/w4v/ - Daily Wait for Vega Thread

File: 62791568_p0.jpg (741KB, 2046x1446px) Image search: [Google]

741KB, 2046x1446px

>>60271822
>AMD will surely release in a timely manner and will definitely blow all of our mobos
ftfy

I'm not even looking forward to the trainwreck anymore

>>60272102

This.

If it doesn't absolutely shit on 10*0 series at every price point it's a fucking waste.

>>60271822
don't worry guys i'll bite the bullet and buy one for the daily /guts/ thread besides all I play is dos games what more could go worng?

its not like i'm a poor fag i'm just bored with Oh its another nvidia card meme.
look nvidia is okay brand but i'm tired of the same old same old.

So you cucks just sit back and wait (tm)
they say it will barely beat a GTX 1070 and some say a GTX 1080.

Only a dumb arse would think it will beat a GTX 1080ti let alone a titan Xp though

besides it will be fun just blow a load of money for no reason other then shits and giggles hehehehe.

>>60271822

do we know anything about TDP?

Will the 1070 competitor draw less than 200W?

>>60272490
Hopefully not. Don't want poor fags on my throne of my meme gpu.
I have to be unique, Can't let the poorfags have all the fun.

File: TK_doom.png (95KB, 215x301px) Image search: [Google]

95KB, 215x301px

>>60271822
http://www.tweaktown.com/news/57418/amd-radeon-rx-vega-less-20-000-available-launch/index.html
>AMD will be releasing its next-gen Radeon RX Vega family of graphics cards in the next few weeks, and now I've had an exclusive industry source who has told me that AMD will have only a handful of Radeon RX Vega graphics cards at launch.
>I've been told that there will be less than 16,000 cards that will ship in the first few months after it launches, something that will come down to the HBM2 used on the card. HBM2 is in extremely limited supply, and is expensive to use - and since there's not enough, that scarcity is driving up the production costs of the card - and will see AMD only having 16,000 cards or so in the months post-launch.
>If this is true, AMD could be in a very rough spot with Radeon RX Vega - especially if it was to deliver on performance. There are hundreds of thousands of thirsty Radeon fans that want a next-gen graphics card, and the hype train for Radeon RX Vega is simultaneously withering out - and burning hotter than the sun. Personally, I want AMD to hit a home run with Radeon RX Vega - but at the same time, NVIDIA has completely secured the high-end market with the GTX 1080 and GTX 1080 Ti.
>We can expect a family of cards, with the latest rumors on performance of what should hopefully be the GTX 1070 competitor in RX Vega form above. It's a prototype card, and a rumored benchmark run - but, if it's true - how many of these cards would fill up that 16,000 quantity if there are other higher-end SKUs made available in the Radeon RX Vega family.

RIP in peace? Did we just get BTFO?

>My dad works at Nintendo: the source

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

156KB, 633x758px

>>60274371
NOOOO

>>60274371
sounds like bullshit to me

>>60272609
no worries, you will take the place of the poorfag ;^)

File: TZ509.jpg (274KB, 1296x968px) Image search: [Google]

274KB, 1296x968px

>>60274371
>tweakertown
>I've had an exclusive industry source

That would be his dealer, amirite?

>>60271822
wait™. We should wait™ lads.

>>60274371
This is the dumbest shit I've heard all day, do these guys know how little 16k chips is? Do they even know how much DRAM wafers Hynix makes in a day?
Even if their DRAM had 1.7% yields and they're bankrupt, there's more than enough to make some 30000 HBM stacks

>>60274371
>I've had an exclusive industry source who has told me
>I've been told that

VEGA is old and busted. Who waiting™ for NAVI here?

>>60276342
If

>>60276342
If Navi is multiple GPU on a MCM, then I'd personally wait for Navi since that's gonna be the largest performance uplift since G80

>>60276022
for 16bg you need 8x 2gb stacks so only 3750... if its 4gb per stack its only 7500 cards from 30k stacks

>>60271822
>Novideo shills overhyping Vega so they can fuck with everyone else when it releases and doesn't live up to this absurd hype
nice try

>>60274371
This is probably bullshit but it brings up a pretty good source of debate which is whether or not AMD has any business trying to pioneer a brand new memory tech like HBM2. How can AMD simultaneously position itself as the value brand and also on the cutting edge of a brand new technology? The answer is it's going to be extremely difficult to do without cutting some corners or charging a high price.

