Because wishful thinking trumps physics and engineering?
Well no, of course not. But I suspect the cards will be here sooner than expected. I think Nvidia is farther along then they are letting on. But we'll know soon enough.
You don't read other people's posts do you ?
Nvidia and AMD for that matter have no control or relation to the release time of 16nm/14nm cards, they only have the 4 bad choices to choose a lesser bad from :
1) release early and with gddr5 2) wait for ggdr5x and/or HBM2 3) release with HBM1 4) do paper launch as soon as no.2 is between 100-1000 ready
none of which of all of these memory technologies can be sped up by neither nvidia or amd, because they aren't developing them in house, ...
all breadcrumbs point to the same style staggered low/mid launch around june computex, and 780Ti/980Ti launch of a "full" big die around christmas.
I hope that u understand that even if you are right that nvidia is farther ahead then they are letting on, you are still wrong, if u don't understand why even after my post, i have a high value bridge-like property in brooklyn to sell you, ...
Because wishful thinking trumps physics and engineering?
Well no, of course not. But I suspect the cards will be here sooner than expected. I think Nvidia is farther along then they are letting on. But we'll know soon enough.
Even if Nvidia has millions of completely flawless Pascal dies sitting in their inventory today, if they want to use GDDR5X or HBM2, you're looking at a fall launch at the earliest. You can't sell cards using memory that doesn't exist, and the GDDR5X and HBM2 that exists today from Samsung, Hynix, and Micron isn't enough to do a hard launch of a consumer GPU--and won't be for quite some time. You can get a few memory chips early to test and debug your own hardware, but that's not at all similar to the many millions of chips you need for consumer GPUs.
Samsung said they started volume production of HBM2 in January, and Hynix a couple months later. Micron started production of GDDR5X in late March. These are all public announcements, not wild rumors. It takes several weeks for chips to go through fabs, and then more time to mount them onto video cards and distribute inventory around the world. And it also takes quite a while to accumulate enough inventory for a hard launch.
So you should think Fall at the earliest for a hard launch of any GPUs on HBM2 or GDDR5X, even assuming that everything goes absolutely flawlessly on Nvidia's part. Which it never does, of course. And as GP100 is by far the most ambitious GPU ever made, one would expect it to have quite a bit more go wrong than most GPUs.
Because wishful thinking trumps physics and engineering?
Well no, of course not. But I suspect the cards will be here sooner than expected. I think Nvidia is farther along then they are letting on. But we'll know soon enough.
You don't read other people's posts do you ? ....
Not all of them, no. Who has time for that shit?
My hunch is based on what I've seen from Nvidia in the past. But I'm not suggesting I'm prescient. Maybe it will be Christmas like you suggest. I'm gonna go with Late Summer, though. 1080 GTX or whatever the hell they'll call it. The highend one.
If that's how you're going to play, then I'm going with a hunch that the Houston Rockets will win the NBA finals this year. Because it's still technically possible, and only takes ignoring a ton of evidence that it won't happen.
I don't think 1080 (or whatever it will be branded as) will be on the new process node. I believe that will be the last iteration of the current technology.
The Titan X already uses a 600 mm^2 chip. Nvidia literally can't make a meaningfully bigger chip than that on 28 nm because TSMC can't physically build it. For Nvidia to create another generation of 28 nm parts would be completely insane unless they think 14/16 nm is still more than a year off--in spite of AMD already promising Polaris (with GDDR5, so the memory delays are irrelevant) around the middle of this year.
If you're expecting them to take a Titan X, clock it a little higher, and call it a 1080, then that's at least possible. I don't really expect them to do that, but the arguments for or against that are purely marketing, not engineering.
I don't think 1080 (or whatever it will be branded as) will be on the new process node. I believe that will be the last iteration of the current technology.
The Titan X already uses a 600 mm^2 chip. Nvidia literally can't make a meaningfully bigger chip than that on 28 nm because TSMC can't physically build it. For Nvidia to create another generation of 28 nm parts would be completely insane unless they think 14/16 nm is still more than a year off--in spite of AMD already promising Polaris (with GDDR5, so the memory delays are irrelevant) around the middle of this year.
