So now that Vega is officially as disappointing as everyone knew it would be...

So now that Vega is officially as disappointing as everyone knew it would be, it would seem that getting a nVidia is the only option.

However, would it be okay to lock yourself to Pascal, an architecture that is two years old and about to be replaced?

I guess I'm asking: how good is nVidia's support for older cards? I got burned when AMD moved my trinity APU 13 months after it released it.

Is the 9X0/10X0 series gonna be like the 7X0/8X0 series and become unoptimized trash in a year or two?

If so, might as well wait for Volta.

Other urls found in this thread:

developer.nvidia.com/opengl-driver
nvidia.com/en-us/data-center/tesla-v100/
nvidia.com/object/tesla-p100.html
youtube.com/watch?v=iq202LKkeHI
blogs.nvidia.com/blog/2017/07/22/tesla-v100-cvpr-nvail/
youtube.com/watch?v=oPRP6C5MlWY
nvidia.com/en-us/geforce/products/10series/titan-xp/
blogs.nvidia.com/blog/2017/07/31/titan-xp-drivers-new-levels-of-performance-for-creative-professionals/
nvidia.com/download/driverResults.aspx/121776/en-us
computerbase.de/2017-01/geforce-treiber-test/2/
computerbase.de/2017-01/geforce-treiber-test/2/#diagramm-dirt-rally
techreport.com/news/32191/nvidia-finally-lets-fermi-gpu-owners-enjoy-directx-12
twitter.com/AnonBabble

Buy 1080ti now and 2080ti when it comes out

B-but what about money?

If I could just buy things regardless of the cost, this question would be moot.

Buy the latest and greatest whenever the my stuff != latest and greatest.

Money? Doesn't your mom have a credit card?

The only Pascal cards that made sense were the 6GB 1060 and maybe the 1070. The 1080 is an absolutely retarded buy and the 1080Ti is too expensive without being the full chip. Vega doesn't look too good, but I'm optimistic in it being another Hawaii after it gets fully worked on. If it packs SR-IOV then you should consider it even if it has Polaris performance, having a virtualized enviroment is a great future. Nvidia are assholes and refuse to cooperate with their consumer cards, even with vanilla GPU pass through. Their *works and g-sync also disgust me, they're a terrible company to the consumer.

I don't know Op. My fake benchmarks show Vega beating the 1080ti in gta5 by 64%.

Shoe me your fake bench so we can compare. You must have it since you declared Vega dead.

Wait for Nvidia Volta GV102 GTX 2080 Ti

The volta V100 nvlink gpu is 50% faster than the pascal P100, it's not even a matchup, volta will absolutely crush this vega shitstorm. Pascal is already over a year old.

>muh big dies

>nvlink
kek

nvidia deprecated as soon as a new series is out.

my hd7770 is still getting driver updates.

>muh GIGANTIC DIE
Jensen i'll start drinking your tears the moment ASICs kill your meme cards.

AMD did fuck over APU owners though, the Bulldozer and Piledriver APUs both used VLIW4, wasn't until Steamroller that they made an APU with GCN in mid-2014, and promptly discontinued support for all VLIW GPUs in mid 2015, with a last bone throw in early 2016 to give them the new UI.

developer.nvidia.com/opengl-driver

Nvidia just added OpenGL 4.6 support to Fermi, anyone that says deprecated should be taken out and shot for shooting their big fat mouths off

Nvidia supports their hardware way better than AYYMD HOUSEFIRES garbage that get thrown out as soon as possible

15 teraflops
nvidia.com/en-us/data-center/tesla-v100/

10.6 teraflops
nvidia.com/object/tesla-p100.html

If the 1080ti is 12 teraflops, the volta desktop equivilent has the potential to be at minimum 15 teraflops, which is 25% faster. Minimum. We haven't seen 25% gains since 2004.

>815mm2 die
Do you even know what that number means you fucking ignorant dipshit?

Die size is irrelevant

The only relevant thing is power consumption and AYYMD just keeps headbutting against the power wall like a retarded fuck

It means it is bigger. So what?