File: sk_hynix_hbm2_implementations.png (121KB, 1191x747px) Image search: [Google]

121KB, 1191x747px

>>60276392
Vega has 8GB, so that's two 4GB stacks, each 4GB stack has 4x1GB DRAM modules.

If Hynix can't make 100k 1GB modules in a few month, I wonder what they're even doing making DRAM.

>>60276451
They can, like fuck, Hynix is big and they are used to pioneer new memory types, see GDDR3 and forward.

File: 8r5a4kwxurux.jpg (11KB, 195x200px) Image search: [Google]

11KB, 195x200px

>>60274371
>(((((((sources))))))

yeah sure let's trust them!

http://prntscr.com/f5gqbn

File: 167419.jpg (148KB, 640x1163px) Image search: [Google]

148KB, 640x1163px

So... I should buy a 580 now or I wait for Vega?

File: bulletin20160106Fig01.png (74KB, 600x264px) Image search: [Google]

74KB, 600x264px

>>60276460
Of course they can, DRAM modules are tiny, there's thousands of them per wafer, Hynix puts out over a million wafers PER MONTH

Honestly this Tweaktown shit is beyond clickbait.

>>60276478
And yout budget is? Vega will have 1070 and onwards competitiors only.

>>60276488
And considering HBM chips are small. Basically it's a shitty clickbait.

>>60276489
Vega is too big for 1070 equivalent.
Unless they underclock it by 50%
Which I don't see happening

>>60276507
Cutdown Vega10/Vega11 fits right there.

>>60276517
Assuming Vega 11 comes out the same time as Vega 10.

Also Vega 10, even cut down is still far from the 1070.
They'd have to gimp its clock by 20% and shaders by 20%

>>60276532
Lisa said there'll be multiple Vega SKUs on launch. I'd rather believe her than any clickbait 'articles' full of 'leaks'.

>>60276551
Multiple SKUs can mean anything from 2 to 6, at most AMD can get 3 SKUs out of one chip.

If they have two chips (Vega 11), then they either have from 4 to 6 SKUs ready.

>>60276572
There was 11 Vega-based SKUs listed according to some 'leaks'. Assuming half are enterprise models, we have 5 to 6 models of Vega-based cards. Probably split in half between Vega10 and Vega11.

>>60276507
Better performance then Rx570/rx470 in crossfire should be min target. same or lower, it a failure

>>60271822
AMD to the consumers is like when as a kid you would say a fuckton of stupid things you'll do to for your parents to avoid getting in trouble and then when the heat is off you're like "oh shit I gotta do that stuff"
"HBM, HBM2,VEGAAAAAA"

>>60276572
I want a Vega Nano.
>10-15% slower than big Vega
>twice as efficient

>>60276642
Vega is already efficient as is, Nano will probably have better bins and lower clocks.

>>60276642
It'll only be twice as efficient if AMD doesn't clock Vega 10 to the moon.

>>60276428
>AMD fags have gotten so deluded they think Nvidia is overhyping vega
Nope, it's all on you retards

>>60276641
other people did this?
thought I was the only one

File: 550.jpg (47KB, 600x1066px) Image search: [Google]

47KB, 600x1066px

>>60276489
Yeah, but it won't lower the price of the 570/580?

>>60276724
...Why in the hell would it?

>>60276724
Why?

File: 633.jpg (30KB, 337x599px) Image search: [Google]

30KB, 337x599px

>>60276730
I'm not an expert in this things, but I thought that a release of a new, better product lowers the price of the old one.

>>60276754
That's only for company's that aren't losing 38 million a quarter

>>60276754
Does VW releasing a new Porche GT lower the price of a VW Golf?

Dude, the 570/580 came out literally less than a month ago, they're mid range cards.

>>60276767
o shit

File: images.jpg (5KB, 224x225px) Image search: [Google]

5KB, 224x225px

>>60276771
>>60276767
Mm, okay, I'll get the 580. Thanks for the help.

File: GTC17-WMFG-GTCsite-2000x483-3-1.jpg (83KB, 2000x583px) Image search: [Google]

83KB, 2000x583px

THANK YOU BASED NVIDIA

>>60276868
>memes and more memes, not even woodscrews
Old man please show some VIDEO CARDS, or at least some woodscrews.

