ITS HAPPENING - NVIDIA VOLTA FOR Q3 2017

File: 2f7.jpg (94KB, 601x508px) Image search: [Google]

94KB, 601x508px

>>59973346
JUST WAIT FOR VEGA YOU IDIOTS

File: amdvega.jpg (156KB, 633x758px) Image search: [Google]

156KB, 633x758px

>>59973346
ITS NOT FAIR GUYS

WHO ARE THEY COMPETING WITH? FUCKING CHILL WITH THE RELEASES.

>>59973346
There was a thread on this retarded rumorshit yesterday. >>59960283

And it's just a dumbass rumor. Fuck off with this wccftech garbage.

>>59973443
Huang is probably fed up with amdrones bullshit and wants them to commit sudoku.

>>59973346
sir please do the needful and wait for vega

File: C7hgE8k.png (18KB, 722x477px) Image search: [Google]

18KB, 722x477px

>>59973346
>read the chinese source article
>article talks about a rumor that it will launch in Q3 and how it would be good financially
>they're reporting it as fact
God damn is tech news trash.

Weren't they also memeing a few months back it would be "12nm"? Or are they just forgetting it like when they reported on this: http://web.archive.org/web/20150413165626/http://wccftech.com/rumor-amd-phenom-iv-x12-170-baeca-25nm-cpu-leaked-features-12-cores-6-ghz-core-clock-am4-socket-compatbility/

https://www.youtube.com/watch?v=b_ZMOn0X6jw

>>59973346
Oh look, another GP100-esque paper launch. Bravo novideo.

Since it's not taped out yet will they showcase it in VHS tape with woodscrews?

File: 7.jpg (62KB, 800x600px) Image search: [Google]

62KB, 800x600px

>>59973784
I miss the VHS

what's the volta

File: screenshot-www.anandtech.com-2017-04-19-16-26-04.png (747KB, 720x1379px) Image search: [Google]

747KB, 720x1379px

>>59973766
http://www.anandtech.com/show/11102/nvidia-announces-quadro-gp100

>>59973459
it's a rumor based on a rumor.

>>59973897
Yes and? Look when Pascal was announced.

>>59973912
Where is Vega (tm) :3

>>59973933
Already out for ML market.

AMD comes late to the silicon party. Again

It's also apparently on 12nm and TSMC can shit out a new node on a 600mm2 GPU half a year after the GP102, yeah, legit.

File: Vega-Final-Presentation-15.jpg (665KB, 2666x1500px) Image search: [Google]

665KB, 2666x1500px

Hey look it's stacked vram next to the GPU,I wonder who had that design idea for that for years now.

>>59974571
Hmmmmmm

File: 1461745848598.png (574KB, 706x706px) Image search: [Google]

574KB, 706x706px

>>59973346
>yfw still uses shitty software scheduling like Kepler/Maxwell/Pascal

>>59974571
Nvidia beat AYYMD HOUSEFIRES to HBM2 by over 1 year and AYYMD still can't launch KEKGA

File: 1491502354030.jpg (73KB, 633x758px) Image search: [Google]

73KB, 633x758px

>>59973443
THIS

>>59974638
https://forum.beyond3d.com/threads/nvidia-pascal-reviews-1080xp-1080ti-1080-1070-1060-and-1050.57930/page-38#post-1974379

>Just another useless youtube video whose author does not have a clue what he's talking about... Can I have 20 minutes of my life back?
>It's hard to even comment on this with people not having a clue about what a scheduler does and which scheduler does what (Yes there's more then one). Fermi vs. Kepler scheduling has been discussed around here quite a bit already. It has to do with how warps are scheduled after kernel dispatch or draw call is issued. That is to say after a shader instruction is executed for a group (warp) of 32 pixels/vertices/threads when can an SM pickup next instruction from this same warp. This was done in hardware in case of Fermi and has been done in software since. The way this software scheduling works is that a shader compiler will emit control codes/hints within shader code. >You can read more about it here. This is compile time job. Once shader is compiled CPU doesn't do anything anymore.
>How this works against his fantasies further on is anybody's guess.

>Still posting this debunked lie
>2017

File: [GJM] Urara Meirochou - 06 [30C4B436].mkv_snapshot_02.06_[2017.02.10_10.41.54].jpg (209KB, 1280x720px) Image search: [Google]

209KB, 1280x720px

>>59973900
>this thread is a rumor based on a rumor based on another rumor

>>59974683
MADrones always shill muh async. It's their version of aplelfaggots retina. Completely clueless and regurgitate amd marketing buzzwords.

