feel sorry for people who fell for the maxwell meme
>the upcoming GPU technology will offer roughly 10 times the performance of Maxwell-powered GPUs
>>51692048
COMPUTE, retard.
Probably gonna get release in two years
>>51692048
watch. They'll release GP204 and then wait for a year and a half to release GP200 because AMD won't have its shit together and be able to provide meaningful competition
>>51692133
You mean GP104 and GP100.
And Arctic Islands is coming next summer.
I'll just a buy a new one and sell my 970
http://www.3dcenter.org/news/reihenweise-pascal-und-volta-codenamen-aufgetaucht-gp100-gp102-gp104-gp106-gp107-gp10b-gv100
GP100
GP102
GP104
GP106
GP107
GP10B
THANK YOU BASED NVIDIA
>>51694671
>970
why would someone want that when the new gpus are 10 times faster?
>>51692091
>>51692048
It will be fixed with drivers
>>51692048
>new GPU more power than old GPU
>>51694743
will you still believe that when benchmarks show only a 5% increase in vidya FPS?
>>51694785
Big Pascal will almost certainly be a 35-40% increase over the 980 Ti, just like that was over a 780 Ti. Especially once Nvidia gimp Maxwell at a driver level.
>3.5
>3.5
>3.5
>3.5
>3.5
>>51692048
>tfw been sitting on 1st gen kepler since launch
should've gotten a 7970 when i had the chance but w/e.
>>51692133
>Rx 400 series GPUs slated summer next year
>Pascal not even slated until late 2016
It's going to be fucking 7000 series all over again.
>>51694886
You're a moron
Pascal will be out earlier, taped out in June while AMD only recently taped out, AMD is months behind Nvidia
>>51694927
Keep telling yourself that. :^)
>>51694935
Keep telling yourself that yourself, LOSER
AMD is months behind Nvidia, period
>>51692048
>he thinks 10x faster = 10x performance
oh, plebs... never change I love to laugh at your blissful ignorance
>>51694949
>...LOSER
Did your shift key get stuck under all your tears and dry semen? I'd clean your keyboard if I were you, Anon. And then neck myself.
>>51694949
But that's wrong, you fucking retard.
http://www.gputechconf.com/
APRIL 4-7, 2016
SILICON VALLEY
THANK YOU BASED NVIDIA
Can't wait for Pascal
>>51694960
>>51694962
>BUTTMAD AMDPOORFAGS WITH NO GPUS & NO DRIVERS OR HOUSEFIRE GPU KILLING DRIVERS
>>51692096
>Probably gonna get release in two years
Q2 2016
but have we seen the wood screws yet? You only know its legit when Huang breaks out the wood screws.
>>51694989
>muh memes!!
>the upcoming GPU technology will offer roughly 10 times the performance of Maxwell-powered GPUs in terms of processing power related to "deep learning" apps
>in terms of processing power related to "deep learning" apps
in terms of processing power related to "deep learning" apps
>in terms of processing power related to "deep learning" apps
in terms of processing power related to "deep learning" apps
>in terms of processing power related to "deep learning" apps
in terms of processing power related to "deep learning" apps
>in terms of processing power related to "deep learning" apps
in terms of processing power related to "deep learning" apps
>in terms of processing power related to "deep learning" apps
in terms of processing power related to "deep learning" apps
>>51692048
>in terms of processing power related to deep learning
>>51694976
>10 core broadwell-e Q1 2016
>pascal GTX with nvlink april 2016
>Intel Optane 3d xpoint drives that make SSDs look like drives from the 80s
fucking hell my anus is not ready, 2016 is going to be retarded
>>51695037
how much does that apply to average consumer or gaymer ;_,
>>51692048
if the next macbook pro comes with a discrete pascal GPU I will jizz my pants.
>>51694808
Eh, I'm hoping that rather than gimp, they just push for "4K gaymen". That way there's no need to gimp the Maxwell cards when you can buy one Pascal card for your 3 monitor setup instead of three Maxwell cards.
Early Nvidia adopters usually don't give a fuck about price anyways.
>>51695054
>average
>10 core broadwell-e
start stealing from millionaires and saving now, because you aren't seeing this in your lifetime otherwise
Nvidia usually starts with the average cards, so might get one of them if you have the motherboard to support it. Going on how much memory bandwidth these will require I don't see them running on PCIE but I guess we'll see
Optane drives, who knows. Maybe for the first 6 months they'll charge an arm, leg, testicle and an ear or 2? I hope it will drop after that.
>>51695054
Not much lol
Games are just starting to take advantage of the extra threads on i7's
>>51695087
It will but it will only be in the $2,500 model and only in 15''.
Apple's top priority is thinner and better battery life. Performance is secondary at best (Macbook Pro) and an afterthought at worst (Macbook 2015). Intel's integrated graphics significantly cut down on power consumption and are sufficient for most normies' use cases.
>>51695101
At this rate, Apple laptops will be sharp enough to kill people with, since they seem to want to have "katana thin" at the molecular level.
