fry178:

@Monchis bitter you didnt buy one? or gun to the head when in an electronic store and forced to go to checkout? #beingforcedtobuystuff

lmao, I just want to see progression. You see, for a 2 node jump Pascals have to be the most awful cards ever, get them off my sight.

Seems to be a very strong batch of tea leaves...the original article https://www.heise.de/newsticker/meldung/Nvidia-Milliarden-Umsatz-und-kraeftiger-Gewinnsprung-dank-Gaming-3887633.html consists of a rumor of a GPU uarch called Ampere'. "End of the third quarter, Nvidia has introduced the GeForce GTX 1070 Ti , which is positioned between the already available gaming graphics cards GeForce GTX 1070 and GeForce GTX 1080th The GTX 1000 graphics cards are all based on GPUs Pascal series. Information on upcoming GPU generation, which will be released under the name "amp", there should be the GPU Technology Conference 2018th ( [EMAIL='mfi@ct.de']Mfi[/EMAIL] )" (Google translation) That is it. Nothing about replacing Volta. All those claims are additional speculation piled on the original rumor.

Very good move from Nvidia's side. The whole HBM thing is just not a good idea out of a business perspective for the mass consumer market.

@Monchis (great, you saw my sarcasm mode was on..) pfft, it could be a 3rd gen maxwell refresh for all i care, as long as its using less power and 30% faster 😉

Isn't this simply the code name for the architecture coming after Volta? The Heise artikel seems to indicate that the next architecture to be announced is named Ampere, which would make sense in the naming scheme that NVIDIA uses: Names of Scientists who worked in the area of electricity. Since the roadmap that NVIDIA has shown for their future architectures hasn't been updated in a while (VOLTA is still the newest architecture on the roadmap slides) it's only natural that with the release of Volta the roadmap will be updated soon.

HeavyHemi:

Seems to be a very strong batch of tea leaves...the original article https://www.heise.de/newsticker/meldung/Nvidia-Milliarden-Umsatz-und-kraeftiger-Gewinnsprung-dank-Gaming-3887633.html consists of a rumor of a GPU uarch called Ampere'. "End of the third quarter, Nvidia has introduced the GeForce GTX 1070 Ti , which is positioned between the already available gaming graphics cards GeForce GTX 1070 and GeForce GTX 1080th The GTX 1000 graphics cards are all based on GPUs Pascal series. Information on upcoming GPU generation, which will be released under the name "amp", there should be the GPU Technology Conference 2018th ( [EMAIL='mfi@ct.de']Mfi[/EMAIL] )" (Google translation) That is it. Nothing about replacing Volta. All those claims are additional speculation piled on the original rumor.

& ampersand ahaha that is awesome

kilyan:

i'll keep my just bought 1080 till it gets rusty lol

Good looking, i support you

Crazy Joe:

Isn't this simply the code name for the architecture coming after Volta? The Heise artikel seems to indicate that the next architecture to be announced is named Ampere, which would make sense in the naming scheme that NVIDIA uses: Names of Scientists who worked in the area of electricity. Since the roadmap that NVIDIA has shown for their future architectures hasn't been updated in a while (VOLTA is still the newest architecture on the roadmap slides) it's only natural that with the release of Volta the roadmap will be updated soon.

Blaise Pascal, Alessandro Volta, and André-Marie Ampère were born in this order so fits pretty well for naming schemes. My guess is this is non news and amp/Ampère is the next gen after Volta.

y'know i said this two months ago and got massively flamed. nvidia is in love with A.I., cars, and deep learning - all of which volta delivers in spades - just not at a profitable price point for gaming as the yields from chip fab are entirely too low for mass market. AND there are entirely too many companies here (silicon valley) doing A.I., cars, and deep learning for nvidia to ignore - especially as they pay many multiples of the price of gaming gpu's. just google (alphabet) alone buys tens of millions of dollars worth of gpu's from nvidia. and that barely scratches the surface of this new market. so what i'm saying is nvidia doesn't need to do a damn thing about making volta for the mass market - they are raking in the cash as the last quarterly report shows. while making a new series with just what gamers need and nothing more makes entirely too much sense. so gamers - nvidia doesn't need you anymore...but they still want you.

is everyone just guessing the next cards are going to be a 2xxx and not just a 11xx? just curious.

JamesSneed:

Blaise Pascal, Alessandro Volta, and André-Marie Ampère were born in this order so fits pretty well for naming schemes. My guess is this is non news and amp/Ampère is the next gen after Volta.

I also think it's the next gen architecture after Volta. The born order is a good idea, it works with Kepler as well, but Fermi/Tesla architecture/names before that were born after most of these other people.

tunejunky:

y'know i said this two months ago and got massively flamed. nvidia is in love with A.I., cars, and deep learning - all of which volta delivers in spades - just not at a profitable price point for gaming as the yields from chip fab are entirely too low for mass market. AND there are entirely too many companies here (silicon valley) doing A.I., cars, and deep learning for nvidia to ignore - especially as they pay many multiples of the price of gaming gpu's. just google (alphabet) alone buys tens of millions of dollars worth of gpu's from nvidia. and that barely scratches the surface of this new market. so what i'm saying is nvidia doesn't need to do a damn thing about making volta for the mass market - they are raking in the cash as the last quarterly report shows. while making a new series with just what gamers need and nothing more makes entirely too much sense. so gamers - nvidia doesn't need you anymore...but they still want you.

