10-12-17

>>62376480
How many GB of that will not be connected via a telegraph?

2050Ti when?

Should I sell my 1080Ti now?

>>62378102
No. The 2080 won't release this year and it won't be much stronger than a 1080 Ti. Ti > 80 is never worth it. Wait for the next Ti.

GTX 2050 Ti, GTX 1060/GTX 980 level of performance for 75W or less TDP, no PCIe power connector required

MUH

the fuck is anyone going to do with 12GB?

>>62379599
Don't know, maybe someone will need it to run all those 4k textures while running theit games in 4k. Never heard of 8gb's not being enough for that, but hey, I guess we should not underestimate how lazy devs can be with their games.

>$700 saved up for Nvidia Volta GV102 GTX 2080 Ti

MY BODY IS READY

>>62376480
So, what Nvidia will call the cards when they get to the 5xxx?

File: 1500151249459.png (2MB, 1200x1505px) Image search: [Google]

2MB, 1200x1505px

>tfw you just bought a 1080ti

Damn, I wish Nvidia weren't the only option. Fuck RTG.

citation needed get fucked

Wtf im so confused why is Nvidea going up by 10's now shouldn't the next card be the 1180?

They already said they're not coming until next year because Vega turned out to be a joke

>>62380298
>buying cards just when new release season is around
Every time

File: JHH2017_1_06Jun2017.jpg (182KB, 1800x1200px) Image search: [Google]

182KB, 1800x1200px

https://www.gputechconf.eu/Home.aspx

>Join us for the opening keynote on October 10, 2017.

Fanless <75W 2050Ti when?

>>62380488
2018

>>62380475
It's probably going to be about deep learning memes.

guess there is no need to upgrade my 980 Ti since I only play on a shitty half broken 1080p tv/monitor. I'll only upgrade when 4K will be the usual res for games

Jensen said Volta won't be released until 2018 and Pascal is already "good enough", since they know that they have no competition with AMD's Radeon GPU.

>12 & 16GB cards
My dong can only expand so much

It's confirmed volta will use gddr6, which isn't even ready for production yet

750 ti install gentoo w3m

>>62380488
Why go for low end? Thats just dumb. 2050ti will be 1060 3GB level.

Meanwhile 2060 will be 1070 level.

>>62380288
GeForce FX5000 obviously

>>62376480
you mean
GTX 2080
GTX 2070 12+4GB
GTX 2070 6+6GB

>>62380499
Fucking this, even jacketman himself admitted Volta isn't coming out any time soon.

>>62380854
Not everyone needs a 1060 or a 1070 level card

>>62379599

Multiple video games. Some people run like 2 or 3 mmorpg in the background while playing another game or working.

>>62378120
>>62378102
You should always wait for the Ti but they're still probably not going to give us big Volta for awhile. Was watching a retrospective on Nvidia's release cycle and in their first like 5 generations there would be ~75% boost in performance from top tier cards each year and pricing was pretty stable. The GTX line started off performing terribly but since the 700 series they've intentionally crippled the XX80 cards so they can overprice the Tis and Titans. They're infamous for rebranding cards and reusing old architectures when they know they're on top. Maxwell and Tesla both had basically no competition from AMD. The return of cryptocurrency didn't help AMDs market share, but they seem perfectly happy making mining ASICs instead.

>>62381031
>They're infamous for rebranding cards and reusing old architectures
And AMD doesn't do that?

>>62380431
Nope, those just marketing tactic to make people buy pascal

http://www.gputechconf.com/

>GTC 2018
>SILICON VALLEY

>MARCH 26-29, 2018

get a hold of your horses, 1080Ti got released only 6 months ago

>>62376480
>Huang said during an earnings call reports PCGamer, “Volta for gaming, we haven’t announced anything. And all I can say is that our pipeline is filled with some exciting new toys for the gamers, and we have some really exciting new technology to offer them in the pipeline. But for the holiday season for the foreseeable future, I think Pascal is just unbeatable.”

http://www.legitreviews.com/no-volta-gamers-holiday-season_197168


Stop being retarded

>>62381244
Yes, Pajeet, Nvidia doesn't have to put anything new out because AMD's brand new NEXT GEN cards are still competing with Nvidia's 18 month old offerings

>tfw GT 2030 for 70$ and 30 TDP with GTX 1050 ti performance

Feels good.

>>62376480
>2010, December, 17th
What did he mean by this?

>>62376480
shouldn't it be 1180 and 1170? i feel dumb

>>62381384
oh yeah because surely they can put a consumer volta out given how fucked up their v tdp is LOL

>>62381603
Best to go with new lineup and start fresh with 2000 series instead of 1XXX series

>>62378120
>>62378137
>>62380488
>>62380854
>>62381390
>all these poor souls thinking that the pascal -> volta jump is going to be as good as the maxwell -> pascal one was

>>62381617
Are you literally retarded?

>>62376480
>>62380288
Wait, it's been confirmed that this serie will be calle GTX 20xx?

File: 1502709377455.png (470KB, 456x600px) Image search: [Google]

470KB, 456x600px

>>62379599
Shut your whore mouth you filthy Indian animal. I love how you fucking subhuman brownniggers were always advocating for muh 8 gigs of VRAM when it was only AMD's selling point but now when they're behind even in this you're backtracking saying it doesn't matter. Kill yourself u Pajeet.

>>62381349

OP BTFO!

/thread

>>62381900
you understand that HBCC literally makes VRAM amount a lot less important right?

>>62380298
>mfw just bought

>>62380982
> Multiple video games. Some people run like 2 or 3 mmorpg in the background
Background minimized vidiyas do not consume VRAM.

