[Boards: 3 / a / aco / adv / an / asp / b / bant / biz / c / can / cgl / ck / cm / co / cock / d / diy / e / fa / fap / fit / fitlit / g / gd / gif / h / hc / his / hm / hr / i / ic / int / jp / k / lgbt / lit / m / mlp / mlpol / mo / mtv / mu / n / news / o / out / outsoc / p / po / pol / qa / qst / r / r9k / s / s4s / sci / soc / sp / spa / t / tg / toy / trash / trv / tv / u / v / vg / vint / vip / vp / vr / w / wg / wsg / wsr / x / y ] [Search | Voyeur Cams | Click for more| Home]

NVIDIA Unveils Pascal GPU: 16GB of memory, 1TB/s Bandwidth

This is a blue board which means that it's for everybody (Safe For Work content only). If you see any adult content, please report it.
Voyeur Cams

Thread replies: 231
Thread images: 45

File: file.png (361KB, 567x797px) Image search: [iqdb] [SauceNao] [Google]
file.png
361KB, 567x797px
http://vrworld.com/2015/11/16/nvidia-unveils-pascal-gpu-16gb-of-memory-1tbs-bandwidth/

Who's hype for $10,000 GPUs?
>>
>>51400741
Good thing I didn't give my dad that spare kidney of mine.
>>
>Pascal is expected to hit the market during the first half of 2016.
AMD BTFO
>>
>“Using 3D memory, not only the memory capacity will go up, the memory bandwidth will go up significantly. With a much faster GPU, and higher memory bandwidth, the existing interconnects in the server are just plain outdated. So, we had to develop our own interconnect called NVLink, five times faster than existing technology.”

>Combined figures are very interesting to compare – 24GB GDDR5 and 480GB/s bandwidth should be replaced with 32GB HBM2 and 2TB/s bandwidth, mutually connected through NVLink rather than PCIe.

Wait what. Are they requiring motherboards to have a whole new god damned physical slot for these things?
>>
File: 1441616158100.jpg (15KB, 206x226px) Image search: [iqdb] [SauceNao] [Google]
1441616158100.jpg
15KB, 206x226px
>>51400788
:^)
>>
>>51400788
That's right. Buying an Nvidia card in the future will require you to also purchase a motherboard that only works with Nvidia graphics cards.

They're trying to create an ecosystem to lock you in. First they started with monitors that require an Nvidia card to be used at their full potential, and now they're working on motherboards that discourage ever switching to AMD due to the increased cost and hassle that would come with needing a different motherboard. Where will they go next, I wonder?
>>
>>51400907
>Nvidia motherboards are Intel only
>>
>>51400788
So doees this mean the card will take up TWO pcie 3.0x16 slots, cause if so, neat.
>>
http://www.gputechconf.com/

APRIL 4-7, 2016 SILICON VALLEY
>>
>>51400939
>>51400923
>>51400907
>>51400788
NVLink is enterprise grade only, at least for the time being.
>>
File: 1446199126153.jpg (8KB, 262x312px) Image search: [iqdb] [SauceNao] [Google]
1446199126153.jpg
8KB, 262x312px
>>51400788
>>51400907
>>
>>51401018
Wasn't it about sharing between GPUs the VRAM?
>>
IT'S OVER, AMD IS FINISHED & BANKRUPT
>>
This is a tragedy for us consumers. AMD will be officially killed (or left with just minimal amount to be technically alive because then NVIDIA can say they're not officially a monopoly, although I suppose that's what AMD is currently).
>>
>>51400907
What's wrong with this? PCI-E can't transfer that much data. What do you want them to do? Be bottlenecked by a port?
>>
Butthurt AMD shills inbound
>>
>>51401158
>amd next gen gpu actually better than nvidia even after goyworks
>no one buys nvidia because you need new mobo
>nvidia sudoku
>>
>nvidia descartes
oh yes, give her the 3D
>>
>>51400907
nvidiots will defend this
>>
>>51401061
So an SLI bridge?

Nvidia link is going to be a bigger failure than their console.
>>
>>51401191
develope an open source, backwards compatible standard. Thats what amd did with cpus years ago, and what needs to remain the practice. No competition is unhealthy
>>
>>51401295
More like amdfags trembles in fear at the thought of it.

Doesn't bother me one bit. I'll buy the best available, nvidia or amd.

Amd knows they aren't the best so they're worried.
>>
>>51401191
>consumer-grade graphics cards
>bottlenecked by PCIe
anon please, don't tell me you fell for nvidia's memes
>>
File: intel_vs_amd_3.jpg (106KB, 1200x706px) Image search: [iqdb] [SauceNao] [Google]
intel_vs_amd_3.jpg
106KB, 1200x706px
So it's finally over, AMD is finished and bankrupt.
>>
>>51401319
That literally only benefits amd. There's no other player on the field.

Sorry but that's not how business works.
>>
>>51401338
jokes on you.
the amd cpu is older.
>>
So you guys are complaining about NVIDIA announcing their most powerful consumer GPU yet....
Cool.
>>
>>51401507
they haven't announced any consumer GPUs, though.
>>
>>51401523
My dad owns NVIDIA. He said low price/powerful GPU's for 8K VR gaming is coming next year.
They will be less than $500 and will come with a 10 year warranty.
>>
>>51401354
It's like you like being fucked in the ass by corporations. Nvidia would love nothing more than to charge 980ti prices for a 950.
>>
>>51401338
uh it clearly show AMD won there SIXTEEN THOUSAND SIX HUNDRED AND SIXTY SIX (16666) points
>>
>>51401586
Good luck prying the 950 out of my wallet.

