>yfw Fiji all over again, only worse this time

JUST WAIT (tm)(c)(copyright: AMD, Calcutta)

Ittiitits for content CREATORRRRRRRRRRRRRRRSSSS

>>61198653
Inb4 >>>/v/
Is amd making this card for mining ? Seems like amd is not giving a fuck for games these days.

>>61198677
The RX vega will be based on the exact same card with one HBM module less

File: 1310483412100.jpg (36KB, 413x395px) Image search: [Google]

36KB, 413x395px

>months of AMD shilling on /g/
>keep warning them to not overhype, remember Fiji
>"It will beat the 1080ti just wait!"
>it arrives
>slower than a 1080 with higher power consumption
Every single time

AMDlets when will they learn

these benches are worse than the apparent performance they showed months ago in doom so im still pretty confident that drivers will have double digit percentage impact on the proper final rx vega cards at launch. if it really is this bad though then yeah this is the worst new gpu architecture in decades and utter dwarfs the fermi disaster

>>61198695
Vega FE shipped the slightly modified Fiji drivers.
If only someone tested ROCm on it.

>>61198695
The aren't worse than what was showed at all

>gamers nexus

>>61198695
What are you talking about? You mean this demo?
https://www.youtube.com/watch?v=Ku5OMWYVKSs

Numbers pretty much. Keep in mind this is avg fps and the results might slightly vary depending on the benchmark method. Vega probably had drivers for months considering the massive delays. This is pretty much what you get plus minus 5 fps with non-reference models depending if they will be clocked higher or not.

>>61198758
pretty much match*

>>61198758
>Vega FE has TBR disabled
>bbut drivers were ready believe me gweilo
(You)

>>61198769
Just wait

>>61198777
Yes, fuck me, at least it's not Maxwell-tier delay, it's only two month late.

>>61198790
>not Maxwell tier-delay
>arrives a year after the architecture it's competing with
You're right it isn't Maxwell-tier delay, it's about 10 times worse

>>61198653
>yfw Fiji all over again, only worse this time

yeah, that's fiji's drivers.

>>61198769
You're really delusional if you think AMD is sandbagging their products and makes official PR statements showing benchmarks and fps without proper drivers. I love how you indirectly call them bunch of incompetent idiots just to defend their product. I love how every shill on /g/ suddenly stopped spamming Adored's videos after he pretty much calculated and predicted the very performance you guys are getting now and gave up on AMD's GPUs trying to compete with Nvidia.

>>61198800
Let's not talk how late 6 series was back then.

>>61198769
>AMD ships pro card with broken drivers
>just wait to finish that project.

>>61198805
At least they are showing something instead of nuking a bunch of SKUs are releasing big Polaris.

>>61198814
It's not a pro card. It's a prosumer one. Also early adopters meme edition.

>>61198823

This shit was supposed to come out Q4, last year. Then Q1. Then H1. What is this shit, it had better be good

>>61198653
see
>>61191977

File: pp.jpg (18KB, 248x189px) Image search: [Google]

18KB, 248x189px

>>61198843
>Insider source at RTG

>>61198841
And motherfucking Maxwell was moved from Q3 to Q4 2013 to Q1 2014 for no fucking reason (besides NVIDIA writing drivers for TBR). And yet no one complained.

FUUUUCCCKKK I HAD THE CHANCE TO BUY A 1080 FOR $350. FUCK YOU /G/. FUCK THJIS TROLLING BULLSHIT i HATE YOU SO MUCH

>>61198830
>amd markets pro card without signed drivers
>bluluminum gets the jackasses
>first public vid is a poorfag build hooked up to a tv in a 1 bedroom apartment
this is par for the course

Err most autocad place is using Quadro, most devs are using Titan. What makes Vega so special that people will switch?

>>61198848
Nvidia has marketing to cover their fuckups.

>>61198890
And AMD has a year full of shiny new products to cover any Vega delays.

File: adoredtv.png (112KB, 330x216px) Image search: [Google]

112KB, 330x216px

>When your's company best friend on YT tries to warn you 5 months ago so you can cool down with the hype and realize that AMD isn't ever going to compete with Nvidia's best and you spam his video with autistic screeching because you belong to the worst fan base in hardware industry.

Friendly reminder that gamernexus was also considered "AMD friendly" before Ryzen release in threads we had made by delusional AMD shills that couldn't take no for an answer.

https://www.youtube.com/watch?v=b_ZMOn0X6jw

>>61198902
Ryzen is good at benchmarks, however when you use shit like 7zip actually unzipping your porn its much slower than an i5 or i7. Also it sucks dick with mmorpgs like GW2 or ESO.

>>61198848

Maxwell delivered. That might be the difference. It's one thing to delay and come out with the top card currently used on Steam, it's another if it's a disappointment

>>61198920
>Maxwell delivered.
And Vega is yet to do it.
>It's one thing to delay and come out with the top card currently used on Steam
970 sold due to aggresive marketing and resurgence of ebin pc mustard race.
Not because 3.5 meme is any good.

>Muh gamedevs don't optimize for AyyMD

>>61198953
Muh gamedevs don't optimize for anything.

>>61198848
>for no fucking reason
There was a reason, the reason was that there was no competing product except their own old shit and they didn't want to cannibalize their sales before the production run finished.

>>61198935
970 sold because for its price it offered the best performance in 1080p gaming that was unheard of earlier and also because x60 cards from Nvidia before Pascal had a joke performance so if you wanted to get 60 fps in Witcher 3 etc. 970 was your cheapest choice.

>>61198935

True, Vega still has a chance, but this first look at it isn't great even if it's muddy.

970 sold since it was at a brilliant price point for the performance. People didn't know about the memory thing for a long time.

>>61198935
>aggressive marketing
What about the fact it still shit on every AMD product for about a year and came at a great price point and didn't consume the sun?

>>61198966
Kekler refresh was not fucking supposed to exist, anon.
It's a result of TSMC 20nm node being shit and Maxwell1 being undercooked turd.

File: pp2.jpg (24KB, 400x400px) Image search: [Google]

24KB, 400x400px

>>61198960
>mfw all things considered the "muh Evil worst company in USA for two consecutive years" EA is doing best job of all AAA publishers at optimizing their games.

>>61198981
Mostly because DICE programmers are really-really talented and almost every EA game uses Frostbite.

AYYMD HOUSEFIRES

AYYMD IS FINISHED & BANKRUPT

AYYMDPOORFAGS CONFIRMED ON SUICIDE WATCH

>>61198977
>was not supposed to exist
It existed because they didn't have any competition and instead of pushing for an early release they decided to rehash their old shit due to the fact it'll still sell because AMD lol.

>Muh undercooked turd that outsold every single AMD card top to bottom.

>>61198999
>denial
Okay.

>>61199004
>denial
>Maxwell was a failure
I don't think you understand meanings of words.

>>61199018
Maxwell1 was a failure. It had only one SKU anyway.

>>61199023
This is the face of autism you guys.....