>>59974683
>>59974765
Whine all you like, Nvidia can't kludge their DX11 performance with driver patches anymore.

>>59974858
I'm laughing at drones like you.

Either another GP100 paper launch or another 750Ti.
Which I guess is nice if you want poorfag-tier GPUs

>>59974651
The best part is that Hynix can't make their HBM2 clock much past 1GHz ATM, so it means that the bandwidth is limited to 408GB/s-500GB/s if they're able to get it to clock a bit higher before Pooga's launch.

It'll straight up have less potential bandwidth than the G5x used in NVIDIA's GP102 cards, and if the Fury X memory controller resembles anything used in Pooga, it won't even get that.

>>59973346
The only way this makes sense is to float out fake rumors to get people to wait for volta instead of buying vega, especially if nvidia knows vega's going to blow them the fuck out

>>59974956
Also pooga only uses 2 stacks of HBM2 limiting it to half the bandwidth of P100 4 stacks. I won't be surprised if gddr5x cards have higher bandwidth than pooga.

Learn to use the catalog before making duplicate threads.

>>59973443
As much as I hate Nvidia as a company, I have to commend them for their aggressive release schedule. I like AMD more, but they're just too damn slow in their releases.

>>59975288
>dieshrink vs major architecture change
Wow. Really makes you think.

>>59974651
There wasn't a functioning (beyond a few hours) GP100 until September, they didn't start shipping engineering samples until December. Actual sales were finally in late January.

>>59974956
Nothing stops them from using Samsung HBM2. The chip also does actually support four HBM2 pads, consumer version will just have two.

File: 1490302558198.jpg (54KB, 305x309px) Image search: [Google]

54KB, 305x309px

>>59973346
>>59960283
Anyone that believes this rumor should kill themselves.
It's way too unbelievable.

>16GB of HBM2
lmfao.
Just like how the 1080Ti was rumored to come out in January 2017 with 8GB of HBM2.

Volta won't come until Q2 2018.
If it comes sooner, it's because Nvidia will gut lots of features they were working on to get it out sooner like AMD did with Fiji. And there won't be a Titan Xv with 16GB of HBM2

>>59973367
>>59973411
>>59973443
samefag

>>59973346

Fake. Now GTFO!

>>59976251
kill yourself dumb amdrone
you can't even tell samefag

>>59973493
>reporting it as fact
Literally the first phrase in the article says:
>There is a very interesting rumor currently making the rounds...
Are you retarded my man?

Nvidia scared and chasing it's own tail.

I fucking love it.

>>59976719
>scared
nvidia has more 1080/70 sold than poo 480s

>>59976738

It also can't run DX12.

>>59976784
>demolished in dx12 and vulkan by 1070
>can't run it
This is why amdrones are laughing stock.
Just wait

>>59976738

I kinda doubt that.

1060 is what Nvidia sells in bulk. 480 and 1070 probably sold roughly the same amount of units. AMD sold mostly 470 and 460s.

>>59976719
Yeah, AMD is so bad this generation that Nvidia has to compete with themselves.
The jump from the 900 series to the 1000 series was so fucking large it's going to be pretty damn hard to come with a followup later this year.

>>59976810

Nvidia wait fags on suicide watch.

>>59973443
They dont want to be like intel

>>59976845

More like trying to clear their slate for a major architecture change that doesn't suck dick at the vastly superior DX12 API.

>mfw 10XX series owners are going to stop having performance updates already

>>59976845

Not really.

The jump from Maxwell to Pascal isn't that massive. It is really just average at best.

The main problem is that masses don't give a shit about high-end performance GPUs. It is a shrinking market. The real money is elsewhere and both AMD/Nvidia know this.

File: Capture.png (359KB, 1077x607px) Image search: [Google]

359KB, 1077x607px

>>59976900
>Not really.
>The jump from Maxwell to Pascal isn't that massive. It is really just average at best.
>literally the biggest jump in over 7 years
>average
I don't think you know what the word "average" means.
>The main problem is that masses don't give a shit about high-end performance GPUs
More people own GTX 1080s than RX 480s, according to the Steam survey. Hell, more people own GTX 980 Tis than RX 480s. Plenty of people appear to care. Not as many as those who own 1060s for instance, but the profit on a 1080 must also be higher. Probably not the most lucrative anyway, but they've made plenty from the thing.