Businesses aren't charities.
>>
>>51401571
Give me things that will never happen for 500 Alex
>>
I just happen to know Nvidia's CMO, he says GP204 will be 30% faster than GM200 and GP200 will be 80% faster GM200
>>
File: 1447631766803.jpg (38KB, 720x720px) Image search: [iqdb] [SauceNao] [Google]
1447631766803.jpg
38KB, 720x720px
>>51401682
>>
>>51401696
Why is some weird girl getting strangled on a Serbian flag?
>>
So if I understand correctly they will be stuffing their new 3Dxpoint memory in these gpus

http://www.dailytech.com/Exclusive+If+Intel+and+Microns+Xpoint+is+3D+Phase+Change+Memory+Boy+Did+They+Patent+It/article37451.htm
>>
>>51401682
>I just happen to know Nvidia's CMO
Does your uncle work for Nintendo, too?
>>
>>51401765
Are you retarded?
>>
>>51400788
I pledge here not to buy nvidia motherboard ever.
>>
When should we expect consumer cards using this 16nm tech?
>>
>>51401824
Fuck off entitled cunt, 28nm more than fine.
>>
>>51401338
Comparing a mobile CPU to it was the fairest thing they could do.
If they compared a desktop Intel it wouldn't be fair since the Intel would win so much that they had to run the CPU on 3 cores for the multi-core test.
>>
File: 1439486156131.png (74KB, 313x133px) Image search: [iqdb] [SauceNao] [Google]
1439486156131.png
74KB, 313x133px
>>51400788
WOOOOOOW
>>
File: X9svTXx.gif (481KB, 400x226px) Image search: [iqdb] [SauceNao] [Google]
X9svTXx.gif
481KB, 400x226px
>>51402934
>>
>>51400907
Don't really mind being locked in bh Nvidia. I only ever buy from them anyway.
>>
>>51400788
No, just more proprietary goyware for nvidia to shovel out to the goyim.
Open standards don't make nvidia money.
>>
nvidia is the new micoshaft taken to the next level
>>
>>51401623
And then the industry stagnates or turns into a very successful monopoly.

You're right! Businesses aren't charities. Charities don't have this kind of potential negative impact to make on their own "industry."
>>
i'm currently using a 630m. should i upgrade to a 970 now or wait for pascal? are the new cards going to be overpriced? i'm on a budget
>>
>>51400907
After keking users out of a motherboard I don't think they need to do much else.
Once AMD dies there will be nvidia only computers just because AMD is dead.
>>
>nvidia says "up to 16GB of memory, 1TB/s Bandwidth" because they're showing off the "potential" of HBM2
>faggot media just runs as a confirmed 16GB HBM2 GPU
You fucking retards, they aren't gonna go forward with that, it would be too expensive.
It's just a marketing scheme that both Nvidia and AMD run every fucking year for hype.
>>
>>51403671
>Once AMD dies
>people still think that's gonna happen
>>
>>51401338
This is pretty bait.
That mobile cpu overclocks to about 3.8ghz and its not 3 years old (clearly says 2014) and the macbook pro is not an ultrabook

The 8370 isnt the fastest desktop cpu amd gas to offer and if you count opterons its not even close
>>
File: ham.gif (795KB, 500x281px) Image search: [iqdb] [SauceNao] [Google]
ham.gif
795KB, 500x281px
10,000, that in 5 months will cost 1 grand.
Can't wait to buy it for 1 grand or less in 5 months after release, gotta love nvidia.
>>
>>51400788
Fuck its going to be like nForce chipsets but a million times worse jesus
>>
>>51401018
underrated post

article isn't about consumer products
>>
Does anyone believe they'll actually have them ready in the market in the first half of 2016? Pretty sure HBM2 was only to enter large scale manufacturing by the end of the 1st half, and even then, AMD was to get the majority of the first batches because they helped develop it.

How can Nvidia hope to have enough stock to satisfy the number of orders worldwide? I can only imagine that even at the end of the 1st half, Nvidia will barely have enough test units themselves.

Are they just gonna skyrocket the price of all their cards that come with HBM2? Are they gonna pull a Fury, and only release it on the top of the top tier, while the remaining cards of the series are Maxwell rebrands? Could we legit be looking at the first consumer-targetted cards with a price tag of 5,000$? Wouldn't surprise me after the blunder that was the TitanX.

Either way, I'm calling bullshit on Nvidia.
>>
>>51400907
I don't really mind. When was the last time AMD make a better card than nVidia anyway?
>>
>>51400907
By doing that they also prevent you from switching TO nvidia.
>>
>>51403855
FuryX is already edging out the 980Ti.
>>
b-but vram is irrelevant XD
>>
>>51403884
>reference 980ti's without OC
>>
>>51403892
It's only irrelevant when it's AMD that has more of it :^)

>>51403921
It is outpacing it though. You can already tell it'll be better than it in one or two quarters.

Nvidia cards come optimized for the current systems and games and can't be squeezed out much more. AMD cards, on the other hand, just keep on giving.
>>
The shilling in this post is intense, brand whoring is a mental illness, get threated immediately.
>>
>ITT imbeciles who think Nvidia will force them to buy dedicated mobos with no PCIe

Did you mongols even read the link? It's for ENTERPRISE, aka not consumers.
>>
>>51403855
R700, Evergreen, Southern Islands, Hawii, and now Fiji is on par with 980ti after its first driver update.
>>
This isn't even for video games you dumbshits.
>>
>>51403823
The whole thing just sounds like marketing hype to be followed by a paper launch next year. Normal consumers will forget these kinds of cards even exist 2 weeks later.
>>
>>51403996
Seeing as how NVLink is something IBM and Nvidia worked on for POWER machines, normal consumers wont even see it.
>>
>>51404023
Eh, they could pull it off as some sort of marketing or to lock users to their standard. But then again, doing it now wouldn't make much sense, they could've done it before.

