NVIDIA Fermi: Architecture discussion
- Thread starter Rys
- Start date
They also said it would launch in 2009, but hey, I never believed that, especially since they were telling people March at the same conference if they signed the right NDAs.
I think they have to launch with A3s, mainly because the problems they need to fix, power, clocks and memory controller issues, are likely to require a full respin, not just metal layer. If they need to do that, it won't be Q1, most likely won't be Q2, and then you have to wonder if they will bother at all?
I am not sure Fermi would make a dandy Windows 8 part.
-Charlie
All good and well, but can they launch with A3? A GF100 with de-activated parts and clocking a merely 550/1200 is not going to beat a 5870 by a margin good enough to warrant the price NV needs to make money with that monster chip. They are screwed.
nvidia is selling a 487mm² Chip on a card for 120 €. nvidia was screwed all the way with the GT200-Series but they made more profit. Very mysterious.
May be they are screwed with the power and clocks. But nothing is over till the fat lady sings. :smile:
Let's pray for it's launch this festive season,
if only because we want real benches.
So I'm in the process of cleaning a significant part of the Charlie love/hate crud, and moving it to a very nice thread I made in RSPC just for the guys that have the irresistible urge to share their feelings on the topic. Go there and flame/one-line/whatever as much as you like, for as long as you like, but if the Fermi thread is polluted with that again some short or not so short term holidays will be happening.
And to the great dismay of many, Charlie won't be the first to go. If someone like silent_guy could take the time to properly address/combat his points, so can the army of posters that seem to have followed him in/seek only to snipe him...or, if they're incapable, they can go to RSPC and keep on doing what they've been doing here with such dedication. A bon entendeur, salut!
And here's the promised thread:http://forum.beyond3d.com/showthread.php?t=55901
And to the great dismay of many, Charlie won't be the first to go. If someone like silent_guy could take the time to properly address/combat his points, so can the army of posters that seem to have followed him in/seek only to snipe him...or, if they're incapable, they can go to RSPC and keep on doing what they've been doing here with such dedication. A bon entendeur, salut!
And here's the promised thread:http://forum.beyond3d.com/showthread.php?t=55901
Last edited by a moderator:
Just for reference could anyone give me an educated guess what the percentages are on functional first silicon for new GPU architectures? I think the delay with the Fermi occurred long before tape out, with everything happening now being pretty much within expectation of the normal development process (ignoring yield problems, which will hit them harder than anyone else simply for having more vulnerable area than anyone else).
Groo The Wanderer
Regular
There is the date TSMC promises a process, the date when it works, and the date that there is volume. For 40nm, that was October 08, August 09, and December 09 respectively. Other recent processes from them haven't been any different. I would expect 28 to not be real until Q2/10 in anything more than sample quantities.
Two years for a new step would put it much closer to the end of 2010 though, possibly into 2011.
-Charlie
Groo The Wanderer
Regular
No they are not selling it. Ask any AIB about allocation, they are getting just enough to make it look like there is stock. They are losing money on each one, so they have to cut it back.
If you stop and think about it, why are there not floods of G200bs again? TSMC has NO shortage of 55nm capacity, they will run wafers tomorrow with a phone call, probably give you a good rate too. NV is blaming a capacity shortage on the lack of G200b parts, but that is simply not true. Call TSMC, they will tell you.
NV isn't making them because they will lose money. FLOWERS, FLOWERS, AND EVEN MORE FLOWERS
-Charlie
Last edited by a moderator:
early february new nvidia numbers will be released and tell the truth.
Volume of sales is also important to hold market share.
If your theory is right, the volume of sales will be much lower.
Oh, then germany must be the only country in the world where you get GT200b cards...
That's only the GTX285: http://geizhals.at/deutschland/?cat...=&asuch=gtx285&v=k&plz=&dist=&pixonoff=on&xf=
And all the cards came in stock between mid-november and today. Your EOL story was wrong.
Last edited by a moderator:
LonelyMind
Newcomer
GTX285 are virtually extinct in Sweden. They are still listed, but most with delivery date in february. Same with GTX275, delivery feb. You can find some odd models/brands I never heard of before, in stock with unknown quantities, but that's it. Even saw the odd batch of never heard of Zotac GTX260, but nothing at all from the major brands left.
May the february date be in anticipation of Fermi arriving and replacing it?
Last edited by a moderator:
thatdude90210
Regular
Could be just no one is buying it. The EOL story has been confirmed by multiple sources. An example:Oh, then germany must be the only country in the world where you get GT200b cards...
