NVIDIA Prepping High-Performance Volta GPU Lineup - GV104, GV102,

>>56046946
You mean that Pascal wasn't their "high" performance?
AMD is fucked.

>>56047021
Better fucked than cucked. Enjoy your driver "support" for Pascal when Volta arrives in six months. :^)

>>56046946
Does that mean that Pascal will only be a single series unlike the other architectures?

>Vega is meant to take on Pascal
>Nvidia is already prepping Volta which has a new architecture and will put an end to the "BUT AMD HAS DX12 AND ASYNC!!!" meme

AMD - Years behind the competition

>>56047021
>bought 390x for 275$
>Volta/Vega is coming soon
>Just in time to have enough money to buy either
More like people who bought 2016 cards are fucked

10 series cards were a fucking joke
Anyone that bought one is a retard

>>56047518
>10 series cards were a fucking joke
W H Y
H
Y

>>56047530

Rushed to market, over priced, throttling, horrible PCB design and layout. No HBM2. Shall I go on?

Enjoy your annual videocard purchases Nvidiots.

Will Volta be the return of the hardware scheduler that went MIA because Fermi was literally a nuclear reactor level housefire

>>56047579
>Rushed to market
It's bad?

>over priced
Pascal is cheaper than Maxwell...

>throttling
?

>horrible PCB design and layout
?

>No HBM2
and?

>nvidiot
I have a 290...

>>56047530
> 900 series getting relegated to legacy status
>”But muh legacy term doesn’t mean it will not be supported!”
>Overpriced Pascal FE editions
>”But muh craftmanship and order of 10 gaming perfection!”
>Suddenly Volta appears out of nowhere and probably priced higher than Pascal with FE edition
>”But muh technological advancement!! I must buy this one next year!”
I hope you guys see the pattern.

>>56047518
>>56047579
>polaris wan't a joke

>>56047683

Apple doesn't think so.

http://www.extremetech.com/gaming/233389-report-apple-planning-major-macbook-pro-refresh-with-amd-polaris-gpus-touchid-and-oled-function-keys

File: perfrel_2560_1440.png (33KB, 500x810px) Image search: [Google]

33KB, 500x810px

>>56047643
And this is why Pascal is a joke?

And you don't talk about the performance, the overclocking, the power draw, the performance per watt?

Where are the AMD cards?

>>56047731
Apple is the joke itself. Overpriced throttling laptops or w/e the fuck retards call them nowadays and bendable phones - such a great company!

>>56047731
It's all about trading profit margin.

>>56047781
You don't my point don't you?...
What's the point of buying pascal this year when I can buy Volta for 2017 which probably better and much efficient

File: 1457152463153.png (222KB, 449x401px) Image search: [Google]

222KB, 449x401px

>>56047579
>Rushed to market, over priced, throttling, horrible PCB design and layout. No HBM2

Oh my God.

>>56047843
>a wait™fag

>>56047579
So a bunch of dumb shit that doesn't matter but AMDrones like to use as an argument anyway?

The fact is the GTX 1070 is the best per/watt GPU in years, and the Titan XP is the most powerful single card ever and wont burn your house down

>>56047864
>He didn’t understand my point
> reiterate properly my point/argument
>sling “a wait™fag” out of nowhere
christ /g/ what happened to you and I thought I’m getting a proper insight

>>56047864
>Enjoy your annual GPU purchase

File: inigo.jpg (74KB, 700x468px) Image search: [Google]

74KB, 700x468px

>>56047040

>>56047919

Nvidia/AMD threads are full of wannabe /g/ posters. They belong in /v/.

File: 1457993499218.jpg (62KB, 576x1024px) Image search: [Google]

62KB, 576x1024px

>>56047579
Shitposting

>>56047864
Seriously AMDtards are so fucking dumb.

They'd rather wait for years than actually buy something good now.

>>56047910

Polaris 150W TDP
GP 104 120W TDP

>MUH FURNACES!

>>56047922
>don't sell your old gpu

File: image.jpg (42KB, 500x500px) Image search: [Google]

42KB, 500x500px

>>56047579
1060/70 wasnt overpriced, me and my buds that bought them haven't had throttling issues, PCB design is subjective and HBM2 is a bitch to manufacture

pajeet plsgo

>>56047961

>memeposting

>>56047579
WHAHAHAHAHAHAHAHAHAHAHA

File: amd kid.jpg (2MB, 8624x3896px) Image search: [Google]

2MB, 8624x3896px

>>56047993

>>56047982

>What Nvidiots actually believe

>>56048013

>Defining shitposting

File: 14nm vs 16nm vs 28nm.png (344KB, 1988x930px) Image search: [Google]

344KB, 1988x930px

>>56047974

>>56048051

And all it took to get there was ripping out hardware features and putting in tiny VRAM bus widths.

Meanwhile a year from now Volta is kicking the shit out of Pascal while Polaris is just catching it's stride.

>>56048080
It's 290x vs 780ti all over again

>>56047781
>1440p
Most people are still below 1080p dude

>>56048095

Just wait for those intrinsic shader functions like Doom has to be wide spread.

File: perfrel_1920_1080.png (34KB, 500x810px) Image search: [Google]

34KB, 500x810px

>>56048119

So, is it worth at all to go for the Pascal cards anymore or should one just wait for Volta?
I was thinking of upgrading my 960 to a 1070.

File: 1perf_oc.png (25KB, 500x490px) Image search: [Google]

25KB, 500x490px

>>56048236
The 1070 is faster than Titan x, go for it.

>>56047579
enjoy buying refurbished amd cards from pajeet when amd files for bankruptcy in four years

>>56048283

Everyone plays battlefield 3!

>>56048373
Show me a game where the Titan X is faster than GTX1070

Someone please tell me what my opinion should be on AMD new onboard SSD (SSG?) GPUs

>>56047021

pascal is a test run filled with crap figures (titan x gp102 not even 3840 core) and lies (twice titan x), shrunk down maxwell, the actual customers who bought this were fucked but the performance is so high nobody gives a damn

nvidia officially moved from tsmc to samsung so this can only make sense for their real new architecture

>>56046946
GP102/Titan X 2.0/1080 Ti will be Nvidia's high end through early-mid 2017.

The only reason for a 384b GDDR5X card to even exist is that they knew they had HBM design problems even before the Koreans announced their production slip with the new memory.

Vega, no matter how decent or not it ends up, won't be facing of Volta for 3 months in the best case, who knows how long worst case.

>>56048451

not to mention 6.7GB instead of 8GB but everyone is so engrossed with post purchase rationalization and confirmation bias, enjoy spending more in a couple of years or get phased out in your cuck resolutions

>>56048448
feature targeted at professional which will probably be integrated in the desktop generation after Vega because AMD will run out of gimmicks to market

>>56048517

Will it be useful for simulations or is it only for rendering grafix?

>>56048478

implying there's no Titan X black lurking with full GP100 and HBM2 at several thousand bucks

>>56047587
>>56047587
This.

Volta actually refers to the amount or voltage these things will require for the extra hardware nvidia needs to add to stay competitive.

>>56048482
>6.7GiiiiB

Okay

>>56047488
>people who bought 2016 cards are fucked

2016 cards are basically Polaris and GP106, which are lower end, plus GP104 in the $400-$700 range.

GTX 1080 was obviously wallet rape, but I can't really see Vega/Volta competing with 1070 or lower GPUs.

>>56048539
will be very good for memory intensive stuff
probably 1-2 irrelevant games
some miners may benefit a lot from it

>>56048558
GP100 has 2 main issues for gaming:

- it's designed around Hawaii-style full 2:1 fp32:fp64 and not at all around gaming

- it's woodscrews tier at this point. they had to show off a "remote live" demo last week at SIGGRAPH

>>56048478
At this point I just feel bad for amd. Maxwell was a 70% performance improvement over 700 series, Pascal is a 60% improvement over Maxwell.
Polaris is only about 20% faster than tahiti from a million years ago. Vega will have trouble getting to 1070 speed, launching a year later, having to compete against sale prices and game bundles.

>>56048013
lovely goy boi

>>56047488
>More like people who bought 2016 cards are fucked

I used my last card a 290 tri-x for a year and sold it 2 weeks ago for net $10 less than I paid. A year from now when new cards come out I'll be so fucked when I sell my 1040 for like $400.

>>56048051
AMD is so fucking shit

>>56047579

>Rushed to market, over priced, throttling, horrible PCB design and layout. No HBM2. Shall I go on?

ironic. poolaris cards have/had all of those problems, especially the pcb part.

>>56048147

shader intrinsics being widespread would be a bad thing. they're architecture-specific ISA that will be obsolete if AMD makes a new architecture instead of rehashing the same GCN core.

>>56049759
Kyle was 100% right

"AMD has created a product that runs hotter and slower than its competition's new architecture by a potentially significant margin. Let the internal finger-pointing begin! The NVIDIA GeForce GTX 1080 and 1070 launch was a surprise to AMD and it simply does not have a harpoon for Moby Dick this time...again. One could go as far as saying that the potential exists for another R600 moment. Now, I am sure AMD will take every step possible to mitigate this gap but the simple reality of the situation as it stands today is that AMD has a loser on its hands and are going to have to pull every trick in the book to spit-shine the turds, and the fact of the matter is pricing is how you do that."

All the while there are miners AMD GPU's will sell well. Even used.

>>56048517

SSG will never make it into a consumer card. It's only used as a stopgap because the SSG card has a fiji die and only 4GB of vram. Once HBM2 is shipping we'll have cards with 32GB of HBM2 pushing >1TB/s in workstation cards.

>>56050021
Their pricing isn't even that good. 480 is still sold out across the board in the US. 470 is $200 compared to $180 380x, and $230 970. 460 cost more and is worse than the 370, gets blown up by the 950 at the same price. Low end new gen cards shouldn't be losing to old stuff in price/performance but AMDs architecture improvements are so bad that that's the case.

Pascal is revolutionary. AMDPOORFAGS are jelly.

>>56047982
>PCB design is subjective
>Engineering is now subjective
I'm a raging nvidiot but you're a retard.

>>56050021

Is he still crying about not being sent a nano?

>>56050797

So revolutionary, less then a week after Titan XP launches Volta leaks start happening.

>>56050020
>What's a GCN 1.1 vs 4.0?

They aren't changing from GCN any time soon. Especially since all their built in shader functions port from consoles to PCs. So in order to take advantage of it they have to write it once for either or, and it's already taken advantage of across every GCN card.

Meanwhile Nvidia still trying to pull their finger out their ass and patch Vulkan.

didnt nvidia have some supercomputer deal due and hence have to get the volta ready soon. the actual consumer voltas will probably be out in 2018 like originally planned.

>>56050001
Source?

>>56051157
>They aren't changing from GCN any time soon.

they will have to if they want to ever compete with NVIDIA again. GCN's per-core throughput hasn't improved since GCN 1, all they've done is pack more CUs into their designs. once they unfuck themselves and put out something new, all those old optimizations will be completely obsolete without relying on implementations of the old ISA in hardware, which turns your architecture into most CPUs are now.

>>56051265
I guess that means no source.

Samsung 14nm FinFET?

File: 61004993.jpg (83KB, 460x259px) Image search: [Google]

83KB, 460x259px

>>56051964

>through put core per
What the fuck are you talking about? Was that supposed to make sense?

They have legitimately TOO much hardware, from superior compute, to superior 32FP and 64FP, to dedicated shaders for the TruAudio DSP's, to the ACE's that give them TEN TIMES the asynchronous compute advantage over Nvidias shit tier Asynchronous compute.

And they STILL manage to keep up in DX11 and curb stomp in Vulkan and DX12.

And if Volta doesn't include a hardware based compute capability Nvidia will be shit tier in VR and the current API's.

So just to sum up, AMD has had the hardware for years. They now have the software to take advantage of it. But keep buying your stripped down and streamlined hardware for more money. Nvidia couldn't be successful with out your ignorance.

>>56048448
Pixar remarked that a big problem in Monsters University's visuals was the fact that they were limited to 64 gigabytes of data per frame.
AMD looked at the issue Pixar had, decided it probably applied to all rendering software used in movies, and made a push to resolve the issue.
If this works as intended a company like Pixar could buy a farm of these and solve that memory issue. If that happens a lot of other studios are going to jump on what's obviously a good thing - removing a technical limitation in their CG.

File: 1467425350256.jpg (61KB, 634x634px) Image search: [Google]

61KB, 634x634px

>>56052771
I think this is the first time I've legitimately recognized damage control.

I'm gonna get a silent, multi-fanned Volta GPU and name it Frances the Mute

>>56047579
You shut your whore mouth. Iam enjoying my 1060 mini

>>56051964
Oh look its this tech illiterate retard again.

IPC has improved with each iteration of the GCN line. Ellesmere massively outperforms Tahiti per clock.

File: LL.jpg (74KB, 637x417px) Image search: [Google]

74KB, 637x417px

>>56051265
Reference Polaris cards have too many VRM phases drawing off the PCIe slot, which can theoretically cause the mobo 12V line voltage to droop a bit when total card draw jumps to/above 150W.

An 8-pin power connector would have been the smart call, but signs point to the RX480 being a last-minute factory overclock job.

>>56053772
On the one hand I can get GTX 1060's with good customer support.

>>56053799
On the other I can get an rx480 Sapphire for £240.

This is doing my bonce in.

>>56053815
There was nothing last minute about it, the narrative of morons regurgitating this on Anandtech is mind numbing.

The only area where 36CU Ellesmere calls behind 44CU Hawaii is in pixel throughput from having half the ROPs. The comparatively high clocks seen in RX 480 are to offset this. The die was made to be small, cheap, and simple. They targeted high clocks from the beginning knowing that 14nm LPP provides a substantial increase in fmax vs prior 28nm HPP nodes.

>>56047843
>What's the point of buying this year when I can buy something probably better next year.

Repeat every year and pray for your hardware to die so you have an excuse to upgrade.

>>56054261
Not taking the piss but since I've been on consoles from 2006 until I got a 750ti.

Would it be such a bad thing to run a 1060 into the ground? My only contention with AMD cards really is the shit AIB partners unless you want to pay a fuck load.

I don't trust XFX, sapphire, or powercolor.