Or they're just going maximum turbo jew to finally attempt to kill off AMD
>>
>>51404107
I don't think it would be super smart of them to go pissing in Intel's territory. I'm sure it's also expensive as fuck, otherwise we would already have something 5 times faster than current PCIE x16.
>>
>>51404203
True that. I'm pretty sure PCI-E is still not enough of a bottleneck to justify new standards though. It's much faster than anything you need right now or near future.
>>
>>51404220
New standards? Nvidia is not interested in making a new standard. They're interested in more consumer lock-in. You need to really keep this in mind whenever you think nvidia.
Is there anything that nvidia hasn't vendor locked?
>>
>>51404235
Poorfags
>>
>>51404254
Still waiting for things nvidia hasn't vendor locked. Or at least attempted to.
>>
>>51404264
Poorfags. They haven't vendor locked poorfags like you.
>>
File: 1481.gif (2MB, 254x275px) Image search: [iqdb] [SauceNao] [Google]
1481.gif
2MB, 254x275px
>>51400741
>1TB/s

pic related
>>
>>51404274
If you think you're being clever or funny, you're not. And you have a extremely inaccurate view of yourself.
>>
>>51404309
Enterprise: 100%
Consumer: 82%
Discreet mobile: 100%

Well who's not using Nvidia these days?

People having financial hardship. That's where AMD swoops in.
>>
1TB/s*

*only for the first 4GB of memory.
>>
>>51404309
He's still obviously superior to you though. Keep crying, poorfag ;)
>>
>>51404338
Error: Most people are poor, ergo market share isn't not displaying poorhood.

You have to not be poor to afford AMD too. If you're $50 less poor than someone who can afford AMD cards and you call yourself 'rich' and them 'poor' then it means that $50 is a large percentage of your capital which would make you objectively POOR.
>>
>>51404365
>isn't not
Great way to fuck yourself over dude
You mean that if the majority of people are poor this means that a marketshare showing that most people use the more expensive items does not show poorness objectively.
>>
>>51404365
That's the thing, if you aren't poor you wouldn't buy AMD.
>>
>>51400907
Haha, if I ever needed a reason to never buy Nvidia again...
>>
>>51400939
Wait that would be fantastic. I mean, it sucks to use 2 physical connections instead of 1, but
a) it beats having some proprietary connector and having to buy a whole new motherboard
b) GPUs wouldn sag as much because they'd have 2 connectors holding them
What a time to be alive.
>>
>>51404398
And if you're poor it means you would buy AMD?
Error: You're not addressing anything.
If the marketshare shows most people buy the most expensive thing then the marketshare is not related to how rich the people are.

And since you're calling AMD users poor (when they are objectively able to afford $300 graphics cards) it means that the small difference in prices (A hundred bux is not a lot to rich people, but it is enough for you to say the difference between rich and poor in a population rich enough to afford graphics cards at all) it means you are poor.
You are objectively poor.
>>
>>51403263
>Run bitch
>>
>>51404410
Most likely it'll have an proprietary chip on the certified motherboards that links together two x16 lanes via software.

The newfags here are forgetting not even 10 years ago we needed specific sli or cfx certified motherboards. pcie is an industry standard. They won't be moving away from it.
>>
>>51404423
Okay, so not even poor people are buying AMD. Is this why they're failing at everything they do?
>>
>>51404423
>saving money
>poor

A shekel a day keeps the shoa away
>>
>>51403871
4GB would've sufficed for the 390. Though 8 is nice for high res textures and mods or crossfiring for 4K.

The only card that could have truly benefited from 8GB, the FuryX is 4GB because of HBM1 limitations, pity. would've been insane to crossfire two 8GB versions of the furyX.
>>
>>51401354
>that's not how business works
I don't give a shit how businesses work. They should be creating open standards. That's how the whole fucking computing world has strived.
>>
>>51404564
This. Anyone who doesn't understand that's how we even got to this point in time with standardization is a fucking subhuman mongoloid.
>>
>>51400741
I'll probably pick one up when people are still paying off their first dibs price like HBM. My 290x has been serving me well for years now anyways
>>
>>51404564
>use open standards!
Translation: Bawwww why isn't Nvidia sharing their research with us
>competition is good!
Translation: Wahhhhh stop buying Nvidia

Are you one of those failures that tried to copy other students work back in school?

Why can't AMD get shit together and stop asking for free handouts?
>>
File: 1444127643430.gif (2MB, 320x135px) Image search: [iqdb] [SauceNao] [Google]
1444127643430.gif
2MB, 320x135px
>>51404675
>Bawwww why isn't Nvidia sharing their research with us
NVLink isn't research, it's money grabbing and monopolizing.
What the hell can it do that PCI can't?
Just make use of the 32x socket if you need more bandwidth you fuckers!
>>
>>51405723
I believe they're realeasing the pcie 4.0 slot standards in early 2017 anyways so what's the point of pouring rnd into something that is a standard for only one side of the table
>>
>>51401708
It's Ronda Rousey getting BTFO
>>
File: 7kChU2K.jpg (40KB, 454x453px) Image search: [iqdb] [SauceNao] [Google]
7kChU2K.jpg
40KB, 454x453px
>>51403972
FUCKING THIS PEOPLE LEARN TO READ
THIS IS BASICALLY THE PASCAL QUADRO ANNOUNCEMENT
FUCKING THIS PEOPLE LEARN TO READ
THIS IS BASICALLY THE PASCAL QUADRO ANNOUNCEMENT
FUCKING THIS PEOPLE LEARN TO READ
THIS IS BASICALLY THE PASCAL QUADRO ANNOUNCEMENT
FUCKING THIS PEOPLE LEARN TO READ
THIS IS BASICALLY THE PASCAL QUADRO ANNOUNCEMENT
FUCKING THIS PEOPLE LEARN TO READ
THIS IS BASICALLY THE PASCAL QUADRO ANNOUNCEMENT
FUCKING THIS PEOPLE LEARN TO READ
THIS IS BASICALLY THE PASCAL QUADRO ANNOUNCEMENT

That out of the way - I'm excited to see pascal finally becoming something we may see a lot more of soon.
>>
>every normie has been convinced AMD has no drivers
It's really simple as to why AMD is dying. They even believe you NEED gsync.
>>
>>51405723
>NVLink isn't research, it's money grabbing and monopolizing.
We just barely know what it is. We have no idea how it'll be implemented.
>What the hell can it do that PCI can't?
Read the fucking article.