>>51694808
>>51695094
My real concern is will any game be able to make a 980ti go under 60 fps on 1080p? no matter how good pascal is i don't think so
and i am about to get one for that reason.
>>51694989
So that shift key is definitely stuck down? You should get on with killing yourself already.
>>51695278
modern games aren't even good.
why worry bout that shit
>>51695311
Not really worried, that's why i am getting a 980ti to enjoy myself now, games can't even beat witcher 3 in graphics right now and a 980ti
will have plenty of room if you play it max on 1080p 60 fps, so thats what i am saying to people who say wait for pascal, no matter how good pascal
ends up being, no game will come out at least next year that can make the 980ti go under 60 fps on 1080p.
>>51695342
Minecraft can make your 980Ti go 0.1fps.
>>51692091
It's not even compute, it's "deep learning" which means they can make it whatever the fuck they want it to be
>>51695015
MAXWELLKEKS ON SUICIDE WATCH
>the next generation is more powerful!
wow, incredible, this never happens
>talking about desktop computer hardware
>apps
it's all over.
fuck this modern pc times.
>>51694956
>tips fedora
>>51692048
So they're finally going to stop reducing the FP64 clocks on their GPUs?
First Kepler with 1/24th and then Maxwell with 1/32nd. When will it end?
>>51695115
>We're slicing for an entirely new audience, the fedora wearing katana enthusiast neckbeard.
Apple pls
>>51694989
so do you have a source saying when pascal will be released?
because i can find sources for arctic islands release dates.
>>51695474
You're a moron, consumer cards will always have low FP64 performance
>>51695487
Did you intentionally ignore the April 4-7 dates?
>>51695095
all the consumer grade cards will be on pcie. nvlink is server grade only for the time being. it's too expensive to implement otherwise and motherboard manufacturers would hate having half their boards taken up by something one side can't use at all. it would limit people to expensive as hell boards that can only do 2x SLI or 2x crossfire.
>>51695403
tbqh senpai I don't like it either but if you still can't handle the word "app" you're an 100% autist
>>51692048
finally 120fps Minecraft
arent the Fury gpu some next level shit already. HBM came out of nowhere like they released 2 generations at once
>>51695498
then they shouldn't be fucking advertising them as such.
>>51695511
so, paper launch then 3 months later actual chips?
k.
>>51695520
>all the consumer grade cards will be on pcie. nvlink is server grade only for the time being. it's too expensive to implement otherwise and motherboard manufacturers would hate having half their boards taken up by something one side can't use at all. it would limit people to expensive as hell boards that can only do 2x SLI or 2x crossfire.
Good to know. It really seems nvlink would need a total redesign on consumer grade motherboards. I was wondering how they are going to accomplish this on a x99 board.
Any idea how much a pascal card would actually saturate a board with 16 pci express 3.0 lanes? Assuming theoretical maximum bandwidth of 32gb/s for 16x pci-e 3.0 and a pascal card trying to needing pull possibly 200gb/s from them in order to operate at 100%?
Do you think we could see 32x pcie? or pcie 4.0 earlier? How about SLI adapters for a single card just to provide additional bandwidth?
>>51695715
PCIe 4.0 hasn't even be finalised
http://www.eetimes.com/document.asp?doc_id=1326922&page_number=1
>The bad news is this is the last turn of the crank for a copper version of the interconnect and, despite progress, a final 1.0 version of the spec may not be ready until early 2017.
>>51695715
for 32x PCIe you would need a slot twice as long. wouldn't happen. PCIe 4.0 would be more likely.
The most likely case is that Nvidia is pulling shit out of their ass and these cards are all going to be 3.0 and won't even saturate an x8 slot.
i suspect the x10 performance is because they're actually increasing clocks for FP64 and the ultra-high bandwidth demand is purely for computing huge datasets which doesn't even apply to gaming performance.
>>51695787
>a final 1.0 version of the spec may not be ready until early 2017.
Oh, dip
I was thinking it was releasing in 2017. PCI-E might come to an end by the time it sees daylight...
>>51692048
yet it will only be 10-20% faster in gaming
>>51694743
Not everyone has infinite money
>>51695801
>i suspect the x10 performance is because they're actually increasing clocks for FP64 and the ultra-high bandwidth demand is purely for computing huge datasets which doesn't even apply to gaming performance.
I suspect the 10x performance is a buzword for a lazy company that cannot come up with viable buzwords. It really seems it is an internal memory bandwidth bump. I'm betting clocks won't even increase, they might even need to go down slightly.
>>51692048
Kill yourself.
>>51692091
It's not even that. They're claiming it's for fp16.
I don't believe them, even if it's just fp16.
>>51695715
They're pulling shit out of their ass.
>>51695801
The performance increase advertised is specifically for fp16.
>>51695278
>My real concern is will any game be able to make a 980ti go under 60 fps on 1080p?
Plenty, especially if you want to start cranking up AA. A 980 Ti only gets 40fps at 1080p in the new Ass Creed using mere FXAA, let alone the TXAA options.