>970 sucks except it sold more than every other card ever but 0.5gb of ram makes a shit lick of difference because autism

>>61198992
True.
>>61198972
I know some retards in here honestly were expecting 1080 Ti/Quadro performance for cheap but Vega can still do really damn well depending on pricing. If they released slightly cheaper alternatives to 1060/1070/1080 they could still take over market since that's what 99% of people buy anyway.
>>61199004
Stop being retarded anon. You can shill and cry about AMD's superior hardware all you want but it doesn't change fact it was during the maxwell that AMD lost another 15% of market share.
>>61199023
So who cares then? 750 Ti still sold among poorfags I think.

>>61199049
3.5 meme is Maxwell2.
>>61199057
>So who cares then? 750 Ti still sold among poorfags I think.
Well fuck, Polaris also sold well enough. Both were still stopgap measures.

>>61199023

And? They came back with literally the most popular GPU series ever sold in history.

You're sitting here like BUBUBUBUBUBU you fucked up with maxwell 1!

>>61199064
>maxwell 2
750ti is maxwell 1 so is 970 and 980, maxwell2 is the 980ti and the titan xm

>>61199077
>GM204 is Maxwell1
Really makes me think.

>>61199064
That's how it works atm and looking at Vega nothing will change. AMD trying to play catch up game and setting fire under Nvidia's ass forcing them to release another 30% faster slightly refreshed lineup putting them ahead for another 2 years. At least due to complete lack of innovation AMD should stay competitive against Intel for years to come.
>>61199077
Nah he is right. Only Maxwell1 GPUs I know of are 750Ti and few 800 series mobile processors.

>>61198812
Yeah lets not talk about the cards that actualy matter right now

File: fKhPPhO.gif (2MB, 443x250px) Image search: [Google]

2MB, 443x250px

>>61198653
I'm still buying one

>>61199057
>I know some retards in here honestly were expecting 1080 Ti/Quadro performance for cheap but Vega can still do really damn well depending on pricing. If they released slightly cheaper alternatives to 1060/1070/1080 they could still take over market since that's what 99% of people buy anyway.

It really can't if these benchmarks are accurate. If it's using a larger die, taking more power and using HBM2 it's going to be more expensive than the Nvidia equivalent.

Nvidia could do a price drop on launch like they did with the 7xx series

>>61199092
aren't the 700 series cards like literally budget crap?

I just associated maxwell with the 900 series. Do morons actually buy x50 nvidia cards?

>>61199092
>looking at Vega nothing will change
Just like Maxwell never met it's perf/watt increase due to lack of 20nm node? Come on, people always find a way.
>forcing them to release another 30% faster slightly refreshed lineup
Now with housefires? Moar ALU is a dangerous road.

>>61199104
aren't the x50 cards*

>>61199105
>It didn't meet the ideal targets!
>still shit all over AMD anyway

ok

>>61199116
/v/eddit please stop.

>>61199130
>please stop
Can you stop your AMD denial? It's looking real dumb for you claiming that any iteration of Nvidia's cards have been failures for anytime in the last 4 years. I know you'll buy into AMD's vega bullshit and I'm not looking to upgrade cards at all, but denying Nvidia superiority top to bottom for the last 2 GPU generations is laughable.

>>61198695
>utter dwarfs the fermi disaster
This is the intelligence of wood screw cucks.

>>61199145
>gp106 is superior to Ellesmere
?

Also why are you shifting uarch discussion into realm of sales? nVidia managed to sell thermi. They can sell gaymen snake oil and people will buy it.

>>61199105
You mean their meme tensor cores? I guess time will tell.
>>61199104
If you look at Steam hardware survey and take it with a grain of salt you can say 750 Ti was really popular probably cause of prebuilds sold to absolutely oblivious people but still.

>>61199179
No i literally mean consumer Volta will be moar ALUs. Tensor memes are irrelevant and meme learning bubble will burst in several years anyway.

>>61199179
>750ti was really popular because prebuilts sold
Then you would also submit that intel integrated graphics are the best selling GPU ever because you're that retarded.

>>61199190
Intel iGPU is THE best selling GPU if you simply compare Intel's mobile revenue and nVidia's revenue in general.

>>61199182
Somehow Nvidia managed to add more cuda cores with each slightly improved alteration of Maxwell while lowering TDP and temps. I have absolutely no idea how the 12nm process will affect temps.
>>61199190
I don't think that makes much of a difference to Nvidia which ends up making cash. Also I don't even understand how we got to the point where you're calling me a retard because people get their GPU's through prebuilds.

>>61199196

.. Why the fuck would you look at revenue in general if you were after items sold when they have different prices?

Does anyone on /g/ have a fucking brain?

>>61199236
> Somehow Nvidia managed to add more cuda cores with each slightly improved alteration of Maxwell while lowering TDP and temps.
Full node shrink and Finfets are doing miracles for perf/watt.
>I have absolutely no idea how the 12nm process will affect temps.
It's a slightly (very slightly, worth for marketing pruposes) better than 16nm FF+. Better power target and almost no (if any at all) density improvements.

>>61199236
>people who buy prebuilts are nvidia's target market for enthusiast GPUs

Ok

File: 1477155822970.png (26KB, 1795x709px) Image search: [Google]

26KB, 1795x709px

>>61198682
You dipshit, it will have two HBM units, just at 4-stack heights instead of 8-stack heights.

This should allow for much better heat characteristics of the HBM modules, in turn meaning higher clock rates (probably stock 1000Mhz and up to 1100 or 1150 on custom boards)

>>61199238
because statistics are only true if they support your argument.

>>61199278
>this should allow for much better heat characteristics

based on your own original research from your own anus?

Thank you come again.

>>61199278
>that pic
just seeing outlines makes me rage

>>61198653
That can't be right, my rx 480 gets 60 fps at 4k in vulkan for the most part.

>>61198805
>and makes official PR statements showing benchmarks and fps without proper drivers.

1 they dont
2 they explcitly stated NOT TO PLAY GAMES ON IT
3 they provided it to reviewers tellng them DONT BENCH GAMES
4 they said Rx will be much better

>>61199400
Fuck, they never sent anyone review samples of Vega FE.

I'm very neutral when it comes to hardware companies but it's still obvious to me that something is simply wrong with vega. It makes no sense. They put all this time into it and have marketing material claiming significant architectural gains and they just do not seem to be in the FE product.
Raja said "flavors of RX vega will actually be faster than FE" which to me sound like maybe a water cooled one and a dual GPU card but that isn't enough. FE is not fast enough to justify the architecture.
Polaris would have been sufficient if they made a larger die. If they just straight up shrunk fiji it would have been enough. So why go through all the trouble?
The answers that make sense to me are; there was a serious hardware bug that resulted in major uarch features being disabled. gaming friendly features were some how counter productive to professional workloads so were disabled. major driver issues. or the arch is significantly better in things like APUs and other non-enthusiast applications.
Any one of these can be the case but of course I have no information in that regard. Unless the card is priced where it performs it has little reason to exist. If I can get 980ti/1070 performance at the same price from AMD I'll sell my 980ti so I can use freesync but that's all this card has going for it.

>>61198693
Honestly I'm just hoping for an amd flavored 1070 at a decent price for Freesync. I'd be happy with that. Hopefully the consumer version doesn't consume as much power.

>>61199410
Vega11 is later.

>>61198902
>gamernexus was also considered "AMD friendly"
It is sponsored by EVGA, what do you think?