If you don't like it don't fucking buy it. You amdindus always dramatize everything.
>>
>>51405858
Well it is The Way It's Meant to be Played™!
>>
>>51404338
>People having financial hardship. That's where AMD swoops in.

the 970 is $100 cheaper than a 390 and outperforms it, so i don't think that's the case

most of the people i see buying AMD are redditors and /g/ retards who get tricked into buying nuclear reactor rebrandeon cards, the majority of people who do their research can see plain as day that NVIDIA is the superior choice
>>
File: perfrel_2560_1440.png (42KB, 500x1050px) Image search: [iqdb] [SauceNao] [Google]
perfrel_2560_1440.png
42KB, 500x1050px
>>51405986
>the 970 is $100 cheaper than a 390 and outperforms it, so i don't think that's the case
Cool story, Onii-chan.
>>
>>51406018
>dishonestly comparing reference cards to factory oced rebrandeon junk

you tried, shill
>>
>>51406033
There's no AIB card besides the 980ti in that chart, they're all reference.
>>
>>51406049

there's no such thing as a reference 390/390x, currently all of them are factory OCed
>>
File: goy knows.jpg (69KB, 345x431px) Image search: [iqdb] [SauceNao] [Google]
goy knows.jpg
69KB, 345x431px
>>51403728
>gas to offer
>>
File: 1447577616579.png (107KB, 500x600px) Image search: [iqdb] [SauceNao] [Google]
1447577616579.png
107KB, 500x600px
>>
>>51406055
There's a bunch of them at stock, reference clocks, which are used in the TPU benches.
>>
>>51406069

if they were reference they'd be performing the same clock for clock, therefore we know they're NOT using reference clocks.

you're trying too hard today, roy
>>
>>51405858
You need variable framerate, not gsync.
>>
File: 67896.png (38KB, 650x400px) Image search: [iqdb] [SauceNao] [Google]
67896.png
38KB, 650x400px
>970 outperforms a $500 GPU
>>
>>51406094
Ah, you're the type that thinks the Hawaii and Grenada are the same silicon.
I see, I see.
Well, obviously its my fault for even bothering.
>>
File: 67899.png (37KB, 650x400px) Image search: [iqdb] [SauceNao] [Google]
67899.png
37KB, 650x400px
>>
File: 67931.png (37KB, 650x400px) Image search: [iqdb] [SauceNao] [Google]
67931.png
37KB, 650x400px
>>
>>51406116
>>51406137
>aftermarket, factory OC'd
>vs reference 290X
>old benchmark
I see we have a legit shill ITT.
>>
File: 67955.png (38KB, 650x400px) Image search: [iqdb] [SauceNao] [Google]
67955.png
38KB, 650x400px
>>
>>51406165
amd shill gets a taste of its own medicine

LOOOOOOOOOOOOOOOOOOOOOOOL
>>
>>51406165

it performs just fine against the factory oced 390x, a GPU that is still $200 higher in price than a 970 for some reason
>>
File: old man huang.jpg (25KB, 475x324px) Image search: [iqdb] [SauceNao] [Google]
old man huang.jpg
25KB, 475x324px
You are serving me well /g/.
>>
>thanks AMD!
>>
>you can get a 2 year old 290 for under $200 nowadays
>it equals a 970
Great job Nvidia, I'm sure once Pascal hits your 'old' GPU will do even better.
>>
File: perfrel_2560_1440.png (52KB, 500x1050px) Image search: [iqdb] [SauceNao] [Google]
perfrel_2560_1440.png
52KB, 500x1050px
>>51406249
>>
>>51406237

>peak power numbers

Once again [H] proves it has no idea how hardware works at all.
>>
>>51406267
Shill site and fake, GTFO AMDpoor
>>
File: perfrel_2560.gif (64KB, 400x1057px) Image search: [iqdb] [SauceNao] [Google]
perfrel_2560.gif
64KB, 400x1057px
>>51406249
>>51406267
>>
>>51406130
>Hawaii and Grenada are the same silicon
But they essentially are. Just changing a few pathways doesn't make it a completely different chip. It's like saying that a car is a completely different model because it uses a different air filter design.
>>
File: capsoc.jpg (168KB, 485x723px) Image search: [iqdb] [SauceNao] [Google]
capsoc.jpg
168KB, 485x723px
>>51406249
You can buy a 970 for $230 RIGHT NOW and it'll come with an aftermarket heatsink, runs cooler, draws half the power, won't damage your hearing, and gets specifically optimized drivers with every game release AND it'll have 10~20% more performance. You also get a free AAA game which you can play or sell for $40.