>buy r9 290 two years ago
>still maxes every game
>still viable for the next 1-2 years due to dx12
>meanwhile nvidia users are still waiting for the 10x improvement promised to them every year while while they buy the annual $600 titan equivalent for nominal performance gains for a year before its neutered by neglected drivers and games optimized for the next architecture.
When will nvidia users learn?
>>51699417
You can't play your video games at 1080p 144Hz on Ultra, let alone 1440p and god forbid 4K.
>>51699417
>still maxes every game
Sure, buddy. At 30fps. Feel free to post a video of you running The Witcher 3, Just Cause 3, GTA V, Arkham Knight or the last couple of Assassin's Creed games maxed out at 60fps+ to prove me wrong. :^)
>>51694949
GF tapped out last year, they tested 14nm mass production last month.
>>51699159
>the new Ass Creed
That series is a horribly coded game that gets rushed for a yearly new release, so it shouldn't even be considered for benchmarks.
>>51695015
>in terms of processing power related to "deep learning" apps
And all of that is down to HBM2 being 5x faster, plus the new gpu being 2x faster, resulting in the 10x increase.
>>51699555
on my 290, at 1440p 96hz, i get 50-60fps on tw3 on ultra, only hairworks, CR, and vignetting disable because they reduce image quality. I don't think a 980ti would hit 96fps+ with those settings to make any perceptible difference in smoothness.
>>51696923
>It's not even that. They're claiming it's for fp16.
So basically they'll be as good at fp16 as AMD was for the past five years. They'll finally be able to mine bitcoin faster than Radeons.
>>51699664
>plus the new gpu being 2x faster
It never worked that way, it never will. Tech is not marketing magic.
feeling sorry for people who are falling for the Maxwell meme
>10x better performance on Volta
>>51699735
>DDR4 will be so fast and so low voltage
>>51699752
DDR4 is fast and has low voltage.
>believing nvidia marketing slides
What the fuck happened to /g/?
>>51699767
marginally, not exactly meaningful for a consumer
Can someone explain the AMD Poorfag meme? AMD and Nvidia cards seem near equal in price
>>51699913
the difference is nvidia users have to upgrade cards 2-3 times more frequently than AMD users.
nvidia users call amd users poorfags because they use cards more than 8-12 months old instead of upgrading.
>>51699962
Well there isn't any point in replacing an AMD card when almost the entire line is rebrands.
>"3.5Gb"
>"DirectX 12"
>"Game Ready Drivers"
>"Gameworks"
>>51699962
Well my gtx280 lasted well for 7 years before upgrading to gtx970
>>51699159
Isn't FXAA just a blur filter over the whole screen? Or was that some other AMD based AA method?
>>51699635
Syndicate runs fine once you turn a few settings down. It's just demanding.
>>51699672
>maxes every game
>with s-some settings t-turned down
Pic related.
>>51700053
Yes. It's shitty post-process AA that looks like trash, but has a tiny performance impact. SMAA is similarly-light and blurs much less, but generally needs a temporal element added to actually look good. And if you fuck up the temporal element because you're incompetent cunts, like Avalanche did for Just Cause 3's SMAA T2X, it creates ghosting and other problems.
>>51700427
Forget my pic baka desu senpai.
>>51699693
>bitcoin
>fp16
pretty sure that's all integer arithmetic.
>>51692048
28^2 / 16^2 ~ 4
Going from 28 nm to 16 nm will roughly quadruple the transistor density. If you expected your old 28 nm tech silicon to stay relevant you were deluded.
>>51700427
>actually liking chromatic aberration or vignetting
Why even care about graphics if you're this blind?
>>51692048
>He believes in the 10x performance meme
>>51700672
I really wish they wouldn't measure in half-gate pitch,
The real feature size (per transistor) is somewhere around 60-65 for 28nm and 40 for 16
That's not a quadrupling of density, it's 1.7x
>>51700800
>1.7
i recognize that number from somewhere
Implying Nvidia wouldn't kneecap it and sell a 'unleashed' Titan XYZ for $15000.
I feel sorry for the people who will fall for the Pascal meme.
>>51692048
performance =/= efficiency
Will the new shit be power efficient or not? I don't want to be forced to buy a 1kw psu and begin proverbially throwing money away for my electric bill.
>>51700753
>deflecting
>>51700800
idk where you got those numbers from but
65^2 / 40^2 = 2.64
It's still a ridiculous transistor budget increase.
>>51700440
Witcher runs at about 55 fps on an OC 290 on max at 1080p. It, crysis 3, and gta V are about the only things that push it. Whatever though filters aren't really "graphics." Were all being jewed by max settings anyway.
>>51700967
So close to infinitely vertical performance leaps. Future looks bright senpai.
>>51694782
underrated post
Will Maxwell allow 4k gaming at 60fps at the $350 range? If not what about Volta?
>>51694703
Wait, I can't read Reich, are these gonna be the card names or just the processor names?