I agree. I'm just saying that if this rumored "1080" comes to pass it won't be Pascal. If they launch a Pascal card I doubt they will brand it "1080" or anything that looks like it's a leap the existing 9xx series. I don't think they'll launch with an x80 series card either. I think they're going to launch mid-range first.
It's all pure speculation on my part. I'm just thinking the way they've historically numbered cards. This next die shrink coupled with the new memory is big. I think they will do some sort of rebranding to make a clean break. Who knows though I'm just guessing.
There is a good chance of that happening.
The 500 series were respun Fermis, because Keplar was late. AMD has been re-iterating on GCN for years and several generations now.
I don't think nVidia is in the position of ~needing~ to push a new SKU. For the 500 series, the 400 series didn't live up to expectations, and AMD had Cypress out that was beating Fermi in pretty much every price point, and was pushing forward with Northern Islands that bumped performance even more and introduced dynamic power management. So nVidia had to push something before Kepler was ready.
Right now, GCN 1.2 based cards are competitive with nVidia's Maxwell, but nVidia isn't in any danger of being out-maneuvered, at least until GDDR5x/HBM2 is out in available quantities.
If Polaris ships this fall (and right now I have no reason to believe it won't), if nVidia is still on track to push the GP100 by early 2017, they will probably continue down that line - perhaps even going so far as to put out "limited quantity" runs before a general availability (such as the 680 had). But if it slips much past early 2017, I bet we see overclocked &/or respun Maxwells released as a stopgap generation.
I don't think 1080 (or whatever it will be branded as) will be on the new process node. I believe that will be the last iteration of the current technology.
The Titan X already uses a 600 mm^2 chip. Nvidia literally can't make a meaningfully bigger chip than that on 28 nm because TSMC can't physically build it. For Nvidia to create another generation of 28 nm parts would be completely insane unless they think 14/16 nm is still more than a year off--in spite of AMD already promising Polaris (with GDDR5, so the memory delays are irrelevant) around the middle of this year.
I agree. I'm just saying that if this rumored "1080" comes to pass it won't be Pascal. If they launch a Pascal card I doubt they will brand it "1080" or anything that looks like it's a leap the existing 9xx series. I don't think they'll launch with an x80 series card either. I think they're going to launch mid-range first.
It's all pure speculation on my part. I'm just thinking the way they've historically numbered cards. This next die shrink coupled with the new memory is big. I think they will do some sort of rebranding to make a clean break. Who knows though I'm just guessing.
There is a good chance of that happening.
The 500 series were respun Fermis, because Keplar was late. AMD has been re-iterating on GCN for years and several generations now.
I don't think nVidia is in the position of ~needing~ to push a new SKU. For the 500 series, the 400 series didn't live up to expectations, and AMD had Cypress out that was beating Fermi in pretty much every price point, and was pushing forward with Northern Islands that bumped performance even more and introduced dynamic power management. So nVidia had to push something before Kepler was ready.
Right now, GCN 1.2 based cards are competitive with nVidia's Maxwell, but nVidia isn't in any danger of being out-maneuvered, at least until GDDR5x/HBM2 is out in available quantities.
If Polaris ships this fall (and right now I have no reason to believe it won't), if nVidia is still on track to push the GP100 by early 2017, they will probably continue down that line - perhaps even going so far as to put out "limited quantity" runs before a general availability (such as the 680 had). But if it slips much past early 2017, I bet we see overclocked &/or respun Maxwells released as a stopgap generation.
The 500 series was respun Fermi because there were such massive gains to make by fixing a 400 series that was so flagrantly broken. When your top chip is too broken to release a fully functional part, and so is the next chip down, and even the third chip down has such horrible yields that you can't release a fully functional desktop part, things have gone horribly awry. Maxwell works well, so there isn't much to gain by trying to fix things that aren't broken.
The G80 worked well when it released in 2006. That didn't prevent nVidia for using that chip and various respins in like the next 3 generations.
