Fermi 'slips into January'

Page 3 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
Yea, ya mean like this?
feature_image05.jpg
 
Well, can't say I'm surprised since we predicted this weeks ago. Though, until I see ATI's cards back in stock, I find it hard to give any credibility to release dates from NVidia, since at this point they are at the mercy of TSMC and the quality of their third revision. And if AMD can't get enough of a smaller and clearly working card, how will NVidia get large numbers of a larger (and still in testing) card. It just won't happen, but I certainly didn't expect things to be looking this dismal now (for both companies) a few months ago.
 
Yes, they are in stock where I live at least. So that is good news for ATI, but the best indicator for Fermi production capabilities would be the 5870, which is out of stock where I am. So, some good news at least.
 
True that. Even the 4770s were eventually ok for stock before the 5 series launch.
I too think its die size involvement.
Its not like if you get 40% from say 100 chips on a slice, smaller chips, ones twice as large youd get 20, its more like 15, as the defects are disbursed, and easier to land and corrupt a larger area, say, having 10 per slice
 
It would be good to know how well the 5700's are selling tbh. They are decent cards no doubt but they aren't completely killing Nvidia at the price point unlike the 5800's.
 


Technically isn't that like 7 month old news...

"Nov 1 now becomes Jan 15. Since Nvidia is 'teh awsum' at making new parts of late,"

http://www.theinquirer.net/inquirer/news/1051818/nvidia-gt300-promised-october

Oh that Charlie, he doesn't like nV, but he is alot closer to the reality than others out there.

Not saying he's accurate, just.... Semi-Accurate. :lol:

Hey I just couldn't resist, really, anyone who thought November is really precious for their wide-eyed optimism, we need them working on solving world hunger and poverty. :sweat:
 
Exactly Ape. I've read some of those articles where they show us that Fermi CAN'T come out by the end of this year. People just say "he hates Nvidia", or "take it with a truck load of salt, he's never right". I kept asking for them to use any real logic and show me how getting first spin silicon in Sept equals a usable card by Dec. JHH wasn't even holding a card! He was holding a heatspreader! If you look at the "card" that was shown, the odds are high there was nothing under the heatspreader.

I really wish TSMC had their $h!7 together so that AMD could be "raking it in" right now. That would really help them. These shortages aren't good, for anyone.

Seeing as this seems to be a popular thread, can any of you who are smarter then me answer something? If GF is going to start making chips for others, how does their different process effect the chips? I'm assuming GF would be using the same 45nm, etc process that they use for CPUs. How would an SOI 6780 be different from a TSMC 6780? Will we need to start worrying about who made the chip?
 
Can't start worrying about who made the chip even if that was a problem it would be hell for pr if the same named chip performed vastly differently over their life spans that the logical choice would be have each production facility be exclusive in their products.

I'd hardly think anyone except nvidia zealots are surprised at the 2010 date December was being hopeful. But when for 2010 march seemly likely + - a month.
 


We'll see, it depends on the complexity of the protein models, if the go with the current largest models, the HD5870 would stay pretty close with a generalised OpenCL GPU / CPU client. It looks to be about a 20% difference under OpenCL. It depends alot on what the proposed hteory of Fermi is and the reality and how it works under the generalized client. There are a few possible bottlenecks in there, and alot of the HD4K+ barriers will be removed like the LDS issue.

Right now I wouldn't put much faith in what Fermi will and won't be, theoretically the HD2K-4K should've been folding beasts, but unfortunately not how it was implemented. :pfff:
 


Gotta go home (been here at work since 6:30 [hate the new network!!] ) but to things to consider;

A) GF is going to start GPUs on 32nm's optical shrink of 28nm sometime mid-late next year (which is why people talk about the new core late 2010 [to also match the LRB launch timeframe]).

B) Even if they do have a semi-mature process by then, it doesn't guarantee that it translates elsewhere, remember IBM's East Fishkill site made the FXs and they were disasterous (although in light of this I wonder how much blame could've been shared in retrospect) despite being a relatively mature process for IBM theoretically, the new fab was terrible initially, only finally being the normal 'solid yield' about a year and a half later when nV switch it's excess to UMC+TSMC instead of IBM+TSMC. I wouldn't guarantee anything on an early 32nm based production line which really hasn't worked complex ICs yet, just memory IIRC. I think they are doing CPUs for ealry 2010 on 32nm , but the optical shrink to 28nm I think will be first on the GPU if I'm not mistaken.