>>62381349
so you expected them to say volta few month away and tank sale ? they just waiting

File: 1360523238504.jpg (100KB, 540x572px) Image search: [Google]

100KB, 540x572px

>expected to have over 5,000 cuda cores

>>62380298
> implying there's going to be 2080Ti in December
Relax. I bet it will appear a year from now at best.

File: Kemmler_exécuté_par_lélectricité-1.jpg (473KB, 1096x910px) Image search: [Google]

473KB, 1096x910px

>>62381851
Volta will destroy execute Vega

>>62381970
Good, cant wait for a working gf simulation.

File: ohh.jpg (61KB, 960x720px) Image search: [Google]

61KB, 960x720px

>tfw volta sounds like "to rape" in a couple of languages
>tfw have to wait another season
I am comfy on my 950 anyway

File: images.jpg (6KB, 166x200px) Image search: [Google]

6KB, 166x200px

Not long until we see pascal tear
Dont forgot to buy this new Volta

>>62381851
>20%
>good
lol

>>62381986
>Volta will destroy execute Vega
Pascal already destroyed Vega six months ago

>>62382017
20% isn't bad but the most important thing is 11 GB of VRAM in 1080Ti.

>>62382028
if you measure your fps in watts, sure pascal won
otherwise still a black box, v56 is good right now though

>>62382030
yay, 0.5% of consumers got 11Gb of VRAM now! fire those code monkey we don't need no optimization!

>>62382030
11GB is good if you need it, for me it's wasted space.

>>62382396
Ti price won't change anyway its imaginary have nothing to do with actual production costs
why not have 11Gb?

>MFW JUST BOUGHT A 1080 LAST MONTH
reeeeeeeeeeeeee

>>62380470
It's the perfect time for buying 1080Ti though, prices are stabilized and big Volta is more than a year away.

>>62380982
I could do that no probs years ago with a mid-range GPU.

>>62382415
Because I don't need it?
Unused RAM is wasted RAM. I'm using 3GB vRAM GPU at this point, and It's plenty enough for me.

>>62382445
Wrong, GTX 1080 Ti was released in March 2017

It will be out in March 2018 or at worst June 2018 for Computex

>>62382442
Never buy nvidia 3 months past initial generation AIB release.

>>62382484
reeee

didnt they say it costs like 1,000 to produce a card

I'm putting $700 aside now but I just know that these cards are gonna be like $1,200 and miners are going to make it $1800 easily

I miss the simple times man.

I still got my 770 GTX being optimistic of the new cards to come

https://devblogs.nvidia.com/parallelforall/inside-volta/

>The new Volta SM is 50% more energy efficient than the previous generation Pascal design, enabling major boosts in FP32 and FP64 performance in the same power envelope.

THANK YOU BASED NVIDIA

>>62382645
for Deep Learning

>>62382381
>v56 is good right now though
The point is that it was supposed to be compete with Volta and be "next gen", when in reality, it's competing with 18 month old cards and the 64 still gets BTFO by a 1080TI.

Too little, too late

I really wanted AMD to win, but Nvidia destroyed AMD with Pascal, the performance/wattage is just too good, Vega can't compete.

If Volta is a nice jump in performance, it's all over. Ryzen won, Vega lost.

>>62382613
i have a 770 too

idk when to upgrade

>>62376480
source?

>>62382947
Actually Vega 56 is quite fine on power efficiency and performance wise compared to 1070.

The problem is Vega 64 and its ability to compete with 1080 and 1080ti.

Its pure garbage. Overall the whole Vega is poisoned due to this shitshow.

>>62379599
Deep learning

>it will not be the 20xx you fucking faggot it will be the 11xx people whothink this piss me the fuck off wotse than those who call characters/avatars toons if its the 20xx im not fucking buying it

>>62383069
he literally has no source, shut up

>>62379599
CUDA

>>62382998
Vega 56 is the only good one, too bad it was 1 year late and still no availability.

>>62383095
no
THIS IS A REASON TO FREAK THE FUCK OUT IF THEY CALL IT THE 20XX MY NAMING CONVENTION IS FUCKED AND I WILL HAVE TO CALK MY NEXT BUILD THE TARDISt400 OR SOME SUCH SHIT FUCK THAT I WILL NOT CHANGE BECAUSE A COMPANY CHOOSES TO BE RETARDED ITS NOT MY FUCKING FAULT

>>62383129
calm down with the autism

>>62381070
Just saying, don't be surprised if a 2070 is still Pascal and the 2080 is a cut down card.

>>62383154
so im just suplsoed to LET this company bully me into change? im just suppsowd to lay back and let them ASSRAPE me? alll i want is my 1180ti is that SO MUCH to fucking ask?

>>62382381
AMD high-end = last year's mid-range.

>>62382030
>>62381851
No it really is a shit improvement in performance when you look at the early 2000s doubling performance each year.

>>62383116
>>62382998
Vega 56 is only remotely efficient because of HBM2. At stock it still thermal and power throttles. AIBs can probably make something half decent with it though. Still with the price hikes and limited supplies, it's not beating 1080s.

>>62383185
Might as well. They can go eat a smothered bean and chicharron burrito with some tamales at Santiago's and shit out cards better than Vega

>>62377997
this

I'm only interested in $100 budget cards and I want the 2050 to be rivaling 1060's

>>62383185
god dammit they will not be 20xx fuck off i shouldnt have to change

>>62376480
ok op i'm still waiting on sauce but i just picked up two shares of NVDA

>>62383423
>trading stocks

>>62382445
No, actually it's ass because GTX 2080 will be faster and cheaper, probably more power efficient too and it's likely to pop up around early 2018.