G92 was a die shrink from 90 nm to 65 nm (and a 256-bit memory bus rather than 384-bit), and then G92b was a further shrink to 55 nm. A die shrink isn't a mere respin on the same process node.
A shrink of Maxwell is certainly possible. Indeed, Nvidia already did that with a 20 nm Tegra. But even if Nvidia did try to make a straight die shrink of Maxwell to 16 nm, they wouldn't still call it Maxwell. They'd call it something different, like Pascal.
From architectural details, the big Pascal GP100 certainly isn't a straight shrink of a Maxwell chip. A lot of the specs actually match a Radeon R9 Fury (non-X): same number of shaders, texture units, register file size, compute units, and memory bus width. Exactly, not just kind of close.
But it's also possible that Nvidia could do something with Pascal similar to what they did with Fermi, where the top chip was very different from the rest, not just with some extra HPC stuff added, but different number of shaders per compute unit, even.
IT IS STILL HAPPENING. Tablet sales
remain in decline and it is Apple's fault for not polishing the iPad or
putting any fresh ribbons on it, according to the company who worked
this out.
Strategy Analytics' Q4 2014 report shows a weakening in tablet sales
growth and an 11 percent decline against the same period last year.
This fall saw shipments plummet to a barely noticeable 69.9 million
units in the last quarter of 2015 and just 224.3 million over the whole
year.
"Apple suffered big setbacks this year as a lack of innovation during
the last several years caught up with iPad sales. Instead, Apple has
focused on its MacBooks, iPhone 6/6+ and Apple Watch releases," said
Peter King, research director for tablet and touchscreen at Strategy
Analytics.
"The launch of the iPad Pro failed to meet expectations during the
quarter. While we see real long-term potential for iPad Pro in the
enterprise and verticals, there was not enough demand to move the needle
in Q4."
Apple's market share fell by 25 percent over the course of the year,
while Samsung took a 17 percent dunk. White box devices currently
dominate sales, but even they dropped seven percent in terms of
shipments over the period.
Windows Tablet shipments grew by 59 percent over the year, Android fell seven percent and iOS dropped 22 percent. This is becoming something of a trend.
"The surest way to corrupt a youth is to instruct him to hold in higher esteem those who think alike than those who think differently."
Which pretty much kills that site's credibility. There's also Betteridge's Law of Headlines, which suggests an answer of "no" to the question in the article title.
Now, it's very possible that they will publicly demonstrate working Pascal cards in June. AMD demonstrated working Polaris cards in January. That's not the same as a hard launch.
Which pretty much kills that site's credibility. There's also Betteridge's Law of Headlines, which suggests an answer of "no" to the question in the article title.
Now, it's very possible that they will publicly demonstrate working Pascal cards in June. AMD demonstrated working Polaris cards in January. That's not the same as a hard launch.
I'm confused, i haven't opened the article, but the link clearly says "will be demonstrated".
While i don't disagree with you that wccftech isn't exactly the most reliable news source (though btw they're the ones who "broke" the pascal async compute thing, which you were on board with if i recall correctly). You're basically agreeing with him. The article says they will demonstrate pascal in june, and you proceed to say "it's very possible". I don't think anyone was suggesting a hard launch in june, well, he might have been, but the article doesn't appear to be making that suggestion.
"The surest way to corrupt a youth is to instruct him to hold in higher esteem those who think alike than those who think differently."
No, it's a pretty crappy article. They are suggesting a hard launch.
Following its possible introduction, NVIDIA's newest GeForce GTX 1080 and 1070 graphics cards may be released as early as May or June. According to Swedish source Sweclockers, as translated by WCCFTech, NVIDIA may possibly demonstrate the new graphics cards during the upcoming Computex 2016.
Which pretty much kills that site's credibility. There's also Betteridge's Law of Headlines, which suggests an answer of "no" to the question in the article title.
Now, it's very possible that they will publicly demonstrate working Pascal cards in June. AMD demonstrated working Polaris cards in January. That's not the same as a hard launch.
I'm confused, i haven't opened the article, but the link clearly says "will be demonstrated".
