[SemiAccurate] Nvidia's Fermi GTX480 is broken and unfixable

[v8envy]

Count me among those hoping Charlie is pulling crap out of his ass.

But as Lonyo pointed out, some of the circumstantial evidence does hint at not all being happy in NV land. After a quarter of GPU drought I'm not looking forward to a full year of GPU shortage.

I'd also be concerned about the viability of EVGA. A great AIB maker with superb customer service, they simply can't survive in their current form on $130 MSRP and under G92 derivatives for another year.

So yeah, hope he's wrong for the sake of all enthusiasts be they red green or blue in allegiance.

 

[Zap]

EDIT - This actually wouldn't be surprising, given the fact nVidia has cut down on R&D and other investments. Read SAs article about cutting costs and investments.

That's part of the problem. Charlie has a habit of posting aritcles, and then referring to his own earlier articles as "proof" or "sources." That's like a murderer creating his own alibi.

 

[CurseTheSky]

I generally don't jump into these debates, but:

G100 is already late. Rumors are the chip is going to be large, hot, power hungry, and absolutely amazing performance wise. However, we're about a month away from the supposed release date, and we have very little concrete information at all.

In my eyes, this can go one of two ways. Either we'll have another G80 on our hands that'll decimate everything before it, or we'll have another R600. With the tight-lipped nature of Nvidia lately, I'm putting my bets on a R600 situation, though not as dire. They'll probably stuff these cards full of features (Stereovision, CUDA, Physx, a few new additions etc.) and use those as a major selling point. If we had another G80, I think we'd have some charts flowing around, or at least some solid info about the power requirements and lengths of the card(s).

Being primarily a Nvidia user, I hope this isn't correct. I hope they can release a decent card at a decent price, but the lack of benchmarks and other information doesn't leave me very optimistic. The current state of these debates in this forum reminds me a lot of the 2-4 months prior to R600's release.

 

[thilanliyan]

If the 40nm process is as bad as rumoured...nV should maybe devote some resources to getting the 28nm process out a little early...if that's possible.

 

[thilanliyan]

Pooled rather than "mirrored" VRAM?

There was a link a long time to an interview with an engineer from ATI who stated that "pooled" memory isn't the "holy grail" it's made out to be. I posted that link but can't find it anymore.

 

[sandorski]

I'm an AMD/ATI diehard and hope this article is not accurate. If it is, Nvidia probably only has a few months to fix the issues before they fall too far behind to ever catch up. I predict that in that scenario, Intel owns Nvidia within 2 years and that's where things get very interesting.

 

[WelshBloke]

I'm an AMD/ATI diehard and hope this article is not accurate. If it is, Nvidia probably only has a few months to fix the issues before they fall too far behind to ever catch up. I predict that in that scenario, Intel owns Nvidia within 2 years and that's where things get very interesting.

I think thats being a bit overly dramatic, at worse theyll have to write of this year and rush out fermi rev2. They may well get that out this year anyway.

 

[Mr. Pedantic]

If the 40nm process is as bad as rumoured...nV should maybe devote some resources to getting the 28nm process out a little early...if that's possible.

Exactly. If Charlie's right, then NVidia is about 2-3 quarters behind, give or take. This is untenable for an industry where a new generation comes about every 18-24 months. And Fermi is already a quarter or 2 behind Cypress. I think NVidia may have a better job of it just cutting their losses and focusing their Fermi team on making sure 28nm goes successfully, as well as the making of such big chips as NVidia seems wont to do.

G100 is already late. Rumors are the chip is going to be large, hot, power hungry, and absolutely amazing performance wise. However, we're about a month away from the supposed release date, and we have very little concrete information at all.

No, it was absolutely amazing, performance wise. Since then they cut down the core count from 512 to 448, and probably reduced clocks to boot. Just reducing the core count reduces performance by 12.5%. This puts the '60% faster than Cypress' down to about 40%. The clock reduction, maybe another 10-20%. We don't really know. And for all that, Fermi is still hot, power-hungry, and expensive to produce.

Count me among those hoping Charlie is pulling crap out of his ass.

I don't really see how hoping this and hoping that and sticking your head in the sand will really help. Maybe for Jen Hsun. But not for us. Not really.

Its either people at TSMC or the card manufacturers (EVGA, XFX etc), my bet would be TSMC as apparently NV has been giving them a bit of a hard time.

The reason everyone (and I mean everyone, AMD, NVidia, the general public...) is giving TSMC a hard time is because this is their fault. The transistor variability in 40nm, the via problems, any number of other untold faults that AMD and NVidia have to work with getting their products to market. I don't recall another fab process with such a bad rep.

 

[sandorski]

I think thats being a bit overly dramatic, at worse theyll have to write of this year and rush out fermi rev2. They may well get that out this year anyway.

Probably over dramatic, but ATI is also coming out with more this year as well. According to the Article, not only does Nvidia have serious issues with Volume, but they also have serious Pricing issues. Are they going to have to take a Loss just to keep Product on the shelf? Will Fermi rev2 not experience similar issues?

Add in the state of the Economy and all these possible problems really compound.

