AMD Vega 11 GPUs Entering Production, Vega 20 Coming On 7nm

>>62329371
What's wrong with company following the roadmap?
Also,
>wccfturd

>>62329390
>release vega 10 on low capacity
>start producing vega 11
>release vega 11 on low capacity
>start producing vega 20

why

>>62329416
Because these are different dies for different market segments?
Are you fucking retarded?

Heck, if Vega20 truly is 7HP, it will use different packagin, even (CoWoS).

>>62329371
if they get the power problems and shitty performance under control ill jump on it

Considering how quickly miners gonna jump on it thou we're gonna have to get in quick

>>62329522
Vega11 is an entirely different die.
And you will never see Vega20.

>>62329371
Is "7nm" really 7nm or is it re-branded 14nm?

>>62329560
>Vega20.
>32gb of HBM2
wew
>>62329604
its just gonna be Vega 10 on higher clockrates with better power

thats it

>>62329604
7HP is pretty similar to 7LP, just clocks lower. And it's ~10nm BEOL

>>62329611
Vega20 is an HPC Vega. It has nothing to do with Vega10.
At all.

They should focus on making Vega work before shitting out MOAR SKUs.

>>62329629
why are amd dumping vega 10 already? Shit yeilds

>>62329829
They are not dumping it you moron.
Vega20 is an entirely different die for HPC. Think of V100 competitor.

>>62329416
They have shitloads of Vega10 and HBM2 on hand
What they don't have is interposers.

Vega11 is GDDR, so it doesn't need any stinkin' interposer.

>>62329848
Interposers are dirt cheap and easy to fab.
They have a LOT of them.
And Vega11 will also use HBM.

>>62329762
>Hardware teams should focus on Writing Software
Que?
I agree AMD shit the bed with their lack of drivers on VEGA, but the hardware people can't do anything about that, not should they sit on their hands whilst the poo-in-loos curry up some new shit.

>>62329857
Wrong on both counts.

>>62329867
But king currynigger is THE hardware dude.
>>62329874
Interposers are dirt cheap, fucko. And Vega11 uses HBM, since SPIL got packaging orders for it.

please, no HBM meme on vega11

>>62330309
What's wrong with HBM?
And yes, Vega11 uses HBM.

>>62330337
price and availability

>>62330350
>price
? No one knows the price for 4-Hi.
>availability
Both HBM vendors are ramping HBM2 production (and cutting G5, he-he).

I guess Vega11 will use Samsung's low cost HBM.

Where did all the shitposters go?

>>62329371
Do Chinese have some physical disabilities their photos always have SUCH FUCKING SHITTY quality? And at the same time
> 454 kb 1620x1080
WOW. Expert level of optimization.

>>62329560
>Implying there won't be a 4x4gb or 4x2gb version with 1 TB/a memory bandwidth as a Halo gaming product.

>>62331333
Vega is so bad than we don't need shitposting.

>>62331463
How is it even bad?
>>62331441
Nah.

>>62331479
Why not? It would be great for mind share. Make like 3 of them and charge out the ass for them. You don't have to sell any, just gain the mindshare of having the most powerful gaming GPU.

File: 1504925577443.png (573KB, 708x521px) Image search: [Google]

573KB, 708x521px

>>62331479
>How is it even bad?

>>62331512
7HP will have like 1.7-tier yields for anything sizeable.
It's 1.7 vs 1.7 competition.
>>62331514
So what's bad about uarch?

>>62331479
>match with non OC GTX 1080
>1.5 years late...
>draw more power
>12.5B vs 7.2B transistors
>484 mm^2 vs 314 mm^2 die
>HBM2 vs GDDR5
>no headroom for overclockers
>fake launch price
>still no AiB
>wait™ drivers and good bios is not even a meme

bitcoin miners will buy them regardless
might as well throw something new out there

>>62331655
>match with non OC GTX 1080
>1.5 years late...
>draw more power
>12.5B vs 7.2B transistors
>484 mm^2 vs 314 mm^2 die
>HBM2 vs GDDR5
Irrelevant.
>no headroom for overclockers
Even more irrelevant.
>fake launch price
Hi Steve.
>still no AiB
Who cares?
>wait™ drivers
Now that's an actual complaint.

>>62329892
Yeah but he lives in Canacuck and has for at least 15 years.
He can actually poo In a loo.

>>62331463
t. huang

>>62331679
>R300 was in 2002
Time sure flows.

>>62331678
>no headroom for overclockers
>Even more irrelevant.

Holy shit, AMD shills are now okay with Anti Consumer practices.
How the fuck is FREE performance irrelevant?

>>62331678
>wait© 1.5 years for the same product than the competions is not irrelevant
>the relative poor performance is irrelevant
>the production price is irrelevant
>the heat, TDP and the power consumption are irrelevant
>the low oc capacity is irrelevant
>good customs cards with better cooling, noise and vrm is irrelevant

What the hell is relevant for you?

>>62331809
Tell that to Boost 3.0.

>>62331824
What's your point?
Nvidia GPU overclock great since Maxwell 2.

>>62331832
You can't overclock Pascal properly because fucking Boost 3.0 already does it for you.
You win like 100mhz tops.

>>62331860
Not him but:

https://www.techpowerup.com/231751/evga-geforce-gtx-1080-ti-overclocked-to-over-3-ghz-under-ln2

File: Screenshot_20170822_at_9.52.53_AM.png (68KB, 607x621px) Image search: [Google]

68KB, 607x621px

>>62331809
>Undervolted and overclocked Vega 56 ties a fucking 1080 FE
>No headroom for overclocked
You fuk'n wot m8?