Who in the right mind would settle with a $200 reference 290? Are you insane?
>>
File: perf_oc (1).gif (30KB, 400x411px) Image search: [iqdb] [SauceNao] [Google]
perf_oc (1).gif
30KB, 400x411px
>>
>>51406190
>>51406116
>>51406137
>>51406155
>>51406171
>>51406190
>>51406197
>>51406237
>>51406282
>>51406307
http://www.bit-tech.net/hardware/graphics/2015/07/10/sapphire-r9-fury-tri-x/4

GTFO, shill.
>>
File: perf_oc (2).gif (28KB, 400x377px) Image search: [iqdb] [SauceNao] [Google]
perf_oc (2).gif
28KB, 400x377px
>>51406334

wow, you posted a source showing a reference 980 beating the more expensive and hotter fury. it's like you're trying to discredit yourself, roy.

now just imagine what the numbers look like when OCed compared to the voltage and oc locked fiji chips
>>
>>51406237
>doesn't show the overclock draw for 980 or 970
Cute. Probably doesn't show because they both draw embarrassing amounts of power under overclocks.
>>
>>51406334
>sapphire
Not surprising that's amd kuck's favorite brand. They will void your warranty if you try to replace the crusty old factory thermal paste.
>>
>>51406371

the 970 draw is with an OC, the 980 they compared it to was reference
>>
Can we at least agree that the 970 is a really good card despite the 3.5? It's as good as a more expensive 2/390X despite having less VRAM and having less compute units.
Plus, it can overclock far better than a 2/390, draws less power even with an insanely high overclock, run most games with better stability (better, more frequent drivers), is far more supported within the industry due to its proliferation, and has a better value for money than a 2/390 right now.

Poor people don't even buy 390s, they buy 750s and 960s.
>>
>>51406384
why don't they show it under overclock instead of full load then?
why not have numbers for both not overclocked and overclocked?
why are they representing their data unfaithfully?

As a reminder, boost != overclock.
>>
File: power.gif (19KB, 802x503px) Image search: [iqdb] [SauceNao] [Google]
power.gif
19KB, 802x503px
>>51406371
But that's wrong you fucking retard. Voltage has no impact on Maxwell overclocking and all Maxwell chips can overclock to their maximum potential without increasing the voltage, thus the overclock power consumption is negligible.
>>
>>51406412
http://www.kitguru.net/components/graphic-cards/anton-shilov/overclocker-overvoltaging-alone-does-not-help-maxwell-to-overclock/

http://linustechtips.com/main/topic/412778-some-words-from-vince-lucido-kingpin-on-maxwell-asic/


http://www.overclock.net/t/1565834/evga-com-evga-to-offer-ability-to-purchase-980ti-kinpin-cards-by-asic/140#post_24201543
>>
File: 1442193039299.jpg (68KB, 640x480px) Image search: [iqdb] [SauceNao] [Google]
1442193039299.jpg
68KB, 640x480px
>>51406412
>600w
>>
>>51406398
I think we can ALL agree that Nvidia just works.
>>
>>51406398
>It's as good as a more expensive 2/390X despite having less VRAM and having less compute units.
It's not though, all tests show the 290X/390X beating it by at least 10%.
>>
File: scaling.gif (24KB, 802x503px) Image search: [iqdb] [SauceNao] [Google]
scaling.gif
24KB, 802x503px
>>51406432
That image alone doesn't put things into context.

The Fury X consumes an extra 250w with a 75hz overclock.

75.

And AMD advertised it as an overclocker's dream. AMD is a very dishonest and manipulative company.
>>
>>51406488
>It's not though, all tests show the 290X/390X beating it by at least 10%.

no, the 390x beats a reference 970. a factory OCed or properly OCed 970 will perform on par with both the 390x and reference 980, in some cases outperform them
>>
>>51406488
>95% of the performance
>75% of the price
Still seems good to me
>>
>>51406494
I don't remember AMD advertising the fury X as a overclocking dream. Infact what I remember is overclocking not even being allowed on it.
>>
File: perfwatt_1920 (1).gif (41KB, 400x615px) Image search: [iqdb] [SauceNao] [Google]
perfwatt_1920 (1).gif
41KB, 400x615px
LOL
O
L
>>
>>51406533
They're nvidia shills, they go back and forth on what AMD's doing wrong every time they open their mouth.
>>
>>51406533
>I don't remember AMD advertising the fury X as a overclocking dream
>Well, during the official launch of Fury X, Joe Macri, CTO at AMD, said this card will be ‘every overclocker’s dream’!
>>
>>51406560
>believing any of the chiefs of any company
>>
>>51406543
>Doesn't have the Fury X
Fury X hangs out right alongside the 960 in perf/watt.

Fury Nano destroys that whole graph.
>>
>>51406574
Words from an executive officer's mouth of any company is taken as advertisement by the press and the public.
Fury X an 'overclocker's dream' my ass
>>
File: Capture.jpg (158KB, 679x927px) Image search: [iqdb] [SauceNao] [Google]
Capture.jpg
158KB, 679x927px
>>51406560
>how do I google

AMD does whatever it can to get you to buy their product, and their after sales support is non-existent.
>>
File: AMD Vs Nvidia.webm (2MB, 960x540px) Image search: [iqdb] [SauceNao] [Google]
AMD Vs Nvidia.webm
2MB, 960x540px
>>51406068
>>
>>51406628
AMDtards are no different.
There was one guy that tried to argue that the Zen hadn't been delayed.

They're complete fucking retards when it comes to AMD's own press releases. For a loyal bunch, they sure don't read up on their company's progress.
>>
>>51406663
I suppose AMD is the one that crashed and nVidia stood on the safe side.
>>
File: 1445605792048.png (35KB, 601x511px) Image search: [iqdb] [SauceNao] [Google]
1445605792048.png
35KB, 601x511px
>>
File: perfwatt_1920_1080.png (43KB, 500x850px) Image search: [iqdb] [SauceNao] [Google]
perfwatt_1920_1080.png
43KB, 500x850px
>>51406603
Here's a more recent one.