>>59973346
>It's another GP100 paper launch episode
Someone pass the popcorn

>>59977174
But we only have peanuts here. The gaymer kind.

>>59977174
You can get Quadros with the gp100

b-b-but where is Vega?

File: 1481994166110.png (544KB, 960x940px) Image search: [Google]

544KB, 960x940px

>>59977134
>comparing G104's and G100's
???

Do a comparison of GTX580, Titan Black, Titan X, and Titan Xp

>>59977242
Yeah, nearly a year after Nvidia claimed it would be available. AMD is still on schedule with Vega.

>>59977242
You know the MI25 is for sale right.

>>59977291
>Do a comparison
Do you think I did that comparison? Holy shit this is like porn comments where people think they're talking to the porn star.

In any case a 1080 Ti is like 85% faster than a 980 Ti at 4K (check the techpowerup review and do the math), that should tell you all you need to know about how big a leap Pascal was compared to its predecessor.

I thought the rumor was that Volta is coming out for workstations by the end of 2017, and consumer Volta is going to be released sometime in 2018. And that Nvidia is going to "refresh" consumer Pascal in 2017 so that they have something before Volta.

If the Volta *70 card is as powerful as the 1080ti like the 1070 is as powerful as a 980ti, I'll he fucking hyped. A 1080ti with proper async compute and other refinements for $450 or under will get me wet.

File: 16173040944l.jpg (175KB, 752x752px) Image search: [Google]

175KB, 752x752px

>>59976810
>6% more fps more from a card that costs twice as much
>demolished!!!!

>>59977495
>And that Nvidia is going to "refresh" consumer Pascal in 2017 so that they have something before Volta.
They released 1080 Ti and refreshed the Pascal Titan, the 1080 and 1060. Volta is next, whether it's this year or 2018.

>>59973346
Amazing. I'll just use that with my Freesync moni-
oh. WAIT.

File: 1492288255025.png (2MB, 1024x5252px) Image search: [Google]

2MB, 1024x5252px

Vega is being sold as the Mi25.
Keep in mind both nvidia and AMD want to sell their GPUs for the enterprise to be used in machine learning, we are the rabble that gets the leftovers.

>>59977869
>Mi25

I never saw a single benchmark

>>59977869
At least AMD doesn't neuter the FP16 and FP64 performance on the consumers. It can actually be useful for gaymes and hobbyists.

>>59977906
>I never saw a single benchmark
Yes you did. 25TFLOPS fp16. There's nothing else to know about that card. Are you blind, or just retarded?

>>59978083
>Yes you did. 25TFLOPS fp16. There's nothing else to know about that card. Are you blind, or just retarded?
The point is the card is not available...

File: AMD_radeon_instinct.jpg (37KB, 675x380px) Image search: [Google]

37KB, 675x380px

>>59977906
>>59978118
It doesn't have a display adapter. It is actually being shipped currently. I can't talk quantities but yeah.

>>59978083
Vega 20 is intended to be the FP64 card, this is just FP32 and FP16

>>59978322
It's variable precision so both fp8-32 and int8-32

WHY ARE THEY CALLING IT THE gtx 20xx SERIES? WHY ISNT IT THE 11xx? THIS IS BULLSHIT my computer naming scheme is fucked in the ass now

>>59976817
Who is buying 460 wtf is this going in laptops

>>59974683
>lie

okay, where is my dx12 patch Huang?

>>59978322
Who got this card beside amd, seem like woodscrew to me

>>59976251
they already made a Quadro with 16GB of HBM2, if they do a non-workstation card though it'll probably be a new Titan priced at $1200-1500.

>>59979532
you could ask same thing in autumn about tesla

>>59979653
they already made new new titan

tesla is not quadro

>>59979310
Yes, Apple

>>59979777
>tesla is not quadro
no shit retarded amdrone
http://www.nvidia.com/object/compare-quadro-gpus.html

>>59973346
>currytech
>chink source
lmao

>>59973443
NVIDIA cards don't support Chill, their drivers are terrible.

>>59981135
>their drivers are terrible
yeah ati/amd can't get proper OGL drivers for the last 15 years if their lives depend on it

>>59973346
>hbm only on titan xv
haha poorfags btfo

>>59981161
At least their driver software doesn't look 15 years old

>>59981167
>glossy transparent shit
literally kill yourself

>>59977134
>can't into graphs
>failed maths

>>59981167

>aesthetics matter more than ease of interface

end yourself rajesh

>>59981161
OGL hasn't been relevant for 15 years

>>59973346
... Does Nvidia watermark their chips with the McDonald's logo now?