>>62383521
*GTX 1180

>>62383548
We really have no idea one way or another, both naming schemes are equally made up and unofficial.

>>62383607
it better be 11xx or I'm not spending a dime on it

>>62383622
That's OK, if it's 20XX I'm going to buy twice as many.

>>62376480
So is it a bad moment to buy a laptop with a 1060?

>>62380298
These cards aren't releasing until Q2 18 at the earliest. And in nvidia world, a new *80 is barely better than last gen's Ti.

>>62379599
only reason to upgrade to newer cards right now is for 4k, 1440p is already beaten by 1070, so now its 4k turn.

>>62380406
because 1180 looks like shit, 2080 looks much better in marketing

>20 series
why did they skip 11-19?

>they'll release 2050 Ti first
>1060 performance
Fugggg

Wanted to get 1030 for my pc, guess I will be getting 2030 once passive cooled one is released instead

>>62383835
the same reason they skipped the 800 series in desktop, autism

>>62383835
Probably the same reason they went from 9xxx to 1xx?

Will they be going with HBM2, or will they carry on with GDDR5X on the GEFORCE cards? The cost of HBM2 is apparently around 150ish a card and needs a redesigned memory controllers, as well as GDDR5X costing only around a third of that amount. The memory cost increase will likely mean they will go with GDDR5X again.

Wait for NAVI

>>62383972
GDDR5X or GDDR6, HBM only on V100

>>62383972
there is no reason for hbm2 on consumer cards, its literally marketing meme.
Its main advantage is smaller size and low W is literally unimportant for consumers.

>>62384014
I know that the primary use of HBM is the lower power consumption as well as the smaller size on the PCB. Is that the only thing that HBM is useful for? Apart from the ease of stacking memory, especially with HBM2.

>>62384093
well it has better latency and obviously it has potential for much higher bandwidth, but those arent the things holding current gen gpus.

Basically memory tech is more advanced than gpu die. GDDR3 was holding back GPUS, and it was obvious just when GDDR5 was released, the performance uplift was immense, thats not the case with hbm2.

http://www.anandtech.com/show/11398/sk-hynix-advances-graphics-dram-gddr6-added-to-catalogue-gddr5-gets-faster

GDDR6 for GeForce Volta cards

>>62381925
This is what AMDfags believe.

>>62384235
Thanks, seems that the GDDR6 is close to par with HBM2 so it seem irrelevant for Volta to have HBM2, apart from the latency and power draw benefits

>>62384366
http://www.anandtech.com/show/11297/sk-hynix-announces-plans-to-ship-gddr6-memory-for-graphics-cards-in-early-2018

>What is noteworthy is that SK Hynix does disclose some details about the first graphics cards to use its GDDR6 memory. As it appears, that adapter will have a 384-bit memory bus and will thus support memory bandwidth upwards of 768 GB/s.

768GB/s, more memory bandwidth than any HBM2 meme from AYYMD

>>62384398 Navi is supposed to have 'Next Generation Memory' and is expected in 2018, that pretty much confirm it will use GDDR6. Why did AMD use HBM2? is it a marketing ploy, so they can bill it as a workstation/miner card, or to reduce the high TDP of Vega. The cost alone was like around $100 extra a card

>>62381962
Are you a faggot? There would be all kinds of leaks about Volta if it were any sooner. Get fucked.

Will Volta finally fix the async compute problem that they have been having for ages.

>>62376513
all of it

>soon you'll need 2000 watt power supplys just for the GPU

>>62384572
It will have compute capabilities, whether it's Nvidias definition of "asynchronous compute" or AMDs is anyone's guess.

>>62384529
Being this dumb, Only full volta not coming this year gv104 will

>>62384602
But enough about Vega 20...

>>62384622
To be fair Pascal had async compute, it was shit, but it did have it, like DX12 'support'

>>62381349
>>62384529
Nvidia never teases half finished products or leaks unlike AMD.
People were expecting Pascal to be delayed even further then Nvidia came out not just dropping the GTX 1080 bomb but also the 1070 too.

>>62380475
Yeah first is always about AI, GTX events come later.

File: [DB]The Disastrous Life of Saiki K _-_16_(Dual Audio_10bit_HDTV720p_x265).mkv_snapshot_08.19.png (1MB, 1280x720px) Image search: [Google]

1MB, 1280x720px

>>62381349
>Not saying buy Pascal over the holiday season
>Volta, launch sometime in Jan to sell as much as you can with the current architecture.
Come on, who would say, "No don't buy a new GPU now, they will be obsolete in a couple of months

>>62380475
MEME LEARNING
MEME
LEARNING
PLS BUY OUR MEME LEARNING GPUs
ASICs WONT KILL THEM I SWEAR?
CREST LAKE? DLU? TPU2?
OY VEY ANUDDA NANKING.

>>62383769
Eh, 1080 was like 30% faster than 980 Ti. It's not worth paying to upgrade if you already have the previous Ti, but if you're upgrading from something older the new xx80 is going to be a much better choice than the old Ti.

>>62376480
Nvidia VAI

>>62385300
1080 was a double node shrink.
Volta isn't one.

Reminder that the 2080 will be a mid-range GPU.

Wait until next year to buy the Ti.

>>62385497
Of course it will be, when was the last time that the x80 card wasn't a x04 chip?

>>62385557
GF110.
You gotta milk when you can milk.

>>62385605
>9 November 2010
Jesus, there really is no competition, not since Fiji & the Titan

Still sitting comfy with my 980ti. Still pushing 1440p 60fps with no issues. I might upgrade to whatever the new TI is just because.

