Sup Forums has fallen for the 1080 meme..
980ti (reference)
>28nm
>2,048 shaders
>5.2 billion transistors
1080 (founders)
>16nm
>2560 shaders
>7.2 billion transistors
Double-generational leap in production, not even 50% more transistors/shaders. No Async shaders. No HBM.
(Pic related)
2x perf in "VR" vs. an unoptimized 'rendering' not gayming card
>relative vr gaming performance
>my fucking sides
Clearly you don't understand how GPUs work.
What's to understand?
They exaggerated their cards performance with misleading graphs
*AMDs*
*shills*
*poos*
>no Async shaders
Citation needed
Shaders are more efficient in 1080.
meanwhile in AMDrones
Wait for Vega,Wait for Polaris,Wait for Zen,Wait for bulldozer,Wait for Phenom, wait for 8core cpus, wait wait wait wait wait wait wait
"We got it right this time" "just wait for..."
fucking joke and disappointment of entire industry. Rolling with Nvidia as usual, although I will get 1080 Ti or equivalent.
This.
Coming from a 660TI I think I'll just grab a partner 1080. I don't think I can wait for a 1080TI.
Looking at AMD, they can't even get driver updates, let alone cool new features like Ansel (which I'm totally excited about).
Nvidia stock surged 15% in a few days gg with the hype Sup Forums, the buzz is real
Nvidia has more to lose than to gain, bubble incoming
Polaris and Vega bowling balls for a strike
>NVIDIA is successful so that means they're going to fail
>just wait for it guys
>AMD answer for the market is to just wait
>market unaware of nvidia proprietary practices
>lies but it's all relative
>deep learning privacy
>google fb car datacenter hype
>owning both stocks and hedging for the underdog just for the lulz
Ppl forget war is entertainment
I'm just waiting until AMD fucks up this year too.
Should AMD become the new open justified tech Apple of the next decade, this board is going to be fun
AMDfags stay one generation behind waiting for new features.
This is why NVidia will always have the most powerful hardware. They deliver.
As Moore's law end, we're looking at a paradigm shift hence the bubble risk, fabs and minion outlets are churning it out
>most powerful hardware
>still no multi-gpu
>still no 295x2 competitor
>>still no multi-gpu
>>still no 295x2 competitor
Because they're completely worthless.
>multi gpu is good
nah dawg
It's like AMD invented the dual-core processor 4 years ago, and nVidia is still churning out faster and faster Pentium 4s instead
The reckoning is coming.
What you're doing is essentially comparing the clock speeds on CPUs from different microarchitecture families.
GPUs are basically processors that are made to do a special few tasks very very well. It's highly ponderable that they could nearly double a GPU's performance at a task relative to old models that weren't designed to perform that task.
We have to wait for the benchmarks to find out.
True
You can expect 5% more efficient shaders. At the best 10%, not more. This efficiency comes from the architectural changes but not from the die shrink.
Doing some math based on the picture:
980 = 2,6 points of performance
1080 = 4,4 points of performance
So we can expect 1,7 times more performance.
Efficiency:
980 = 165W
1080 = 180W
1,1 times more power draw, that means 1,54 more power efficiency considering that it is 1,7 times faster.
Shaders:
980 = 2560 CUDA cores, 1607MHz
1080 = 2048 CUDA cores, 1126MHz
So the 1080 has 1,25 times more CUDA cores and 1,43 times higher frequency. That means that the 1080 should have 1,79 times higher performance over 980. If to take the architectural efficiency increase in account then it should be over 1,79. If the architecture improved 5%, it should be around 1,88 times faster.
But yeah, look back to the image and you can see that NVidia shows that it is 1,7 times faster. 1,7 is less than the expected 1,88.
The 1080 might be disappointing, I think it has some flaws and bottlenecks. One of those bottlenecks must be the GDDR5X memory. No HBM means bad news.
Meanwhile AMD has smaller dies and GDDR5. Those GPU-s will be cheaper to produce. I think that AMD made a good move with the shader performance and memory bandwidth configuration. They didn't push too far and their cards seem to be more balanced. They also launch Vega in october with HBM2 for the richer gamers.
Lets see if AMD will have anything to counter this
If all you people seriously believe the shit nvidia throws at you then holy shit. I'm not defending AMD here, but believing that a 1080 will be faster than 2x980? Bullshit. It may not even be as fast.
That graph shows the 1080 performing in VR with the simultaneous multiprojection, which will probably be available on Maxwell as well, which will increase the performance in a similar way.
so how much will this thing cost and when can we expect third party versions of it?
Yet in DX12 and Vulkan/Mantle games it will barely be better than 290x 8gb OC
KEK thats a 3 year old car
/thread
10/10 pasta (if not pasta, is now pasta)
>AMDrones grasping at straws
Where's your high end card this generation buddy?
Oh right, you only get mainstream shit
>proprietary practices
Its like you think people care about that in the real world. lol
it's actually other way around
after 1080 paper launch we've seen 5 anti AMD articles/rumors spread everywhere based on pure speculation from some forum posts/nobody knows who.
Meanwhile any negative rumor supported by known facts and numbers about nvidia is shush-shush.