>>62331884
>shuntmodded card under the LN2
We're talking regular OC here.

File: 1449868285351.jpg (86KB, 600x399px) Image search: [Google]

86KB, 600x399px

>>62331678
>still no AiB
>Who cares?

>>62331892
By "headroom for overclocking" he means BIGGUR NUMBERS, while in reality it's like 100-150mhz over boost 3.0.

>>62331911
No one cared about that when Pascal launched.
So who cares?

File: untitled-1.png (59KB, 682x833px) Image search: [Google]

59KB, 682x833px

>>62331892
This is nothing...

>>62331860
BOOST is and always has been very conservative you can easily get 100 to 200mhz extra out on top of it.
Do you fucking idiots believe BOOST pushes the GPU to the maximum? don't be fucking stupid use your god damn brain. it's like you've never overclocked a GPU yourself, if BOOST was such an aggressive implementation it would cause instability, crashing and artifacting as each GPU needs to be fine tweaked.
Don't go with the fucking you can only get 100mhz! because that's not how it works and we can go both sides, a AMD R9 290 AIB comes clocked to 1040 mhz and at best if you're lucky can get it to 1150mhz, the average OC for a 390 is a 1170 mhz with 1200mhz being really good and the same goes for the Fury X, meanwhile Nvidia Maxwell are clocked aroung the 1300mhz and can get up to 1500mhz in average, Pacal clocked to 1500mhz can boost up to 1700+ maybe and can easily go for 1900mhz.
The only thing limited Pascal is the power limits.
Sometimes I wonder why the fuck you people who don't actually buy PC hardware as a hobby or are enthusiast at all come to /g/

>>62331929
>super ebin OC on aftermarket card yields 400 FS points
GPU OC is a fucking meme.

>>62331939
>Boost 3.0
>conservative
?
It literally causes reference shitblowers to throttle.

>>62331927
Because the GP104 is much more efficient than Vega64/56: you don't need a bigger cooling system or more voltage regulator module on the PCB

>>62331969
>"efficient"
It's a smaller fucking die.
You can trim Vega all the way down to 2.5K ALUs and it'll perform the same.
And it will be as "efficient".

>>62331966
>HURR DUUR WHAT IS POWER LIMITS
Now you really showed us how fucking ignorant you are about this subject.

>>62331947
>now GPU OC is a fucking meme

b-b-b-b-but what about Vega 56?

>>62331985
LITERALLY
THROTTLES
REFERENCE
SHITBLOWERS
>conservative

>>62331999
Meme.
It's not working properly anyway.

>>62332000
Stop posting you litearlly retard and go learn something, power limer throttling happens because the GPU is power chocked by the BIOS, a AIB model with higher power limit will keep the clocks higher.

>>62331984
>It's a smaller fucking die

This is my point>>62331655

Vega die is too much big, this is a very relevant point and I don't talk about the price.

>>62332032
Vega10 competes with GP102, not GP104.
Vega11 (probably ~310-320mm^2) will compete with GP104.

>>62331860
>You can't overclock Pascal

http://www.overclock.net/t/1624521/official-nvidia-gtx-1080-ti-owners-club
Plenty of reference 1080Tis going to 2000mhz and over and this is Nvidia's bigger Pascal chipset.

>>62332043
If Vega10 competes with 1080ti, this is the worst card ever made by AMD

>>62332030
It's not hitting the power limit.
It literally throttles after several minutes of Boost 3.0.

>>62332065
It only competes with P6000 right now.
It doesn't work properly anywhere else.

>>62332069
Holy shit you are really retarded.
If it's not throttling from power limit then it does for temperature, which an AIB with higher power limit and better cooling can handle better.
Do you /v/ morons think you can just come to /g/ and try to discuss topics with enthusiasts?

>>62332096
>moving the goalposts
It throttles because of fucking BOOST.

>>62332110
lmaoing at you right now.

File: SAPPHIRE-Radeon-RX-480-NITRO-4.jpg (60KB, 800x639px) Image search: [Google]

60KB, 800x639px

>>62331947
>GPU OC is a fucking meme
Say than to the NitroSapphirefags.

bump for more butthurt

File: Stupid.webm (2MB, 960x540px) Image search: [Google]

2MB, 960x540px

Why would they waste any money on an architecture they know is shit instead of moving to the multi-die navi?

>>62331947
pic related

File: feelman_hands.png (8KB, 509x619px) Image search: [Google]

8KB, 509x619px

>>62333507
>multi-die

>>62333534
That has to be what Navi is, since the original map said it's "Scalable" and we know they're likely using the infinity fabric in their raven ridge APUs.

>>62333507
Vega, specifically the way it chews geometry, is the basis for Navi. It needs to scale with moar ALUs.

>>62331677

>bitcoin miners
>gpu

kys

>>62332069
Pascal throttles hard hased on temp and without BIOS modding, which is locked the fuck down, you can't change the temp limits.

I have a pascal card and every 5C it drops a 12mhz clock bin up til about 60c. This card hits 1963mhz on stock voltage. It clocks all the same regardless of voltage or power limits, in fact only in the most extreme cases of synth tests, OCing the ram to 9000 does it even break the stock 100% power limit. By 1%
Pitiful user control and pitiful locking down.

can someone break this down for a lay man

>Not enjoying your undervolted/overclocked RX Vega 56 flashed to 64 giving GTX 1080 performance for £100 less.