Sauce: http://www.thinq.co.uk/news/2010/5/20/geforce-gtx-470-hasnt-been-cancelled/
Looks like they will be changing the name to GTX465 then.
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
GeForce GTX 470 hasn't been cancelled
- Thread starter willhub
- Start date
Sauce: http://www.thinq.co.uk/news/2010/5/20/geforce-gtx-470-hasnt-been-cancelled/
Looks like they will be changing the name to GTX465 then.
Link isn't tasty.
Associate
- Joined
- 15 Jun 2007
- Posts
- 1,201
- Location
- Liverpool
Um, doesn't this just mean that NV aren't taking orders for reference 470 cards? The partners purchase just the GPU and RAM, hence new non-reference cards coming out. It says nothing of renaming the card to a 465.
Associate
- Joined
- 15 Jun 2007
- Posts
- 1,201
- Location
- Liverpool
Yeah, but that's not what the article says.
Associate
- Joined
- 15 Jun 2007
- Posts
- 1,201
- Location
- Liverpool
Nah it doesn't say that, the GTX465 has 352 cuda cores, it's a different card.
Again it just says NV have stopped taking orders for the reference 470 cards, ie PCB/GPU/RAM/Cooler. Partners can still buy the GPU/RAM for a non-reference 470 design, as we are starting to see. They seem to be concentrating on building reference 465 cards now.
Or they don't have enough GPU's capable of running on a GTX470 so they have no other choice than to disable more shaders and sell it as a GTX465 ?
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
What is with the OP, the title is it hasn't been canceled, and it links to an article that says it was being canceled but has been updated to say its not and its a mistake.
But the OP's own words sounded like he thought it had been canceled, so why the title saying it hadn't?
Either way, the wording of the leak was that Nvidia had STOPPED TAKING ORDERS for the 470gtx, not "stopped PLACING orders" for the 470gtx, the later could have been refering to the pcb and not ordering any more 470gtx base pcb's, the former sounds to me like Nvidia aren't letting AIB's order 470 cards from them, not pcb's, a card is the whole lot.
Its perfectly possible, its just one source right now and its completely unconfirmed.
As lots of people said ages ago, stock, sales numbers and the WHOLE situation won't be known in regards to Fermi till MONTHS after launch.
Because you could produce 100k in the first two months and never make a card again, trickle out 10k a month, put out 10k and not ever another one, who knows, till its happened its just guessing.
See what happens, and take the pee out of Nvidia later when we see the numbers Fermi pushed and if it stays in production for very long.
But the OP's own words sounded like he thought it had been canceled, so why the title saying it hadn't?
Either way, the wording of the leak was that Nvidia had STOPPED TAKING ORDERS for the 470gtx, not "stopped PLACING orders" for the 470gtx, the later could have been refering to the pcb and not ordering any more 470gtx base pcb's, the former sounds to me like Nvidia aren't letting AIB's order 470 cards from them, not pcb's, a card is the whole lot.
Its perfectly possible, its just one source right now and its completely unconfirmed.
As lots of people said ages ago, stock, sales numbers and the WHOLE situation won't be known in regards to Fermi till MONTHS after launch.
Because you could produce 100k in the first two months and never make a card again, trickle out 10k a month, put out 10k and not ever another one, who knows, till its happened its just guessing.
See what happens, and take the pee out of Nvidia later when we see the numbers Fermi pushed and if it stays in production for very long.
What he says.this makes no sense to me either.
This can only mean a 32nm GTX495 is about to come out, hurrah!
32nm is dead, no one is going to be using it except intel.
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
AMD are using 32nm SOI for the next 18-24months, its expensive and uber quality, the only gpu's going near that are intergrated GPu's on llano and a few other fusion products over the next couple years.
Discrete gpu's won't ever make it on 32nm SOI, its VERY expensive for the quality and simply too expensive for a GPU, remember a 480gtx/5870 use WAY more transistors than a CPU, llano will probably be a low profit part because its going to have , from recollection, around a Phenom 2 with a couple hundred shaders stuffed on it using a LOT of extra transistors, compared to a cpu it will be pretty damn big.
realistically the way it should have happened is 40nm should have been a year earlier, 32nm should have been around late last year or beginning of this year, and 28nm towards the end of next year.