>it's actually other way around
This
And i own a 770 4GB its not exactly a powerhouse been based off a 5 year old chip but it does alrite for itself
Fury X obviously
>AMD answer for the market is to just wait
Yeah like you still wait for Pascuck. Enjoy the 700$ price and general unavailability for the next half year.
I dont need your fucking 700$ GPUs. I want to spend less than 300 since newest games are shit and i barely play them.
What is Vega?
>Fury X obviously
I said THIS generation not last generation
A card that won't be released until near the end of this year.
>600$ price tag for a card that beats the 980ti at almost half the power draw is an unreasonable price
Enjoy your cheap mainstream Polaris cards so you can boast to your friends about buying a new card with last years midrange performance
Not waiting for Vega
Instead buying a card that was promised to have 16GB of HBM but is actually just a grossly overclocked card with 3 point of relative VR gaming performance
what is the function of their curve?
>just wait Amd will be better next year!
I've heard this before
600$ is MSRP. It will be sold for 700$ like founders edition.
>so you can boast to your friends about buying a new card
Thats what nvidikeks usually do though. Nice projection.
Why are you even bringing up the founders edition? Only retards will pay 100$ extra for a reference card.
>Thats what nvidikeks usually do though. Nice projection.
No, what Nvidia owners do, is buy the fastest single gpu you can have at that moment while AMDrones are still stuck with outdated performance from last year
>AMDrones are still stuck with outdated performance from last year
That is going to be the case for almost an entire year, until Vega hits. How is AMD expecting to survive?
>4092 CU's
>Last Generation
Yeah OK buddy 'last generation'
nvidia has a stanglehold on graphics cards and it's best to just ignore all the rebranding and hype.
buy only midrange cards, 4k won't be relevant for years and vr is stillborn.
/thread
>nvidia has a stanglehold on graphics cards
Only on desktops and servers of which Desktops are a shrinking market
Everything is moving towards more what AMD are doing
>Consoles
>Mobiles
>AIO's
>Laptop
Low watt high power in a small package.
Something Nvidia cannot into
>tegra
LERL ok
When you cut down size from 28nm to 16nm, you're looking at not a 12/28 drop, but a 16*16/28*28 relative size. And that equates to resistance.
That's 0.184 if you didn't know.
So going by raw size; you could have hit up to a 5x energy efficiency increase.
Now, that's a fairly out-there assumptuous proposition.
Point is that however much of the power goes into the cores; it could've improved by up to 5x. 28nm to 16nm is a big leap.
An overall halving of energy usage is really really believable when it comes to such big changes in transistor size.
It is last generation because it still gets shit on by the 980ti which is last generations top card
>that OC just through the roof
980Ti is literally just a 780TI with more memory and cores slapped on it with a Die Shrink
Its hardly new tech and runs DX12/Vulkan like CRAP
Hence why they rushed the 1080 to market
Well explained
>Desktops are a shrinking market
Except for PC gaming, the only area where you need a GPU.
>Consoles
That's a declining market, and there isn't much profit in it for those who make the hardware, the money is made on the expensive games. But yes, it's AMDs market
>Mobiles
Qualcomm, samsung, apple, huawei's market.
>AIO's
>Laptop
Intels market.
There's a reason AMD is only used in the lowest end, because their power efficiency and performance is terrible compared to intels offers.
>980Ti is literally just a 780TI with more memory and cores slapped on it with a Die Shrink
No it isn't, Maxwell is a completely different architecture you fool
>that graph
Damn I didn't think the console market was sinking like that. Good on Sony for actually selling an increasing amount of playstations
>Why are you even bringing up the founders edition? Only retards will pay 100$ extra for a reference card.
just like every nvidiots
>No, what Nvidia owners do, is buy the fastest single gpu you can have at that moment
Not everyone needs to play newest CoD on ULTRA settings.
Intel officially announced they will stop producing CPUs in 5 years.
Yeah they just did what people actually want. A as powerful console as possible for a lower price than the xbone, Without the junk gimmicks of kinnect, wii u, etc.
>the only area where you need a GPU
Forgetting 4K HEVC video decoding, and any other kind of video decoding. I know people who are having trouble decoding 4K on newer Intel APUs.
>everyone buys the 1080
>wait 2 more months
>Polaris pushed out
>get more FPS than their gayvidia card
Most of people wont buy 1080 m8. They're just like fanatics need high end cards to meme on Sup Forums.
It's not sinking really. It's Nintendo and MS tanking hardcore.
The newest cod actually runs amazing on two 8GB 290x. I just turned off sub surface scattering or whatever that was that makes faces look better or some shit and I get usually mid 100s for frames at 1440p 144hz much higher than a 980ti even.
A billion trillion years in gimp
true. Gaymers will wait for the poo-card
People who do actual 3d card work are waiting for the 1080-Ti or titan though. First generation pascal is basically just a cheaper 980 Ti.
Titans are slower than 1080/ti
Look at 980ti vs titan x. Titan just has more memory but overclocks less and is overall slower.