>>62385724
It's not about lack of competition and more about TSMC always having shit yields on a new nodes.
You see, when nVidia tried to make a big die on a new node it resulted in Thermi.

Got a 1440p 144hz monitor, 970. Was going to check out Vega, then miners got them and the price fuckery with the $100 extra after a month happened.
I will wait for Volta, and pray it's not shit.

File: 1467515288227.jpg (560KB, 800x1046px) Image search: [Google]

560KB, 800x1046px

>>62376480
>held off on buying a video card but temptation is rising
>gotta keep waiting For The Next Big Thing

when will this ride end>

I got a 1080ti a couple of weeks back so I gave my gf my old 980ti and all she uses it for is youtube and hearthstone

>>62385891
It never does. Whne you finally buy a new card, you will feel regret evenually.

>>62385474
Maxwell was 980 was also a better card on release than 780Ti. Only marginally faster, but much more power efficient and coming with extra VRAM too. I don't expect Volta to be such a huge leap as Pascal was, after all Pascal was the largest generational leap we've had since before Fermi, but xx80 Volta will probably be cheaper and 10-20% faster.

>>62380630
>>62380316
>>62378120
>>62380816
>>62381267
>>62381339
>>62381349
t. butthurt buyer's remorse 1080ti subhumans

File: 1502430519727.png (503KB, 700x560px) Image search: [Google]

503KB, 700x560px

Can't wait for the cards to sell out quick and be overpriced within the first week

>>62386097
>Pascal was the largest generational leap we've had since before Fermi
Evergreen was right about 80% coming from RV770.
>Maxwell was 980 was also a better card on release than 780Ti.
It introduced TBRD.

>>62386097
The die shrink from 16nm to 12nm should ensure that.

>>62380887
It already comes out, but not for consumers.
>>62381031
I expect the next X80 card to have a maximum of 30 % performance jump and similar overpricing like the 1080.

The Ti variant will NEVER release first. That would be retarded for NVIDIA. They may release a Titan first, like usual, for 1300 € (European price here) and the richfags will buy them, just to have another Titan and Ti follow a year later.

You could be waiting forever if you always wanted the newest generation. I got my 1080 Ti close to release and think it will be the top dog for a while, even when the new X80 gets released, it will probably have similar performance and more memory bandwith.
>>62381851
The jumps were the same all the time. 780 Ti > 980 was the same performance in the beginning. The 980 got faster in newer games because they ceased driver support for Kepler.
>>62381970
My dick is ready... 4800 CUDA cores clocked at 2.6 GHz. That should even max Mass Effect Andromeda.
>>62382613
1000 $ means the chip of that new Tesla monster card. They won't use this full fat chip for GV102. And yes these cards are going to cost 1300+ €, but there is still hope. The 1080 initially costed as much as the 1080 Ti now, which has a much better price point.

>>62386165
>first week
You mean first month, don't you?

>>62386165
Buddy of mine makes some serious money doing that. He bought a mess of 1060s and 1070s when they first launched knowing they'd be the most popular consumer card. Waited for the inevitable selling out, then started listing his stash on eBay for a $100 markup. More often than not he was able to sell off his entire stock of 10+ cards.

>>62382613
Remember the fab process from TSMC will improve over time, as they can refine it. Plus the xx80 Volta will be a x04 chip anyway, not Full-fat Volta

>>62386337
>1000 burgers
>serious money
at least he's trying to stop being poor

>>62382469
>I- I dont even need more vram! its ok!!!

>>62386240
>3% density improvement
t. (((TSMC)))

>>62386430
It's a rebranded 16FF+ with marginal improvements in power and HUEG reticle limit.
It's not a new node.

>>62386185
>Evergreen
I'm talking about NVIDIA.
>It introduced TBRD.
Sure, you can say that now in hindsight, but nobody knew they were going to do that before Maxwell was released, nor do we actually know anything about Volta gaming performance.

>>62386641
>Sure, you can say that now in hindsight, but nobody knew they were going to do that before Maxwell was released, nor do we actually know anything about Volta gaming performance.
And nobodt knew Vega kills traditional definitions of non-compute shaders and yet here we are.
Anyway Volta is all about cumpute.
It even fixes deadlocks.

>>62386240
""""""""12nm"""""""" is a more refined 16nm. at least intel calls theirs 14nm+, ++ and +++.

File: 1495578764091.png (242KB, 510x346px) Image search: [Google]

242KB, 510x346px

>>62386794
So it's a blatant lie? 12nm isn't actually 12nm. How the hell are they allowed to say that it is?

>>62386973
Different foundries measure feature sizes differently.
There's not standard for that.

Wasn't there something about going under 10nm and that it fucks with the gates at that size?

>>62387163
That was before FinFET.
Of course you can't make planar 10nm.

File: capture3uu72.jpg (32KB, 629x309px) Image search: [Google]

32KB, 629x309px

>>62387223
Ah, thanks. I wondered how they solved it.

I'm guessing the 2060 will probably have at least 8 GB VRAM

If this is true, Vega is done

>>62391255
What Vega?
Vega10, Vega11 or Vega20?
Or Vega12?
Anyway Vega is not working properly yet.

>>62391209
great job nostradamus

2080 Ti will have 18 GB of VRAM... Predicting it now

File: 1292704891564.png (817KB, 720x540px) Image search: [Google]

817KB, 720x540px

I wanted to buy a 1060 (6gb) but with Etherium they've become way too expensive so I haven't saved up for it

I hope Nvidia makes them bad for mining on purpose

>>62391573
G6 offers no density impromenets, only higher speeds.
So, 12gigs it is.