> Die size is irrelevant
TSMC will thank you later Jensen.
If they trim meme shit and FP64 they can fit ~5k ALUs into ~550-600mm^2.
The margins will still tank and shareholders would murded JHH six times in a row for that.

You're stupid ignorant fuck as usual

youtube.com/watch?v=iq202LKkeHI

>Perf per watt equals perf

>Transistors are free, power is not

Nvidia was right, bigger die sizes are irrelevant because it gets you to where you want to be

>what are yields
>what are prices
>why are you so fucking stupid

>Transistors are free
HAHAHAHAHA
TSMC will totally thank you later, Jensen.

Yields are just fine, stay mad though :^)

blogs.nvidia.com/blog/2017/07/22/tesla-v100-cvpr-nvail/

>Yields are just fine
Just like thermi yields were just fine?
According to nVidia, that is. :^)

Fermi yields were fine and Nvidia delivered it after fixing the interconnect problem

Stay mad though :^)

Spoken like a true fanboy. Kill yourself already.

Oh yes, it was TOTALLY the interconnect that was the problem.
Totally.
:^)

No it is not.

I expect the volta high-end GPU to probably be around 18 teraflops, considering cooling features like fans allow for higher clock speeds. Not having a bunch of bullshit like HBM2 or FP64 features make the die much smaller. P100 wasn't even the highest performing FP32 card for pascal, the P6000 was, by far.

youtube.com/watch?v=oPRP6C5MlWY

STAY MAD, AYYMDPOORFAGS

You ignorant fucks don't know anything as usual :^)

>trusting leather jacket man of denial sub-20% yields
Hahahahahahaha.
HIlarious, that is.

One point seven.

>Huang received his undergraduate degree in electrical engineering from Oregon State University in 1984, and his master's degree in electrical engineering from Stanford University in 1992.

Jen-Hsun is an engineer, I trust him more than a mother's basement dweller such as you :^)

Go back and suck on mommy's nipples :^)

Yes, we're all totally trust woodscrews-man. :^)

See, a mother's basement dweller like you with no master's degree from Stanford can't argue for shit because you have nothing :^)

Yes, i like woodscrews. :^)

>he believes consumer cards gets the full chip

I will love to see that cut out V104 being sold at 799

>1000 bucks a silicon wafer
>1 working die per yield

This will be fun

>Nvidia supports their hardware way better than AYYMD HOUSEFIRES garbage that get thrown out as soon as possible
Both AMD and Nvidia support their hardware well. Unless it's trash and the latest one that qualifies as that is Fermi for Nvidia. Can't say the same for reds though.

>he believes industry cards gets the full chip
I will love to see any industry card that had higher performance than its desktop equivilent.

The 1080ti had nearly 20% better FP32 than the P100.

Yep that's why Kekler is still around and kicking ass. Oh wait.

Abd yet it has less FP32 perf than P6000. :^)

The P6000 had no desktop equivilent, very first nvidia industry card to ever not have one. If it did, it would be same specs, overclocked, and have better performance.

>The P6000 had no desktop equivilent
Titan Xp?

It's called Nvidia TITAN Xp

It has full GPU like P6000, 3840 CUDA cores and it's clocked much higher in core and memory

>Poor Volta

nvidia.com/en-us/geforce/products/10series/titan-xp/

Crushes AYYMD HOUSEFIRES like the garbage that is it and runs much more efficiently

3584 CUDA vs 3840 CUDA

Titan Xp, not Titan X (Pascal).

Are you fucking retarded?

TITAN Xp has 3840 cores and clocked much higher than P6000

nvidia.com/en-us/geforce/products/10series/titan-xp/

>So now that Vega is officially as disappointing as everyone knew it would be, it would seem that getting a nVidia is the only option.

t. leather jacket man

>$1200
Nah, thanks.

>$1,200
Fuck off Huang.

AMD literally has no competition; what is Nvidia's solution to this, bros?

