NVIDIA GF100 & Friends speculation

[DavidGraham]

I have many doubts it is just a respin , respins don't give you 20% more performance (if that is true) , I would believe it if the clocks are insanely high , but even at 800/1600 , it still doesn't make sense .

Either they used a GF104 derivation , or they slightly modified GF100 architecture , diluting some bottlenecks here and there .

I wouldn't put much faith on Charlie news these days , ever since GF104 and he has been wrong many times , on both Nvidia and AMD fronts .

 

[Florin]

Redesigned/replaced. That's separate from it being a viable product.

Hmm, I haven't said anything about that so I'm not sure why you keep spinning this to me.

 

[Bouncing Zabaglione Bros.]

Hmm, I haven't said anything about that so I'm not sure why you keep damage control spinning this to me.

Because this came up due to Charlie and his claim that Fermi is "broken", I'm addressing the various ways that Charlie has used that term. GF100 in it's current form is a product, but it's not the originally promised 512 core product. The rumours are that it's not making them money, and yields were bad enough to have to overvolt a cut down version of that product with it's associated heat and noise.

If you want to claim that GTX580 is a fixed GTX480 (and therefore Charlie was wrong), then you have to ask where the originally promised configuration of the GTX480 has gone - it was obviously not a viable product for Nvidia to ship for technical and economic reasons ie broken.

Or are you saying that any time a company introduces a new product to replace a problematic older one, that somehow it proves the previous problematic version was not faulty/problematic/broken? Did FX35 prove that FX30 was not broken?

 

[CarstenS]

If you want to claim that GTX580 is a fixed GTX480 (and therefore Charlie was wrong), then you have to ask where the originally promised configuration of the GTX480 has gone - it was obviously not a viable product for Nvidia to ship for technical and economic reasons ie broken.

I wonder what that is (my bold).

 

[Florin]

Because this came up due to Charlie and his claim that Fermi is "broken", I'm addressing the various ways that Charlie has used that term. GF100 in it's current form is a product, but it's not the originally promised 512 core product. The rumours are that it's not making them money, and yields were bad enough to have to overvolt a cut down version of that product with it's associated heat and noise.

If you want to claim that GTX580 is a fixed GTX480 (and therefore Charlie was wrong), then you have to ask where the originally promised configuration of the GTX480 has gone - it was obviously not a viable product for Nvidia to ship for technical and economic reasons ie broken.

Or are you saying that any time a company introduces a new product to replace a problematic older one, that somehow it proves the previous problematic version was not faulty/problematic/broken? Did FX35 prove that FX30 was not broken?

Ah yes, at the end of the day all things in life just boil down to that pivotal moment in history, NV30

Charlie does enough spinning of his own - I doubt he needs your convoluted interpretations why he was actually right when he said GF100 couldn't ever be fixed until 28 nm.

 

[Bouncing Zabaglione Bros.]

Ah yes, at the end of the day all things in life just boil down to that pivotal moment in history, NV30

Way not to address anything else I wrote.

Charlie does enough spinning of his own - I doubt he needs your convoluted interpretations why he was actually right when he said GF100 couldn't ever be fixed until 28 nm.

Are you still suggesting that because there's a GF100b, therefore GF100 was not broken? Then why bother with a b chip at all? I'm pretty puzzled why you're spinning the line that GF100b proves that GF100 wasn't broken, so hah, Charlie gets it wrong again.

Why launch it months ahead of shipping products at an analysts call? Pretty puzzling for a company that doesn't comment on unreleased products.

 

[Florin]

Are you still suggesting that because there's a GF100b, therefore GF100 was not broken?

GF100 actually seems to be holding the performance crown quite nicely so far. But I hadn't addressed whether it was broken or not at all, so the answer to your question would be no. All I'm observing is that it is apparently fixable now.

Then why bother with a b chip at all?

Why bother with any improvements at all? That's not a serious question is it?

I'm pretty puzzled why you're spinning the line that GF100b proves that GF100 wasn't broken, so hah, Charlie gets it wrong again.

Why launch it months ahead of shipping products at an analysts call? Pretty puzzling for a company that doesn't comment on unreleased products.

Hmm what's puzzling about any of this? I'm fairly sure you already have some explanations ready.

 

[SimBy]

GF100 holds performance crown, but as far as market goes it's a flop. I'm not sure GF100b can change that. I was expecting some sort of GTX460 scaled up.

