we are looking at 28nm "small" maxwell, successor of GK104, "next GTX 680". There is not ARM inside, and numbers of CUDAS is different then source said ... about Big one and ARM core, i am not sure, but should be only HPC GPU, in gaming sku will be disabled maybe
That would make sense.Originally Posted by PedantOne
A built-in ARM core would be useful for GPGPU, and pretty useless for gaming.
Donate to XS forums
The short version is that the Maxwell releases are coming in two waves with four GM204 based SKUs in each one. The first wave is set for late Q3 or early Q4, that would be September or October in human-speak, sources say it is likely the latter month though. There will be four Maxwell based parts all of which are 28nm GM204s like we told you last year. These would be, from top to bottom, the GTX 880 Ti, GTX 880, GTX 870, and GTX 860. All four replace the Kepler based GPUs that are 100 GTX units smaller in model number.
The real interesting part is the second phase of releases that come in mid-Q1/2015. These are claimed 16nm Maxwells that look to be simple shrinks of the GM204′s that come out later this year. Nvidia is cunningly calling these new parts GTX 880 Ti, GTX 880, GTX 870, and GTX 860. We strongly suspect they replace the GTX 880 Ti, GTX 880, GTX 870, and GTX 860, respectively.
In short there will be two releases of the same part, one 28nm this year and one 16nm next year with 4-5 months separating them. The timing and volume depends on TSMC more than anything else, at least for the second wave, and so does the performance. If you are a risk-taker, you should probably wait for the 16nm parts, they should be cheaper, cooler, faster, and better in most regards unless TSMC really screws thing up.
As a side note, we said that during Computex Nvidia was blaming the delay from Q4 -> Q1/2015 on excess inventory. We very much doubted this explanation and said as much, technically speaking it was bull. This latest update brings the roadmap back to where we said it would be in October of last year, and it also adds the 16nm shrinks. From the outside it looks like Nvidia was on the fence about taping out and producing 28nm GM204s for <6 months before a shrink. For one reason or other they seem to have un-pulled that plug and it is now a go again. Quite the expensive stopgap part there.S|A
16nm in Q1 2015? That sounds incredible. 28nm to 16nm means equivalent chips will become up to 77% smaller...
Last edited by zalbard; 07-06-2014 at 11:29 AM.
Donate to XS forums
Dude it's too incredible to be true. Do you have any idea how difficult of a problem electron leakage already is? Barring a major breakthrough in dialectric materials, there's absolutely 100% no way TSMC can have that process ready half a year after the infamous 20nm comes out.
You can quote me on that. There's just too many blatantly wrong rumors going around that will spread false hope.
TSMC 16nm is actually 20nm with Finfets. Everyone was expecting it to be ready-ready in 1H '15.
Originally Posted by motown_steve
Every genocide that was committed during the 20th century has been preceded by the disarmament of the target population. Once the government outlaws your guns your life becomes a luxury afforded to you by the state. You become a tool to benefit the state. Should you cease to benefit the state or even worse become an annoyance or even a hindrance to the state then your life becomes more trouble than it is worth.
Once the government outlaws your guns your life is forfeit. You're already dead, it's just a question of when they are going to get around to you.
You sure about that? Where does it say that 16 and 28 nm actually are gate lengths? Remember Intels nomenclature of their processes? They are "random" for a new set of process parameters which have nothing to do with physical dimensions. It will be smaller, yes. But how much smaller...only TSMC knows.
Good to know.
This is why I wrote "up to". In fact, if what LordEC911 says is correct then area reduction will be much more modest.
Donate to XS forums
Yeah, if you read any of TSMC's PR about 16nmFinFet or 16nmFinfet+, they mention no density benefits over 20nm SOC just power consumption or speed benefits.
The other hint is the fact that 20nm SOC, 16nm FinFet and 16nm Finfet+ design rules are all basically the same, majority of the tooling is the same. Obviously there are some substantial changes for actual FinFet manufacturing but if the above article is correct TSMC should be able to ramp production very quickly throughout 2015 for each process.
As far the as the density, I'm sure there will be some density benefits and AMD/Nvidia will be capable of increasing density a bit more by sacrificing a bit on other benefits but I don't expect to see the typical 1.6-1.8x density improvements of past nodes.
Found this article/explanation about the difference between 20nm, 16nm and 14nm with FinFets.
Last edited by LordEC911; 07-06-2014 at 03:03 PM.
Originally Posted by motown_steve
Every genocide that was committed during the 20th century has been preceded by the disarmament of the target population. Once the government outlaws your guns your life becomes a luxury afforded to you by the state. You become a tool to benefit the state. Should you cease to benefit the state or even worse become an annoyance or even a hindrance to the state then your life becomes more trouble than it is worth.
Once the government outlaws your guns your life is forfeit. You're already dead, it's just a question of when they are going to get around to you.
All I can say that the new cards better have HDMI 2.0...there will be many highly annoyed people if it doesn't...
Server: HP Proliant ML370 G6, 2x Xeon X5690, 144GB ECC Registered, 8x OCZ Vertex 3 MAX IOPS 240GB on LSi 9265-8i (RAID 0), 12x Seagate Constellation ES.2 3TB SAS on LSi 9280-24i4e (RAID 6) and dual 1200W redundant power supplies.
Gamer: Intel Core i7 6950X@4.2GHz, Rampage Edition 10, 128GB (8x16GB) Corsair Dominator Platinum 2800MHz, 2x NVidia Titan X (Pascal), Corsair H110i, Vengeance C70 w/Corsair AX1500i, Intel P3700 2TB (boot), Samsung SM961 1TB (Games), 2x Samsung PM1725 6.4TB (11.64TB usable) Windows Software RAID 0 (local storage).