I haven't checked, but I don't know the extent of the 32nm fab for GF, whereas TSMC looks to be trying to catch up to intel (although still about a generation and a half back high high-k gates).

Anywhoo, I'd double check, but I gotta run. One would hope it will be better, but I wouldn't bet anything on it.

T.
 
Izzy lets pretend that SOI 32nm chips will be better then TSMCs. TSMC could make all the low/mid range cards, while GF makes the higher end ones. That would fix the PR nightmare. Or if SOI is worse just flip it around. I've been wondering what more then two companies making chips means, as AMDs and Intel 65nm processes were very different. Their 45nm processes are different as well, but not to the same degree.

If Ape or anyone else has some links on the subject, I'd be happy to read them. I've been wondering, and looking forward to educating myself.
 

I'm quite impressed with Charlie this year, as much as he dislikes Nvidia, he is very often right. For example when he said (last year!) Nvidia will be exiting chipset business, I though he was crazy, who would waste a market worth more than a billion? A year later it all makes sense now, Intel cut Nvidia out, its not worth for them to spend on R&D on AMD chipsets too - AMD makes as good if not better chipsets, cheaper. Bumpgate and Fermi being late didnt suprised me as much, yet almost everyone jumped on Charlie guts, just to find out he was right all along 😉
 

And a 5870 should be able to open a can of whoopass on a GTX295, but it doesn't, because the client is so obsolete 🙁

Oh, and congrats on having the first silver badge (next to your username) I've seen :) I'm still waiting for a bronze one :cry:
 
It's a really complex subject actually, but yes it's facinating.

http://cleanroom.byu.edu/everything_wafers.phtml

You can get some good starting info there, a bit dry for me but useful.

There is also a certain mysticism around it because it is a multi-$billion yearly industry. You won't get real figures because they guard them but from what I understand AMD have better yields than intel do at every process, although intel are always first to a new process.
 


Charlie is sometimes right...but in the times he is right, he is right for the wrong reasons. So that still makes most of his arguments still totally wrong...
 
I wouldnt say that BSD. Certainly in the case of Fermi he has practically written the story months in advance.

Of course that takes a bit of luck, but in this case he's been pretty stunningly accurate for the past 8-9 months.
 


For example, if Nostradamus says that the world will end due to nuclear war, but it actually ends due to a giant meteorite....the result is the same, but the reasons are wrong. So Nostradamus is still incorrect.

Accurately predicting the end result without accurately predicting the cause...

Besides, saying Charlie makes accurate predictions is just cherry picking the articles that are correct while ignoring the articles that are totally wrong. It's not exactly hard to foresee Fermi's delays given the heaps of trouble Nvidia has been going through...
 
I have to admit I started to question some of his later articles, putting them down mostly as 'wishful thinking'.

But tbh...those 1.7% yield figures are starting to look pretty plausible right? At least you can believe it now that TSMC are back below 40%. Take a scratch-built chip that big, with Nvidia's lack of expertise on so many fronts and you can quickly come to the conclusion that 1.7% might even be generous.

http://www.semiaccurate.com/2009/09/15/nvidia-gt300-yeilds-under-2/

It really doesn't look so far off now does it?
 


Yeah here's hopping for the OpenCL client ASAFP (the F stands for folding, not FU.... :evil: ), it's sad the amount of lost potential there, not from a BS marketing standpoint, but just from the spinning your wheels in neutral standpoint. It's not like if there was an OpenCL client today ATi could possibly sell more cards with their current shortages, but unfortunately alot of science is being left in the double calculation. At work we still can't use the new cards cause no one is an OpenCL guru yet, and I'm not working in that department anymore, so I'm not busting my huump learning F'in code, when I can play with shiny toys (currently getting ~20Mb/s on my wireless connection :sol: ). I have to say ATi dev team needs more guys, but I bet there's a hiring freeze there too. I just wish we were already a year into the future when people were more adept than we are now. It's all promise and so little to really work with right now, and no one has spare cash to hire young students yet to become the future experts. So much wasted potential. 🙁
Funny thing speaking of Charlie's article(s), did you read Theo's @ BSoN about Batmangate, really a good read, even if you don't belive whatever part of it, and interesting attempt at balance in a very contentious issue.

Oh, and congrats on having the first silver badge (next to your username) I've seen :) I'm still waiting for a bronze one :cry:
Click to expand...

I'd like to thank the Academy, and my Parents Hannah and Barbera. [:thegreatgrapeape] [:thegreatgrapeape:3] [:thegreatgrapeape]
 
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom