Nvidia is bankrupt and finished

Nvidia is bankrupt and finished

Other urls found in this thread:

ve.ga/
ve.ga
videocardz.com/65406/exclusive-amd-vega-presentation
youtube.com/watch?v=p4YvbwAHJrc
youtube.com/user/amd/videos
youtube.com/watch?v=Ku5OMWYVKSs
youtube.com/watch?v=TLM3LNpEYXA
youtube.com/watch?v=M3chULm4glM
youtube.com/watch?v=4DEfj2MRLtA
twitter.com/NSFWRedditGif

poor volta

>slower than a 6 month old GTX1080
>about to get shit on by the 1080Ti
>bankrupt

have fun with your RX490
AKA 2x RX470 stuffed into a single chip.

...

...

next rebrand, why hype?

Skynet confirmed for running on Vega

what time is the presentation and how can I watch it ??

ve.ga/

AYYMD HOUSEFIRES

AYYMD IS FINISHED & BANKRUPT

>rebrand

ve.ga

Leaked presentation:
videocardz.com/65406/exclusive-amd-vega-presentation

They'll preview the architecture a litle, no actual gaymen GPUs will be shown.

thanks

>GPU for the Immersive Era
I wonder where I heard that before...

Well, since they sell the best performing GPUs i suspect they aren't

OMG!!! GRAPHS with LINES pointing UP!!

NVIDJEWS SOOO DONE AND GASSED NAO!!

No Zen prices will be announced, no gaymen Vega cards will be revealed and yet another shit CES.

3:57 h

they had ryzen+vega playing battlefront apparently outperforming the 1080, why wouldn't they show it again?

>memecardz.com

Find a more credible source, faggot.

how do these fuckers get any info if they are always leaking shit

They had Vega shown on their last livestream, if there was anything else leaked by now we would've known.

That slideshow is both hyping and disappointing.
On the first hand, it shows plenty of great architectural improvement which will result in a great GPU, although I'm fairly sure they overhype a lot the memory efficiency requirements we have for typical consumer load, it's definitely more oriented toward pro user. On the other hand, it's basically a bunch of nothing as far as actual information go, since there's no actual GPU announced, no card announcement of any kind, so basically it's exactly what was written on the tin, an architecture preview.

But at that point I'm really eager for more than that, and I'd love to see an actual GPU, and not in a framerate capped game.

Probably because their source knows that they won't get caught, so they won't mind much leaking stuff. That, and the fact that they usually leak positive stuff, so I guess AMD doesn't give too much of a damn.

they always leak shit 2-3 hours before...

which is why no one cares

the madman he actually did it..
check the geometry throughout

Poo in the loo, Currytech fag.

>Thinks Nvidia could die
>Nvidia just announced GeForce Now which will make more money PER HOUR than AMD can make in a decade.

who wants to take a bet we won't see this gpu till march, AT LEAST

thats a pretty safe bet tho

It'll launch just like how Polaris did.

what does this mean?

>he didnt download robinhood and buy $AMD

Its like you dont enjoy making passive money

They've got Vega demos running live on the show floor.
youtube.com/watch?v=p4YvbwAHJrc

Guessing by the numbers it's faster than what 4096shaers@1500Mhz (Fury X at 1500Mhz) would suggest, so their architecture improvements seem to be doing something.

>Dissing anandtech

Fite me irl mate

basicly vega has finally the advantage over the gameworks shit now..

they will force all of the workloads through their new "ncu's" plus having a bigger and faster primitive rasterizer and a deeper and better color compression(those 2 were the main areas that nvidia had an advantage and plus the inability of amd to fully use their shaders )
and a safe bet given the configuration of the chip the cards will be on 1.5ghz base clock

>thinking I was dissing Anandtech
>not even thinking about WTFtech
I don't know whether I should give you a pat on the back or if I should be yelling at you.

You'll see some gains just from increased number of ROPs anyways.

intel and novidia ded

samefag

videocardz is pointing that the card on the 4k video is the small vega with the 8gb hbm2

So frustrating we have to wait so long for it to come out

Intel and Nvidia r&d must have been getting bored

I honestly forgot about them, they're banned everywhere and Reddit and not even retards go on that site

The real interesting bits (with end-note info analyzed):
>Draw Stream Binning is based on AMD's recent patent published in Dec 2016. It's a hybrid Tile-based Rasterization. In effect, it will boost Rasterization performance as well as reduce bandwidth requirements. This feature is great for both performance & efficiency and it's what makes Maxwell/Pascal so good vs Kepler/GCN.
>NCU confirmed new Compute Units (CU). Not based on GCN as the end-notes are distinctly bringing up the fact that GCN is composed of CUs with 64 Shaders (4x 16 ALU SIMDs). NCU is not going with this config anymore. The config is flexible and per their patents, it can be 1x 16 ALU + 3-5 x 8/4/2 ALU SIMDs & 2x Scalar ALU (twice as many as GCN, really fast for certain ops). From the looks of their diagram, I would guess, twice the # of CUs, hence, twice the Ops per clock when compared to GCN (which AMD is comparing against Fiji/Fury X). This means per CU, ALU count will drop to 32. The derived SIMD config will be potentially: 1x 16 ALU & 1x8 & 1x4 & 1x2 + 2 Scalar ALU = 32.

The 512TB virtual address space is more for MI25 and HPC usage, as it can link with Zen for HSA over the entirety of the memory.

The "Render Back Ends are now Clients of the L2 Cache" basically means less trips to and from memory for game engines that use deferred rendering. Bandwidth savings can lead to performance gains and efficiency.

Hopefully some more gameplay demos, a die-shot of Vega so we can see how big it is (~400mm2?)... and potentially release info during the presentation. This is just a slide-deck, what AMD says will be equally important.

One more thing..
>New Geometry Engine. Compared to GCN in Fury X, the Geo Engines can do 4x polygons per clock. Vega can do up to 11x polygon per clock, a very nice boost.

Typical AMD, only a bunch of useless buzzwords

>NDA ends in 3 hours
Wow, it's fucking nothing.

The front end of the GPU among other things contains the bits that process geometry. Polygons. The types of information that tessellation uses. Nvidia GPUs perform so well here because they have more dedicated hardware for it, and AMD has historically been lacking in this area.

AMD is basically saying: We tessellation kangz

Still better than nvidia meme presentation with $25 for 20 hours geforce now and fcking audi car

AMD admits Nvidia will always be better

>Typical nvidiot too stupid to understand technical specs

good luck paying $900 for the 1080ti

Stupid AMD shill

Robinhood is chink shit because they went to China before they went to literally any other first world country. Fuck those pieces of shit.

Hi Pho, you are lost, please return to reddit immediately.

Stream when?

ve.ga

ve.ga

2 hours 28 minutes

2.5 hours from when I posted
ve.ga

Chance of Nvidia having a dedicated stream?

They did hours ago. Didn't show 1080Ti or anything remotely interesting.

...

Who here shorting $AMD?

So these slides are showing fp8bit quarter precision calculations as well? Get ready for ~50tflop marketing bullshit.

Site down?

Marketing slides bug you so much?

doesn't seem to load for me

fucking pajeet quality

Not really, it's just how clickbait sites are going to spam the fp8 figures improperly.

Don't read clickbait sites?

>980ti is still 800$ in my country in the current year


fucking kill me already

>good luck paying $900

what

Good goy.

Technically speaking, its the same price on newegg here in the States. They quit making it, supply is gone.

what do you mean rabbi, i don't really have a choice...

>any other first world country
you make it sound like china was ever part of the first world.

I find it very strange that people keep using the 1st/2nd/3rd world partition. the soviet union dissolved some 25 years ago and people didn't get the memo

>no card launch

Guess I'll have to settle for GeForce Now

Fucking spoilers. I'll wait for the presentation and ignore this thread

B-b-but isn't AMD for retards?

25 min, GUISE

the epected price of the 1080ti is 850$ minimum.

nvidia is not scared of what amd is doing, or at least weren't, but amd is showing their gpu first again sadly, so nvidia will react to the price like they did last time, only real difference is amd either fired the person leaking or they just have that tight a lip on it this time.

>ve.ga countdown just ended
>no steam

wtf AMD?

ve.ga down

youtube.com/user/amd/videos

ve.ga unreachable, just great fucking start already

youtube.com/watch?v=Ku5OMWYVKSs

youtube.com/watch?v=TLM3LNpEYXA

youtube.com/watch?v=M3chULm4glM

where are the specs?

> but i just love beserk mode, UHHHH

>AMD
>making a countdowntimer for 5 short youtube video's you upload

WOW ITS FUCKING NOTHING
AMD IS FINISHED

amd did nothign wrogn
amd dod nothign

why do pajeets always ruin everything ?

>check the geometry throughout

> up to 11 polys per clock with 4 geometry engines

typo, or just a weird design?

He looks like poo in the loo Jen Hsun

youtube.com/watch?v=4DEfj2MRLtA

SHOW STARTS IN 14 MINUTES!!!111eleven

fuck off retard

That dip in fps in their own fucking video.

Did AMD fuck up? Not making a 600mm GPU when your competition is around 400mm? It would be pricier, but the performance would leapfrog the competition.

>Streamed live on Dec 13, 2016

>The "Render Back Ends are now Clients of the L2 Cache"

holy shit, that wasn't already the case?
the entire point of ROPs/RBEs is to sit as close to RAM as possible and to make consolidated writes so they don't steal all the bandwidth.

Nvidia tried making a 600mm^2 GPU.
It's so rare you have to beg Nvidia to be able to buy one and it's $10.000 a pop.

Yeah, but that's a enterprise beast with a shitload of fp64 with nvlink.