Funny how a slightly overclocked 980 ti is more efficient than a stock one, which goes exactly in line with what I said here >>51406412
>>51406430

Maxwell is so efficient you can overclock it by 50% (from stock 1000mhz) and you'll only see a 10-20w increase in power consumption. The scaling is nearly 100% too.
>>
File: perfwatt_3840_2160.png (43KB, 500x850px) Image search: [iqdb] [SauceNao] [Google]
perfwatt_3840_2160.png
43KB, 500x850px
>>51406936
>>
ITT /g/ fails to understand 1) how the laws of physics work 2) how power consumption works and 3) how to read charts and graphs.
>>
>>51403930
Take a look at the 980ti strix or the matrix compared to the r9 fury. The fury can't compete with a fully oc'd 980ti.
>>
>>51403673
this.
they're essentially just showing investors how much growth potential they have with the new tech.

in laymans terms this means
>releasing gimped hardware for expensive as fuck
>potential for 32gb vram 2tb bandwidth
>means they can sell expensive as fuck hardware year after year until they reach that point
>only ever incrementally increasing the power of the GPUs to milk the tech instead of trying to progress the hardware further.

kind of like how they double the storage capacity of SD cards every five years even though it's been proven to store 2TB worth easily and cheaply.
>>
NVIDIA
FINDS
A
WAY

>AMD ON SUICIDE WATCH
>>
When would the mobile line get released? I'm looking into buying a gayming laptop.
>>
Anyone notice that blurb on ECC out of nowhere? I wish there were more details about that. Sounds like ECC will be required.
>>
when will pascal be released approximately, i'm still rocking a 4850 and need an upgrade badly
>>
Hi there!

You seem to have made a bit of a mistake in your post. Luckily, the users of 4chan are always willing to help you clear this problem right up! You appear to have used a tripcode when posting, but your identity has nothing at all to do with the conversation! Whoops! You should always remember to stop using your tripcode when the thread it was used for is gone, unless another one is started! Posting with a tripcode when it isn't necessary is pants on head retarded. You should always try to post anonymously, unless your identity is absolutely vital to the post that you're making!

Now, there's no need to thank me - I'm just doing my bait to help you get used to the anonymous image-board culture!
>>
>>51400788
Don't they mean that nvlink is like a second sli bridge, but for vram.
>>
>>51407408
>doing my bait
>>
File: 1420862366287.jpg (17KB, 200x199px) Image search: [iqdb] [SauceNao] [Google]
1420862366287.jpg
17KB, 200x199px
>>51400788
DOA
O
A
>>
>>51400788
>mutually connected through NVLink rather than PCIe.
>mutually connected through NVLink
>mutually

What two things are being connected here?

Sounds to me like they're talking about the connection within the card itself, between the processor and on-board memory.
>>
>>51401354
No one? Really? What about OpenPower they are involved now? Not to mention, how is thunderbolt faring against USB? For that matter, what made the PCs what they are now if not IBM being defeated when challenging the BIOS reverse engineering?
>>
>>51405986
>the 970 is $100 cheaper than a 390 and outperforms it
Fuck off shill
>>
>>51401338
That's not even the most powerful desktop one
The AMD fx-9590 is
>>
>>51408017
He probably lives in Narnia, where Nvidia is cheaper than amd. And Intel actually makes progress without competition.
>>
>>51410225

i hope this is b8, the 970, 980 and 980ti are all cheaper than AMD's equivalents

>980ti - $600 vs $650~ fury x (which is usually price gouged due to poor supply)
>980 - $500 vs $550 fury
>970 - $250 vs $330 390 and $430 390x (it outperforms both)
>>
>>51410555
Fuck off with that 970 pricing, would be nice to actually have something comparable in my country. And comparing stock 390 to max OC 970.

Not fuck you there is only one 390 with actual overclock and it's 1050mhz core from MSI rest is either stock like Gigabyte that sometimes says it's like 1010 core but defaults to 1000 or is Sapphire tier overclock of fucking 10MHz on core which is pretty much nothing. And fucking MSI 390/390x has a lot of headroom for OC.
>>
>>51410555
Not in Belgium. 980ti costs €750 980 costs €600, 970 costs €400 I'm not sure about ayymd.
Just look at pic related, and somehow they seem to have run out of half their cards even though most cheapasses would probably order from amazon.fr or .de (there's not amazon.be sadly)??
>>
>>51410760
>Ayymd
>overclocking
Don't even choose to be honest mate
>>
>>51410816
MSI can do 1200MHz with some luck on 390 because it's overvolted. There is also that 1200MHz core guy with his 290x.
>>
>>51410836
>Meanwhile 970 easily goes over 1500
Nvidia was stupid enough to make the 970 so overclockable, I think the 970s are just 980s with the 3,5/4 meme.
>>
>>51410836
You'll get a 100mhz oc at best and the gains are absolute shit. An oc'ed hawaii also consumes more power than a fucking 980 ti.
>>
>>51410920
Because they are
When the 3.5 thing came out, their press report was:
>We make the high end ones and then from there we just cripple stuff down to make the lower ones
>>
>>51410920
And 970 doesn't scale so linearly. And no 980 is as much of a 970 as 290x is of 290. Which means they probably castrate some parts of 980 and sell it as 970.
>>51410955
I guess those guys with 200MHz+ are liars. And no shit 980Ti is power efficient. Good card for winter.
>>
>>51410958
Kek'd. I don't know if this is for real, but this is basically what a lot of big companies do.
>>
what a waste of money that will be
>>
>>51410975
>I guess those guys with 200MHz+ are liars
Not liars, just pumping 1.3v and watercooling that shit while the card sucks up 600w at full load.
>>
>>51411154
http://www.overclock.net/t/1561704/official-amd-r9-390-390x-owners-club
Fucking air stock everywhere.
>>
>>51400788
Didn't intel do this in the 90s?
>>
Desktop computers are going to be obsolete in 5-10 years to home users. It will be used only on servers and workstations.

