https://www.youtube.com/watch?v=bhG AS_oGN3c

>>61144993
Damage controlling what? A card working in Fiji fallback mode?

We already know the drivers are completely unfinished. This doesn't tell us anything. Wait for Drivers(tm).

>>61145179
Well the werkstation performance seems decent for the price. I wonder how much actual WX Vega will cost.

>>61145196
Over $2000

it throttles to 1440 mhz. goddamn blower type coolers.

100fps average in dirt rally on 1440p
THAT'S BARELY BETTER THAN FURY X WTF

>>61145662
titan Xp was over 150. something is very wrong with vega.

>>61145662
Superposition test was 5500 compared to 9400 on Titan XP, that's almost HALF


Clearly something's wrong, this isn't even Fury X scores.

>>61145662
>>61145695
>>61145705
It's throttling due to the shitty blower style cooler, and the drivers are literally Fiji. If they can't pull out a miracle, RX Vega is going to be a dud on launch with these drivers. Or they're going to delay it again. Which would cause butthurt levels of a galactic scale.

>>61145734
>the drivers are literally Fiji
amd said these are vega drivers.

File: radda.gif (399KB, 320x239px) Image search: [Google]

399KB, 320x239px

>>61145734
>It's throttling due to the shitty blower style cooler
FURY X IS AVERAGE 90FPS
THIS THROTTLES DOWN TO 1400-1500MHz
HOW IS IT ONLY 10% BETTER THAN FUCKING FURY X WITH 40-50% BIGGER CLOCK!?!??!?!?!

>>61145780
Okay, so it took AMD 2 years to release Fiji with worse performance per clock?

Makes sense, why didn't they just scrap this and use a bigger Polaris?
These results make zero sense.

>>61145780
Source?

>>61145802
I agree, it makes zero sense. but then why did amd say these are vega drivers?
>>61145821
the fucking stream.

HITMAN HITS 85-90
THAT'S LITERALLY HOW MUCH FURY X HITS!
WHAT THE HELL!

Okay, so what the fuck is going on?
AMD could have literally released Fury X on 14nm and do better than this, why didn't they?

>>61145879
>why didn't they
they tried to fix the front end. allegedly.

Why would Raja do this to himself? Who's gonna hire him after this mess?

> what should I expect?

1070 level performance

These results are nonsensical, 15-20% slower than a 1080ti? Fine.

Fury X? Literally what

File: 14672963061.png (91KB, 653x726px) Image search: [Google]

91KB, 653x726px

heh.. don't worry guys.. HBM and DX12 will fix everything.. heheh..

RTG for life!

>>61146079

just wait for drivers, heh.... all will be well

Man it's a good thing their CPUs are bailing them out.

this is what happens when you fire all senior personnel and replace them with indians. AMD put every cent they had into zen and couldn't get a knockout win there, so now everything else is going to suffer from the technical debt of letting incompetent poos design hardware on a super hamstrung budget.

vega 75 fps in rise of the tomb raider
titan Xp is 115.

So what do you think the issue is?
Or has AMD really managed to release Fury X with 20% less IPC?

RAJA IS A SNAKEOIL SALESMAN

>>61146221
>So what do you think the issue is?

the issue is that AMD isn't innovating at all and is just rehashing the same design(s) from 2012.

and we're stuck with 2 more years of poolaris and vega until navi is ready.

>>61146210
that's like 15% better than fury x....

>>61146221
honestly either vega is completely broken or the drivers are actually in some sort of legacy mode that decreases the performance massively.
I don't see any other explanation.

sir pls do the needful and delete this thread

Anyone remember the Vega 1200MHz results from 4 months back? They literally score better than FE.

Seriously, someone explain what's going on here

>>61146079
HBM actually does look pretty promising long run

just not now...with this...

>>61146278
all lies naive kid

>>61146278
>someone explain what's going on here
we can't. there is no explanation for this. besides the drivers.

>>61146259
Completely broken? Why would they release like 4 different product lines for it then? Why waste money on making new drivers, making new enterprise/datacenter gear if it literally is worse?

>>61146306
For gaming it looks like a dud, but it's some 25% slower than a P6000(which is TitanXP quadro) in workstation shit without certification, so it's clearly doing fine there.

But the gaming results are all too confusing.

it throttles like hell in witcher 3 to stay under 280 watts.

Power usage is 280W
.....

