GTX 2000 Series Volta Cards Discussion

File: 1494242831041.png (98KB, 261x300px) Image search: [Google]

98KB, 261x300px

Also, when I said
>Same power as a GTX 1060
I didn't mean that it had the same power draw, as the RX 580 actually has MORE. I was trying to say that it had the same performance as the 1060 with more power draw and more heat.

you should get Vega instead.

>>62093432
knew someone would say it. lmao.

am physical nvidia fanboy. Will never switch off my high power and reliable temps.

I've had some AMD products through my gaming life starting at 2006, had the first one gigabyte card they made, I believe. It burnt out within the first month as it was brand new, sent it back to the company and they gave me a new one. Three months later it burnt out.

Then in 2014, I got a build with an FX-8320. That also overheated with great airflow and a reliable build. It stayed at the temp of around 80 degrees max. Somehow that temp melted the pins into my motherboard.

Will never get AMD again. Pinless Intel processors and ice cold ASUS Geforce cards for me for life.

>>62093477
that has nothing to do with AMD. its the board partner. you should switch manufacturer.

what i meant was the upcoming vega 11.

im upgrading too so ill have to chooses between gtx 2060 or vega 11.

i might chooses the one with less power consumption. because i live in the country with the most expensive electricity in the world

>>62093353
Bet it'll be great for mining.

>>62093353
>amd cards are toasters
meanwhile in the tech world
https://www.hardwareluxx.de/index.php/artikel/hardware/grafikkarten/44084-amd-radeon-rx-vega-56-und-vega-64-im-undervolting-test.html

oh wait i guess its magic

>>62093505
vega 11 is a meme
HBM2 is too goddamn expensive

>>62093550
>hardwareluxx
>the amd arab shill is at again

>>62093353
>What is /g's opinion on the GTX 2000 series GPU lineup?
probably not coming out for awhile since Vega shit the bed and the 1080Ti is going to be the fastest card on the market for the next 12 months.

>>62093353
From what I'm reading on it, Tesla will be having a hand in the chip development. If this is true, then that'll be great. It means that the cards will be really premium and high end, of course with high end pricetags attached I will assume. The 1080ti has 12bn transistors and they expect the 20xx series to have at least 21bn with their involvement, almost double with a rumour of 16gb gddr6 graphics. Volta will certainly be the most anticipated card I will look forward to. Instead of needing two 1080Ti's, most people can do a lot with just one 20xx card. Future's bright boyos, wonder what fAilMD will do?

>>62093558
nobody know for sure.

File: Z.jpg (9KB, 219x230px) Image search: [Google]

9KB, 219x230px

>>62093353
Not 1100???????

>>62093517
I actually gave a separate mining computer that has a low profile GTX 950 with a 200 watt PSU. 16 gigs of laptop DDR3, and a 1 TB hard drive. A little server.
>>62093505
Could of sworn the only think I'm wishing to change is the GPU, if it's a Vega GPU it's AMD. Got bad experiences with AMD.
>>62093550
Never said an AMD card was a toaster performance wise, I meant heat wise. I want less heat for more active time, due to my very small main PC being too low profile for watercooling and an ASUS mini GTX 950,
>implying ASUS makes a mini 2050

>>62093573
one of the most famous german site
>arab shill
oh well since you want meme tier see this https://www.youtube.com/watch?v=OZkfYDWwo7s&t=11s

>>62093593
The name isn't set in stone. Could be 2000, could be 1100

>>62093594
have*, not gave. My mistake.

>>62093594
vega from amd has a top tier pcb with a shit tier cooling system if you have a watercooling solution get it and slap a block on it otherwise wait for the aib's in september

ATTENTION!
This is friendly reminder:
1050 Ti already exists and you DO NOT need anything more.

File: 8-year-old-fed-bust.jpg (41KB, 575x509px) Image search: [Google]

41KB, 575x509px

>>62093618
But what if I just wanna shitpost how rich I am in the BST threads?

>>62093618
But I have a WQHD 166hz g-sync panel so I actually do need more.

