Daily reminder: wait™ Vega

Anon remembers the HD2900 almost exactly 10 years ago? I don't think this will be any better.
Just wait.
I think I will finally throw my principals overboard and get myself cucked by Nvidia

Welcom welcom welcom

my name is Raja

>>60608840
3 DAYS TO REVEAL
REEEEEEEEEEEEEEE

File: 69f9979993d63791c84fdee415e942c505249c02056f0330732fa7ec80881829.gif (3MB, 476x341px) Image search: [Google]

3MB, 476x341px

>>60608840
:^)

File: download (3).jpg (9KB, 312x162px) Image search: [Google]

9KB, 312x162px

>>60608937

Like that's a problem...

Overclocked HBM and 1500MHZ clockspeed already put it on Titan Xp performance in Firestrike extreme.

Now throw in architecture refinements, faster IPC and primitive culling with tiled rendering, and it's going to be ugly until Volta launches.

>>60608840
When are we getting more info ?

>>60609759

30th at Computex.

File: 10k.png (646KB, 1414x1095px) Image search: [Google]

646KB, 1414x1095px

>>60609712

File: MKW.png (428KB, 1349x801px) Image search: [Google]

428KB, 1349x801px

>>60609814
:(
Think it's time to OC the 8350 a few hundred.
The card is at 2070Mhz. +420 on the memory.

>>60609982

What cooling do you have for the FX? If yiy need to get a good cooler, it might be worth it to just start saving for Zen+ and Vega 2.0/Navi.

>>60608840
#justW8 op

>>60610109
H100i V2. I'll be getting zen+ and OCing the shit out of this build with peltier plates and a hard mod on the 1060.
8000Mhz being the record for a 1060 on ln2. Should be impressive what it can do.
Only issue is the chip might be fucked due to a bad mobo it was first on, getting to stupid hot levels without post. Have tried some OCing, but couldn't get anything that stable. But I was pushing north bridge stupid hard.

>>60608840
>AMD still doesn't have a cuda like framework

Don't wanna go nvidia but they are so damn stupid for not developing a framework which does exactly what cuda does.

>>60610283
>what is
>ROCm
I dunno.

>>60610388
>no neural network framework uses it
And I didn't see that tensorflow and co are planning to integrate it. Maybe I should have mentioned that in my post.

>>60610438
>uses it
Yet? AMD is launching MIopen/ROCm with Instinct line launch. Aka Just Wait.

Yesterday, a friend asked me to install a new Nvidia GPU in his PC.
I had to log in with an e-mail address just to update to the latest drivers.

Was thinking about switching to Nvidia for my next GPU, but not anymore.

>>60609712
The Vega cards will barely be able to keep up with a 1080.

They have no hope of competing with a 1080TI or a Titan.

AMD are finished.

>>60610583
How so?

File: 1481185222348.jpg (45KB, 293x301px) Image search: [Google]

45KB, 293x301px

>>60610583
That read with such low energy. Almost like a broken person wrote it.

>>60610583
Personally I'm expecting 1080ti performance, not that it matters, volta is just around the corner.

>>60610643
Volta is bigger Pascal.

File: 1478668306532.gif (1MB, 480x480px) Image search: [Google]

1MB, 480x480px

>tfw you were waiting for Vega for over a year
>G1 Gaming 1080 goes on sale for $360 with coupon
>willpower gone
I tried so hard lads.

>>60610501
>I had to log in with an e-mail address just to update to the latest drivers

this is bullshit

>>60610632
That's because I am a broken person.

My last three GPUs were a 7950, an R9 390 and a Fury. I even have a Freesync monitor.

I've given up hope of AMD ever releasing a GPU that can compete with Nvidia's high-end again.

>>60610643
nah between 1080 and 1080 ti

>>60610666

Six months you mean?

>>60610643

Titan XP performance at 1080ti prices.

>>60610283
OpenCL?

>>60610743
lol

File: GPU-Z.2.1.0_2017-05-27_11-11-22.jpg (157KB, 392x480px) Image search: [Google]

157KB, 392x480px

how much longer am i gonna have to wait? i'm dyin over here.
was gonna get a 480 but then 580's were just around the corner when i decided i needed to upgrade and then the cryptofucks ruined that.
i might actually buy into the jew and buy a 1060 at this point.

File: 1481126306019.jpg (133KB, 1024x1024px) Image search: [Google]

133KB, 1024x1024px

Doesn't their $600 card have to come close to the 1080ti to even be in the running? I don't really get how they compete with their $500 card either, given that it's the current MSRP for the 1080, and sales around $450 are excessively common

How does AMD swing this?

>>60610801
>How does AMD swing this?
By curling up in the fetal position and bawling their eyes out.

>>60610801
By giving better price/perf, Evergreen-style? Fuck Vega is designed by the same pajeet who made Evergreen.

>>60610467
I still doubt that older cards will get support. If I knew that the current 500 series will get support for deep learning then I'd get a rx560 immediately.

>>60610840
>I still doubt that older cards will get support.
Why? It's a fucking software stack.

>>60608840
Vega 10 will be a $700 card (HBM alone already costs $160, and the chip is large + has an interposer = poor yields, higher cost).

I can't afford that, I'm already waiting for Vega 11.

>>60610801
What's she from? It's posted everywhere.

>>60610761

AMD HIP compiler.

https://github.com/ROCm-Developer-Tools/HIP

Take CUDA software that recompiles to C++ code with zero performance loss.

>>60609814
Is this a nitrogen suicide run? Can you do this with just a custom water loop?

>>60610906
>Is this a nitrogen suicide run?
Yes. It's also probably the first time LN2 stats are actually fucking useful.

>>60610778

I would laugh too.


Wait for Volta? No thanks.

>>60610779
if you play @1080p you will be just fucking great off with a 580 holy shit.
The real reason to actually buy Vega is to play our gay men @4K which i want to do.

>>60610801
>How does AMD swing this?
Better 4k performance thanks to the massive memory bandwidth

>>60610933
the 580's are currently like $40 overpriced because of low stock so i ain't gettin one.

>>60610877
>What's she from? It's posted everywhere.
Kronprinz original char or maybe someone from his manga,

>>60610906

Suicide run on LN2.

http://forum.hwbot.org/showthread.php?t=142320

He took a base Fury, unlocked it to a Fury X, and overclocked the hell out if it

We already know Vega hits at least 1500mhz clock speeds as well.

>>60610978
Not to mention that HBM can clock up too.
HBM can run at 600-700 clocks

>>60610643
>volta is just around the corner
>Q2 2018

Yeah, ok.

>>60610501
Don't forget to disable the telemetry service. lol

>>60611061

His LN2 run overclocked HBM ans gavebit an effective 1TB/s bandwidth.

Sounds an awful lot like A Vega pre-architecture change to me.

File: 1359407888020.gif (3MB, 800x800px) Image search: [Google]

3MB, 800x800px

>>60610779
>cryptocucks completely ruin the mid-range GPU market
>AMD can't ramp up production in anticipation of cryptobubbles they can't even predict
>they still do nothing to stop it because they're that desperate for sales

>>60611177

Why stop it? A sale is a sale. Especially to spoiled ass gamers who give AMD no respect. They only got involved when Newegg was selling 290Xs for 700 dollars and wanted a cut of the action.

>>60611177
I heard Vega has programmable shaders which would allow them to block most shader computation with firmware if they wanted to.
I hope they cuck the buttcoinfags like Nvidia did.

>>60611272

Polaris has the same programmable shader tech.

>>60611230
There are plenty of reasons
1. Those sales don't appear in GPU benchmark and Steam surveys, making it look like AMD has less marketshare than they really have (technically it's true, nobody will ever see them outside of second hand sales years later)
2. Game developers won't be convinced to optimize for or give a shit about Radeon GPUs when they know half the cards are in the hands of cryptofaggots.

>>60611284
it doesn't though.

>>60610671
Post probably meant "had to create account with email to use GeForce Experience".

It's definately a shit practice, but not mandatory.