>>62391209
Wrong, 2060 is 192bit memory bus just like 1060, so it can only have 6GB or 12GB of GDDR6

GTX 2080 is 256bit memory bus so it can have either 8 or 16GB of GDDR6

>>62383381
unless it's on 7nm, it won't be.

>>62380298
>Bought one back in June. Looking to build my first rig after dealing with a crappy laptop with a gt650m for 4 years.

Not really bummed if nvidia puts out anything ""better""

>>62391610
Pray it uses GDDR5x. altough faster than regular GDDR5 its actually worse for mining which is why the 1080 is barely used for it.

>>62391573
22GB

>>62391644
it's on 12nm (16nm)

>>62391713
1050ti is on 14nm, so it's nothing.

>>62391713
It's 12nm (16nm (20nm)).

Anymore info. OP?

File: 1391014141608.jpg (163KB, 857x536px) Image search: [Google]

163KB, 857x536px

>>62391352
>Anyway Vega is not working properly yet.
A month later. Sad.

>mfw they justify it by calling it Early Access

>>62391953
shareholder pressure is the fucking plague

>>62380470
always go for the previous gen flagship

>>62391573
Depending on the memory bus, it might have 16 GB HBM2.
>>62391611
For the GV104. We don't know which memory type the big chip will use.

>>62392207
You never go for old flagships when the new generation is avaiable.

>>62392226
GDDR6, do you not know anything?

http://www.anandtech.com/show/11297/sk-hynix-announces-plans-to-ship-gddr6-memory-for-graphics-cards-in-early-2018

>What is noteworthy is that SK Hynix does disclose some details about the first graphics cards to use its GDDR6 memory. As it appears, that adapter will have a 384-bit memory bus and will thus support memory bandwidth upwards of 768 GB/s.

>>62392308
Titan release first? It would be wondering for me. They always release the X80 first, then follow by the rest of the lineup. I think they want to make people tempted to buy the more expensive cards instead of waiting for the ones on their price target.

The Ti will be released in the end, so richfags buy the Titans and enthusiasts buy X80s first, then switch to the Ti, so NVIDIA sells twice in a generation.

>I bought two Pascal cards new
>The 1060 I had first doesn't count it was used

>>62392207
>buying previon gen flagship before price drop/people selling their old card to flood market

>>62392510
No, the big chip aka GV102 will have a 384bit memory bus with GDDR6 16GB/s for 768GB/s of memory bandwidth

GV104 will have 256bit memory bus with GDDR6 but we don't know the speed it will use for now, it could have about 512GB of memory bandwidth with 16GB/s GDDR6 or 448GB/s with 14GB/s GDDR6

>>62392510
>I bought two Pascal cards new
>The 1060 I had first doesn't count it was used
Idiots like you is why Nvidia is making so much money

File: 1487438433841.png (314KB, 486x455px) Image search: [Google]

314KB, 486x455px

>everybody itt ignoring that miners are ruining the market

Don't know why I would upgrade from my 770 anyways but fuck there are some delusional people in this thread

>>62376480
Dont expect huge performance bumps as there is no competition atm. +20% compared to Pascal max. 12nm TSMC is a meme as it is essentially 16nm++.
Just have a look at Intel 14+/14++ brought literally no improvements.

>>62392690

Alright, I'm the person who made the original post on this thread. I haven't posted much since, however, I'd like to go ahead and say that you will not have to worry about mining "ruining" the market soon enough. That's all I'll say about it.

>>62392741
Wrong

>The new Volta SM is 50% more energy efficient than the previous generation Pascal design, enabling major boosts in FP32 and FP64 performance in the same power envelope.

Stop talking about things you don't understand

>>62392781
>Stop talking about things you don't understand
oh the irony, I hope it was intentional

File: Screenshot_20170813013434.jpg (355KB, 887x1325px) Image search: [Google]

355KB, 887x1325px

>>62392777
Man I can't believe a terrorist is going to blow up all of Bitcoin

>>62392781
That's slapping moar ALUs.
If you consider that than Fiji is like 80% more efficient compared to Tahiti.

>>62385880
Yeah, but is there really a reason? 1080 can handle 1440, 1080ti can handle 1440 144.

>>62392781
Only in HPC market, moron. Regular custumers will still recieve a shit update and get milked for that.

>>62392895
Wrong again

FP32 is for consumer gaming

Please stop posting, you're just trolling at this point

>>62392956
if FP32 was for consumer gaming then vega would shit all over the 1080ti, but that's clearly not the case

>>62380270
that won't nearly be enough *cough*

>>62392956
It's simply throwing MOAR ALUs.
And that doesn't scale linearly.

>>62392979
Vega is a raw performance beast. But it needs better drivers.

Also games are often optimized for NVIDIA.

>>62392956
Oh yes, you are right. NVIDIA is a honorable company with only the best in mind for their customers and will provide me a 65% performance uplift with Volta over Pascal for 100$ less.

Now BTFO kid, thats not how the world works.

GTX 2080 Ti will have 5120 CUDA cores based on Volta architecture

It's a huge jump over 3584 CUDA cores on GTX 1080 Ti and will finally power through 4K games

AMD getting utterly crushed and destroyed since they can't scale past 4096 cores

>>62392587
Uh, they don't make money on used cards

>>62393069
If those are ever to come?
RTG is in radio silence mode atm. Not even a roadmap or a "We're working on it". Their PR dept. is obv. run by autistic monkeys.

>>62376480
Are these new graphics cards AMRAAM compatible? If not i ain't buying.