>>62093477
>my gaming life
>>>/v/
you're welcome

no point in releasing Volta when AMD is years behind leather jacket man

File: dank.png (525KB, 835x1055px) Image search: [Google]

525KB, 835x1055px

>>62093618
Dude I am a competitive gamer. Sure for the casual that is what is needed, but I want to push my processor to its limit. i7-3770 can handle a 1060 and a 2050 should be the same power as that. I just want at least 100 fps on future competitive games. On overwatch with tweaked settings my 950 stays around 130 fps average, but you know. There is a future for competitive titles.

>>62093353
>bigger dies when AMD is pulling a Larrabee that werks
Lmao.
Just make a bigger die.

>>62093657
sir, pls see this: >>62093646

Also:
- not your tech support

Now pls fuck off this board, this isn't for manchildren like you.

>>62093646
>sends OP to /v/
>the same amd g*rm shill was shitting /v/ too
really makes you think

>>62093594
Intel + nVidia is still the top choice for a stable and reliable build.

>>62093353
but nvidia is already bankcrupt at that point

>>62093678
Of course, man! Get a full AMD build and an Intel+Geforce build, wait 10 years with consistent use, which one is more likely to be running?

>>62093675
I don't get it.
Go to your containment board.
>>>/v/

>>62093702
The same hardwareluxx poster was shitting up /v/.

>>62093701
that would depend on the usage, environment and the build quality.

If done properly both will survive more than 10 years.

>>62093701
You'd have to get 1000 AMD builds and a 1000 nvidia builds, and then maybe one of them would fail.

>>62093747
I asked which was more likely, not which will go first, because anything could happen at any time.

>>62093585
Tesla? Is Jim Keller cucking us now?

File: GF-GTX950-E2GB_OC_LP_01.jpg (54KB, 500x365px) Image search: [Google]

54KB, 500x365px

GPU for my mining PC if anyone wanted to know.
Positive hash/power draw ratio.

>>62093782
No, Tesla as nvidia's compute line of cards.

>>62093782
tesla got keller BECAUSE nvidia was so bad at it

they literally cant design shit

>>62093353
It's going to shit on all of AMD's offerings by a lot. It's also probably going to be twice as expensive as current GPUs because of that.

>>62093353
it will be awhile before it comes out, during one of their Q&A sessions they said it costs just under a grand to manufacture a volta die on TSMCs 12nm process. So my guess is they're waiting on manufacturing costs to go down before they port it to consumers. So if your opposed to AMD, I would recommend either getting a 1060, or wait till around may for volta.

File: Untitled.png (180KB, 696x576px) Image search: [Google]

180KB, 696x576px

>>62093782
If it was the architecture then I doubt Tesla would be written on the actual card. I think the actual company Tesla might be involved now.

>>62093842
this dude calling nvidia bad.
HAH.
Good luck with your amd bro.

>>62093865
https://electrek.co/2016/01/28/tesla-jim-keller-apple-processor-architect-2/

whatever you say kiddo

>>62093897
lemme get this straight.
AMD CPUs hide in the shadows for what, 4 years?
Now we're expected to be like "OH SHIT, AMD IS BACK AGAIN? LET ME KISS YOUR ASS RIGHT QUICK AND TALK SHIT ABOUT YOU OVER THE INTERNET!"

File: 1501944059454.png (174KB, 696x678px) Image search: [Google]

174KB, 696x678px

>>62093930
i cant hear you over the market share im currently gaining

>>62093930
You do understand that CPU R&D cycle is fucking LONG, and it's especially long for a totally new from scratch design?
gb2/v/

>>62093794
How is getting gpus from Japan anyways. Is it cheaper?

>>62093738
How the fuck it's related that /g/ isn't /v/? I don't care who is shilling what on /v/, just get the fuck out, trash.
>>>/v/

>>62093943
Just wait for their mobile gaming APU, I'm sure it will be great! *rolls eyes*

Going to laugh when their high end APU gets fucking wrecked by a year old mobile 1060 in both battery life and power.

>>62094107
There's no "high-end APU", RR tops as 11CUs.
That and kill yourself you retarded attentionwhoring tripshit.