>>60276868
Volta or self crashing meme?

>>60276895
Memes. Self-crashing memes, meme learning, meme everything.

>>60276895
I'll be glad if they even mention the former exists, they've literally said nothing about it.

>>60276868
>Memechine learning
>Self-memeing memes

Thank you based Nmemea

>>60277085
What? You want VIDEO CARDS? Enjoy the memes goyim.

>>60277096
Why not VIDEO CARS?

>>60277111
Fuck you Huang.

>>60277111
Russia pls go

>>60276360
>>60276342
Navi is 7nm.

first available in 2019, if we are lucky.

>>60278343
GloFo is going 7nn FinFET DUV in 2h 2018, so we may get Navi in q4 2018.

>>60276451
That's 4 stacks plus the base die, and all the while you have to hope that they are still functional while stacked.

Silicon on interposers have always had shit yields due to that.

>>60278518
Seems way too early if GloFo only starts production in 2H.

>>60278518
>GloFo is going 7nn FinFET DUV in 2h 2018

Which means they'll have it ready for production by 2019 at best, and then you have to assume that the process is actually viable for larger chips.

>>60278596
Nope, actual production, not risk production in 2H 2018

>>60271822
All Vega needs to do is offer 1080ti performance at the $599 price bracket

>>60271822
Why the fuck do idiots still bother with AMD

Everything they release is shit

>>60278704
No, it needs to be at $400 or it'll never sell

>>60276439
>>60274371
This reeks of retard.
AMD has the patent for HBM.

>>60278596
That's actual production. Risk production is earlier, probably q2 2018. Remeber that it's basically IBM's 7nm process, the one they were developing for quite some time.

>>60278796
Except nvidia uses HBM2 in their Tesla cards? I don't think nvidia is willing to give and anything in licensing fees, are you sure AMD owns HBM IP?

>>60278796
Reminder that NVIDIA backed HMC. The one that bombed. Hard.

>>60278836
Yes, AMD and SK Hynix own HBM patents. Samsung licensed it.

>>60278811
Risk production is Q3-Q4 2017 for Glofo
It was originally supposed to be Q4/Q1 but got pushed a quarter earlier due to favorable results

>>60278887
Oh well, things look even MORE grim for Intel.

>>60278837
They backed it because they're retarded.
But look at the millions and millions of HBM stacks that are made. From AMD's patent.

>>60278860
We might see HBM in Kabylake-G according to Hiroshige Goto

>>60278971
>We might see HBM in Kabylake-G according to Hiroshige Goto
It'll probably use EMIB, which is interesting. Also Knights Mill will use HBM.

>>60279004
He said it will, he also said it might feature a radeon GPU.

>>60279028
>Intel GPU with Radeon GPU
PAJEETED.

>>60279028

I doubt that. What i'm more inclined to believe is it will use technologies AMD holds licenses and patents to (rather than going to Nvidia) because generally speaking anything new on the GPU front is under ownership of either AMD or Nvidia.

>>60279077
>might


http://pc.watch.impress.co.jp/docs/column/kaigai/1054618.html

>>60279142
AMD licenses its GPU core though? Matrox for example uses a certified AMD GPU with a Matrox sticker.

So what will Intel do without process lead?

>>60279193
Try to hold out until 2021

>>60279142
Goto knows his shit and is a Japanese Kyle Bennet, the package size of Kabylake-G is very large so there's clearly something going on there.

Also nobody said Intel would get Vega IP, Polaris IP would be enough.
At least they'll have working 3D drivers

>>60279216
Then what if their new arch is second Itanic?

>>60279263
Refresh Skylake again

>>60279275
But even the best goyim won't buy another Skylake in 2021.

>>60279290
Refresh Icelake again!

""""New Architecture""""

>>60279299
Wow Intel so innovating, much better than fucking poojeets from poomd.

>>60279299
Is it gonna be another 25% IPC improvement Skylake was touted to be?

>>60279319
No, only "15%" for you, good goy.

>>60279339
>>60279319
http://www.redgamingtech.com/intel-skylake-benchmarks-surface-15-ipc-gains-on-skylake-s/

>>60279077
EMIB is superior to TSV and easier to implement.

