The GT300/Fermi Thread - Part 2!

[saaya]

You can't add bus width like that. AFR mirrors the memory for GPU1 in GPU2 so the bandwidth is still 256-bit effective, iirc.

is it? what really matters is the available bandwidth per frame rendered, isnt it? and for a dualgpu card that bandwidth IS double...
if you look at how dualgpu cards perform compared to single gpu cards of the same model and single gpu cards of the next generation with almost double the memory bandwidth per gpu and double the gpu horsepower per gpu... youll notice that dualgpu cards are a pretty reliable perf indication for next gen single gpu cards... not 100%, but a pretty good indication...

a dual gpu card is basically gpu raid0, you dont gain memory space by adding a second unit, but you gain bandwidth...
a dual fermi card thus has effective bandwidth of a 768bit gddr5 bus...
i venture a guess here that fermi2 and even fermi3 single gpu cards wont beat that :o

Nvidia, get the molasses out your ass. Seems the older your company gets, the slower it gets, the dumber it gets; beat you with the inside of a horses' ass.

i heard nvidia is notorious for not promoting staff, only if people leave and then get hired back they have a chance to sit in a more comfy seat or make more money... plus they are known to drive their employees pretty hard, 14+ working days... sooner or later that starts to hurt the employees performance...

Being informed/educated is a consumers responsibility, to think that that the consumer is not capable of making an informed decision by reading the box or ask questions of the product they intend to buy is ridiculous.

hey i totally agree... it IS ridiculous... but its the sad truth

About AMD copying Intels naming scheme I think it's because they want people to think that their 965 is equal to Intels 965.
Just like they did with the 3000+ etc trying to convince people their 3000+ equals a Pentium 4 3000mhz.

for the + rating from amd... no idea what their itentions were, but the idea behind the + rating during the a64 days was really good! amd never talked that much about it, but they actually ran a series of benchmarks, sysmark, office, excel, games, photoshop and a few others, and they based the + rating on the performance of a cpu in relation to an original athlon 1ghz (where applicable since it didnt even have sse)

so when amd launched their dualcores, those + ratings were pretty good... they didnt relate to games and other applications all that much, but for productivity they were pretty spot on iirc

I remember one episode of this "vicious circle" a couple weeks ago. Somebody was making claims about the performance of the Fermi without having any documentation/benchmarks.

He was claiming boldly that he knew more about the Fermi-performance back in July than most us Know today, because of his "insider knowledge".

At the end he wanted to use the picture of his beer belly to prove his groundless claims about the performance of an unreleased product. Just because the claims for his "insider knowledge".

the sad part is that i know you wont even apologize for these remarks once it turns out he was right...

I wish more people would do that, then this would be a 1 page thread.

yes and forums wouldnt exist, there would only be 1000 chatrooms on the whole net and 5 news websites... is that really what you want?

back on topic:
what do you guys think how fermi will scale cpu wise?
will it need as much cpu power as a 295 to max out, less, more?
and how tdp/temp limited will it be when overclocking it?