That's pretty damning - already promising how the next part will be better. All from a company that doesn't talk about unreleased products.
The cost rises non-linearly with die size.
A 10-15% win with a 58% larger die, is an architectural smack-down in my book. To be fair, fermi is the beginning of a new architecture (with a lot of headroom) while ati needs substantial improvements in the near future to remain competitive. Considering fermi's gestation, it is not at all obvious that amd can pull them off without missing their initial clocks/yields/power/area targets either. Overall, Fermi is a nice architecture. GF100 is it's meh implementation. The heat/noise and missed clocks only serve to add insult to injury.
No idea about voltages.
Care to elaborate? Specifically for yields instead of binning.
Charlie said ~10K 480s and ~20k 470's. Let's make it 50K total chips from 9K wafers @$5K/wafer.
That makes it $900/working chip.
Yeah, there's no way those are the correct numbers.Charlie said ~10K 480s and ~20k 470's. Let's make it 50K total chips from 9K wafers @$5K/wafer.
That makes it $900/working chip.
Let's see what Fuad will do this time around after the shaft he received at the hands of nvpr @gf100.
Er, isn't the GF104 supposed to be a cut-down version of the GF100? It might be "good" in efficiency terms compared to the GF100, but I sincerely doubt it will be higher-performance.
I guess the "other" campaign that tried to make GF100 worse that it turned out to be, has nothing to do with the bleak expectations for GF104 ?
Isn't that surprising
Which campaign? GTX480 is the most power hungry graphics card ever made (exceeds TDP even in current games), it's also the most noisy graphics card ever made, it's overpriced, still unavailable, and its performance is 15% better than single chip competitor, despite it was planned to compete with dual-GPU part. What exactly is better on this product, than the most pesimistic predictions stated?
Can't exactly say I am overwhelmed or stoked about the Geforce GTX 400 series, rather the opposite but it does add one interesting thing. It delivers better minimum frames per second than the Radeon HD 5800 series as far as I can see. Not that I am going to spend €499 to get in what may be known as a very exclusive club, as my Radeon HD 4890 still fairs well and plays nice in Mac OSX .
You've forgotten the dust-buster?