>Rahja says it will get better
>muh reviewers say "it will CERTAINLY not get better goyim we know it"
>Rahja says the biggest challenge is the drivers
>completely ignored

>>61199396
That can't be right, my r270 gets 60 fps at 4k in vulkan for the most part.

>>61198981
Yeah but they leave half the game as DLC so it's a moot point.

>>61199409
>Raja said "flavors of RX vega will actually be faster than FE" which to me sound like maybe a water cooled one and a dual GPU card but that isn't enough.
It means his software monkeys will enable TBR for RX Vega launch drivers.

>>61199335
Vega FE can overclock its HBM to 1100 as told in this video:
https://www.youtube.com/watch?v=YflRl9np70k

"Memory can be pushed to about 1075"
http://www.overclock.net/t/1633446/preliminary-view-of-amd-vega-frontier-edition-bios#post_26197966

Vega FE with 8-hi HBM stacks can apparently OC from 945Mhz to 1075-1100Mhz

Besides your inability to logically reason that more stacks of heat producing dies decreases the rate of heat transfer for the lower stacks,
Screencap this post and laugh at me when RX Vega is released and overclockers can push their 8GB of memory higher than Vega FE owners can push their 16GB

>>61199426
Everyone knows that you can clock 4-Hi stacks higher than 8-Hi. That's why nVidia uses 4 4-Hi stacks for P/V100.

>>61199416
I know you are trying to meme but it's true, vulkan is amazingly efficient on amd gpus.
Heck when I tried ghost recon wildlands at 4k I got a rock solid 30 fps on it, was pretty comfy.

>>61199406
They gave one site an exclusive review sample to run CAD, 3D, and other workstation shit.
It was pcworld or something.
Go find the very first hands on review n the internet.

They have FORBIDDEN them to publish game results.

>>61199432
Let me argue with an idiot for my own sake, there's already one person involved pointing out the obvious.

>>61199437
Too bad they also sold FE cards for public and here we are.
I think RTG marketing team is due for something good spanking from mommy Su.

>>61198653
I hear'd it was a hybrid card a sort of gaming and sort of prosumer card.

the fuck if I know what is going on.

>>61199477
As of now it works like an overclocked Fiji with worse IPC (probably due to gaps in graphical pipeline).

File: 1470303001763.png (7KB, 1795x709px) Image search: [Google]

7KB, 1795x709px

>>61199278
>That pic

THANK FUCK, I FOUND ONE MORE DETAILED THAN THIS

>>61199477
AMD thought it was a good idea to release card with no drivers to public.

>>61199448
Perhaps they REALLY REALLY had to deliver on that "2017 Q1" promisse.

Although it would be better if they had only released Mi25 as 'Vega'.
Or maybe the investors got sick of waiting.

In fact OPs theory actually does make sense if the first batch of Vegas is broken then to sell them as Instinct and Solidworks cards is the best option.

>>61199507
Certain people have MI25 already.
For a considerable amount of time.

>4 more weeks
Fuck this is so frustrating

>>61199519
Wel it had OFICIALLY been released on Epyc day.

>>61199531
And we know precisely fuck about Vega besides some high level details.
This nu-AMD way of saying nothing and then dumping everything on one presentation is frustrating.

>>61199548
I wish someone benched FE for some meme learning stuff.

>>61199563
Like Deep Bench?
They did.

>>61199587
And how was it?

>>61199551
I want the damn Vega white paper and programming guide so I can scour through it to find the ALU-level changes in the ISA for hints about exactly how much has been sacrificed in IPC to attain those clock targets.

Have they lengthened the pipeline two stages? Has it all been extreme tuning of the existing design framework? Is GCN still utterly incapable of division ops in any reasonable time frame without resorting to specific fine tuned code?

>>61199415
>the guy working for the company trying to sell a product says it will get better
>independent reviewer says it won't

Oh boy what a hard choice, who should I trust!

>>61199278
>Just wait it will get better!
Every single time

File: actual.jpg (208KB, 1067x600px) Image search: [Google]

208KB, 1067x600px

>>61199605
>how much has been sacrificed in IPC
they claimed the opposite though

File: 1474230183008.png (54KB, 815x370px) Image search: [Google]

54KB, 815x370px

>>61199641

>>61198653
>tfw they still got out of stock quickly because "muh ethereum"

fuck, nvidia for gaming, amd for mining then

>>61198693
the relevant vega isnt out yet senpai

>>61199605
And I want to know how the fuck is AMD doing TBR. But fuck, no whitepapers for you, goy!

Side note: Linux Kernel 4.12 with initial Radeon RX Vega support (headless) and nvidia Pascal 1000 optimizations was released today.

>>61198679
Miners don't like HBM

>>61199704
RX Vega or Vega FE?

File: cb-vega.png (78KB, 920x1367px) Image search: [Google]

78KB, 920x1367px

literally worse than fury x in some compubech opencl tests

File: cb-vega2.png (116KB, 920x1911px) Image search: [Google]

116KB, 920x1911px

>>61199800
I wonder if it's hardware related, since opencl doesn't seem like something that needs driver optimizations

>>61199632
>independent reviewers say it wont
How the fuck do they know?
Their argument is "this is the final silicon, this is the vega chip"
Well shit it is, but if its not utilized by the software and all those hardare features like tiled rasterizer, geometry shader, pipeline bubble remover etc just dont fucking function theres a good reason to assume they will get better.

Try running Polaris on pre-polaris GCN drivers, it does work but feels crippled in all ways.

>>61199774
headless = no display driver, so only good for calculations. But at least they are working on it.

>>61199866
You're delusional m8

We've had many releases like this card before and even if there is some room for improvement it won't be great, 10% tops.

That pits it just above the 1080 in gaming performance.

History has shown time and time again that you can make an accurate guess about the gaming versions performance based on cards like these, like with the Titan series in Nvidia's case.

But sure, keep believing that it will be different this time, overhype the card again

>>61200174
There's a difference between driver optimization and simply enabling uarch features in the drivers you dolt. Fucking /v/eddit rope yourself.

>>61200186
Stay delusional you dafty, you overhyping retards will never learn.

>>61200192
Yes yes leather jacket man I am totally ready to give you my money (no).

>>61200201
Leather jacket man, lol wut

>>61199396
>my rx 480 gets 60 fps at 4k in vulkan for the most part.
Yeah, now try maxing out the settings and try again

>>61200207
The magical man that brings you woodscrews, 1.7% yields and 3.5 gigabytes of memory who also happens to love leather jackets.

>>61200222
>AMDlets still mad that AMD hasn't been able to touch Nvidia in performance/watt since Maxwell because of that sweet ass tile rasterizing

The 970 3,5GB debacle was definately embarassing but Maxwell was an amazing architecture, anybody who denies that knows nothing about gpu's

>>61200174
>ou can make an accurate guess about the gaming versions performance based on cards like thes

Yoy cant PREDICT what improvements would enabling all those hardware features would have.

Even if it was simply a FURY X with 60% core clocks and double memory speed it is supposed to be much faster than this.

But there's also fuckton of sick hardware features.

Vega FE right now is way under it's reasonable projection.

>>61200250
>getting mad about nvidiots getting cucked by leather jacket man himself
It was hilarious.

>>61200259
At least it works.

>>61200259
None of that matters when it doesn't translate to actual gaming performance.

