Is this really due to 130nm process?

[Bopple]

I've been waiting for GFFX for very long with anticipation. But this is something wrong.

Ridiculous heating, cooling system, amazing(?) clock speed, whopping(?) performance,
release delay, and then this immature drivers for that long preparation time.
The whole sum of these are self-contradicting. What's happening here?
I'm making my 2 cents here...

While nVIDIA was babbling about 130nm process all the time, i'm feeling there is something behind.

I mean...they were agrandizing themselves with arrogant contentment, underestimating ATi.

This leads to being overjoyed out of no reason, implementing those immense overspec of DX9
and some confusing new technologies like dynamic pipelines.

But with this, they wasted their precious transistors on nothing and lost too much performance power.
Even worse, 9700pro was more than expected and almost upto its paper specs.
So they couldn't beat 9700pro with prototype GFFX.

However, they couldn't just scrap FX and re-design all over again at that point especially when
9700pro was out already. So their only option left was to overclock it.

That's ok. But it generated too much heat to overclock enough to outperform 9700pro.
So they needed some heat reducing technologies. Didn't they mention about such things?
But they failed to get those techs to fruit.

I see there's no or little of such techs on GFFX considering the current heat generations.

Anyway, they failed and alternatively / inevitably chose this grand eyecandy cooling system. But even with that,
they couldn't handle the heat problem enough to get things right on 130nm wafers. What's the temp? 150F? or 70Celsius?

Wasn't it a bit weird while TSMC could handle other 130nm stuffs good, they couldn't FX for what? about 6months?

Considering these things, then i can gather together why its driver is so immature contrasted to their previous drivers.
So to say, They had no real GFFX board on their hands regardless of 130nm process. And therefore they had
no good enough time to mature the driver.

If i were CEO of nVIDIA, i would have sent some kind of DX9 benchmark program of their own with the package sent to review sites.
For it is the only thing left that GFFX can boast its muscles.

But they didn't. Why?
Overly pressed, they perhaps missed the timing to release it, or its results no better than current ones.
I don't know which is true. But the latter seems more likely to me to be the reason.

Am i assuming too much? maybe. it's all due to 130nm. and i hope so. for i've waited for this damn thing too long.
But something deep inside tells me this is not the case.

 

[tart666]

Like Anand said, the decision to use 0.13um process was made 2 years ago, way before the process was actually qualified.

The other part of the problem is the brand new design

(edit) Everyone will just have to accept the first generation of FX is going to suck, like the first P4 core sucked (remember sock 423?), like Itanium 1 sucked, etc. It's gen 2, 3, and higher that matter.

 

[Bopple]

(edit) Everyone will just have to accept the first generation of FX is going to suck, like the first P4 core sucked (remember sock 423?), like Itanium 1 sucked, etc. It's gen 2, 3, and higher that matter.

i agree with you. still, that's not allowable excuse for they wasting their very precious transistors on nothing and overhyping.
some say it's the review sites that hyped. but do you really believe there was no nvidia pr team behind all those hype rumors?
i feel they cheated us thoroughly.
still it has to be seen the real launch, 3dmark2003, doom 3 and so on...but frankly speaking, i have no faith anymore.

 

[tart666]

Well, you can't tell PR people to stop trying to hype up a crappy product. The capitalist world as we know it would stop right there.

Bottomline is, only a few dim-witted individuals with too much cash will be duped into buying FX. We will all be better off for it.

 

[chizow]

The card doesn't suck, its just 6 months late and has been totally trumped by the R9700pro, which raised all our expectations. The 130nm process definitely needs work, as these chips run too hot and draw too much power (smaller should mean cooler with less power reqs.) Aside from that, the core looks to be a real winner, just too hot atm. The real shortcoming of the GF FX is the 128-bit memory interface, something DDR-II has no way to overcome. DDR-II only raises potential/theoretical bandwidth, but if the bottleneck is between the GPU and the Memory (the 128-bit interface), it'll rarely reach its maximum peak bandwidth.

Chiz

 

[Keysplayr]

Maybe Nvidia just jumped to far too fast. They forgot what it was like to be pressured by being easily on top for the last five years. I just know that when the Radeon 9700 published its first benchies, Nvidia was there watching. It probably took a week to mop up all of the saliva off the Nvidia boardroom floor because of all jaws hitting it. Does anyone know the difference in the transistor count from GF4 to GFFX? I think it was said that the GFFX has something like 120million transistors. More than double what a Pentium 4 has. Maybe this is why they cant tame the heat even at 13u.

Keys

 

[Bopple]

Nods. all of them do hype.
but there should be limit.
look how enormous it was. how much misleading...how many ppl got influenced...
and the hell...for more than 1 year?

maybe they had no other choice really. but even with that in my mind, i feel cheated and am angry.

anyways...i still wish it proves itself to be a better product for us all.
even though it were improved by 10~15%, i'd say it's still inferior product to 9700pro.

 

[Bopple]

Originally posted by: keysplayr2003
Does anyone know the difference in the transistor count from GF4 to GFFX? I think it was said that the GFFX has something like 120million transistors. More than double what a Pentium 4 has. Maybe this is why they cant tame the heat even at 13u.

Keys

9700pro has about 108 millions unless i'm wrong. and it's on 150nm process. and that uses conservative cooling system.
what about GFFX? got 120 millions on 130nm and that hell of a joke cooler. and even with that monster cooler, it's at 150F. LOL
It's definitely overclocked. and to the limit.

 

[chizow]

Originally posted by: Bopple

Originally posted by: keysplayr2003
Does anyone know the difference in the transistor count from GF4 to GFFX? I think it was said that the GFFX has something like 120million transistors. More than double what a Pentium 4 has. Maybe this is why they cant tame the heat even at 13u.

Keys

9700pro has about 108 millions unless i'm wrong. and it's on 150nm process. and that uses conservative cooling system.
what about GFFX? got 120 millions on 130nm and that hell of a joke cooler. and even with that monster cooler, it's at 150F. LOL
It's definitely overclocked. and to the limit.

Probably has a lot to do with the flaws in the 130nm fab process. Don't forget, the reason the GF FX was delayed so long was b/c yields were not high enough to begin mass production (may still not be). Also, everyone who has gone to a 130nm process has had serious growing pains, the P4, the XP T-bred family, its all the same, they had major issues before core and fab process revisions. I honestly expect this core to be a winner once yields and the fab process improves...look at the numbers, this thing is cranking out some serious polys and does extremely well in all the shader tests; the 128-bit interface is strangling it though. NV35 should be a nice part if they can address the heat issues and give the thing a 256-bit memory bus.

Chiz

 

[Bopple]

Yeah it'll get better as we saw on Intel and AMD. But likewise, as we have seen, not on this FX section.
This FX is already fixed one. Nothing like improved mass production atm can save this FX.
I'm sure they'll make much cooler and faster one on the next gen.
But nothing can change the fact at your face now.

And as my original article says, i don't believe they're delayed mainly by 130nm process.
They failed because they used good-looking techs like over-DX9 spec, some strange but useless techs, and DDR-2 (and 128-bit bus due to it)
I'm thinking if they didn't do those bragging arrogance, they could have beaten 9700pro easily or just with mildly overclocked FX unlike this joke.

And for shaders, i think you haven't read reviews thoroughly. 9700p is better than FX at advanced shading and vertex shading(15%, 50% faster)
FX is better at pixel shading(about 10%). But ppl keep saying FX will crush 9700p when 3dmark2003 or DOOM3 comes. LOL

It's not shader which lets FX on par with 9700p. It's overclocking and 9700p having only 1 texture unit per pipeline. Though the latter is not that significant.