40nm was soooo late and has taken so long to get to this point that if 32nm came out around now, it would be so close to 28nm it would be a waste of cash, it takes literally billions in research and billions more for the equipment to outfit a fab for a new process. So skipping 32nm and waiting 6 extra months for 28nm saves TSMC and glofo a good 2-3billion on 32nm equipment that very few people will use.
28nm, because of the fairly large jump, is likely to bring us some pretty hugely cheaper gpu's next time around and really thats the time to question if AMD are still going after the value market. 40-28nm is pretty huge. If 32nm was on course you'd be looking at a double power, double transistor core at around the same size as now. Move that same core to 28nm and its going to end up significantly smaller for only double the transistors as now, meaning higher yields, good clocks and tiny size. That leaves a shedload of extra room for making something with more than twice the shaders a 5870 has.
28nm could be very interesting, the normal plan would be 800 shaders on 55nm, 1600 shaders on 40nm, 3200 shaders at 32nm, 6400 shaders at 28nm. 28nm is likely to be around for quite a long time though. I'd only expect a 3200-3600 shader part when 28nm is new, and it should be tiny, low power and cheap as chips with very cheap X2 cards, then a year later still on 28nm, a much more powerful part, maybe double the size, but won't be as cheap.
EDIT:- a current i7 is 731million transistors, a 480gtx is 3billion +, a 5870 is 2.15billion transistors, its tiny in comparison, and why an expensive higher clock speed capable far higher quality process can be used for CPu's, but not gpu's.
Discrete gpu's won't ever make it on 32nm SOI, its VERY expensive for the quality and simply too expensive for a GPU, remember a 480gtx/5870 use WAY more transistors than a CPU, llano will probably be a low profit part because its going to have , from recollection, around a Phenom 2 with a couple hundred shaders stuffed on it using a LOT of extra transistors, compared to a cpu it will be pretty damn big.
realistically the way it should have happened is 40nm should have been a year earlier, 32nm should have been around late last year or beginning of this year, and 28nm towards the end of next year.
40nm was soooo late and has taken so long to get to this point that if 32nm came out around now, it would be so close to 28nm it would be a waste of cash, it takes literally billions in research and billions more for the equipment to outfit a fab for a new process. So skipping 32nm and waiting 6 extra months for 28nm saves TSMC and glofo a good 2-3billion on 32nm equipment that very few people will use.
28nm, because of the fairly large jump, is likely to bring us some pretty hugely cheaper gpu's next time around and really thats the time to question if AMD are still going after the value market. 40-28nm is pretty huge. If 32nm was on course you'd be looking at a double power, double transistor core at around the same size as now. Move that same core to 28nm and its going to end up significantly smaller for only double the transistors as now, meaning higher yields, good clocks and tiny size. That leaves a shedload of extra room for making something with more than twice the shaders a 5870 has.
28nm could be very interesting, the normal plan would be 800 shaders on 55nm, 1600 shaders on 40nm, 3200 shaders at 32nm, 6400 shaders at 28nm. 28nm is likely to be around for quite a long time though. I'd only expect a 3200-3600 shader part when 28nm is new, and it should be tiny, low power and cheap as chips with very cheap X2 cards, then a year later still on 28nm, a much more powerful part, maybe double the size, but won't be as cheap.
EDIT:- a current i7 is 731million transistors, a 480gtx is 3billion +, a 5870 is 2.15billion transistors, its tiny in comparison, and why an expensive higher clock speed capable far higher quality process can be used for CPu's, but not gpu's.
Last edited:
Since the price of the gtx465 has been leaked now and is only £10 less than the gtx470, doesn't that mean it is quite likely that we'll not see anymore gtx470?
Or perhaps just "premium" gtx470 cards from partners which would have enough price difference to the referenace gtx465.
Otherwise who in their right mind is going to buy the gtx465 to save themselves a tenner?
Or perhaps just "premium" gtx470 cards from partners which would have enough price difference to the referenace gtx465.
Otherwise who in their right mind is going to buy the gtx465 to save themselves a tenner?