>>62392777
Please elaborate

>>62393124
Source or won't happen.

>>62393124
My guess would be same shader count or 4096 max. or the die would get too big (yields!). Probably some more ROPs too in order to push them pixels and GDDR6 but thats pretty much it.

>>62381851
>he doesn't know about processes and die size

>>62393160
https://devblogs.nvidia.com/parallelforall/inside-volta/

>FP32 Cores / GPU 5120

GTX 1080 Ti had the same 3584 FP32 CUDA cores as GP100

TITAN V will have 5376 FP32 cores, fully enabled just like the 3840 FP32 CUDA cores on TITAN Xp

>With 84 SMs, a full GV100 GPU has a total of 5376 FP32 cores

>>62392690
Only because we all know it's fucked, so now we just get to argue about the cards by their merit instead of price:perf

>>62393205
12nm == 16nm++

>>62393201
ROP setup for GM200/GP102 is already retardedly overkill.
>>62393124
AMD can throw however much ALUs they do want.

>>62376480
>GTX 2080
>most likely will be GV104
>$899
>GV102 to cost north of $2000
>you'll have to sell your house and kidney to have the privilege of knowing how much the GV110 card will cost
Why do they keep falling for this?

>>62393160
>source
it's a new nvidia card so it will actually revolutionize? That's really the only source you need.

>>62381851
>Maxwell to Pascal
>big jump
>same architecture, smaller process
They're basically selling you higher clocked, cheaper to manufacture chips for a higher price. I don't think they could've kiked you any more.

>>62393363
Spending your parent's money doesn't hurt quite as much as having to actually work for months in order to be able to afford one of these.

>>62393430
>work for months in order to be able to afford one of these.
You should probably stick to 1080p 60fps

>>62393509
I'm actually going to get a small loan and sell my old gpu to afford the newest line. Of course this option isn't available to the majority.

>>62393542
You're funny

>>62393573
It's no joke sir, I know a little basement dweller like you living on your mommies money wouldn't understand the first thing about economics. But please try to act grown up for a moment.

>>62393430
Or NEETing until you can finally afford your desired flagship

My 1080 Ti will be still worth 300-400 € then so I only have to save half of the amount

Man, those GPU are getting some crazy amount of VRAM, holy shit...and here I am sitting on my Fury X with 4GB of VRAM, didn't encounter any memory issues so far.

>>62386268
>The jumps were the same all the time. 780 Ti > 980 was the same performance in the beginning. The 980 got faster in newer games because they ceased driver support for Kepler.
Cunt way of doing business.

>>62394086
Non-HBM GPUs won't move past 12/24Gigs for quite some time.
I mean PCBs for G5X are already log-thick.

>>62376480
Nvidia has no competition, they're not in any hurry to release Volta

>>62394335
Oh, they do.
Radeon Pro looks bretty angry and eager for blood.

>>62378120
>1080 comes out
>waitfags say wait for Ti
>1080Ti doesn't come out for 6 whole months
>immediately one-upped by Titan Xp

>>62376480
Navi won't even beat Pascal

>>62393134
Their PR department is too busy shilling for drastically overpriced Polaris cards on Twitter.

>>62394439
It's not like they can hope to control prices with mining craze.

>>62376480
you people are fucking retarded

this whole thread is fucking retarded

volta isn't coming out for at least a year and a half

>>62394613

You sound mad.

>>62394677
you're damn right i'm mad

i'd buy that shit this instant, i need a new computer

what's the purpose of this whole fucking thread when we don't know jack shit about volta?

>>62394086
Because you probably play at low resolutions. 1080 Tis are not meant for 1080p. I average around 8 GB VRAM usage in AAA titles maxed at 4K. So 11-12 GB is a nice amount while still having overhead.
>>62394393
It usually takes one year for the Ti versions to show up.

File: 1496527749800.jpg (272KB, 449x449px) Image search: [Google]

272KB, 449x449px

>>62380298
>mfw I still only have a single 1080p monitor with my 1080Ti.

Total overkill, but I'm not sweating it. I really would like a nice 28" monitor so I could stick this 1080 on its side as a text monitor, but I'm a fucking poorfag, so idk when that'll happen. Until then, pretty okay with my setup.

so am I replacing my 970 with this?

>>62392573
>the big chip aka GV102 will have a 384bit memory bus with GDDR6
Nvidia will not unevenly pair memory controllers to VRAM on a high-end card ever again after the 970 fiasco. If it's got a 384-bit memory bus it's going to have 12 or 24GB of VRAM.

>>62394821
GV102 will have 24GB of GDDR6, GV104 will have 16GB of GDDR6 and GV106 will probably have 12GB of GDDR6

>>62383797
so they are going to 2xxx -> 3xxx -> 4xxx -> ... -> 9xxx (again) and then reset again to 100??? WTF!?!@?!

>>62376480
That would be utterly retarded, stop posting

>>62394757
>poorfag
>1080Ti
I think you mean you just aren't smart with your money

>>62395028
I wouldn't argue that.

>>62395050
So why would you buy a 1080ti for a 1080p display? a 1050ti is plenty good enough given the broken GPU pricing atm, even the RX 560 is pretty solid

>>62394613
GDDR6 availability from SK Hynix starts in Q4 of this year and TSMC is reportedly already ramping up 12nm production specifically for a Volta "GeForce" chip that will arrive very early in 2018, which gives some credence to the rumor of a planned March 2018 launch.

So... it's possibly only ~6 months away now.

>>62394873
GDDR6 has no density improvements.

File: blog_image_gddr_4.png (34KB, 1047x401px) Image search: [Google]

34KB, 1047x401px

>>62395163
You're wrong, GDDR6 allows for 16Gb and 32Gb densities

https://www.micron.com/about/blogs/2017/june/what-drives-our-commitment-to-16gbps-graphics-memory

Please don't post if you don't know anything

>>62376480
any eta on 2050, 2050ti, or 2060?

>>62395067
Why wouldn't you?
>You won't be outspec'd for a while
>Can always upgrade monitor and know your graphics card can still perform
>NVidia so you know the gimpworks™ won't affect you much
You're just wasting money but for some benefits. Also 1080TI will be obsolete, hell it already is. 1060 is for your 1080p60 gaming.

>>62395373
That depends on when GV104 launches, GV106 and GV107 won't launch before GV104

>>62395067
I bought it because it would max anything I wanted to at that resoultion, and probably still will at 1440p. I also know that having anything running at 120fps at Gsync on a 4k monitor at maxed the fuck out settings is still a bit off from now, so I'm not worried about it. None of the games I play atm really need that kind of juice.


Honestly, this was more of a e-peen thing.

>>62394757
Is it at least 144 Hz?
>>62394821
I hope they continue to do it like on the 1080 Ti, better have 11 GB instead of wasting 1 GB you couldn't use anyway.
>>62395028
I'm a NEET too, its just about priorities, but you can make money easily with selling used hardware.

>>62395458
And you are why nvidia makes so much money despite being anti-consumer

File: 1481924374219.jpg (29KB, 400x465px) Image search: [Google]

29KB, 400x465px

Please let mining crash before volta comes out. the reason i skipped Pascal a few months ago is because the prices are fucked.

>>62376480
its 1180 you dumb fuck

>>62395503
I should note that I upgraded from a 9800GT, so yeah, they're still anti-consumer, won't disagree there, but I'm not really doing my part to line their coffers.

>>62395503
You're a bigger poorfag than he is.

>>62380856
>FX series

Nightmares with my FX5500.

>>62395315
That's 16 chips.
A FUCKING lot of PCB space.

>>62395490
Lmao no, this monitor is from like, 2012. Idr if there were monitors above 60hz at that point, but it's hardly my biggest concern at the moment.

So are they going to sell their entry level 90mm2 gpu as a 2080 this time?

File: 978.jpg (63KB, 340x565px) Image search: [Google]

63KB, 340x565px

>>62395547
>Be me
>Pascal comes out
>Everyone tells me to wait for a price drop which usually happen a few months after launch
>Mining shit happens
>50% more for a 1060/1070

>>62395163
>>62395315
youre retarded. initial will be 8gb for both 80 and 70. ti will be 8x2(16gb) in a two die package. plebs. also hbm2 is dead

>>62381851

Pascal is maxwell you dolt.

File: 2011-11-16-Shmuck-Therapy~01~01~01.jpg (27KB, 285x267px) Image search: [Google]

27KB, 285x267px

>>62394757
Downsampling

>>62395684
>Nvidia refresh
>50% performance improvement
>AMD refresh
>5% performance improvement

>>62382030
>the most important thing is 11 GB of VRAM in 1080Ti

Name some games which use 11GB of VRAM.

File: dCoHEeAl.jpg (79KB, 611x640px) Image search: [Google]

79KB, 611x640px

>>62395638
No, you will get a <300mm2 die, and it will still beat AMD

File: perfrel_2560_1440.png (52KB, 500x1090px) Image search: [Google]

52KB, 500x1090px

>>62395716
The throttling, air-cooled Vega 64 is 22% faster than a water-cooled Fury X though.

>>62395790
>double node jump
>22% improvement
Bravo AMD!

>>62395591
The density improvement is per chip.

You can do a clamshell configuration where two chips are mirrored on both sides of the PCB but that wont be necessary unless you wanted to QUADRUPLE the capacity compared to current GDDR5/X cards.

File: PNY-Professional-Graphics-Cards-Quadro-P6000-back.png (497KB, 800x800px) Image search: [Google]

497KB, 800x800px

The GV102 Quadro professional card will have 48GB of GDDR6 and memory can be soldered on the back of the card too, if you still don't realize that

>>62395832
>triple node jump
>5% improvement
Bravo Intel!

I JUST WANT A GOOD AND NOT OVERPRICED GPU GOD DAMNIT

File: ptlbr_noa0016.png (273KB, 683x507px) Image search: [Google]

273KB, 683x507px

>bought 1070 in May
Should I sell it off on craigslist and use my old 960 till Volta releases? I doubt I'll even be able to get my hands on 2070 or whatever it'll be till january anyway because Canada always gets shafted on stock.

>>62395697
Way ahead of ya. I'm not sure how to get all my games to do it, but the games that have the option are all running at 120~150% flawlessly, and I get a little bit of performance back from not needing AA at that point.

>>62395881
Yes. Then sell the Volta card three months after you buy it, because the next big thing will be right around the corner!

File: 1501735648299.png (174KB, 329x298px) Image search: [Google]

174KB, 329x298px

>>62395879
>nvidia raise 10 series prices because of some fabricated shortage
>nvidia base their 20 series prices on the raised 10 series prices

File: 1464311829450.gif (541KB, 480x228px) Image search: [Google]

541KB, 480x228px

>>62395790
>getting beat by an 18 month old card in the same price bracket

>>62395746
AMD pleb detected

PC gaming is dead and its the gpu industry that killed it

>>62395959
I own a 1080. Name some games that I don't have the VRAM to play.

I'm out of the loop. Is it coming out Q3 this year or next year?