>>61146278
>Seriously, someone explain what's going on here
as I said in >>61146259 it could be drivers in legacy mode. That's literally the only explanation.... Or they managed to brake it somehow between that run and FE release...

>>61146306
probably saying "we're cancelling a product" would result in their stock going back to 2$ per share

>>61146258

Shitty front-end and geometry engines.

>>61146344
>>61146306

oh and like >>61146332 says, the workstation shit isn't TOO bad. But still a bit underwhelming desu.

>>61146350
Workstation is actually in line what I expected, that's all in the certifications, and Vega won't have them this early because they're done by other companies.
It's the reason Quadros are like 4-5x faster than Geforce cards in that stuff.
It's not all too different for Pro AMD cards as well.

I would have loved a 14nm Fiji that fixed the original's bottlenecks.

They've done absolutely nothing, hell even regressed. It's like they expected HBM to do all the work for them.

RTG wasn't created to focus more on graphics. It was created to keep Raja from pooing on the rest of AMD

It can't do this bad without literal sabotage, something is not fucking adding up.

Two rx470's, or just Fiji with higher clocks, BOTH would do better, use less money, would be out a fucking year ago, do better in literally everything and not need new drivers.

It's like going into the woods to find a rabbit but the rabbit has drowned in an ocean 50 miles away.
Something's fundamentally wrong here.

clocks at 1440MHz through the tests
40% higher clocks, performance 15-20% over fury x
really hoping it's actually broken drivers

>>61146443
there are diminishing returns for clock speed so it might be correct

File: rtg_india.jpg (179KB, 750x868px) Image search: [Google]

179KB, 750x868px

>>61146437
>It can't do this bad without literal sabotage

What software does he use for benchmark

So what fucked up the architecture? http://www.anandtech.com/show/11002/the-amd-vega-gpu-architecture-teaser/2

Am I dreaming, have they managed to actually regress while literally using the same core layout?

What happened to those Sniper Elite and RoTR benches where it showed it performing like a 1080ti? Who's fucking who here?

File: current info.png (27KB, 763x670px) Image search: [Google]

27KB, 763x670px

What he's basically said so far.

File: 1496473914580.jpg (68KB, 500x552px) Image search: [Google]

68KB, 500x552px

>>61146552
>280W

I think if there's a few things where it performs "like it's supposed to" we can chalk it up to shit drivers. If everything is shit, it's probably doomed.

>>61146528
Drivers most likely, It should be at minimum 50% faster than a Fury X

>>61146552
Correct me if I'm wrong, but wouldn't you get better performance and the same power consumption if you just strip two rx480 together?

I seriously hope those pro drivers are gimping performance, because it does mostly fine there.

>It's clear we have a thermal issue

pony up that $500 extra for the CLC version goy!

>>61146736
It was hitting 58c at 100% fan speed and still throttling down to 1440mhz.

Seems like it'll hit a thermal limit first, but its also power limited. 300w needed just to sustain 1440mhz... jesus.

File: lisa su welp.png (332KB, 500x593px) Image search: [Google]

332KB, 500x593px

RAJA YOU HAVE ONE MONTH TO FIX THIS SHIT OR I SWEAR TO GOD

What's up with that 16 ROP bug?

>>61146902
Probably fallback compatibility mode for the drivers when they don't work (which is always)

1070 level performance

1070 level performance

>>61146963
>>61146985
BTFO as soon as drivers are ready

Good, hopefully they shut down their gaming division and focus solely on compute/datacenter/pro

I want these gaymer shits to pay $1000 for a 350mm^2 die

this is why amd is mess

start @ 39m
https://www.youtube.com/watch?v=YgGQmna-J3Q

Some AMD guy on some forum about the drivers: " It's not gimped (that would be completely ridiculous), but it is older. "

https://forum.beyond3d.com/threads/amd-vega-10-vega-11-and-vega-20-rumors-and-discussion.59649/page-119#post-1989522

Miners saved RTG.

>>61147189
They are right,

If the AMD drivers are not optimized for 3DMark (DX11), we have a problem.
Also the GTX 1080 have a TDP of 180w vs 300w for Vega...
And HBM2 was a huge mistake
What the point of this card?

whens Volta

>>61148167
Q2 2018.

Typical RTG launch day travesty.

Fuck Raja.

Performance per watt is worse than Polaris. WTF.

>>61148827
Shit was not ready for launch, at all. Only the workstation results aren't a complete disaster. One month, Raja. ONE MONTH.