>>62093772
that would depend on the manufacturer. AMD and nvidia only supplies the GPU. Other components are made by the OEM.

>>62093353
It will be Pascal with HBM, maybe on a newer node process, just like Tesla or whatever Volta they've already released.

>>62094167
>HBM2
Haha no.
Nvidia is not a really good friend with Hynix or Samsung to do that.
GDDR6 it is.

File: Begone_thot.jpg (183KB, 640x828px) Image search: [Google]

183KB, 640x828px

>>62093585
21bn on a 815mm2 die, not gonna fly in the consumer space senpai.

GV104 will have 3584 shaders, clock higher, use a bit less power, and be a bit smaller than GP102. GV102 will be a big 600mm2 die, 5376 shaders but cut for yields. TSMC 12nm isn't a new node.

>>62093353
You will be waiting for a long time for GTX 2xxx series. I would assume two years from now, nVidia has no real reason to release anything.

I got a bitcoin layin around

should I get two 1080tis or should I """wait""" for volta

>>62093794
>https://www.amazon.co.jp/%E7%8E%84%E4%BA%BA%E5%BF%97%E5%90%91-%E3%83%93%E3%83%87%E3%82%AA%E3%82%AB%E3%83%BC%E3%83%89GEFORCE-%E3%83%AD%E3%83%BC%E3%83%97%E3%83%AD%E3%83%95%E3%82%A1%E3%82%A4%E3%83%AB%E5%AF%BE%E5%BF%9C-%E3%82%AA%E3%83%BC%E3%83%90%E3%83%BC%E3%82%AF%E3%83%AD%E3%83%83%E3%82%AF%E3%83%A2%E3%83%87%E3%83%AB-GF-GTX950-E2GB/dp/B01G1VE1HE

Is there any reason why you went moon rather than buy it here in the mericas

>>62093859
Fucking retard.
Nvidia has been calling their server/compute GPUs Tesla since long before Elon Musk even thought about starting a company named Tesla. They have nothing to do with Elon's company at all. They probably use Tesla GPUs, but not for being "involved" with Nvidia in the way you're implying.

Nvidia has three lineups: GeForce GTX, Quadro, and Tesla. Gaming/consumer, workstation, and server/compute, in that order.

>>62094350
You forgot GRID but whatever.

GTX 2050 Ti will be great

75W no PCIe power connector, performance at GTX 1060/GTX 980 level allowing you to play 1080P games at Very High/Ultra settings

Can't wait

Why do people expect HUEG performance gains from Volta when it's not a double node jump?

I'll join 8700 + 2070 master race.

>>62094373
No, I left it out because it isn't a GPU lineup. It's a hypervisor that manages distributing Tesla GPUs to VMs.

>>62094394
They sell it as separate SKUs so whatver.

>>62094388
Volta is biggest arch since Fermin, maxwell and pascal are medium and low change architecture,paperwhite about Volta SM say Performance/watt begins 50% over pascal.

Volta is big deal.

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

>>62094410
>Volta is biggest arch since Fermi
No.
>paperwhite about Volta SM say Performance/watt begins 50% over pascal
*only when doing meme ops
Literally everything that Volta brings is for compute, even larger L1$ that can work as a part of unified memoryu

>>62094383
>1080P games
>1080P

lol

File: 54036839.jpg (51KB, 500x404px) Image search: [Google]

51KB, 500x404px

>>62094424
>muh 4K meme

>>62093558
it will use low cost hbm, not hbm2

>>62094433
>muh 2006 resolutions

>>62094417
>SM
What a tricky way to word it.
>>62094437
This one doesn't officially exist yet anywhere but Samsung's HotChips2016 presentation.

>>62094417
>Data Center GPU
you're posted this like 5 times already

>>62094445
and samsung will deliver. there is a long time till december

File: NVIDIA-Volta-GV100-SM.jpg (106KB, 800x442px) Image search: [Google]

106KB, 800x442px

THANK YOU BASED NVIDIA

>>62094466
The volume is extremely important though.
That and 2.5D packaging is a VERY slow process.
>>62094475
I can add moar ALUs, do a half-node jump to offset the clock loss and claim 50% efficiency too.
But it's boring and not entirely fair.
That and Vega pulling a Larrabee is still more interesting than moar ALUs/cachez/meme ALUs.