It's the same old story, people have been saying AMD would dominate in games because of their hardware based scheduler for ages now. And yet for a few games nothing has changed, Nvidia still dominates.

And for a simple reason, AMD might have cool features on their card, but they leave it up to the developers to actualy make use of them, and many will not, because it takes alot of time, it's expensive and not Always worth it.

>>61200285
>developers will directly use TBR or primitive shaders
Oh come the fuck on.

>>61200268
Confirmed tech illiterate who can only spout 'le epic boogyman' memes, instead of actualy talking about gpu technology
Top kek

File: 580 1060 game breakdown.png (505KB, 1920x1080px) Image search: [Google]

505KB, 1920x1080px

>>61200285
It's already happening and will continue to happen as more devs move towards DX12/Vulkan.

>>61200297
You can't expect NVidiots to actually know what developers use. Unless it's Goyworks features.

>>61200299
>nvidia literally lying about 970 spec
>memes
Class-action lawsuit was also a meme?
>>61200302
He's also shitposting, since you can simply wrap primitive shaders around usual vertex code using drivers.

>>61200308
The only good goyworks feature is HBAO+ anyway.

>>61200302
It might, it might not.

I'm not falling for /g/ memes anymore, they have been saying AMD will dominate for years and now the best they can come up with is a card that's slower than the three top cards of Nvidia one year old architecture.

AMD 'might' be faster in the future
Nvidia is faster right now and might still be so in the future and way more energy efficient

>>61200309
Yeah just continue to stick your head into the sand and deny that Maxwell was a fantastic leap forward in terms of power/performanc.

>>61200331
>way more entry efficient
Wat? GP102 is fucking powerhungry because there's nothing you can do when you're throwing more ALUs at the problem.

>>61200250
> Maxwell because of that sweet ass tile rasterizing

Did anyone test draw stream binning rasterezation that AMD was talking about?

>>61200335
PowerVR did these leaps back in 1994.
>>61200351
Yes and it's disabled.

>>61200336
Top fucking kek, I'm out of here, you people live in an alternative reality

File: NVIDIA-Quadro-P6000-Power-Test.jpg (79KB, 480x408px) Image search: [Google]

79KB, 480x408px

>>61200357
We sure do

>>61200357
>300w on almost every AiB card is efficient
The fuck? The moment you throw more ALUs at the problem efficiency goes out of the window.

>>61200366
That's full system power, though.

>>61200373
And that's not in any way "efficient".

>>61200285
Well Rahja did say it will change.
There has to be a reasont they were showcasing gay ming without FPS counter, and didnt allow pcworld to bench games on their exclusive review.

It means the do expect it to change.
Othervice they would probably just scrap it entirely and sell the whole bunch as Radeon Pro and Mi25.

>>61200391
They'll shill MI25 and Vega SGG to no end on SIGGRAPH anyway.

>>61200285
>but they leave it up to the developers to actualy make use of them,
Tiled rasterizer and new geometry pipelne have nothing to do with developers.

>>61199810
>since opencl doesn't seem like something that needs driver optimizations

Lol

File: 3dsmark.png (46KB, 1313x387px) Image search: [Google]

46KB, 1313x387px

what's up with the driver version of vega?
like if you look from oldest to newest entries the driver version increases then keeps decreasing then increases and then decreases again for the final version??

>>61200667
forgot source: http://www.3dmark.com/compare/3dm11/12249366/3dm11/12118513/3dm11/12085688/3dm11/12070976/3dm11/12148225/3dm11/12213436/3dm11/12196389/3dm11/12220458#

>>61200667
Because there's effectively 3 big driver now.

FE, Firepro, and RX drivers.

>>61201212
And neither works.

>>61201267
Firepro/pro drivers seem to work fine most of the time, but even they're still far from finished.

https://www.pcper.com/reviews/Graphics-Cards/Radeon-Vega-Frontier-Edition-16GB-Air-Cooled-Review/Professional-Testing-SPEC

It beats the p5000 in some production software so we know the raw horsepower is there. It's a matter of whether in-game performance is due to poor software or a hardware flaw.

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

156KB, 633x758px

>>61198653
>tfw waited for this

File: 1462727743869.png (93KB, 216x249px) Image search: [Google]

93KB, 216x249px

>muhh wait for drivers™
>heeerppp Vega use a Fiji driver (WTF?)
>is to a gayming card
>heheheh compare with a Titan card
>wait for optimization
>top finewine kek
>wait for DX12 and Vulkan
>amd will release a driver for performance improvements in 200%
>wait for vega gayming card
>wait a little more time, please sir
>it's a workstation card
>just wait™
>opencl need driver optimization
>wait 6m~1 year for vega to release all its power

>.
>..
>...


>....

>i already asked to wait?

amd failed so hard yet again, nvidia continue to milk its customers...i mean nvidia is literally using its mainstream cards as highend and amd topend pro card fails to even overtake it.
With Volta nvidia will just release some 200mm mainstream card as the new 1180ti...

>>61202561
>another month of /v/ crossboarders shitting the place up with spongebob memes

I want RX Vega to release just so these shitheads will fuck off again (until the next GPU shitposting wars)

>>61198917
it's shit at compressing 7zip, not decompressing. It's 6900k level for decompression, mong

File: 1477263009035.png (45KB, 1047x855px) Image search: [Google]

45KB, 1047x855px

>>61202824
>I want RX Vega to release just so these shitheads will fuck off again

Just wait, sir.

File: smug horror.png (121KB, 354x347px) Image search: [Google]

121KB, 354x347px

This thread is a genuine proof that /g/ is overrun by amd fanboys and shills who bend reality much like macfags did.

>>61202866
amd ryzen vega mac pros when

Of those drivers which driver is the latest RX driver do you think? :)

>>61202866
this post is proof /g/ is overrun by feminine cock lusting sissy faggots who want to be girls.

File: incoming miners.jpg (83KB, 1280x720px) Image search: [Google]

83KB, 1280x720px

The entire GPU market resembles a planet that's been raped by Tyranids. No prices anywhere make any sense and even if RX Vega launches and flops, it's not like there's any better options.

tl;dr GPU market is fucked for the time being anyway.

>>61202939
I want to be the little girl so I can pleasure myself all day.

>>61202953
Don't worry, looks like even nVidia joined ebin multichip race.
http://research.nvidia.com/publication/2017-06_MCM-GPU%3A-Multi-Chip-Module-GPUs

Inb4 miners dont buy hbm cards and 1070s are still 600$ forcing people to buy the only decent new gpu for the cost

>>61202972
you would enjoy being pleasured by big strong men more though.

>>61202984
What can they realistically do with this?

>>61202998
Sounds like a win win to me

>>61203015
Exactly

>>61203002
Nawww I'd get a dog instead. Men are horrible. A cute cuddly doggo who is loyal is much better.

>>61203004
Say "we AMD nao" and make retardedly huge and powerful GPUs.

>>61203035
To your luck im a dog.

>>61198972

Actually, the memory issues was discovered shortly after the GTX 970 was released when a bunch of geek stress test the shit of the hardware and noticed odd behaviors in their tests when the application utilized more than 3.5GiB of video memory.

They posted in their finding on various fourms and imageboards and the rest is history.