>>60279375
Ah, leaks"

>>60279388
On paper? Yes. We don't know it's real performance yet.

>>60279388
EMIB is point to point, interposer/TSV can wire lines anywhere on and off the die.

But no doubt it's cheaper, but will you ever see that as a consumer?

Any Intel's "innovation" looked good on paper but it sucked some giant horsecocks in real life so I am very skeptical about EMIB.

File: 6.jpg (178KB, 1333x750px) Image search: [Google]

178KB, 1333x750px

>>60279440
They wouldn't use it if it was shit, but its superiority is no doubt Intel blowing hot air, like everything it'll have its drawbacks against TSV

This is Intel's own slide, who knows how much that silicon bridge can provide bandwidth, how scalable it is.

>>60279503
Doesn't look very scalable though, while you can make a interposer as silly size as you want(Fiji 1000mm^2 give or take a few) since it's a metal layer with no logic, to connect more stuff with EMIB seems like can get ugly quickly.

>>60279503
>They wouldn't use it if it was shit
>what is Optane/MIC/Netburst and successors
Intel is FAMOUS for promising glory and then fucking off to suck gigantic amounts of raw cocks.

File: 7.jpg (105KB, 1000x563px) Image search: [Google]

105KB, 1000x563px

>>60279540
And here we have EMIB in action(on paper), it's nice for these higher node parts will improve the margins of the product but you'll be crazy if you think the chips will be cheaper to buy.

>>60279564
Ghetto CPU

File: rageakko.png (41KB, 194x255px) Image search: [Google]

41KB, 194x255px

>>60279564
>Let's just slap a bunch of random shit together and call it a new architecture

Fucking. BRILLIANT.

>>60279648
Hey, it's way more interesting than Skylake +2% IPC + 5% faster iGPU

>>60279663
Still boring. But hey, AMD is going to MCM GPUs.

>>60279663
Drying wallpaper is more interesting than that.

File: 1.jpg (426KB, 1569x841px) Image search: [Google]

426KB, 1569x841px

Think it would be possible to connect external memory with infinity fabric on steroids(200GB/s?)

>>60276868
So basically Novidya intend to steal AMD's thunder again. They are shit scared that AMD will do what it has already done to Intel with Ryzen. They really are a bunch of faggots.

>>60279727
What thunder? That novideo conference is another rounds of self-crashing memes.

>>60279727
They're going to announce their self-driving car went 15 more feet before running into a tree/pedestrian/building.

>>60279689

> MCM GPUs

Fantastic ideas on paper until you realise the issues with building a fabric fast enough to handle the communication. Plus thew fact your groups of cores are now physically further away which gives you a hard lower limit of how fast communication can be.

>>60279727
GTC is a Nvidia sponsored event, AMD doesn't appear there from what I recall.

>>60279778
It's just the timing seems odd. Nvidia and AMD always seem to hold events around the same periods.

>>60279767
CPUs are far more sensitive to latency and AMD stuck 4 of them on a MCM, they have no problems with latency.

GPUs are much more bandwidth hungry though, so I imagine a solution to that would be some big, fat, metal layer with plenty of bandwidth with control nodes like the GMI links on Zen for sync

File: IMG_2268.jpg (709KB, 1920x1080px) Image search: [Google]

709KB, 1920x1080px

#readyforvega

>>60279767
>Fantastic ideas on paper until you realise the issues with building a fabric fast enough to handle the communication.
They designed IF for a reason. It's EXTREMELY scalable.

>>60279828
Look at the total bandwidth of a CPU and a GPU, IF in its current form (50GB/s and probably over 100GB/s+ on Naples) cannot handle two GPU dies due to massive bandwidth constrains.

IF is the solution for it, but not the one we currently see, there's probably a higher latency/higher bandwidth version of it as well or on 7nm they can clock it to Netburst speeds and increase its data path sizes

AMD's Fury/Fury X/Nano GPUs mostly GONE from newegg.
Either taken down or out of stock, it's likely a EoL notice was sent recently.


https://www.newegg.com/Product/ProductList.aspx?Submit=ENE&DEPA=0&Order=BESTMATCH&Description=amd+fury&N=-1&isNodeId=1


Vega is dropping very soon.

>>60279896
Good.