>>62094423
To all /v/ mindset, people buy DLC,microtransactions,skins,pay to win and loves call of duty,Dota,LoL,loves consoles,muh normies, and Buy laptop gaming, simple nVidia getting power efficient GPU will makes billions.

>>62094484
There is nothing interesting about AYYMD HOUSEFIRES

Now fuck back off to your AYYMD HOUSEFIRES threads

>>62094502
>nothing interesting about fucking Larrabee but an actual GPU
u wot

>>62094510
It it not a Larrabee and it's not power efficient at all, enjoy putting it in your datacenter and wasting power

leather jacket man said its not coming out anytime soon

>>62093477
>ice cold geforce
>gtx 480 reaching 100 degrees


UMA DELICIA

>>62094518
>It it not a Larrabee
Oh yes it is.
They quite literally implied the possibility of manual pipeline assembly.
That's quite literally it.
>power efficient
Meme.
>enjoy putting it in your datacenter and wasting power
Meme, since it's usually bundled with 1S EPYC, which means supreme perf/watt for system overall.

all I want is something noticeably stronger than 750 Ti while being a relatively low power draw

>>62093657
overwatch is not a competitive game

>>62094543
so, 1050? 1050 ti?

>>62094543
Wait for GTX 2050 Ti

>>62094585
more leaning on the side of 1050 Ti, but yeah, something like that.
>>62094601
well, j have no ram in my system, and being a jobless gag it's going to take a while to save up for 16gb, so I might have to wait for 11xx/20xx series GeForce cards, or if AYYMD can manage something similar by then

>>62093657
>overwatch
>competitive
pick one

I am so close to ordering a 1080ti

Holy fuck volta hurry up

>>62094107
Fuck off, faggot

>>62094762
>hurry it
It's late Q1 2018 at very best.

>>62094617
So get a fucking job. Jesus Christ. Why are you concerned with computer parts when you should be getting a fucking job.

File: average amdrone.png (524KB, 920x900px) Image search: [Google]

524KB, 920x900px

>>62094522
Still draws less power than the housefire known as Vega.

And it was stiill the top dog GPU when released. Vega doesn't even compete vs. a super cutdown die from NVIDIA while being 50% bigger than said die.

>>62094803
It wasn't.
5970 crushed 480 in basically everything.
So please stop shitposting.

File: You and two (You)s.png (312KB, 640x480px) Image search: [Google]

312KB, 640x480px

>>62094107
So a RR APU will get beat by a mobile i7 and discrete mobile graphics. No fucking shit? Yet they'll do so by %10 perf, tons more heat to dissipate and double the cost.

You do realize their APU line is economy, possibly economy+, whereas a laptop with a good, discrete card like the 1060 is far beyond economy+ in terms of categorization.

>>62094784
not him, but but don't just assume everyone in the world can get a job as easily as breathing
no job means no money
no money means no car
no car means much harder to travel
I have to live in a communal home or just on the streets
no jobs in my town means I need to walk about 5 hours each day, or get extremely lucky someone else is willing to drive there, and there's no guarantee they'd accept my application
the only money I get is from doing odd jobs around town
fuck off you entitled twat. remember that not everyone is living in the exact same situation.

Please stop shilling for Intel in Nvidia threads. I hate AMD GPUs with a burning passion but their Ryzen shits on anything Intel has to offer. I want Intel oven dodgers to stop jumping on Nvidia train just to make AMD look bad.

>>62093353

They just said that they won't release it because amd Fucked up.

Also the next gen nvidia will use gddr6 or 5x.

File: 20170802_201209.jpg (3MB, 4032x3024px) Image search: [Google]

3MB, 4032x3024px

>>62093477

After t he Vega Fuck up i bought the Evga ftw3 1080ti.

Tbh I'm not impressed with thermals and noise.