Basically, Nvidia marketing fucked up bloody. They failed to properly disclosed the nature of GP104 binned. Engineers knew the issues from Day #1 and was similar to the "amorphous memory" found on GK106 chips and Nvidia disclosed that shit on Day #1 via white papers.

>>61202984
>nvidia master plan
>tfw Huang watched adored video and copied all the sycret amd tech

>>61203037
Well, would like to see a working product with this. Who's going to have the first MCM GPU? NVidya? AMD?

>>61203984
Of course it's going to be AMD.

>>61203074
honestly if nvidia just disabled the slow ram and made the card 3.5gb, it would have had 97% the performance, yet zero drivers issues or the strange performance issues.

>>61203944
Don't worry, they need working interconnect first.

This makes me think the amd shills are lying about ryzen too. Who do I listen to?

>>61204086
>best selling CPUs everywhere
Hi Brian.

>>61204086
No one cares what you do you dumb nigger

>>61204100
What are the best selling cpus? My name is Brian but not that one lol

>>61204126
Yes, you're totally not that Brian.

>>61204050
strange thing is amd roadmap shows Navi as being scalable gpu in 2018.5-2019, which makes sense.

But nvidia road map ends with Volta and we are few month away from paper release of it, yet there isnt any official announcement of new architecture. Judging this i think nvidia is unsure what next gpu will be mcm or just even bigger single die, it should hit around 2019.5 .
But if they didnt even start work on mcm yet, i doubt they can design proper mcm in 2years.
Honestly it will probably be another diestrink to 7nm.

>>61204178
It will totally be a dieshrink.
Raja and ATi was late to unified shaders for consumer market.
Now it's time for revenge.

>>61204178
Does Nvidia fabs actually have magic IBM technology that allows of 7nm node?

>>61204215
TSMC's high power version of 7nm node is somewhere in 2019.
The bad thing is that you can't fight IBM wizards.

Said TSMC's 7nm is not going to use EUV. That's really-really bad if GloFo manages to incorporate EUV into their 7nm node in 2019.

>>61204241
whats the differnce?

>>61204315
Absolutely hilarious amount of reduced complexity due to EUV cases much better yields.

File: 1499104870271.jpg (162KB, 1462x1462px) Image search: [Google]

162KB, 1462x1462px

>>61198653
Another
Massive
Disappointment

>>61204225

>IBM wizards.

IBM is the sort of company that decides "hmm, today we're going to break the laws of physics" and manage to do just that. POWER8 and 9 are proof what happens when you bend physics over the table and go in raw.

>>61204333
well if they are getting high enough yields it doesnt matter for them then.

>>61204355
Anon with 7nm DUV being as complex as it is, the yields will never be 'high enough'.

>>61204368
so why they dont use it then?

>>61204352
>jewish intel faggots say 5nm is imposible?
>REV UP THE OVENS BOYS ITS TIME FOR ANUDDA SHOAH

>>61204381
Ask TSMC. Both GloFo and Samsung (members of Common Platform Alliance) will use (GloFo will start with DUV) EUV for their 7nm node.

>>61204384

>5ghz across 6 million threads?
>sure

>>61204215

Nvidia don't have fabs, they contract out to the usual suspects.

>>61204394
didnt nvidia do some test run on glofo or samsung silicon 1050 chips?
They might just jump the gun if samsung/glofo silicon is superior with 7nm.

>>61204420
GloFo will 146% have all the capacity used by AMD and Samsung's node is low power and i doubt they'll producte some 300mm^2 chips for nVidia.

>>61204429
well if nvidia offer more money then whats stopping GloFo from taking nvidia order first?

>>61204461
Wafer amendment that they have with AMD.

>>61204471

Didn't AMD have to pay a shitload to GloFo recently for changing the agreement since they weren't selling enough chips?

>>61204501
Too bad they are selling so much of them GloFo is increasing Fab8 capacity by 20% soon.

>>61204471
didnt amd literally paid buttload of money to not have to order from them? And even then im sure GloFo have the final say if what order they want to do. If nvidia order is juicy enough they will take it.

>>61204528
You can't have order more juicy than Zen2.

>>61204540
> 7nm 5ghz clocks

I'm signed up for the just wait club

>>61204540
Inb4 Intel just orders a button of chips on GloFo (they do have a ton of cash) and thus they put AMD orders on wait, while getting 7nm ebinmail processors.

>>61204569
You literally can't have order MORE juicy than Zen2. It's ALL the capacity you can muster will be put to a good use.
Like really, Zen is simply good.
Zen2 with astronomic clocks will win them even fucking HPC markets.

>>61204589
They can't since they have their own fab. Intel is no fabless company.

>>61204589
what in the fuck is an ebinmail, my dude

>>61204608
Intel's new mesh interconnect.

>>61204604
lol what? whats stopping them? have bilions in bank account, whats stopping them from running there own fab and then ordering bunch of shit from pajeets? Nothing, You are just projecting. Ofcourse intel would never order, since its signaling that they have lost the die war, but technically they could order from pajeets.

>>61204528
No they paid for one year. AMD and GF are bound to each other

>>61204637
anti trust agreements, maybe? I'm just assuming here.

>>61204638
glofo would be dumb to shit on a loyal customer like AMD. It's in their best interests to compete with intel, as if intel buys chips from them and AMD gets royally fucked, glofo would have one less major customer

>>61204637
>IBM
>pajeets
BEGONE KIKE BEFORE I START REVING UP MU OVENS

>>61204637
>GF which has a new campus in fucking upstate NY is pajeets
>not h1btel

>>61204652
>>61204650
>Glofo Ceo Sanjay Jha
>not pajeets
srsly

>>61204680
>7nm node was presented by Nazi IBM engineer
srsly i am getting my oven ready

https://www.globalfoundries.com/news-events/press-releases/globalfoundries-track-deliver-leading-performance-7nm-finfet-technology
BEGONE, KIKE

>>61204695
what the fuck does IBM have do to with Glofo you dumb fuck
they are ordering from glofo not IBM

>>61204716
not him but glofo is just a fab, they use design specs from whatever is given to them. e.g. AMD using IBM processes. All they have to do is not fuck up in making it

>>61204716
IBM PAYED
FUCKING PAYED
5billion dorra
for GloFo to absorb their fab business along with staff and IPs.

>>61204729
>>61204735
still GloFo is pajeets just check there CEO

>>61204745
>company FULL of IBM nazis
>pajeets
Begone kike.

>>61204748
t. Pajeet

>>61204756
t. Kike

>>61204647

AMD are contractually obligated to get a certain amount of chips from glofo, they are not just some 'loyal customer' since AMD used to own them

>>61204745
>>61204748
>>61204756
>>61204758
NO U

>/g/

>>61204765
regardless of contract it is still fucking dumb for glofo to fuck over somebody buying chips from them. Almost moreso if they're guaranteed to buy them.

>>61204774
t. Faggot

>>61204789
its not like amd have somewhere else to go, also most likely nvidia will just order chips from tsmc and samsung, Huang is still to butthurt about anything amd

>>61204812
Man i still remember nForce.
Ruinz literally killed bff's.

>>61204790
Shouldn't it be Fa/g//g/ot?