>>60279896
Just consumer Fiji stopped production, enterprise cards are still there

>>60279806
IF had a 2:1 debug mode on engineerings boards, so it's very much possible they'll clock the hell out of it

>>60280070
They WILL clock the shit out of it in Naples.

File: 1492136279500.jpg (282KB, 1366x2048px) Image search: [Google]

282KB, 1366x2048px

>>60276439
Look at the Ryzen CPUs. AMD isn't a "value" brand anymore.

In the era of the Athlon XP it absolutely crushed the Pentium 4 in terms of performance per dollar. The same was true for the Phenom II.

The Ryzen's on the other hand are barely equal in performance per dollar compared to Intel.

I wouldn't be surprised if VEGA is priced like a nvidia 1080 while providing the performance of a 1070.

>>60276451
>>60276460
Keep in mind that any new memory type will be expensive and in limited supply early on. Ramping up production takes time, specially these times where smartphones are the priority. Just look at those expensive DDR4 PC memory prices we have right now. DDR4 is no longer a brand new technology but PC memory isn't the priority and I'm guessing special graphics card memory isn't either.

File: AMDRyzen1800XIntel6950X1_0.png (29KB, 617x837px) Image search: [Google]

29KB, 617x837px

>>60280329
>The Ryzen's on the other hand are barely equal in performance per dollar compared to Intel.
Oh, an 1800x is the same cost as a 6950x now? No one had informed me of this.

File: Screenshot_20170418_142236.png (55KB, 804x1036px) Image search: [Google]

55KB, 804x1036px

>>60280329
>>60280354
Or to a 6900k?

File: [Nofunloli] Milky Holmes Fun Fun Pearly Night ~The Gift of Ken and Janet~[5E6E6D13].mkv_snapshot_18.58_[2017.03.05_18.25.20].png (710KB, 863x720px) Image search: [Google]

710KB, 863x720px

>>60280329
>The Ryzen's on the other hand are barely equal in performance per dollar compared to Intel.
>1600 is best price/performance CPU seen in half a decade
ok then.

File: LRM_EXPORT_20170507_091219.jpg (2MB, 1485x4363px) Image search: [Google]

2MB, 1485x4363px

>>60280329
>>60280354
>>60280393

File: Sys-plus-1.jpg (78KB, 748x469px) Image search: [Google]

78KB, 748x469px

Okay okay, no need to shit up the thread spamming benchmarks, everyone knows Ryzen's MT performance is astounding even against equal core count Intels

>>60280329
HBM has been in production for a good while, HBM2 has also been licensed by Samsung so naturally price is gonna drop.

On the other hand, for the high end HBM vs GDDR price isn't all that much of a difference, we're not talking about $200 SKUs here where $30 is a big deal.
While at the highend you get a smaller card, lower power consumption, and less space spent on the memory controller.

File: 1492344707790.jpg (226KB, 800x1198px) Image search: [Google]

226KB, 800x1198px

>>60276532
>>60276572
VEGA 10 will be released soon, nothing called VEGA 11 is even remotely close to release.

You can tell by looking at drm-next and MESA. There's code for VEGA 10 in place which means it will be available pretty soon. There's no code for any other upcoming card in or anywhere near the Linux kernel or MESA.

>>60280521
Then there's no GP104 competitor, Vega 10 is far too large to compete with those.

>>60280521
Who's dick do I have to suck to get at her feminine penis?

File: 1492148142468.jpg (459KB, 1280x1919px) Image search: [Google]

459KB, 1280x1919px

>>60276724
No, not if you go by history. AMD has a history of releasing a generation of rebrandeons and a top-end card that is way more expensive, just like they are doing now. They are not about to let you or your family have access to 570s or 580s any cheaper. And you may or may have noticed that 470s and 480s are not on sale anywhere or priced cheaper than before just because the rebrandeons came out.

>>60278704
They are not going to let you have that.

When is Intel gonna announce a GPU deal with AMD?

>>60280559
As late as possible.

What if Intel and AMD cooperate and create a consumer 32 core cpu that gets 6Ghz on air, reaches 200 degrees C on stock, ships with water cooling, has a backdoor, costs $250, and that doesn't work with Nvidia GPUs?

>>60280623
>who's in charge of the project

Cooperation failed.
Also impossible on silicon unless you're talking A53-tier cores

>>60280329
>The same was true for the Phenom II
no