While i don't disagree with you that wccftech isn't exactly the most reliable news source (though btw they're the ones who "broke" the pascal async compute thing, which you were on board with if i recall correctly). You're basically agreeing with him. The article says they will demonstrate pascal in june, and you proceed to say "it's very possible". I don't think anyone was suggesting a hard launch in june, well, he might have been, but the article doesn't appear to be making that suggestion.
Well, the previous rumor from "sources" claimed those would be unveiled at GTC and lunced in May even.
Now "sources" claim they will be unveiled at Computex and launched in June.
Its all a bunch of rumors.
As far as Async goes, "rumor" is far younger than those who follow the matter started doubting Pascal will have it because of NVidias DX12 capability shenannigans over whole Async matter. Pascal so far doesnt have Async (it has been unveiled at GTC). Maxwell obviously cant do Async as it would be working by now.
PC sales and PC gaming is not the same thing. Unless you are ready to suggest that there are more than 1 billion PC gamers out there I dont think falling PC sales has dick to do with video cards
Please do not respond to me, even if I ask you a question, its rhetorical.
Neither AMD nor Nvidia have any real presence in the mobile graphics market, and the PC market is shrinking.
This guy has no clue of what he speaks! So Nvidia have no presence in the mobile gpu market huh? Tell that to me as i write this post on my Nvidia sheild tablet, the most powerfull gaming tablet on the planet!......Idiot!
Comments
Nvidia and AMD for that matter have no control or relation to the release time of 16nm/14nm cards, they only have the 4 bad choices to choose a lesser bad from :
1) release early and with gddr5
2) wait for ggdr5x and/or HBM2
3) release with HBM1
4) do paper launch as soon as no.2 is between 100-1000 ready
none of which of all of these memory technologies can be sped up by neither nvidia or amd, because they aren't developing them in house, ...
all breadcrumbs point to the same style staggered low/mid launch around june computex, and 780Ti/980Ti launch of a "full" big die around christmas.
I hope that u understand that even if you are right that nvidia is farther ahead then they are letting on, you are still wrong, if u don't understand why even after my post, i have a high value bridge-like property in brooklyn to sell you, ...
Samsung said they started volume production of HBM2 in January, and Hynix a couple months later. Micron started production of GDDR5X in late March. These are all public announcements, not wild rumors. It takes several weeks for chips to go through fabs, and then more time to mount them onto video cards and distribute inventory around the world. And it also takes quite a while to accumulate enough inventory for a hard launch.
So you should think Fall at the earliest for a hard launch of any GPUs on HBM2 or GDDR5X, even assuming that everything goes absolutely flawlessly on Nvidia's part. Which it never does, of course. And as GP100 is by far the most ambitious GPU ever made, one would expect it to have quite a bit more go wrong than most GPUs.
The Titan X already uses a 600 mm^2 chip. Nvidia literally can't make a meaningfully bigger chip than that on 28 nm because TSMC can't physically build it. For Nvidia to create another generation of 28 nm parts would be completely insane unless they think 14/16 nm is still more than a year off--in spite of AMD already promising Polaris (with GDDR5, so the memory delays are irrelevant) around the middle of this year.
The 500 series were respun Fermis, because Keplar was late. AMD has been re-iterating on GCN for years and several generations now.
I don't think nVidia is in the position of ~needing~ to push a new SKU. For the 500 series, the 400 series didn't live up to expectations, and AMD had Cypress out that was beating Fermi in pretty much every price point, and was pushing forward with Northern Islands that bumped performance even more and introduced dynamic power management. So nVidia had to push something before Kepler was ready.
Right now, GCN 1.2 based cards are competitive with nVidia's Maxwell, but nVidia isn't in any danger of being out-maneuvered, at least until GDDR5x/HBM2 is out in available quantities.
If Polaris ships this fall (and right now I have no reason to believe it won't), if nVidia is still on track to push the GP100 by early 2017, they will probably continue down that line - perhaps even going so far as to put out "limited quantity" runs before a general availability (such as the 680 had). But if it slips much past early 2017, I bet we see overclocked &/or respun Maxwells released as a stopgap generation.