That's only the GTX285: http://geizhals.at/deutschland/?cat...=&asuch=gtx285&v=k&plz=&dist=&pixonoff=on&xf=
And all the cards came in stock between mid-november and today. Your EOL story was wrong.
Anandtech http://www.anandtech.com/weblog/showpost.aspx?i=654
"Meanwhile NVIDIA started the end-of-life process for the GTX 200 series some time ago, with production of the GT200 GPU ramping down. So NVIDIA doesn’t need to play pricing games with AMD, as they’ve already planned on selling out anyhow."
Why should any etailer brought new cards when nobody wants them? The GTX285 was down to 5 or 6 stores in october.
And where is the confirmation? I throught the same thing but it was false. There is no quote from nVidia.
nVidia said in their last CC that they lost between 50-100 Millionen $ because they did not supply enough chips to their partner. And this is a problem of the geforce business.
Last edited by a moderator:
FrameBuffer
Banned
.. where is YOUR proof that "it was false" ?
It would seem to me that no matter what information or data provided, unless it paints nV in a positive view or light, your choice is to refute it with equally unsubstantiated claims. I'm all for people being able to express their opinions however it gets tiresome with the constant nV-bias.
Do you have some sort of affiliation with any entity that would profit from nVidia's support or are you just blindly biased ?
Edit: To moderator, feel free to edit/move/remove if deemed too aggressive or otherwise inappropriate.
Last edited by a moderator:
Sometimes some threads/members call a little aggression on themselves, but it appears you're only being aggressive on sourcing which seems fair since he was too.
Did you not see my link? I can only speak for germany but there is no sign of a EOL for the GT200 cards.
I like the truth.
Can you show me how a product is EOL when you have over 100 shops which can deliver one? And the EOL information is nearly 3 months old.
FrameBuffer
Banned
just a very very short list from a 3 min search:
ATI 3850
http://www.newegg.com/Product/Produ...0048 106792462 1067933728&name=Radeon HD 3850
http://3btech.net/paatirahdpro1.html
http://www.ewiz.com/detail.php?name=AT-3850AGP
http://www.buy.com/prod/visiontek-r...56-bit-pci-e-2-0-ati/q/loc/101/210692756.html
http://www.pcrush.com/product/Video...sionTek-Radeon-HD-3850-Graphics-Card?bsrc=sli
http://www.pcrush.com/product/Video...sionTek-Radeon-HD-3850-Graphics-Card?bsrc=sli
http://www.pcrush.com/product/Video...hnology-Radeon-HD-3850-Graphics-Card?bsrc=sli
http://www.amazon.com/dp/B00145CQUC...e=asn&creative=380341&creativeASIN=B00145CQUC
http://www.buy.com/retail/product.asp?sku=211031672&listingid=42037873
http://www.tigerdirect.com/applicat...CODE=GOOGLEBASE&cm_mmc_o=VRqCjC7BBTkwCjCECjCE
http://www.walmart.com/catalog/prod...0000003142050&ci_src=14110944&ci_sku=13214355
http://www.compuvest.com/Search.jsp?Search=100228L&advsite=froogle&dtm=20091218&sku=331003782-00
Intel 775
27 @ Newegg
http://www.newegg.com/Product/ProductList.aspx?Submit=ENE&N=2010340343 1051707842&name=LGA 775
180+ @ pricewatch
http://www.pricewatch.com/search?q=Intel+775&start=0&mn=Intel&gallery=0&cn=CPU
Yeah your right, you never find EOL products in any real quantity in retail stock because once they are officially "EOL", all products in the channels dissapear overnight /sigh
Last edited by a moderator:
jimmyjames123
Regular
For the C2050 to have 416 active ALU's is one way to differentiate it from the more expensive C2070 (in addition to amount of lower amount of memory and lower clock speeds), so it seems logical to me. I just don't know why NVIDIA's C2050/C2070 pdf document (the one at their website) would list 448 ALU's in the document, with no mention of any other # of ALU's in the document, when they mean to disable 32 ALU's on the lower priced Tesla variant. There is nothing wrong with disabling some units for lower priced variants, we all understand the need for lower cost and more differentiation. Maybe NVIDIA's doesn't want an SA article with a big headline "NVIDIA Tesla C2050 has 20% less functional units than promised, buyer beware XXX"
It's going to be a while before these Tesla cards are available anyway, not before 2Q '10 it seems.