>>61204812
AMD not having a choice in the matter has little effect on glofo's business. If AMD goes under, a large portion of their fabs will go unused for too long and they'll likely have to consolidate. It's in their best interests to work with AMD, at least for now.

>>61204812
>its not like amd have somewhere else to go
Samsung.

>>61204812

AMD can and have bought from other foundries, but for 14/7nm they have to pay a penalty to Gloflo to do it.

>>61204844

They're literally contracted to have to work with AMD, there is no choice

Probably gonna be more expensive to buy one here in aus over a 1080 with a decent oc and cooler.

Since we know and stuff runs as fast as it can out of the box I'm gonna have to jump ship.

It's been fun and Bros but time to admit defeat for now.

>>61205516
Hi Huang.

File: Mr.kown_0.gif (1MB, 268x274px) Image search: [Google]

1MB, 268x274px

>>61205570
I've been waiting since early 2016 and when the rx480 dropped it was slower than my 390x

WTF are ayymd doing they can't even beat their own cards anymore and vega is fiji 2.0 but fucking almost 2 years late.

Hbm2 is a meme for gaming it was proved last time round.

I guess amd could get lucky with the button mining boom which I reckon Vega will be good at.

For the mean time I need a GPU that actually doesn't suck in99% of games not some special snowflake that needs dx12 and vulkan to perform like it should.

>>61205687

>Hbm2 is a meme for gaming it was proved last time round.


Anybody who says this clearly has absolutely zero technical understanding of what HBM is, what its designed to do and why it was a non-factor on fiji's performance.

>>61205766
/v/edditors are stupid.

Nvidia went the correct route, GDDR5X is cheaper and good enough

>>61205859
"Good enough" is not an answer when you can slap more shaders/ROPs insted of fucking DDR PHYs.

>>61205766
All I see is endless blablabla while this supposed "great advantage" is no-fucking-where to be seen. Nvidia made the right call, deal with it.

>>61206010
>nvidia cucked us out of superior memory to increase their margins
Deal with it.

>>61205892

Plus you have the advantage of when really trying to kill your bandwdith the physical distance the memory is from the gpu - after a certain point the electrical signal simply cannot travel the distance any faster and the only solution at that point is shorten that distance.

>>61206010

Nvidia uses HBM on their srs bsns cards for a fucking good reason - on a technical level HBM is superior to GDDR5(X) in 99% of ways.

>>61206044

Nvidia are free to use HBM2, they probably realize that it's pointless at this point for mainstream cards as memory bandwidth isn't starving the GPU.

It's like having 64 gigs of ram over 16. Is it better? Yeah, would you use it for gaming? Probably not.

>>61206044
Yeah I will deal with unmatched performance that amd can't even come close to while you keep clinging to your buzzwords and meme features that don't provide any real world advantage worth a fucking shit, alright. It's always hilarious with you amd shills. You are CONSTANTLY churning out endless excuses and damage control as to why amd gpus provide dogshit performance. Maybe it's highline you deal with the fact that the company you shill for is incompetent and that this is all there is to it, but I get, it's always something else's fault and other people "just don't get it", right? Enjoy playing ashes of nobody.

>>61206122
>>61206127
Look, it's /v/eddit! Truly it is the summer time here.

>>61206143

No you're right, HBM is a fantastic technology that allowed the Fury/Fury X to beat all the competition, and all tech companies are banging down their doors to get it in their products since it's so good and not at all expensive for a needless gain, and in no way will it increase the costs and decrease the availability of AMD cards due to supply issues.

>>61199765
wasn't that only because it was 4gb?

>>61206172
You are simply showing your tech illiteracy, anon. Please stop, it's not /v/.

>>61206186

I'm saying you're right friend, all those things happened. I'm singing the praises of HBM.

>>61206206
And you should do it. HBM is the best thing that happened for DRAM since forever.

Is HBM the RAMBUS of AMD?

>>61206223
No, since HBM is a widely used JEDEC standart.
Even Huang uses HBM. And he backed abortion that was HMC.

File: aitd2.jpg (174KB, 1280x720px) Image search: [Google]

174KB, 1280x720px

>>61198653
RE7 on vega leaked
>>61204735
paid

>>61206219
this
igpus will actually be viable bcos of hbm
ddr5 will be useless because of hbm
soon stacked processors will be a thing too

>>61206327
>igpus will actually be viable bcos of hbm
Yes.
>ddr5 will be useless because of hbm
Nah, on-package memory is not everything.
>soon stacked processors will be a thing too
Maybe. We should start by MCMing GPUs.

>>61206327

How soon? HBM has been out a long time and I don't see many companies flocking to it. Are they waiting for prices to come down?

>>61206347
>HBM has been out a long time
4 years is a long time?
>I don't see many companies flocking to it
Not everything needs on-package memory.
>Are they waiting for prices to come down?
Prices will come down once HBM becomes at least somewhat mainstream.

>>61206373

4 years in tech.. yes that is a long time?

But if it doesn't become mainstream by few companies using it then it won't come down?

>>61206421
>But if it doesn't become mainstream by few companies using it then it won't come down?
Guess why AMD uses HBM2 and why Samsung is developing low-cost HBM.

>>61206435

I'd assume AMD are using it because they developed it..? You did know that right?

Samsung is developing a slower cheaper HBM yes but I'm after actual usage and it's effect on the tech market

>>61206502
>I'd assume AMD are using it because they developed it..?
They use it because it saves die space and in general better.

>>61206526

Yes and in the cards that use it that still can't save them in the GPU race, back to square 1. It might be a great tech, but right now it's adding cost to their GPUs while Nvidia are putting along and beating it with GDDR5x, which is the original point - that memory bandwidth isn't a problem for GPUs that HBM needs to solve right now.

>>61206575
HBM allows you to put moar ALUs/ROPs/whatever in the same die space though.
Anyway GPUs are fucking irrelevant toys i am so fucking tired of /g/ getting spammed with fucking GPUs.
Fuck it.

>>61206596

>Anyway GPUs are fucking irrelevant toys i am so fucking tired of /g/ getting spammed with fucking GPUs.
>Fuck it.

You're posting in a fucking GPU thread what do you expect?

AMDs gigantic dies are already busy trying to keep up with Nvidias year+ old generation and spoiler: they can't

>>61206630
Vega10 is 13mm^2 bigger than gp102.
Also, no drivers, mr. Huang.

>>61206657

And power usage? And that's with HBM2 right?

So we have bigger die, fuckton more power, advanced HBM2 memory, coming out over a year later, and we're relying on driver updates to save it? With all that it can't beat year old shit?

>>61206683
>So we have bigger die
It's about as big as gp102.
>fuckton more power
It's 50watts more than throttling reference turd and about the same as any AiB gp101.
>advanced HBM2 memory
And that's bad?
>coming out over a year later
Sometimes you have to do the fucking R&D.
>we're relying on driver updates to save it
We're relying on driver updates to make it work.
>With all that it can't beat year old shit?
Like anything Huang will shit out in the next 3 years will be anything else than Maxwell: moar ALUs edition.

>>61206716
also hardware scheduler doesn't help on the power issue.

>>61206747
Don't worry Huang will bring it back sooner or later.