>>59981245
>>>/v/
and even vermin would tell you that shit like emulators require ogl and ampoo is no go

>>59981198
>glossy
I don't think you know what that means
>>59981236
>ease
Yeah NVIDIA's control panel sure is easy to use when it has zero functionality. Still uses more resources than Crimson though.

>>59981370
Crimson tends to hang on open for a second on my system, though I do have a Core 2 Quad still.

>>59981698
NV Control Panel literally takes longer to open than Crimson did for me, and I went from a 480 to a 1080 ti

>another paper launch

>>59981708
I haven't used Nvidia drivers since 196.75 killed my 8800 GTS so I wouldn't know. I haven't seen a reason to get something newer then my 5870

>>59981739
On the bright side you missed another 2 or 3 driver releases that bricked cards again.

>>59973346
>>59973367
>>59973411
>Posting currytech
Fuck off Pajeet, it's time to go back to the shitting street

>>59976496
proof?

File: 1484757497575.jpg (8KB, 250x247px) Image search: [Google]

8KB, 250x247px

>when gpus advance faster than games and displays that can make effective use of it

>>59981854
8k60 capable gpus when

>>59981854
It's great. $100 budget cards are enough for 1080p now.

>>59981995
except for the 970

>GTX 2070 with 11GB of GDDR5X @ 12Gbps
>GTX 2080 with 11GB of GDDR5X @ 14Gbps
>TITAN Xv with 16GB of HBM2 @ 16Gbps
Pascal rebrands and a Volta Titan.

>>59982142
>16GB HBM2
>NVIDIA Card
Sounds like the TITAN Xv will be $2500, they need their juicy profit margins.

>>59982117
The 970 isn't a current gen card. I was referring to the RX 460/470 and GTX 1050/1050ti. Both are suitable for 1080p, especially if you don't need to max out everything.

>>59981721
Where is v-v-vega?

>>59981854
Game engines anon,won't be another next gen engine for awhile.

>>59983136
wait for it

>>59977134
>steam hardware survey
Last time i did it 'twas 2011/12 or something like that.
You can only do it now if you yourself decided to do so, and trust me, 1080/ti owners like to measure their SUPERIORITY much more than 480 owners.

>>59981854
8k monitors exist today. Good luck finding a card to run 8k ultra anytime soon.

>>59986940
won't matter since buying an 8k monitor will leave you broke

fmd they're expensive

>Volta 2050 Ti will have same or better performance than GTX 1060 which has same or better performance than GTX 980
>1080P gaming high settings at 75W bus powered without needing to plug in any extra power plugs

THE FUTURE IS SO BRIGHT, I GOTTA WEAR SHADES

>MADDRONES BTFO
>NVIDIOT ON SUICIDE WATCH
Jesus Christ this thread is shit.

>>59977134
Raw performance jump is huge, but the hardware jump isn't some revolutionary thing.

They increased cuda core count by 25% and clockspeed by 50% while shrinking to 16nm. Other than that and a few feature implementations, it is essentially a smaller and faster Maxwell. It's similar to the difference between 28nm GCN and Polaris (14nm GCN).

File: 7ypFu4W_BGFEsDcD_Slem67BfOBlRvfUW9PuhoKX7cw.jpg (115KB, 1024x683px) Image search: [Google]

115KB, 1024x683px

>>59973346
###POORVOLTAge
<------ lmao that was a dank meme i r89 r9001 390/3.5gb of v for vendetta ram, novidiots be like do a 1080 and walk awayy lmao, 144 wayy sly novideo grillforce gtx 480 clocked at 3.5hz 1080p 3.5 fps gimpworks edition, human eye can see only 3/5 ram anyways, nivea water boiler titanal xp at 3.5p, envidia fryforce shitan rascal 1080shiTi at 87.5% GTX(GetTheXtinguisher) (ngreedia pisscal xp "malware experience" thermi burning sillayy bridges edition with GreedSync support) + krappy fake $hilltel gore i$emen at 7700K (Kelvin) maybelake i0 delidlake cancer core killowatt lake jewel "cooler not included in a box" coffin lake "end of life" edition at 0.90 GHz (a.k.a. Inferno Bridge). 8GiB? What do you need 4GiB for? I'm giving you 3.5GiB! mumble mumble 3.5!?!?!? mumble mumble mumble 390?!?!?!

>>59988786
saved