 

[Piotrsama]

EDIT - This actually wouldn't be surprising, given the fact nVidia has cut down on R&D and other investments. Read SAs article about cutting costs and investments.

That article was later corrected to say it's only a 10% drop, not 33% as originally stated.

 

[Mr. Pedantic]

http://www.semiaccurate.com/2010/02/17/list-questions-nvidia-doesnt-want-answer/

 

[SlowSpyder]

I know a lot of people here don't care for Charlie, but he has been pretty much correct about Fermi so far. The Nvidia specs released said 512 shaders, I'm pretty curious to see what Fermi launches with.

And as far behind as Nvidia is right now, I don't see them not being able to catch up. Look at where AMD was with the R600 launch, I doubt any of us thought that two and a half years later AMD would be dominating Nvidia in the gaming world regarding performance and newer/better technology.

 

[WelshBloke]

Probably over dramatic, but ATI is also coming out with more this year as well. According to the Article, not only does Nvidia have serious issues with Volume, but they also have serious Pricing issues. Are they going to have to take a Loss just to keep Product on the shelf? Will Fermi rev2 not experience similar issues?

Add in the state of the Economy and all these possible problems really compound.

Oh I'm not saying they are in a great place, I just don't think it'll spell their doom.

 

[Voo]

As usual Charlie probably vastly overstates the problems, but he points to some valid sources and Anand said similar things in his RV870 article so I fear there could be some valid points in his rambling


We better all hope the best if we want cheaper chips in the near future..

 

[sandorski]

Oh I'm not saying they are in a great place, I just don't think it'll spell their doom.

Time will tell for sure.

 

[alcoholbob]

Charlie's been pretty much spot on about Fermi so far. He initially claimed that there were 7 working samples of Fermi from the first production run. Nvidia did display a Fermi card with the number "7" labeled on it, so either Charlie's source was accurate or his guess was astronomically lucky.

 

[GodisanAtheist]

Is it just me or were there some comments from Nvidia recently about the Fermi refresh still being slated for 2010 despite how late Fermi is? If that's in fact true, then Nvidia might very well end up doing exactly what Charlie mentions at the end of his article: Soft-launch Fermi 1 just to say they got it out the door while plugging away behind the scenes to get the "refresh" out later this year.

 

[Schmide]

Charlie makes a pretty good methodical argument, we can only hope nVidia makes some good tweeks to invalidate his proposed scenario.

 

[cbn]

There was a link a long time to an interview with an engineer from ATI who stated that "pooled" memory isn't the "holy grail" it's made out to be. I posted that link but can't find it anymore.

Sounds like a good article to read.

 

[mm2587]

The reason everyone (and I mean everyone, AMD, NVidia, the general public...) is giving TSMC a hard time is because this is their fault. The transistor variability in 40nm, the via problems, any number of other untold faults that AMD and NVidia have to work with getting their products to market. I don't recall another fab process with such a bad rep.

TSMC's process might not be the best (ok its certainly far from the best) but some blame must be laid on nvidia.

Both ati and nvidia knew the process wasn't great. Ati engineered around it. Nvidia just complained about it.

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

 

[gorobei]

The real question is how many swings at this is NV going to take.

even if they take back the performance crown, it wont be profitable. the amd-smalldie-mainstream vs nv-bigdie-enthusiast strategy has failed once with rv770 v g200. if fermi repeats this then it will be strike 2.
how long will the board at nv put up with this strategy? could JHH even put out a small g400(or whatever follows fermi)?

 

[SolMiester]

http://www.semiaccurate.com/2010/02/17/list-questions-nvidia-doesnt-want-answer/

Those are questions Charlie will never get answered by nV. They refuse to talk to the guy, haven't done for several years.

ALL of Charlie's article's on nV are either second-hand, rumour's or just pulled out of his ass!

 

[StrangerGuy]

The real question is how many swings at this is NV going to take.

even if they take back the performance crown, it wont be profitable. the amd-smalldie-mainstream vs nv-bigdie-enthusiast strategy has failed once with rv770 v g200. if fermi repeats this then it will be strike 2.
how long will the board at nv put up with this strategy? could JHH even put out a small g400(or whatever follows fermi)?

Agreed. AMD already proved that the absolute performance crown doesn't matter with the 4xxx series. Sure, the GTX 285 is faster but it might as well be non-existent when I can get a 4870 for a quarter of the price.

 

[taltamir]

wait wait wait... did fudzilla get renamed into semi accurate?
that name is only semi accurate, they should be called "fully inaccurate"

Charlie is so pathetic. Don't click on the link, don't bother to read what he says. Forget about idiots like him... besides, we don't want to support that website (ad revenue from visitors)

 

[Mr. Pedantic]

Those are questions Charlie will never get answered by nV. They refuse to talk to the guy, haven't done for several years.

ALL of Charlie's article's on nV are either second-hand, rumour's or just pulled out of his ass!

So NVidia won't answer them just because Charlie asked them? What about if Anand asked them? Editors from Tom's Hardware, Bit-tech, [H]ard|OCP, B3D, etc? What about if I asked them? Just because Charlie asked them does not mean they are any less worthy questions, nor does it mean that the public don't need to know the answers.