>>61206716
>will be anything else than Maxwell: moar ALUs edition.
Tbh, we don't even know name of architecture beyond Volta. Much less its features or lack of them

>>61206575
totally agreed, if amd went gddr, people would have rx vega in hand.I think Raja pushed for this to try and make up for the fury failure.

>>61206774
And consumer Volta is moar ALUs. And GV100 is moar ALUs with moar fixed function elements (meme cores).
At least AMD listed "scalability" for Navi.
>>61206781
What? The problem is drivers you cockgobbling retard. Vega10 was taped out AGES ago.

>>61206770
The "Joule" architecture?

>>61206809
He-he.

>>61206805

They delayed Vega for 8 months because of drivers? Did they manufacture them last year and just waiting on their coders?

>>61206839
Vega10 was taped out in ~november 2016, since they showed us one on New Horizon even.
Yes, writing drivers for new GPU uarch is ass.
Especially if it bring truckload of features like Vega. And at least THREE of them require driver effort.

>>61198653
It's not a gaming card.

Fiji had massive gains in drivers

Just wait for the actual RX Vega release and for early bugs to be worked out

This card really exists so AMD can find the issues before the full RX Vega launch and fix them in software as much as possible.

>>61206875
It's not about GAINS from drivers.
It's about drivers enabling fucking uarch features for fucks sake.

>>61206716

No, the 1080 that it is competing against is 314 vs 484 for the AMD.

It has twice the power consumption of that 1080.

No I'm saying that HBM2 is great, but even with that advantage it can't do shit.

They didn't just start making GPUs, this one has been delayed again and again.

How are you shitting on Maxwell/ALU when this card literally couldn't beat a Fiji overclocked to the same speed? And this company that recycled the 7970 to the 280 to the 380? Then from the 480 to the 580 rebadge?

>>61206925
> They didn't just start making GPUs, this one has been delayed again and again.
Just like motherfucking Maxwell2 was delayed TWICE.
FOR TWO QUARTERS.
AND NO ONE CRIED.
LITERALLY NO ONE CRIED ABOUT GM1xx GETTING AXED.
SO SHUT THE FUCK UP AND HANG YOURSELF.

>>61206886
Well duh, but obviously it's not that simple, especially when you have 1/10th the R&D budget and staff of the other companies.

Vega should be competitive at it's price points but it likely won't be the fastest thing around

UNLESS
BY SOME MIRACLE

AMD comes out with some experimental $1500 top tier Vega card that is basically early NAVI with some kind of infinity fabric or something, that uses 2 GPU dies without the need for crossfire or game engine scaling.

But that's a long shot, they will likely talk about NAVI at least at the Vega launch.

>>61206869

You're literally saying they delayed over 6 months over drivers? I'd expect good availability on launch then, they'd have to have a few million in the warehouses by now.

>>61206955
It's ok when Huang does it.

>>61206186
>y-you just don't understand
>i-it's good b-but I can't explain why

Either it gives a performance advantage or it doesn't and right now AMD has been eating nvidia's dust for years.

>>61206969
>You're literally saying they delayed over 6 months over drivers?
Why the fuck else would they delay working silicon. That everyone eagerly awaits.
>>61206968
>AMD comes out with some experimental $1500 top tier Vega card that is basically early NAVI with some kind of infinity fabric or something, that uses 2 GPU dies without the need for crossfire or game engine scaling.
Why would they? nVidia just published whitepaper about MCM GPUs. They only published fucking paper.
They are at least a year away from developing functional interconnect.

>>61206955

Fuck somebody's angry about AMDs fuckup

Maxwell is forgiven since they released the most popular enthusiast consumer graphics card ever released that still tops the steam charts today

AMD have released this so far, and it doesn't look good

>>61198653
pretty decent performance if the price is right, but the price/performance probably won't be good enough to justify picking it over gtx 1070, and it's too late

>>61207020
>it's okay when Huang does it
Okay.

>>61207046

Yes you've said that about 400 times this thread, usually agreeing with your own posts.

>>61206683

Just wait for the drivers anon. I'm sure when the magical driver comes out it will suddenly halve its power usage and double performance. Just wait!

File: davidstall.jpg (1MB, 2560x1920px) Image search: [Google]

1MB, 2560x1920px

who am I kidding i will still buy it because freesync. just hoping they can make a power efficient version before navi.

>>61207017
>Why would they?
Because it's their only answer to beating the titan Xp/1080ti in performance

They already have a working infinity fabric, just need to make it work for GPUs, NAVI should be the "Ryzen" of GPUs assuming that's what they meant by scalability. Pretty much like 4 480s slapped together on a die with perfect scaling.

>>61207101
>Because it's their only answer to beating the titan Xp/1080ti in performance
Hah aaa no. MCM GPUs are totally in the league of their own in terms of performance, anon.
It's like before and after unified shaders.

>>61207067

The question is how the fuck this happened, I've read a fury X overclock on the new process would have beaten this, how did their new architecture fail so badly

>>61207090
It should undervolt as well as fury could and going by the Vega FE PCB they will likely have a nano as well, hopefully with AiB nano models as well that don't have to throttle.

>>61207126
NO
MOTHERFUCKING
DRIVERS
DO YOU KNOW WHAT TBR DOES?

>>61207145

Your all caps HUANG bullshit doesn't change the benchmarks today

>>61207145
as efficient TBR is, I doubt it can make up for the extra TDP produced by having a hardware scheduler built into the chip. Nvidia's overhead is less because of pushing that onto the CPU. We will likely see a 10-20% increase if TBR and other optimizations are added but this chip is inherently limited by a higher TDP and lower clock speeds. Maybe, with luck, rx vega will have a second stepping of the chip or it'll run cooler from some optimization.

>>61207166
Huang please, hurry up with MCM research. I mean you have zero (0) expirience dealing with complicated NUMA setups, but you have a year and a half to figure it out.

>>61207188
It's a fucking 4096ALU chip.
It's inherently more powerful than even the full gp102.
The faggots only need to make it work.
Also you forgot increased geometry throughtput from primitive shaders.

File: 1457106085553.jpg (232KB, 1900x1267px) Image search: [Google]

232KB, 1900x1267px

>>61206781

It's the classic AMD strategy. They can't develop products for shit so they have to rely on gimmicks that sound intimidating and powerful.

Bulldozer with with its MOAR (half)COARS. Bulldozer and Vishera with its MUH GIGAHURTZ MUH HIGH CLOCKS (and half the IPC of an i3). The r9 390 withs EIGHT GIGABYTES while the card wasn't powerful enough to actively use even 6 gbs, let alone 8, but it fooled the clueless retards who think you measure a GPU's power by the amount of memory you throw into it and that all you need for 4k gaming is a retarded amount of memory. Fury with its meme HBM memory that achieved FUCKALL and water cooling to provide you with the OVERCLOCKER'S DREAM, even though it couldn't OC worth a shit. Then we have mantle/directx 12 and endless Ashes of Singularity which literally no one has played because it's the only case scenario where all this worthless garbage provides any actual advantage. Now we have the return of the meme with HBM 2, because yeah, we are totally so starved for bandwidth, amirite? But hey, don't lose faith, all this very efficient tech totally shows off! But please ignore the fact that you can't see this supposed "revolution" in benchmarks, can't see it in any real world performance test, can't see it in any day to day scenario, can't see it in literally fucking nothing. But Believe™ & Wait™, good goy!