I was going to build a new rig with a 1060 6GB, do I wait for the next cards?

>>62395910
>Way ahead of ya. I'm not sure how to get all my games to do it
GeDoSaTo works with way better results than the in-driver way but it requires custom game profiles and only works with DX9-11.

>>62395972
LMAO
Poorfag

>>62376480
I'm not expecting too much in terms of gaming performance based on previous comments from nvidia

>>62395978
Next year.

>>62396015
Just Wait™

>>62396043
UMA

>>62396022
>GeDoSaTo
I don't think I have any games running DX12, so that's okay, I'll check that out, thanks!

File: UMADELICIA6.jpg (87KB, 728x1007px) Image search: [Google]

87KB, 728x1007px

>>62396068
Seriously, should I wait? Are GPUs cheaper right when they come out? Hopefully the mining shit dies.

>>62396208
Always wait. When the thing you're thinking of buying comes out, remember that something better will be along in six months and continue to wait. That GTX 570 isn't so bad and you can make do until then. Never make the mistake of buying anything. That's what stupid people do.

>>62396015
>>62396208
Don't wait. Unless there is literally confirmed news of something actually good right around the corner, don't wait. You'll be waiting forever for the next best thing.

Buy the damn card and play your games. The 1060 6GB is the only card worth buying right now, as it's in stock, kills at 1080p, and its price is only inflated by ~$50 max.

>>62396283
>>62396296
I'm in no rush and I don't know shit about GPUs and how they're marketed, calm down lads.

>>62396330
FINE. WAIT. I'LL BE OVER HERE PLAYING THE NEW QUAKE WHILE YOU PLAY WITH YOUR PETER

>>62396381
I'll go back to playing my PS4 with all these great games, humans can't notice anything above 24 fps.

>>62396624

top kek

>>62396624
K. Play the new Gran Turismo for me

This is just speculating right?

It won't come out this year will it?

>>62397126
nigger the 1080ti came out 6 months ago
hold your fucking horses, this is 4chan

>>62395629
You could supersample your games... Looks sharper even on 1080p.

>>62397126
Feb2018

>>62380298
>tfw want to buy a Ti but can't because of prices and no stock

>>62397380

Why? The 1080 and Ti has the best price point of the whole lineup at least in Europe. 1070s are ~500 € while you can get 1080 Ti FTW3 for 829 €. I paid 899.

I can be glad, because I bought a 1070 last christmas and thought it was expensive. I got the complete price refunded for my used card because of this mining madness. At all, it was more like a loaned 1070 because I got the full price back and could buy a 1080 Ti.

>>62397126
the chip is already in production for the professional market, Nvidia is just waiting at this point, they no competition in the high end market at all

>>62381925
>makes VRAM amount a lot less important right
kek

>>62398148
>kek
kek

>>62397902
A lot of people are on 1000€ budget and won't shell out 800 on a gpu

>>62382998

You are an idiot. Vega 64 shits on GP104 by a good mile in GPGPU related stuff and professional graphics. The problem is that software isn't tuned for silly gayming yet.

Nvidia threw all of its 16nm magic on Pascal. Don't expect Volta to be a magical jump. It is probably be 10-15% faster at gayming but cheaper to make. It will rival NCU at general-compute though (this is what Nvidia is focusing on with Volta).

>>62398529
>the software isn't tuned yet

Ah yes, AMDrone with his magical finewine at is finest

lads is ripjaws as good as corsair
also is 16gb x2 a good idea or should I go wih 8gbx4

>>62398901
>is G.Skill better than Corsair?

Absolutely

>>62398497
It was a direct answer to >>62397380

I know my GPU cost as much as other people pay for their whole setup. But owning a 1080+ is pure luxury, it isn't necessary. If you can't afford it now wait for the GV104 X70, it will deliver 4K performance for around 400 €.
>>62398529
Vega 64 reaches similar GFLOPs like GV102, but its still 30 % slower in games.

The typical jump is 50 - 60 % per generation. This means, a 980 Ti will be 50 - 60 % faster than a 780 Ti, this counts for every performance level.

IThe next X70 will most likely have the performance of a 1080 Ti.

>>62398989
>>62398989

What

>>62398974

It is because the primitive shaders aren't being used in "gayming" yet. Expect a 20-30% gain which allow Vega to outpace the 1080 but they will never catch-up with 1080Ti.

Nvidia already went through all of easy hanging fruit with increasing gaming performance. They can't easily cram more shit into the chip either = shitty yields which is not profitable for their gayming SKUs.

Pascal is practically a die-shrink Maxwell with mass clockspeed roids throw at it with faster memory to keep up. They already used up all of 16nm node magic. That's what gives the "illusion" of that performance jump.

The reality is that days of rapid performance jump from moving to a new architecture have been over since Kepler/Fermi (GCN 1.0 for AMD).

The laws of physics have caught up. That's kinds why Nvidia is trying to move away from gayming as their bread and butter.

>>62399117
I know the gaming products are just trash, but performance still increases, also because of the faster memory. We are currently at around 512 GB/s, the next generation will reach close to 1 TB/s bandwith. Plus a few % more efficient ALUs paired with a much higher clockspeed.

Didn't it work like this every generation? Fermi was around 600 to 800 MHz, while Pascal reaches 2000 MHz without a problem.

>>62395547
If it doesn't crash, things could end up being real shitty with Volta as far as availability is concerned.

Volta will have loads of bandwidth and is getting power efficiency improvements on three sides (process node, new memory and the architecture itself), so it's power/performance ratio could make it a preferred choice and at that point... good luck getting it for a reasonable price.