It is space, power and time inefficient to 95% of the consumers; smartphones are also more social-friendly
>>
>>51411615
Genius as aways Ahmed.
>>
>>51400788
No. You CAN if you want to, but the Pascal architecture still is compatible with PCIE, it'll just use NVLink to communicate with an SLI card, then
>>
>>51400741
>16GB of memory
christ. i just recently got 8gb ram.
>>
>>51404397
you didn't have to try to translate it shithead
>>
>>51401338
TO BE FILLED BY OEM
>>
>>51400741
AMD BTFO
>>
>>51406758
Other way around
>>
you have to remember, this article just regurgitated the same information that was released way back in march 2015 about pascal.

information that is primarily talking about their professional line. the tesla / quadro line. which has been stuck in keplar due to maxwel's lack of support FP64.

the 16gb of ram will not most likely be sitting on consumer grade $300 - $500 cards, but most likely reserved for their next flagship $1,000 card and quadro / tesla cards.

nvidia has already coming out saying they are not abandoning gddr5 and there will be a lot of pascal cards running on gddr5 instead of hbm.

memory bandwidth is important, but not that important. see amd's current fiji release. hbm didn't really help it out at all even though hbm1 which is used on fiji, while pascal will be hbm2, is already faster than gddr5.

nvlink will be tesla only from it what seems like. nvlink is nvidia's proprietary connector that replaces pci-express completely. if intel doesn't support it, you won't see it on consumer hardware. nvidia cannot make chipsets for intel since intel doesn't license out anymore for third party chipsets.

hbm2 most likely will only be on $350+ nvidia cards. the next 970 equivalent will most likely be 4gb hbm2, 980 most likely 4gb, and the 980 ti replacement most likely 8gb. while the next titan will be 16gb.

really, the 980 ti most likely already provides the same gpu performance as the next 970 pascal replacement. only it offering better ram bandwidth.

ram bandwidth being more important for 1440p and above rather than 1080p or lower.

outside of ram bandwidth the next big upgrade for pascal is die shrink, but seeing the last die shrink, fermi to keplar, and seeing the die shrinks for cpu's, it will provide a boost, but not that much. power usage will go down of course.

which is most likely while most of nvidia's flaunting of pascal has more to do with memory bandwidth and not so much flaunting about gpu performance.
>>
>>51400741
i wanna buy one and hang it on my neck with my gold chain.
>>
>>51400741
>it's actually 16gb
>it's not 14gb
>>
>>51400741
>"unveils"

>exactly what everyone's been saying since finfets and HBM2 were known to the world
>everything /g/ has been saying for +8 months

>still a shitstorm of a thread

never change, boys
>>
>you're a shill
>no, you're a shill
Good conversation guise
>>
>>51406367
there isnt even a fury on that chart...
>>
>>51406758
joke is amd has no drivers and that nvidia runs hot and explodes

These are Fermi memes. https://encyclopediadramatica.se/Fermi
>>
Nvidia couldn't even get away with this crap if they made good products, AMD is better anyway for christ's sake.
>>
>>51405986
pretty much

its just faggots who get on their high horse online saying that they support open standards and get a shitty card to do so
>>
>>51406307
where the hell are you getting that pricing on a 970 from?
>>
>>51416098
America

970 will be $180 on black friday
>>
>>51410555
Not in Switzerland, 980Ti and 980 are overpriced as fuck over here.
Even with poor supply, the Fury X is way cheaper, got mine for 620 bucks in release day.
>>
Jesus, how can amd even compete

$570 after rebate, comes with a free game worth $40 on ebay.

If you have an amex card (such as bluebird prepaid) you get an extra $25 off, making it $545. Sell the game and it'll be $505.
>>
>>51400741
Tfw we are living in the age of 3D and Waifus are real

https://a.uguu.se/ekrxzn.webm

Great to be alive
>>
>>51404489
There is nothing wrong in saving some hard earned shekels, if you don't really that 5 - 10% performance increase.
But straight up saying that someone, who spends $50 less, is a poorfag is fucking retarded.

And I'm saying that as a Nvidia user.
>>
>>51400741
AMD IS KILL
>>
I don't get it. After reading this thread, I feel as if some people actually try to derive a sense of worth and superiority from their fucking GPU purchases. "Nvidia shill" this, "AMD shill" that, it's sad as fuck. You guys literally suck the cock of a business for free. No, even worse, you suck their cock after they make a profit on you.

Fucks sake. NVLink, if it remains closed source, would be terrible for consumers. Locking people in to or out of whole ecosystems benefits only the companies doing so; all it does is remove current and future choice from the user, while clearing the ground for monopolistic practices to start peeking their ugly heads around the corner. Remember Intel prices before AMD?

Supercomputers and industry? Yeah, it doesn't matter if NVLink stays proprietary if it will only be used in those cases, shit there has always been pretty much custom-solution-by-specific-company-with-no-cross-compatability since Day 1.

That being said, AMD needs to step up their game. While it's pretty impressive what they're able to do with three-generation old re-brands, it's getting very long in the tooth.
>>
>>51416098
newest Nvidia viral trick. They promise great deals, but when you actually check them they are out of stock or no longer available.
>>
>>51404365
Are you literally retarded? If you make 10k a year then 50 bucks matters. If you make 200 you're still poor but can afford 50 bucks for nvidia. But then if you make 140k you might buy amd to save 50 bucks?

Please blow your brains out.
>>
>>51417411
http://www.newegg.com/Product/Product.aspx?Item=9SIA24G3KF9215&cm_re=06g-p4-4993-kr-_-14-487-141-_-Product

It's $714 if you actually go on newegg. Clearly that was a price mistake.
>>
>>51420667
Wrong. If you make 10k a year, then that $200 you spend on the card matters. The other $700 on the computer itself matters.
We're talking relative sums here, not absolutes.

