AMD VEGA

>>60320994
>will draw less power than Pascal
Since the rx580 uses the same amount of power that A GTX 1080ti, I'd say not a chance in hell.

AMD would need to more than double performance and cut power draw in 1/2.

>>60320994
Like I said, Vega is DOA

File: amd.png (42KB, 653x726px) Image search: [Google]

42KB, 653x726px

>>60320994
We'll just wait™ for navi.

File: 1494293505786.png (2MB, 1418x1160px) Image search: [Google]

2MB, 1418x1160px

>>60320994

IT'S OVER

AMD IS FINISHED

It took 3 posts for a wojak to be posted. When are you going to get banned again?

File: paladin.jpg (230KB, 1280x720px) Image search: [Google]

230KB, 1280x720px

>>60320994
Vega is DOA. Only retarded fanboys will buy it.

File: 1489487066862.png (316KB, 882x758px) Image search: [Google]

316KB, 882x758px

IT'S NOT FAIR BRO'S

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

156KB, 633x758px

>>60323874
>yfw

Volta is bigger Pascal. Wtf NVIDIA.

>>60320994

"""""""Coming out""""""""

Next year.

Why is people pretending Volta is anything but beefier Pascal?

>>60323964

It's actually Q3 and oem designs in Q4, but still, why release so much information this early? I'd think nvidia either

a) wants as many corporate pre-sell deals as possible
b) set up something to shit on vega's party right before it comes out

Still, an 800mm2 behemoth means a lot of chips not fitting the bill. I'm predicting sizable profit margins to make up for the defective ones.

>>60324034

B. The answer is always shit on AND.

They will always try to take the wind out of AMDs sails. It's why we have had to wait so damn long for Vega because AMD was playing a game of brinksmanship with Nvidia to see who would show their hand first on the high end. Raja Koduri called their bluff and forced Nvidia to show everything full GP100, and 1080ti. Vega is either going to be revolutionary or it's going to be a massive failure. There is no in between at this point.

>Built on shitty Samsung process at GloFo that is well known to be power inefficient, just look at RX 580 consuming more power than a GTX 1080 Ti
>draw less power

TOP KEK

>>60324989
So you know what TBR means?

>>60325018
Yes. To be retarded.

See: you.

>>60325060
(You)

>>60321068
Polaris was just a minor tweak from the previous generation. They pretty much put no effort into it.

Now Vega on the other hand is a massive architecture change.

Volta
> die size as big as a mansion
> power consumption of an entire nuclear reactor
> costs several cubic acres of money

>>60325142
wot? I thought polaris was a custom arch designed for high yield, medium performance, low price chips.

>>60325218
It was half-Fiji with a bit better front-end.

>>60325218
Polaris is a minor GCN tweak on a new manufacturing process.

I think that gaymers are poisining this board.

>>60325154
Amd dual GPU will beat it in power

>>60323735
>>60323888
>>60323902
It's time to banned again, samefag

>>60325267
I think you enjoy having sexual intercourse with people of the XY chromosomal orientation.

>>60325218
The tweaks were on the manufacturing side, made with the experience derived from cheap and efficient 28nm APUs

File: 1494487855012.png (1MB, 1228x1502px) Image search: [Google]

1MB, 1228x1502px

>>60320994
GUYS TELL ME MY WAIT IS JUSTIFIED

>>60320994
>Volta is coming
In 2018?

>>60320994
>Volta is coming

In the form of another 750Ti?
Can't wait, I'll play many games at 720p with that.

>>60325607
>>60325615
No, in Q3 2017. Enjoy your 3 months of shitposting about muh Vega AMDfriends, then it's back to Waiting™ and Rationalizations©.
How long until Waitvi™? 14 months? 18?

>>60325708
So you want me to wait some 4-5 month for bigger Pascal. Nah.

>>60325717
Rev up the excuses, you'll need them come October :^)

>>60325726
Ye ye bigger Pascal is probably longer bibelines of this generation.

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

Right, Q3 2017.
Even 12nm isn't ready yet.
You must be retarded.

>pascal 1070/1080/1080ti come out
>w-wait for v-vega
>next gen volta comes out confirmed to btfo vega
>w-why would i w-wait 5 month for a g-gpu

lamo. amdrones on consistent damage control

>>60325758
>march 2018
>5 months

Can't make this shit up

>>60325758
Volta is Pascal.

>>60325762
>>60325755
They could do another 750ti "launch"

>>60325775
This makes more sense, neither 12nm FFN or GDDR6 are ready for mass production until Q1, unless consumer Volta is still 16nm, then they might make it out a bit earlier but that still leaves GDDR6 availability, so no point in 16nm designs.
More importantly, the 1080ti just released 2 months ago, there's no way they're only having it in the market for 6 months.

File: warp_pascal_volta.png (78KB, 1639x682px) Image search: [Google]

78KB, 1639x682px

So AMD's asynchronous compute advantage is gone as far as Volta's concerned?

Also I'm wondering how will the increased instruction issue throughput and reduced instruction issue latency caused by spiting up the FP32 and INT32 cores effect the performance of upcoming GPGPU oriented game engines. Seems like it could be significant.

>>60326809
>So AMD's asynchronous compute advantage is gone as far as Volta's concerned?
Yep wavy lines >>>>>> straight lines as far as I'm concerned. It's marketing BS until proven otherwise.

>>60322130
>>60323735
>>60323859
>>60323878
>>60323888
>>60323902
>>60325290
>>60325461
>>60325708
>>60325726
>>60325758
43 posts in this thread, 12 of them are trolls/shitposters. This entire board is dogshit.

>>60324884

Looking at Voltas arch it looks like a a step above Pascal for sure, and the Tensor cores look to be pretty damn good, a die shrink as well. The problem I'm seeing is the chip is fucking huge at 815mm holy shit, and we were making fun of Vega at ~515mm.
I think it's going to be a fucking behemoth for AI and deep learning, but the problem will be price. A die the big? Like shit the top end Volta for HPC is gonna be $10k at least. I don't know how they will cut costs to sell Volta for consumers, maybe a different die without tensor cores or something? I don't know we'll see how it pans out.

Vega HAS to be good. I mean top end NON DUAL GPU solution has to be better than 1080ti by 10% or something. I really want it to be good because from an architectural standpoint it looks much more interesting than Volta, and seems like you could do some neat things with it. If Vega is at 1080 level, AMD might as well sell the graphics division.