Beater: Xeon E5-1680 V3, NCase M1, ASRock X99-iTX/ac, 2x32GB Crucial 2400MHz RDIMMs, eVGA Titan X (Maxwell), Samsung 950 Pro 512GB, Corsair SF600, Asetek 92mm AIO water cooler.
Server/workstation: 2x Xeon E5-2687W V2, Asus Z9PE-D8, 256GB 1866MHz Samsung LRDIMMs (8x32GB), eVGA Titan X (Maxwell), 2x Intel S3610 1.6TB SSD, Corsair AX1500i, Chenbro SR10769, Intel P3700 2TB.
Thanks for the help (or lack thereof) in resolving my P3700 issue, FUGGER...
I know it's not the best source of info but still... http://videocardz.com/51009/nvidia-p...ell-gm204-skus (Semi(yea right)-Accurate "original" source)
Here's hoping for 1920 in the 860 & 2560 in 880 Ti.Various sources claim that full GM204 chip would feature between 15 and 20 SMs (1920-2560 CUDAs). Maxwell core is obviously much more power efficient than Kepler, thus the next flagship will most likely have less cores than full GK110 processor. This has not happened before.
3200 in 880ti, I am hoping.
I'd like to see more efficiency; give me 30% more power than 780 ti using only 2560 shaders and I'll be ecstatic.
That'd work...........
Wonder if this release will make the 290x cheaper
I'm guessing 290x would be EOL'd, so temporarily cheaper I suppose.Wonder if this release will make the 290x cheaper
It was a design that basically crams as many shaders as possible without the gpu catching on fire. I highly doubt that AMD ever planned on releasing it when the 7970 launched; however, TSMC not having theirtogether (yet again) forced them to wait on their scheduled refresh ... and you know the rest
So is the SLI bridge location due to Nvidia trying to shorten the copper traces, how much bandwidth do you think they'll be aiming for?
It is a prototype, they have to move stuff around due to test equipment/connections on the PCB.
Originally Posted by motown_steve
Every genocide that was committed during the 20th century has been preceded by the disarmament of the target population. Once the government outlaws your guns your life becomes a luxury afforded to you by the state. You become a tool to benefit the state. Should you cease to benefit the state or even worse become an annoyance or even a hindrance to the state then your life becomes more trouble than it is worth.
Once the government outlaws your guns your life is forfeit. You're already dead, it's just a question of when they are going to get around to you.
Do you guys think this chip is going to be a leap over previous gen. Nvidia has been promising something special with "Maxwell" for many years now. Wonder if they are close to achieve that or are we getting some gap filler till the real deal shows up.
Depends on what you consider a "leap".
The first chip, let's call it GTX 880 could be 20% faster than the 780 Ti while at the same time consuming less power.
Could we call this a leap ? I guess so.
Titan II will be even faster, more of a real leap over the previous generation, but will definitely come later on.
Coding 24/7... Limited forums/PMs time.
-Justice isn't blind, Justice is ashamed.
Many thanks to: Sue Wu, Yiwen Lin, Steven Kuo, Crystal Chen, Vivian Lien, Joe Chan, Sascha Krohn, Joe James, Dan Snyder, Amy Deng, Jack Peterson, Hank Peng, Mafalda Cogliani, Olivia Lee, Marta Piccoli, Mike Clements, Alex Ruedinger, Oliver Baltuch, Korinna Dieck, Steffen Eisentein, Francois Piednoel, Tanja Markovic, Cyril Pelupessy (R.I.P.), Juan J. Guerrero
I think it's safe to assume they're probably going to release their new hardware in a similar fashion as the previous generation. Get the customers who are eager to upgrade from Fermi or older and the suckers who can't stand not to have to latest and greatest. Then, get that last group again in the refresh cycle when the "big dog" GPU is released
Totally agree with you... I hope Titan II will not be a real leap over previous generation when it comes to price...
When i'm being paid i always do my job through.
Bah its 2014 big dog nvidia gpu's gone from 500-600$ to 1000$ dammit. But heck it would be great to see some big dog gpu titan 2 40% faster than 780ti.
20% faster than 780Ti in December for 600$~ nvidia have a winner. 290x wasnt that long ago wonder where ATI are on their next gen.
We need to go back in time and undo the brilliant Titan marketing strat this
Sadly it looks like the $999 MSRP is here to stay.
GPU makers got jealous, in the past only Intel & AMD's flagship CPUs "could" ask for that amount of money... nowadays everybody's doing it.
If somebody in the SSD sector for example manages to double the current Random 4k RW speed records, I wouldn't be surprised to see it asking three times the price of the current performance king.
Coding 24/7... Limited forums/PMs time.
-Justice isn't blind, Justice is ashamed.
Many thanks to: Sue Wu, Yiwen Lin, Steven Kuo, Crystal Chen, Vivian Lien, Joe Chan, Sascha Krohn, Joe James, Dan Snyder, Amy Deng, Jack Peterson, Hank Peng, Mafalda Cogliani, Olivia Lee, Marta Piccoli, Mike Clements, Alex Ruedinger, Oliver Baltuch, Korinna Dieck, Steffen Eisentein, Francois Piednoel, Tanja Markovic, Cyril Pelupessy (R.I.P.
Now, if there were actually noticeable real-world gains from this...
Donate to XS forums
Posting Permissions
Reply With Quote
Bookmarks