>>60611082
Why would they wait another year? They already have a 12nm volta card rolling out in Q3 as the GV100

>>60610662
thread execution, thread scheduling, core layout, memory controllers, and ISA are all completely new on volta

>>60611458
>12nm
lol

>>60611458
>thread execution, thread scheduling, core layout, memory controllers, and ISA are all completely new on volta
Lmao.

>>60608840

vega won't even beat intel HD graphics.

>>60611458
NVidia marketing squad is here.

>>60611501
as far as market share goes, Intel HD graphics beats Nvidia too.

>>60611542
Endless flood of crappy laptops to play Mineycraft and ASSFAGGOTS on.

>>60611539
I've got an 8350 with a fury x, ill be buying vega no matter how it turns out. I'm just basing this off the information ive seen released for volta. I mean the GV100 is an accelerator card not mean for consumers, but it shows that the volta architecture is ready for production, and that TSMC 12nm FFN is ready also, so shouldn't we assume that nvidia will rush there consumer cards to market if vega turns out to be a threat?

Heh

>2 8pin 375W

AYYMD HOUSEFIRES

>>60611659
They need GDDR6. Which may be ready in Q1 2018 but means they won't be able to ship a product until Q2.

File: thermi.gif (592KB, 500x500px) Image search: [Google]

592KB, 500x500px

>>60611676
>pin count equals max load
This is what retarded NVidiots actually believe.

I have a GTX 750 ti with a 6 pin connector. Guess that means it's a literal housefire, lol.

Pic related, the real king of GPU housefires.

>>60611659
>12nm FFN
It's 16nm FF+ rebrand.
>it shows that the volta architecture is ready for production
Of course it's fucking ready, it's a bigger Pascal with ASIC strapped onto it.
>>60611676
Wow so every 2x8pin card is now a housefire?

>>60611659
You should really consider upgrading your CPU before anything else.

>>60611734
Not him but all high end reference Nvidia cards are 6+8

>>60611765
And the Vega (actual Vega card) Pajeet showed also had 6+8 fucking pins. Then said Pajeet said they strapped moar pins for whatever funny reason.

>>60611765
>>60611816
It's extra headroom for OCing you dips.

>>60611843
No shit, faggot.

>>60611843
Oh, thanks but why they did not show the real one?

>>60611873
Which real one?

>>60611888
The 8+6 version

Why is it so hideous?

>>60611959
Why is your face so hideous? Shut up.

>>60612019
t. Rajeesh

>tfw used 1080s are approaching $550CAD (~$400USD)
I'm an AMD fanboy, but fuck even I'm thinking of switching teams

>>60611676
I like how there's literally just one retard spamming this shit on /g/ for months. He must be really desperate to make it into a meme. https://archive.rebeccablacktech.com/g/search/text/%22AYYMD%20HOUSEFIRES%22/

>>60610467
>Yet? AMD is launching MIopen/ROCm with Instinct line launch. Aka Just Wait.
Elaborate please. Will AMD really release a framework that caffe and tensorflow for example can adopt which will also work on the gpus of their polaris architecture? Will they launch it with vega?

File: 1492998382933.jpg (223KB, 555x555px) Image search: [Google]

223KB, 555x555px

>>60612211
>tfw i have a 980 ti

>>60612269
980tis are still $350+ CAD where I am, makes no sense considering a 1070 is $100 more new

>>60612211
There's no way AMD beats these sale prices for a long time after launch. It's the lower models that are going on sale close to $400, but most of the decent ones are mid 4s

Do they even win price/perf with this lineup?

>>60612211
>ever buying anything with someone else's cum in it

>>60612718
didn't stop me from buying ur mom

>>60612711
current as in rx480/580? you can get a 480 for as low as $200 if you buy the shittiest version, so I think it wins there (slightly faster than a 1060)

when I was younger, I didn't care about perf/watt, but now that I actually have SOME money and pay for hydro, AMD's lack of efficiency is really hurting its appeal. the extra few dollars in electricity isn't a deal breaker, but when the gpu increases the temp in my room by a few degrees, that's where I draw the line

>>60612627
The 980 Ti is the exact same as a 1070.

>>60613028
>worse efficiency
>worse overclocking
>less VRAM
>slower in most games (if not all)
cmon bro

>>60613130
https://youtu.be/wmfMfhRI6To