Nv40 16 full pipelines- The Inq.

btw, does someone know where I can find the link to that tremendously stupid video nVIDIA's ballistic PR-staff did on the nV30 soundy fiasco ?

 

This whole transistor argument is pretty silly.

Its meaningless to compare transistors to get anything more than a rough measurement of complexity or functionality.

It doesn't translate well to die area, either (as cache is much higher density than standard cell logic).

It just doesn't mean much.

 

Surely you know the funniest use of the word 'grapefruit' in LGOP?

(Yes, I know I'm OT, but do you really expect me to be on-T, in this thread? )

 

http://www.fileshack.com/file.x?fid=2923

 

Expect? No.

Urge? HELL YEAH!!!

 

I praise you PaulS

 

Not only do I not know the funniest useage of the word, I don't even know what LGOP stands for. Could you please enlighten me?

(Sorry for the OT, but Dio got me all curious. )

 

Man, the horse in this thread is really getting a beating. People seem unable to discuss the rumor, unlike the ATI thread and have spent most of the last 4 pages beating horse.

 

Problem is, the horse ain't dead yet.

But getting on topic then...what would you like to personally "discuss" about this rumor?

 

There's some speculation over at R3D that half of the 16 pipelines are "crack pipes", lol.

One thing that's been quite clear for some time is the potential raw bandwidth they have available - up to 51.2GB/sec. This certainly goes someway towards supporting the 16x1 idea (and possibly useable 8x/16xFSAA modes?).

0.13u, 205-210 million transistors @500MHz+? Hmm...

 

I just read the thread and was going to post the same thing. Personally, I'd like to discuss something other than the stereotypical, "Yeah, yeah. We've heard it all before - a couple of months before NV30, and then NV35. It was bullshit then, it's probably BS now." comments. I mean, we all know that ATI is one hell of a company and the undesputed masters of the universe, but must we repeat this in every thread instead of discussing technical aspects?

Or, my personal favorite was an early responce which attempted to turn the 16pipeline rumor into an nVidia negative, of course, by stating that nVidia failed to hit their clock target, so within the period from the "second to last stepping" to today they basically redesigned the IC! That's freaking amazing, there's no way ATI can compete with that design speed.

And beyond the amazing redesign/synthesis speed, nVidia "lost 10%" clock but gained a "net 20%" overall with this late-breaking redesign. My Gawd, if only nVidia ad people who actually do the cost-benefit and statistsics before starting a multimillion dollar, several year project. But nVidia is a bunch of fucking morons - they're no ATI for sure... altough with the speed they redesigned that with, they don't need to think first.

 

Watch your language

"But nVidia is a bunch of fucking morons - they're no ATI for sure"

You lose your argument and make yourself look like a pathetic child with your profanity and stupidity with such comments.

 

What do you expect? Nvidia is simply reaping the same kind of respect as they have shown to the market.

This rumour makes very unlikely claims, and based on Nvidia's history of miscounting and relabelling it's pipes to make them appear more than they are, is it any wonder this story is being treated with scorn?

If NV40 does arrive and is labelled as 16x1, I bet you it will prove not to be that once it is analysed, just as NV3x is not 8x1 as claimed by Nvidia.

This is just the kind of spoiler that Nvidia puts out as unsubstantiated (and deniable) rumour every time a competitor launches a product. The reason why no-one can discuss is sensibly is because no-one believes it for a second.

 

Not that there isn't any justification for those comments.

When you hear a rumor, the first thing you do is evaluate the liklihood of it (or part of it) being correct. And whether you like it or not, the past is going to be used as a factor in evaluation.

What did you think would be discussed?

You're right...I just don't know why nVidia can't seem to find a "reliable foundry" to make their chips...TSMC!...no...IBM!....no...TSMC!

Seriously, my best guess would be like what most others are guessing: If there is any truth to the rumor, that it's the same kind of "pipelines" that we see in NV3x, which would make NV40 8x1 / 16x0 operation. I of course never rule anything completely out though.

The transistor count just does not look reasonable to me though, if it is an "apples to apples" comparison with the NV3x transistor count.

 

Re: Watch your language

Why it's true, read this thread. At every turn nVidia is diminished as company with the ability to compete in the high-end discrete market, I'm merely restating the obvious.

Thoughts about how this 3rd party rumor could be true (of course, never admitting because they actually did something adventageous):

How is nVidia anything but useless? They're a bunch of idiots.

 

So the # of "transitors" still means hot?
I read: Heat+Loud= no OEM contract.

 

Re: Watch your language

Certainly shows how Nvidia has alienated the early adopters and potential buyers of their new high end cards. People just don't trust Nvidia to tell the truth about their products. Nvidia's actions over the last year to prop up NV3x are now coming back to haunt them. Nvidia have spent their good name and pissed away lots of customer loyalty.

The piper always needs paying sooner or later.

 

Re: Watch your language

I think the question you should ask yourself before getting all worked up and indignant Vince is this, "Has nVidia earned the contempt/distrust/scorn of the graphics community or is it just misplaced frustration by fanboys?".

If it's the former I think the comments aren't a problem and are just the natural result of listening to total BS out of nVidia for a year and a half or so, if you feel it's the latter I guess you'd have a point.

Which do you think it is Vince?

 

Re: Watch your language

Vince, what I see in this thread is a lot of Justifiable Skepticism when it comes to anything related to specs.

Even you have to admit that nVidia's PR with specs and hype, cheating, in the NV3x was just dreadful. nvidia did that for obvious business reasons, and I can't even say I blame them for that from a business perspective. However, going that route doesn't come without it's downsides...namely, you lose a degree of trust.

OK then, what leads you to believe this could be true? Or that it's more likely to be true than false? Is there something in nVidia's history that shows they can drastically increase their tranny count, maintain clock-speeds, and keep power consumption all on the same process?

I'm not saying it can't or won't be true. But you're asking people to speculate how it could be true, when we have little basis for doing so.

No, their PR is a bunch of smokescreen artists.

 

Re: Watch your language