A shrink of Maxwell is certainly possible. Indeed, Nvidia already did that with a 20 nm Tegra. But even if Nvidia did try to make a straight die shrink of Maxwell to 16 nm, they wouldn't still call it Maxwell. They'd call it something different, like Pascal.
From architectural details, the big Pascal GP100 certainly isn't a straight shrink of a Maxwell chip. A lot of the specs actually match a Radeon R9 Fury (non-X): same number of shaders, texture units, register file size, compute units, and memory bus width. Exactly, not just kind of close.
But it's also possible that Nvidia could do something with Pascal similar to what they did with Fermi, where the top chip was very different from the rest, not just with some extra HPC stuff added, but different number of shaders per compute unit, even.
~~ postlarval ~~
http://www.businessinsider.com/tablet-sales-down-in-addition-to-ipad-2015-10
http://www.theinquirer.net/inquirer/news/2444426/tablet-sales-are-still-in-decline-and-all-eyes-are-on-apple
IT IS STILL HAPPENING. Tablet sales remain in decline and it is Apple's fault for not polishing the iPad or putting any fresh ribbons on it, according to the company who worked this out.
Strategy Analytics' Q4 2014 report shows a weakening in tablet sales growth and an 11 percent decline against the same period last year.
This fall saw shipments plummet to a barely noticeable 69.9 million units in the last quarter of 2015 and just 224.3 million over the whole year.
The Strategy Analytics Preliminary Global Tablet Shipments and Market Share by Operating System: Q4 2015 is the source of the figures, and it is not all bad news. 2-in-1 tablets and Windows tablets remain resilient, and it is apparently the traditional slate that has slipped. Apple has done the market no favours at all, it seems, and has had a lacklustre time in the shipping department.
"Apple suffered big setbacks this year as a lack of innovation during the last several years caught up with iPad sales. Instead, Apple has focused on its MacBooks, iPhone 6/6+ and Apple Watch releases," said Peter King, research director for tablet and touchscreen at Strategy Analytics.
"The launch of the iPad Pro failed to meet expectations during the quarter. While we see real long-term potential for iPad Pro in the enterprise and verticals, there was not enough demand to move the needle in Q4."
Apple's market share fell by 25 percent over the course of the year, while Samsung took a 17 percent dunk. White box devices currently dominate sales, but even they dropped seven percent in terms of shipments over the period.
Windows Tablet shipments grew by 59 percent over the year, Android fell seven percent and iOS dropped 22 percent. This is becoming something of a trend."The surest way to corrupt a youth is to instruct him to hold in higher esteem those who think alike than those who think differently."
- Friedrich Nietzsche
"According to WCCFTech, a reliable NVIDIA source"
Which pretty much kills that site's credibility. There's also Betteridge's Law of Headlines, which suggests an answer of "no" to the question in the article title.
Now, it's very possible that they will publicly demonstrate working Pascal cards in June. AMD demonstrated working Polaris cards in January. That's not the same as a hard launch.
While i don't disagree with you that wccftech isn't exactly the most reliable news source (though btw they're the ones who "broke" the pascal async compute thing, which you were on board with if i recall correctly). You're basically agreeing with him. The article says they will demonstrate pascal in june, and you proceed to say "it's very possible". I don't think anyone was suggesting a hard launch in june, well, he might have been, but the article doesn't appear to be making that suggestion.
"The surest way to corrupt a youth is to instruct him to hold in higher esteem those who think alike than those who think differently."
- Friedrich Nietzsche
Now "sources" claim they will be unveiled at Computex and launched in June.
Its all a bunch of rumors.
As far as Async goes, "rumor" is far younger than those who follow the matter started doubting Pascal will have it because of NVidias DX12 capability shenannigans over whole Async matter. Pascal so far doesnt have Async (it has been unveiled at GTC). Maxwell obviously cant do Async as it would be working by now.
guys..
PC sales and PC gaming is not the same thing. Unless you are ready to suggest that there are more than 1 billion PC gamers out there I dont think falling PC sales has dick to do with video cards
Please do not respond to me, even if I ask you a question, its rhetorical.
Please do not respond to me