 

[DarthShader]

Hmm, I haven't said anything about that so I'm not sure why you keep spinning this to me.

And Charlie didn't say anything about Fermi being fixed, so I'm not sure why you keep spinning this to him.

He did mention bug fixes, but those are often not enough to call something "fixed". I am sure one could post many software examples here.

 

[Bouncing Zabaglione Bros.]

GF100 actually seems to be holding the performance crown quite nicely so far. But I hadn't addressed whether it was broken or not at all, so the answer to your question would be no. All I'm observing is that it is apparently fixable now.

Not so much so that Nvidia is willing to announce a poor quarter, and cannibalise sales with an unreleased product only six months after the shipping of GF100.

Hmm what's puzzling about any of this? I'm fairly sure you already have some explanations ready.

I'm puzzled as to why you are being so obtuse. Charlie obviously wrote of GF100 being broken metaphorically as a viable product, referring to performance, die size, yields, power/heat/noise issues, selling price in the face of AMDs smaller, better yielding parts, profit margins etc.

You seem to be acting as if Charlie was talking literally, that every GF100 part was physically broken, that you would put it into a PC and nothing would happen. That's obviously not the case, so I don't understand all the rah-rah that a year after their competitors (more likely nearer eighteen months if they don't ship this year), Nvidia finally gets a revised GF100 working as it should be and is being rebadged into a new product line. That's really not something to be patting Nvidia on the back about.

The fact that it's being announced at investors day, well in advance of shipping products to bury bad financial results, cannibalising current sales, and in preference to acting as a Cayman spoiler speaks volumes.

 

[neliz]

The only performance crown the GTX480 currently holds is on the matter of it's sometimes ridiculous power consumption.
People saying that it currently holds the performance crown have already applied their own personal preference on it.

Maybe the title "It's the fastest graphics card between $350 and $450" would be a more realistic view on things.

I wonder what that is (my bold).

When you CEO announces that the new chip has "up to 512 Shaders" .. people will think that there will actually be products available with 512 Shaders.

 

[BRiT]

Can we please stop beating the "Broken Charlie GF100" horse? Or at least push the love/hate into Charlie's dedicated thread?

 

[rpg.314]

I have many doubts it is just a respin , respins don't give you 20% more performance (if that is true) ,

Why not? Especially for a silicon respin. Also, Charlie said 1.5GHz was the target for GF110, then how is this clock 20% higher than gf100.

 

[Sinistar]

I wonder what that is (my bold).

From here http://www.hardocp.com/article/2010/01/17/nvidias_fermi_gf100_facts_opinions

 

[Florin]

The only performance crown the GTX480 currently holds is on the matter of it's sometimes ridiculous power consumption.
People saying that it currently holds the performance crown have already applied their own personal preference on it.

Maybe the title "It's the fastest graphics card between $350 and $450" would be a more realistic view on things.

Well, I wouldn't want to argue that point with one of the authorities on applying personal preferences to GPU discussions, but I do note that you're starting to talk about GTX480 now. What I said was that GF100 holds the performance crown, as in it being the fastest GPU out bar none. That is likely to change any day now, but then half a year going on 9 months ain't terrible.

 

[BRiT]

When you say "fastest GPU" you're talking solely from the single chip perspective and not from the single product perspective?

 

[Florin]

When you say "fastest GPU" you're talking solely from the single chip perspective and not from the single product perspective?

Well of course. Would you, say, claim Hemlock is the fastest GPU?

 

[BRiT]

No, but from a consumer product perspective it doesn't matter. I see the title of single fastest GPU even though it's not the single fastest product similar to being described as "She has a great personality".

 

[Florin]

No, but from a consumer product perspective it doesn't matter. I see the title of single fastest GPU even though it's not the single fastest product similar to being described as "She has a great personality".

Hmm well I would say it does actually matter now and then. I mean that other 'fastest product' isn't actually always the fastest at any point in time, due to it being a bit of a special needs kid that requires a lot of love from the software side of things. Which is often late to arrive in the first place, as well as having a tendency to come and go.

Oh and of course that other product is indeed a lot more expensive as well.

 

[neliz]

Oh and of course that other product is indeed a lot more expensive as well.

*cough* you're making up excuses *cough*

In other words, you're spinning.