Its loud and hot. I expected that the ftw3 would be cool af. But no. Even with a custom fan profile and no oc it gets close to 77c during pubg. And it's noisy too

Im considering upgrading. My case fans but never had this issue when I had my 390x

>>62095494
>Im considering upgrading.
with what?

nvidia is too expensive here.
amd to go

>>62095572

My Samshits auto formatting is dumb.

I was trying to say. Upgrade my case fans

Poo fans are expensive tho.

>>62093353
why not just 11xx?
Makes more sense, unless they have completely changed their lineup.

>>62095494
>buing EVGA 10xx series
it's like you don't read the news

>>62095494
there's no reason not to watercool it if you can spend this much for a consumer/gaming card.

>>62093505
>that has nothing to do with AMD. its the board partner. you should switch manufacturer.

Nvidia
>Partners dont matter, a card is a card
AMD
>Some partners will melt the card to your motherboard, NOT AMDs FAULT :(!

File: 1437600843130.jpg (129KB, 1104x1104px) Image search: [Google]

129KB, 1104x1104px

>>62093657
>He thinks gaming is the most stressful thing a computer can do

>>62096577
>Some partners will melt the card to your motherboard
opposed to
>some partners will blow up your motherboard
?
it's like you missed whole evga debacle

>>62095494
should have bought amp extreme or strix

>>62094814
>5970 crushed 480 in basically everything.
amdrone history revisionist

>>62095959
good job buying into hysteria

>>62093616
realistically october, that's when miners should hit a new hash tier that puts it out of consumer gpu range for mining.

>>62093353
>Got that 400 watt PSU feel me?
i ran a i7 2600 and a R9 290 reference card in a mini itx off an 450w , no issues whatsoever

>>62094350
>Nvidia has been calling their server/compute GPUs Tesla since long before Elon Musk even thought about starting a company named Tesla
The first Nvidia "Tesla" GPU was G80 based, which launched in 2006. Tesla, Inc. was founded in 2003.

Anyway, who gives a shit about who chose what name first, both are just ripping off the name of a dead Croatian. There have been other companies that used the Tesla name before Nvidia and Elon did it.

File: 1502447076690.png (386KB, 960x960px) Image search: [Google]

386KB, 960x960px

>>62094871
This desu. My current setup is a Ryzen 5 1600 and a 1060 6GB, and it's been going great so far.

File: 1452799837787.png (650KB, 512x800px) Image search: [Google]

650KB, 512x800px

>>62093477
>my gaming life

>>62095494
>air cooling a flagship gpu
>it's loud

No shit.

I have a feeling the Pascal architecture was kinda like Sandy Bridge, way ahead of its time with performance surpassing current needs.

I sincerely doubt volta will be a huge jump forward.
Maybe just small 5%-10% gains over each model.

On the other hand i don't really know much about the specs of volta, 12nm does sound promising

>>62100539
You mean maxwell, pascal is ivy bridge with it's process shrink and volta is adding useless shit like haswell (avx2, meme learning tensor).

Think is nvidia have a lot of room to play with in their margins, the gp104 and 106 are really small for the price point they hit. Volta variants could just be 20% larger with a small clock jump combined with gddr6 and you have your standard accross the board performance increase (2060 ~= 1070 stuff) and hey ho another year of sales.

If any of that's even needed of course, there could just be a round of rebrands for most things and the new titan/2080ti is the only new silicon above the pascal family.

>>62100539
No, GPUs don't work like that.
You can always go wider/taller.

>>62094297
3072 shaders for GV104 seems more likely IMO. Nvidia's special 12nm FFN doesn't have a substantial density improvement over 16nm FF+. Volta has a lot more memory on chip from the increased cache and register files and that memory takes up die space. Plus you have more transistors needed for the expanded feature set and the new SM design. I don't see Nvidia letting their 104 class chip balloon past 400mm2.

Given the pipeline changes I also don't see Volta chips clocking that much higher than Pascal if at all, but that won't matter much as Volta's IPC will likely be substantially improved.

Even with only 3072 shaders it will still probably blaze past a GTX 1080 Ti, especially in DX12/Vulkan games.

The case is similar with GV102, it's entirely possible it will have considerably fewer shaders than GV100.