>>61149005
Nothing ever is. But AMD GPUs just never seem to get it right. They have had 80% of their driver team working in crunch time since Polaris launched to fix this shit hardware.

Inex-fucking-cusable.

This is incompetence up and down Radeon Technologies Group.

File: deliverontome.png (35KB, 916x197px) Image search: [Google]

35KB, 916x197px

>>61144952

>60% more transistors than Fury X
>300W from the card alone
>performance regression

AMD accelerated Vega's demise, and sent the card supernova--and by card, I mean star; by launching the card into the star, and making the core chemistry to become imbalanced. Leading to a core collapse supernova. Which resulted in a blackhole, which is responsible for the performance regression noted.

>>61149005

>not ready for launch

Fucking how?! They had two years. TWO YEARS.

>>61149005
gaming drivers do not work, how long will it take you all to understand this?

workstation results are very decent specially for the price, gaming results are very driver dependent

they never tried to sell this as gaming card, watch financial analyst raja

File: building quad sli.gif (475KB, 912x782px) Image search: [Google]

475KB, 912x782px

>>61149602
>he bought only one

>>61150506

Only on a 6600k nig.

this is just extreme sand bagging. you will see ;)

>>61150621
>sandbagging meme
No publicly traded company is insane enough to do this on purpose.

How many fps does the fury x get on doom at 4k?

>TFW sitting pretty on a custom loop cooled Fury X
If Vega turns out to be a flop, I'll be fucking laughing and crying at the same time. Laughing because it'll save me the hassle of upgrading. Crying because Vega was meant to push AMD back into the high end market, and will put doubt in the minds of many that they can get back there.

>>61150689
If I owned doom, I'd tell you. Unfortunately, I don't.

The reason its not doing tiled rasterization is it has not been "turned on" yet. Vega is using the fallback rendering method instead of tiled.

When it was first discovered that Maxwell used tiled based rendering there was talk about a lot of software that needed to be written or rewritten in order to utilize it correctly and Nvidia implemented that in their drivers.

Vega is using a predominantly Fiji driver and this feature has not been "turned on" actually all but one of the new features in Vega is not functional right now the exception being the pixel engine being connected to the L2 cache as that is hardwired. I tore apart the new drivers in IDA today and the code paths between Fiji and Vega are very close and only differ slightly.

This arch is a massive change from anything they have released with GCN. They built in fallbacks in the hardware because of the massive changes. Its a protection against poorly written games and helps AMD have a starting point for driver development. Hell even architecturally Vega is essentially Fiji at its most basic thats why it is performing exactly like it because none of its new features are enabled or have a driver published for them yet. It is performing like a Fury X at 1400 MHz because that is exactly how every computer is treating it.

>>61149666
never underestimate poor management

RTG stinks of poor corporate culture and likely needs to be gutted completely. They need a keller level reboot

>>61150779

I hate to sound like a broken record, but maybe they should put out a statement saying something along the lines of what you just said.

Because right now, the market is basically seeing a card that runs far hotter, consumes way more power, is massively throttling, has 60% higher transistor density, and performs on par to WORSE than a Fury X, and overall sits between a 1070 and a 1080--and is 2 years fucking late to the market.

And them being mum about this particular issue, basically is telling anyone who might be considering this for a gaming card to go buy Nvidia instead.

>>61150779
they REALLY need to release those new drivers then, this is such a bungled release that people are going to treat it as the death of RTG for the next year.

fuck i want to fuck ryan shrout in the ass

Could it be gimped silicon?

>>61150921
No, it's using a driver that enables 0 Vega features.

Where are the uarch slides? SIGGRAPH?

>>61151002
The uarch slides are literally on Radeon website.

File: woozy.jpg (258KB, 1920x1080px) Image search: [Google]

258KB, 1920x1080px

I told you faggots. 1070 levels. now suck my cock

>>61151022
Those aren't uarch slides, there is no numbers.

>>61151054
Well that's about as much technical information about Vega as we have right now.

Way too fishy, AMD with its current cash wouldn't release a new Fiji.

Could be that they want content creators to get their hands on Vega first so they're dettering gamers from it? At least I fucking hope so.

>>61151290
It's not a new Fiji. But it's not working either. It uses what is basically a Fiji drivers.

Hi im vega engineer vega very powerful yes vary thank you thanm you for buying vega is very power and good