>>61207212
>let's remember bulldozer but totally ignore Zen
(You) tried.

>>61207211

>Its so powerful
>It has all these acronyms
>It still benchmarked like complete shit after coming out 14 months after the competition

>>61207211
I mean sure the theoretical increase in performance from the primitive shaders is impressive but it's yet to be seen on a GCN chip of any kind, so how are we supposed to actually judge real world results? I'm not holding my breath about the benefits of it but we'll see.

>>61207145
>>61207067

>>61205766
How the fuck does hbm even matter when the gcn/Fuji doesn't scale well over 4096 shaders anyway? AMD barely have anything faster than a 290 oc and that's a 4 year old card
>>61206127
Sadly this.

AMD at least are making good CPUs again but Vega is dogshit just like every other GPU they have made lately
>>61206875
Muh wait for drivers...

Look I have a 390x and the driver's over past year have been good it's just I can't see drivers alone making up20-40%+ perofrmance deficit

Everyone said the same about the 480 and turned out dead wrong

I have literally waited for amd to make a fast card that isn't a house fire since 2013 and I'm sick of waiting.

350 watt for gtx 1070/390x/fury level performance is not acceptable and it's more expensive to boot

Once the miners Jack up the price it will be the biggest floppy outside Fuji and fermi

>>61207256
They listed up to 11triangles/clock with 4 shaders engines.
We'll see. GCN certainly needed better geometry throughtput.

>>61207231
>b-but t-this one time they haven't delivered a complete POS, please ignore all the other times they have failed astronomically!

Wow, everything is saved, pajeet!

>>61207288
Wow, Huang, remember to look at Magny-Cours and EPYC for complcated NUMA setups.

>>61198653
They can't even sell them cheap because the die is fuckhuge and HBM2 is expensive. What a disaster.

File: file.png (1MB, 1438x852px) Image search: [Google]

1MB, 1438x852px

>>61207240
No you don't understand anon, AMD cards are not like the other gir- I mean, cards. They are special, they have hidden power that only the truly smart can see. You're smart, aren't you?

File: bliss.jpg (173KB, 960x912px) Image search: [Google]

173KB, 960x912px

>>61207274
>fury X
>dogshit
>regular fury
>dogshit

they were good cards but were hamstrung by high prices because of HBM and high TDP, but they were nowhere near bad. They were competitive with the 980 and 980 ti, the top dog cards at the time, and the lower VRAM was not the bottleneck. The marketing of "overclocker's dream" was a fucking mistake though.
>t. fury nitro owner, got one for 250 dollars for some ridiculous reason

Also polaris was a good mid tier card, great perf/$ and decent thermals/wattage. Don't be delusional.

>>61207283
Like I said, hypotheticals are great, but if there is no actual benchmark to back up the actual benefits of said feature I'm skeptical.

>>61207305

And in the end Nvidia could do a price drop again like they did with the 700 series vs the 200 series. It wouldn't have to be a big one either.

>>61207305
So why did they even release it? Rikes...

If it wasn't for Zen and poolaris they would be dead
>>61207315
Lel

As a amd gpu fan sadly I agree.

Well back to the toilet and gpu division.

Just wait for navi 2020 you'll see!

By then volta will have ass fucked amd into Oblivion.

Let's not joke around here they probably won't ever overtake the 1080ti or a 1180 let alone a year old 1080 at stock clocks
>>61207335
Big whoop my oldass 390x stomps over 980s but it's too hot and too old and amd don't have anything to replace it with except a 580 or a rx590 vega which are just as slow

I am not deloosiinal it is. U who are saying Vega isn't shit.

Look at fiji ffs it literally almost sunk amd.

My next build will be a 1800 and 1080 because Intel and AMD GPU are horrid ATM.

So weird it's like it's 2000-2006 again

>>61207335
>they were good cards but were hamstrung by high prices because of HBM and high TDP, but they were nowhere near bad

More expensive than competition? Check.
Worse drivers than competition? Check.
WAY more power consumption and heat generation? Check.
No OC headroom whatsoever? Check.
Dropped like a fat turd by AMD itself which basically ignores its existence? Check.

But yeah, totally a good card.

>They were competitive with the 980 and 980 ti

Yeah, in ashes of nobody maybe.

>>61207418
>2000-2006
Oh yes R300 was not TOTALLY stomping nVidia into the ground. Come on. Shitpost properly. Say that 5 series was great.

>>61207418
did I say vega wasn't shit m8? R e a d i n g comprehension is important. Currently, vega is DOA for gaming and should not be an option for anyone unless the prices make sense.

>>61207432
https://www.youtube.com/watch?v=AkE8mtVv_yg

Decent card, 980 ti and fury x are similar at stock, and when OC'd the 980 ti consumes plenty of power (but is almost as fast as a 1070)

The value of the fury x IS less than the 980 ti but they are comparable cards and it by no means is a shitty card.

What is really the outlier is how good the pascal cards are.

>>61207432
http://www.anandtech.com/show/9421/the-amd-radeon-r9-fury-review-feat-sapphire-asus/17

also, check these graphs for more broofs of the dumb misinformation you're spouting.

Vega can't even beat a 1080 a fucking year old card and has double the tdp

How the fuck is this not a failure?

>>61207288
Yeah let's ignore every good product AMD/ATI ever released, you sad sack of shill trash held together with keyboard warrior jizz.

>>61207667
>every good product AMD/ATI ever released

You mean those from the literal past decade aside from ryzen? Amazing! Some top tier quality out there!

Why the fuck GPU threads are full of totally tech illiterate people?
Also why everyone fellates nVidia so much when they never did anything of value besides inventing fuckton of catchy slogans?

>>61207687
Because they have been consistently curb stomping ayymd since 2015

>>61207706
That's the reason everyone felllates lazy never-do-the-R&D shits like nvidia? Whoa. I'd better fellate IBM, at least these dudes do R&D.

>>61207687
>Why the fuck GPU threads are full of totally tech illiterate people?
/v/ spillover retards. Always happens when new GPUs are around.

>>61207418
>If it wasn't for Zen and poolaris they would be dead
if it wasn't for intel ensuring they don't get cucked by monopoly laws AMD would be dead
FTFY.

>>61207687
>>61207706
>>61207730

Hi HUANG, totally not obvious at all :)

>>61207771
I wish I was Jen Hsun Huang. I too love leather jackets.

>>61207687
>y-you just don't understand!
>>61207315

>>61207730
So what? The pascal arch works well for games.

Gcn doesn't anymore and scales like shit.

Amd literally flushed 2 years of Dev down the toilet on a dead arch and I can say this with confidence as I am a AMD GPU fan
>>61207762
Fuck Intel most of their cpus in the last 3 years are garbage netburst 2.0 tier house fire crap
>>61207771
Get fucked

>>61207211
amd has proven time and time again that their specs on paper aren't as good in reality compared to nvidia

>>61207687
>they never did anything of value

You mean aside from GPUs with performance AMD can only dream of coming close to, let alone surpassing, that don't double down as a heater? Oh you're right, they should have implemented some expensive worthless tech that doesn't provide any actual performance advantage whatsoever like AMD so we could get that nice 1~2 year old performance with double the power usage we've all been craving for.