>>62100923
You can't apply "IPC" to GPU, and larger cachez are there only to simplify programming model.
It's Kekler Refresh tier improvements.
Moar ALUs for everyone!

>>62093353
If Nvidia launched it before the end of this year AMD would go out of business.

>>62101000
You need AMD CPUs to power meme machines.

File: 1501811741079.jpg (70KB, 425x312px) Image search: [Google]

70KB, 425x312px

>>62094107
Friendly reminder to sage and hide the triptard

>>62093353
>Honestly don't understand though, due to their release schedule it should come out very soon. Not years, but months.

They have market monopoly. If anything, it shouldn't come for at least 1 or 2 years, since they'd just be competing with their own products.

>>62100945
IPC seemed the most appropriate term, but there may be a better one.

From the Volta Whitepaper

>Unlike Pascal GPUs, which could not execute FP32 and INT32 instructions simultaneously, the Volta SM includes separate FP32 and INT32 cores, allowing simultaneous execution of FP32 and INT32 operations at full throughput, while also increasing instruction issue throughput. Dependent instruction issue latency is also reduced for core FMA math operations, requiring only four clock cycles on Volta, compared to six cycles on Pascal.

The changes to cache design simplify the programming model, the increased cache size and "register file to shader ratio" however WILL improve performance.

>>62101269
Marginally so. 4clocks/cycle is very GCN. And int32 is bretty fucking irrelevant for gayms since the bibelines are fully FP since DX9.
So it's all compute.
What's really interesting is Vega's programming model for their geometry pipeline.
That and L1$ being part of unified memory and Volta.

>>62101325
>*4 clock cycles
Fix.

>>62093353
>the absolute state of Nvidia shill threads
>people still replying

>>62101365
There's nothing wrong with discussing Volta.

>>62101365
>Buttmad AYYMDPOORFAGS with no GPUs detected

>>62099281
>1080 Ti
>flagship
lol

>>62101475
GTX 1080 Ti is the flagship that people can afford to buy

You fags bitch about the TITAN Xp all the time

>>62101494
That doesn't make it a fucking flagship. 1080 Ti is just a high-end GPU, TITAN is the flagship regardless of whether you cunts can afford it or not.

>>62101503
1080ti is a flagship.
Titan Xp is a "prosumer" card.

>>62101503
>>62101531
Titan cards are an exercise in how deep the consumer will take dick. They're a joke that's been taken too far and it for some reason gave nvidia more sales for charging 3x as much as what used to be a high end graphics card.

>>62093576
Q1 2018 announcement
Q2 2018 release

Basically they won't rush production and keep going according to their roadmap. Had Vega not shit the bed Nvidia would've rushed Volta for Q4 2017

>>62094522
Monkey Brazilian memes are intel buddy

>>62101893
Don't like it
Don't buy it

If suckers do buy it then it's only more money to put into development

>>62093477
>Then in 2014, I got a build with an FX-8320. That also overheated with great airflow and a reliable build. It stayed at the temp of around 80 degrees max. Somehow that temp melted the pins into my motherboard.
That is just complete and utter bullshit unless you did something completely retarded such as fucking up the thermal paste. Vishera is bad in a lot of ways, but running too hot is not one of those unless you got one of the 220W 9xxx series chips.

>>62095494
blind guess - you are a burger.
Should have gone with MSI(GamingX)/Palit(JetStream, not 4 fans disaster)/Zotac/Asus(Strix)/Gigabyte(Extreme, not G1) - all of those are better than EVGA, which practicaly doesnt exist outside US for a reason - its overpriced for what you get

>>62095494
Maybe the card has some problems, bacause 1080Ti has lower TDP than 390X, so overall it should generate less heat.
Core temp doesnt matter, its neither the hottest point on the card, nor what makes card overall generate a lot of heat.
TDP under load is how much heat your whole card generate, whether you core is 70 or 90C.

i have a 1080ti, i don't care

>>62103799
same, the first version of volta will probably be the cut down volta like the regular 1080 / 1070 making the 1080Ti similar to the 2070 or whatever it will be called. Then 6 months to a year later they'll release big volta.