AHAHAHAHA. 10/10. Me and all my friends were watching the stream laughing WTFing at that graph. Shit comparing "Relative gaming performance" to power to try to make it look like some new gaming revolution. Even my NVIDIA fanboys had a good laugh at how stupid that shit was. I'm saving your shit because in 10 years we can all look back and remember some newfag oldfag shit.
>Forgetting 4K HEVC video decoding, and any other kind of video decoding.
Any decent CPU can do that.
You just prove my point, gaymer friend. I rarely play new games, mostly emulate PS2 and other consoles. I need a good mid tier card mostly for high quality video playback.
Titans are always slower than Ti.
But have more CUDA, vram, and TMU units. These are vital for non-gaymers.
Who cares about that. FPS is all that matter. Graphics cards are only made for games. Unless you game on one, you don't need high-end graphics for video rendering as you're just some poorfag who probably can't edit for shit. Professionals use Quadro or FirePro, and I'm pretty sure nobody around is a professional.
>FPS is all that matter.
Kill yourself.
7 what?
"I'm too poor for a video card that can push at least 60FPS: The Post."
how much can one expect the price of last gen cards to drop after the release of the 1080?
980 =/= 980ti fucktard.
"2x VR performance" is based on single benchmark of a game that's not even released yet.
It's complete bullshit.
Nvidia reference cards come with a reference cooler.
Like an Nvidia designed one with all the best bells and whistles and vapour chamber.
Maybe some people like that and want to spend $100 extra on it.
Not me but they're not forcing you to get the founders edition are they?
Maybe in the +$500 range but I'm talking about average people.
It says relative performance.
Also it's a launch marketing event. It's not like Nvidia doesn't have a history of fudging numbers either do they?
Oh right.
Just wait for benchmarks like every other sane non-retarded fanboy for either side before poking holes in imaginary charts with imaginary scales made for media photo opportunities.
I've never been impressed with nvidia generational leaps in gaying.
The 780 Ti felt exactly like a 980 Ti. But the Titan X actually helped in autocad.
Anyway it's much cheaper to buy a ~$100 GPU to decode it.
>28nm flagship feels like 28nm flagship
>they been called reference edition since before 1080/1070
>1080/1070 announce
>early access founder edition
>$100 more
they trying to scam people to buy the card with "inferior" cooler design by raise the price and calling it founder edition. those uninformed buyers will fall for it.
They say which game it is on their website.
Doesn't even have a release date yet.
The founders edition is literally the only reference cards they're going to make looks like.
The rest are going to be 3rd party cooling solutions.
I don't think you understand how GPUs are sold.
>early access founder edition
It's just a fancy name for reference card with vapour chamber cooler.
It's still just a useless metric without any actual detail.
Relative performance may as well be the number of donkeys it takes to pull each card.
Nvidia reference coolers are pretty much the best cooling solutions available this isn't even debatable.
i fixed the graph
>see titan x
That chart is so purposefully misleading...
>no real numbers, just relative performance, compared to what?
>maybe compared against the gtx 960 except for the fact that the gtx 960 isnt at y = 1, it is not the baseline
>Gtx 1080 label appears above the dot, instead of to the right as the rest do, this makes it seem the gtx 980 uses less power than the gtx 1080 when it does in fact use 20 W less.
>not to talk about tdp, which nvidia has been caught lying time and time again with fake values
>also 2x performance in that chart, when in real benchmarks it is only 25% faster than the previous cards
>this makes it seem the gtx 980 uses less power than the gtx 1080 when it does in fact use 20 W less.
Sorry got this backwards, i meant that the gtx 980 uses 20W less than the 1080, but the label of the gtx 980 makes it look like it consumes even more
...
>titan x
If they ship it with the same titan x cooler it would have so much headroom it would be laughable considering how much lower the tdp is.
Also the founders editions are out early because they're fucking reference cards. Go call your 3rd party board partners if they're 2 slow to release new cards especially when they're literally going to be half the TDP and only use one 8 pin connector.
I would literally just saw off a 980ti cooler and stick it on top of a 1080 and charge full price.
>he doesn't believe in the kami-sama called vapour chamber
I'm coming from a 570, so, a 1080 is going to blow my mind.
Why the fuck would I care about benchmarks compared to a 980?
Why would anyone care? If you have a 980, why are you even looking to upgrade this generation? Get a grip you crazies.
>sub surface scattering or whatever that was
You don't know what subsurface scattering is and you're talking about graphics cards like you know anything.
Please stop.
>I'm coming from a 570, so, a 1080 is going to blow my mind.
Who invited this poorfag to this discussion?
I'm not even referencing him directly because I don't want to catch his poor disease.
They have no hardware async. Its all software with the regular cores being the carrier to the "performance".
I only know what I know. I don't play many new games. I'm more of a Quake 3 freezerail, CSS, CSGO player. Twitch skill based stuff which don't really use gpu power at all. I'm not a fan of single player games unless its late 90s turn based stuff. I like to play GTA5 and COD:BO3 every week or so though. The new COD was the first time I ever came into sub surface scattering options but its the first time I've played COD since 2 minus playing with non-pc friends on XBOX 360 together.
Use a decimal, faggot. Those commas are confusing.
Nothing until Vega.
They have already stated that.