If you make 140k a year, then the price isn't important to you, high OR low.
>>
>>51420578
>NVLink, if it remains closed source, would be terrible for consumers
No, because NVLink exists due to insufficient bandwidth in the existing PCIE3 standard. It would be a good thing for consumers who can afford it because they'll be getting a product that is potentially far more capable than any other graphics card using PCIE.
If NVLink is not a PCIE replacement, then it will be no different than an SLI bridge. As long as a motherboard supports NVLink, then there shouldn't be a big deal.

>Locking people in to or out of whole ecosystems benefits only the companies doing so
And how is this a bad thing when it provides a much more stable and consistent environment for developing firmware? Why do you think Apple's OSes has seen such huge success and stability over Windows? Why do we, the end users, have to suffer from potential incompatibility problems with hardware that cost in their hundreds of dollars?
We paid good money for a product and it should just work the way it was meant to. We shouldn't suffer because there are a million different combination of software and hardware that could conflict with a driver update.
>all it does is remove current and future choice from the user
All it does is narrows the options consumers have in exchange for better operating experiences. The more choices you have, the more problems and incompatibility you introduce to the system. It's why Windows is a buggy, unoptimized POS OS compared to Apple's OS.
I'd be willing to pay that price premium for a much better user experience and I'll be damned if I'm the only one who thinks so.
>>
>>51420700
see
>>51414535
>nvlink will be tesla only from it what seems like. nvlink is nvidia's proprietary connector that replaces pci-express completely. if intel doesn't support it, you won't see it on consumer hardware. nvidia cannot make chipsets for intel since intel doesn't license its out anymore for third party chipsets.
>>
>>51420723
Then Intel better do what's in their best interests and make NVLink compatible with their future chipsets.
>>
>>51420738
intel won't.

intel already stated the next chipset will be pci-express 4
>>
>>51420738
intel isn't in the business of using other peoples prosperity nonsense on their chipsets.

it was intel after all that forced amd and nvidia to play nice and allow both crossfire and sli support on intels own chipset.

intel won't use nvlink since its not open. they will only use it if nvidia gets rid of its license for it.
>>
>>51420742
NVLink is probably compatible or will be compatible with PCIE4.
>>
>>51420761
proprietary *
>>
>>51420770
http://www.kitguru.net/components/graphic-cards/anton-shilov/pci-express-4-0-with-16gts-data-rates-and-new-connector-to-be-finalized-by-2017/

http://www.theregister.co.uk/2014/06/10/cheap_backwardscompatible_pcie_40_on_track_for_del2015del_2016/

no it won't

again, nvlink is for the professional market. the tesla crowd where they are connecting 20+ gpus together for work orders and need the bandwidth for that many cards running together.

not your typical gamer running one or two cards to shitpost on 4chan and masturbate to cartoon porn games.

even right now pci-express 3 isn't even fully saturated even in dual card configs.

pascal needing more bandwidth than pci-express 3 16x? no. no where close.

maybe volta, but thats not slated till 2017 - 2018 and pci-express 4 will be out by then.

this has all happened before, and will happen again, and again, and again.

from pci-express 1 to 1.1 to 2 and now 3.

from tesla, to fermi, to keplar. the last three die shrinks.
>>
>>51420668
http://www.newegg.com/Product/Product.aspx?Item=N82E16814487142&ignorebbr=1&cm_re=EVGA_GeForce_GTX_980_Ti-_-14-487-142-_-Product

it sold out already. not a price mistake.
>>
>>51412754
Means it's probably in amd's own test rig, and therefore it's probably as high as it can score.
>>
File: roadmap_large.png (32KB, 580x266px) Image search: [iqdb] [SauceNao] [Google]
roadmap_large.png
32KB, 580x266px
>>51420840
how could it be? its a competitor to pci-express 4.

anyways, nvidia even said it out of their own mouth, nvlink is for the tesla market where that sort of bandwidth is needed. the cards linked up to their own proprietary connector and connected to a server farm over possibly some sort of fiber since fiber is the only thing that could handle the bandwidth.

don't get me wrong, pascal will be great, but i feel like people are ignoring the last 10+ years of history of new generations of cards being released on new die shrinks. if anything 980 ti would most likely already offering the next $350 card replacement. it lines up just great with the last 10+ years of card releases.

the only thing new this time around is the new memory standard, hbm 2. but unless you're gaming on 4k, i can't see hbm 2 benefiting much for a user on a 1080p screen or lower.

i see volta the next generation thats worth it and looking at nvidias own slides, volta has the highest "boost" of any previous generation over the previous generation.
Thread posts: 231
Thread images: 45


[Boards: 3 / a / aco / adv / an / asp / b / bant / biz / c / can / cgl / ck / cm / co / cock / d / diy / e / fa / fap / fit / fitlit / g / gd / gif / h / hc / his / hm / hr / i / ic / int / jp / k / lgbt / lit / m / mlp / mlpol / mo / mtv / mu / n / news / o / out / outsoc / p / po / pol / qa / qst / r / r9k / s / s4s / sci / soc / sp / spa / t / tg / toy / trash / trv / tv / u / v / vg / vint / vip / vp / vr / w / wg / wsg / wsr / x / y] [Search | Top | Home]
Please support this website by donating Bitcoins to 16mKtbZiwW52BLkibtCr8jUg2KVUMTxVQ5
If a post contains copyrighted or illegal content, please click on that post's [Report] button and fill out a post removal request
All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.
This is a 4chan archive - all of the content originated from that site. This means that 4Archive shows an archive of their content. If you need information for a Poster - contact them.