I remember you posting about this, I don't remember you getting massively flamed. I remember disagreeing with the "yields too low" part because all the current Volta yields are based on a 800m2 chip that would never be manufactured for the consumer in the first place. It would be like saying "Pascal is not at a profitable price point for gaming" right when they announced GP100 - when they shipped GP104 a month after. GV104 for consumers isn't going to be 800mm2, it isn't going to have HBM2, it most likely won't have tensor cores, the yields will be significantly higher than anything out of Volta currently. The wait for consumer based Volta is most likely due to GDDR6 availability, which AFAIK is still slated for Q4/Q1 - GDDR5x was available when GP100 launched, so the launches weren't staggered, GDDR6 isn't and thus you get a staggered launch. I've been saying for a while that I expect Nvidia to split it's datacenter/consumer architectures slightly more going forward, Pascal kind of started it, Volta moreso, it's possible that they just start renaming consumer architectures to something else, but I think in this case Ampere, if that name is accurate, is whatever architecture is after Volta in probably late 2019 and not just a consumer name. Remember, it takes 3-4 years to build these architectures, 2020 GPU designs are probably just starting to take shape.

Next architecture surely is 'Ca$hGrab', then 'Milkit' followed by '$uckerPunch'?

Denial:

I also think it's the next gen architecture after Volta. The born order is a good idea, it works with Kepler as well, but Fermi/Tesla architecture/names before that were born after most of these other people. I remember you posting about this, I don't remember you getting massively flamed. I remember disagreeing with the "yields too low" part because all the current Volta yields are based on a 800m2 chip that would never be manufactured for the consumer in the first place. It would be like saying "Pascal is not at a profitable price point for gaming" right when they announced GP100 - when they shipped GP104 a month after. GV104 for consumers isn't going to be 800mm2, it isn't going to have HBM2, it most likely won't have tensor cores, the yields will be significantly higher than anything out of Volta currently. The wait for consumer based Volta is most likely due to GDDR6 availability, which AFAIK is still slated for Q4/Q1 - GDDR5x was available when GP100 launched, so the launches weren't staggered, GDDR6 isn't and thus you get a staggered launch. I've been saying for a while that I expect Nvidia to split it's datacenter/consumer architectures slightly more going forward, Pascal kind of started it, Volta moreso, it's possible that they just start renaming consumer architectures to something else, but I think in this case Ampere, if that name is accurate, is whatever architecture is after Volta in probably late 2019 and not just a consumer name. Remember, it takes 3-4 years to build these architectures, 2020 GPU designs are probably just starting to take shape.

all of which i consider true. but in 2014 the (profit) center of gravity began to move within nvidia (to industrial from consumer) as the heretofore unintended consequence of their gpu development began to take shape in the A.I. and automotive markets. just the rush to develop autonomous cars by itself would be a major new market, but A.I. and Deep Learning are becoming vast new markets as every university, major industries, and militaries are snapping up high value gpus by the bucket load. for 2018, the growth estimates of A.I., Deep Learning, Autonomous Vehicle (including military drones), and Big Data markets are in double digits for each. all of which require "gpus" and are not cost sensitive...unlike the mass market. i do not doubt in any fashion the ability of nvidia to take from what they learned w/Volta...i count on it. likewise, the gaming market doesn't need the capabilities of Volta either (desire is not equal to need)... all gamers need is a single card to give them 120fps @ 4k and they will sing praises 'til the end of days. nvidia can do that with existing (volta derived) technology - but if they strip off extraneous (to gaming) functions and features they can have a smaller die size with a higher yield at a smaller process... which is to say with the new TSMC process leapfrogging Samsung all they have to do is milk Pascal for another 6-8 months...which is exactly what i perceive tho' i may be wrong.

QUOTE ...current Volta yields are based on a 800m2 chip... holy crap, is that a big chip. whats the PFLOPS on those 🙄

Well seems nvidia to love ai and deep learning nothing strange there every product they release is targeting profit and they profit the most there , meanwhile while they are not challenged in gpu for gamers they have no reason to rush . But i do not think they are willing to abandon a market that made em even more when they have the perfomance crown !

fry178:

@Monchis (great, you saw my sarcasm mode was on..) pfft, it could be a 3rd gen maxwell refresh for all i care, as long as its using less power and 30% faster 😉

I'd also like it to cost less than a left nut so I can buy one without feeling like I've been violated.

Why would anyone pay another $300 for x60 level cards when such cards already need to turn details out of the box to play smoothly at 4 year old console resolution, so short lived... $300 for another 30% is not gonna cut it to play games at $400 ps4 pro´s 1800p resolution smoothly, what a joke.

Monchis:

Why would anyone pay another $300 for x60 level cards when such cards already need to turn details out of the box to play smoothly at 4 year old console resolution, so short lived... $300 for another 30% is not gonna cut it to play games at $400 ps4 pro´s 1800p resolution smoothly, what a joke.

Yayyyyy Nonsense post for the day has been made! GTX 1060 3GB/6GB can play most games, even todays games, at 1080p, on highest quality.

I miss named cards, the Banshee, the Voodoo, the Viper, the Riva.......miles better than i got a GTX1070. lol

@Monchis cause not everyone likes to play up to date games at the low texture/detail settings/and 30fps the consoles do. im running a 1060 right now that can do R6 Siege at 1080p/60fps with LOD on ultra, texture and shaders on high, shadows and reflection maxed out, HBOA running on quality, when the game is actually horribly coded/optimized. the console release is limited to med/low. besides the fact that i have multiple times more options and even things to switch on/off that consoles don't offer. i can easily swap image quality for frame rate and the other way around.