NVIDIA GF100 & Friends speculation

[DegustatoR]

It's just a shame they weren't planning on releasing it with Less™ though

I wonder how could they plan otherwise...

 

[Ailuros]

Well it doesn't in any case chime in well with the latest 3.2b transistors.

 

[Rys]

Why's it suddenly gone up the size of 200 Voodoo1s? I thought 3.0B was understood.

 

[Silus]

Why's it suddenly gone up the size of 200 Voodoo1s? I thought 3.0B was understood.

It was the number thrown in an article concerning "zero defects" that neliz posted a few pages back.

 

[Ailuros]

Why's it suddenly gone up the size of 200 Voodoo1s? I thought 3.0B was understood.

http://www.semiconductor.net/article/438968-Nvidia_s_Chen_Calls_for_Zero_Via_Defects-full.php

 

[mczak]

yep

It's just a shame they weren't planning on releasing it with Less™ though

Scratching head... I can't interpret this any other way than meaning it'll launch with either only half the clusters enabled, or the launched product actually won't be GF100 based but a lower end chip... Both ideas don't seem to make much sense to me...

 

[Mize]

Scratching head... I can't interpret this any other way than meaning it'll launch with either only half the clusters enabled, or the launched product actually won't be GF100 based but a lower end chip... Both ideas don't seem to make much sense to me...

The correct conclusion is neliz likes to promote chaos.

[muhwahahahaha]

 

[neliz]

The correct conclusion is neliz likes to promote chaos.

[muhwahahahaha]

I would like to say "you'll see for yourself shortly" but that seems to not be the case either (smells like A4 to me)

 

[Bouncing Zabaglione Bros.]

Jeez. At this rate we'll be buying the AMD refresh product before we find anything out about Fermi. HOW WILL 5870 PRICES COME DOWN!? <shakes fist at sky>

 

[Mize]

I hope you're joking neliz.
I assume the implication is A3 can't hit 700 but needs to or the yields are so low it isn't manufacturable (seems more likely since they are showing hardware). Would suck.

Chaos or leak? Hmmm...

 

[Rys]

This thread can't help itself

 

[Silus]

I hope you're joking neliz.
I assume the implication is A3 can't hit 700 but needs to or the yields are so low it isn't manufacturable (seems more likely since they are showing hardware). Would suck.

Chaos or leak? Hmmm...

As silent_guy often explained, a metal spin will not miraculously increase clocks. Yields seem to be a problem yes, but I'm not sure that is the cause for what neliz is saying, if it's true of course

 

[Sxotty]

Jeez. At this rate we'll be buying the AMD refresh product before we find anything out about Fermi. HOW WILL 5870 PRICES COME DOWN!? <shakes fist at sky>

Maybe larabee will get restarted Otherwise you can just enjoy the brave new world where competition no longer exists.

 

[Silus]

This thread can't help itself

Heh, I'm sure you can clear/explain some things

 

[MfA]

What they can with metal respins is quite immaterial, they have had enough time to do a new tape out.

 

[Novum]

Something like what ATI did with 5870, or going even further?

ATI changed nothing in the texture sampling of HD5870 in that regard. The base filter is still bilinear.

 

[MfA]

ATI changed nothing in the texture sampling of HD5870 in that regard. The base filter is still bilinear.

The hardware filter is bilinear ... I assume though the 4 "read address probes" per SIMD can all address the texture cache individually when you don't use the filter, could be wrong though. Have you measured the point sampled throughput?

 

[FrameBuffer]

Haha, oh dear. Its not the number of texture units, its how you use them?

Is that what she said ?? ((sorry had to))

 

[Squilliam]

Is that what she said ?? ((sorry had to))

Im sorry you had to find out.

Neliz and I are more than GPU buddies, we have been Crossfired...

 

[trinibwoy]

Wow, we are in dire need of something material to talk about