Lol, I never realized the Titan Xp was different than the Titan X pascal. I would always call the Titan X pascal the Titan Xp, because it was easier than saying pascal after. Lol, so prove the P6000 is better FP32 then.

>launches Titan X for $1200
>Nvidroids eats it like guacamole
>launches 1080ti faster and almost half cheaper
>launches Titan Xp and pretend Titan X never happened

I love NVidia, they do know how to treat people like shit.

We pretend it doesn't exist.

>NVIDIA CEO Gives World’s Top AI Researchers First NVIDIA Tesla V100s

Yeah, 90% of then looks like they will use it to porn VR.

>headless cards
>porn VR
?

You can't use it for VR, there's no output connectors on Tesla cards

How they will use it then?
>He said the V100 would be used for new research on autonomous driving and virtual reality, among other areas.

Pure compute board.

>AMD Schills detected

Get ready for the house fire.

The prices of the other cards did fall down though and with Volta the same thing will happen again.

>damage control
Wew lad.
How do nVidia is going to compete against SSG?

SSG only has 12.7 Tflops and costs 7 grand. What am I missing here?

>What am I missing here?
Dunno.

Gonna leave this here

Who's deciding their card is suddenly obsolete because of 5 fps?

...

>MORE power= better!!!!!!
>House fire happens
>Nvidia wins again

You know what go ahead and buy it. Darwinism is at work here.

>He doesn't even know what Radeon Pro SSG is
Stay clueless, Sup Forumsermin.

>Turns out I'm retarded and wrong
>So hey prove me something else hurrr

I'm not sure if I'm interpreting this correctly. High settings on Nvidia has some lighting effects turned off? Or is it not even possible due to differing architecture/instructions so the game will ultimately look different whether you go green or red?

Also, the right side looks a lot cleaner even if it is lacking some effects, although that might be a design think (who the fuck implements fog in the middle of a battlefield in a top-down RTS, seriously?)

>SMEGMA / W
wait what?

blogs.nvidia.com/blog/2017/07/31/titan-xp-drivers-new-levels-of-performance-for-creative-professionals/

nvidia.com/download/driverResults.aspx/121776/en-us

>Our latest driver — available today — delivers 3x more performance in applications like Maya to help you create and design faster than ever.

Nvidia DESTROYS AYYMD HOUSEFIRES even before it's available, TOP KEK

>like Maya
Why do you even care Sup Forumsermin shitposting trash? Kill yourself.

U MAD, FAGGOT?

>how good is nVidia's support for older cards?
>Is the 9X0/10X0 series gonna be like the 7X0/8X0 series and become unoptimized trash in a year or two?
Well it's been nearly 3 years for the GTX980 and it's still getting performance updates. "Finewine" isn't limited to Ayymd.

computerbase.de/2017-01/geforce-treiber-test/2/

um no honey.
computerbase.de/2017-01/geforce-treiber-test/2/#diagramm-dirt-rally

Ah, I remember this guy. He screencapped frame drops to make it look like nVidia was intentionally gimping their older cards lmao. Course we have tons of evidence now that #TeamRed was just spreading disinfo as usual. Sad!

how long till 580s come back to regular price? Surely the gains from mining have lessened in the last weeks

Never.

>even before it's available
FE Vega still beats the crap out of this shit in workstation and creative tasks, and it was out a month ago. Retarded housesfires shitposter should kill itself as per usual.

>this salty that AMD puts out nothing but trash

I just don't understand how you keep falling for their marketing. How can people be this stupid?

Kill yourself.

No it does not

Keep on lying though :^)

>we haven't seen 25% gains since 2004
Off the top of my head, the 5870 was around 40% faster than the 285

...

Come back when they add DX12 and Vulkan support.

t. nvidia employeer

Disappointment or not, but do you think it'll be physically possible to get one with all the miners out there?

Are you retarded or what?

techreport.com/news/32191/nvidia-finally-lets-fermi-gpu-owners-enjoy-directx-12

Vulkan isn't gonna happen on Fermi

They already have.