Realistically speaking is VEGA even going to be able to compete

Yes. AMD always competes in price.

>>59899915
No. Give up, buy a Quadro GP100 if you want HBM2 and use GeForce drivers.

>>59900165
but what if i want freesync 2 you dumb kike?

AMD ruined ATI. Before you could buy a hd4850 for $150 that would reck Nvidias high end 8800gts now their best cards are mediocre midranges that have to compete with both gimpworks and jewvidia $600 cards.

>>59899915
Rumors put it flogging 1070's. Possibly nipping at 1080Ti's heels.
I doubt AMD will release something that doesn't at least almost match 1070.
They've got the 580 now that does in the 1060.
There's no market for something that's a midpoint between that and a 1070-80.

>>59899915
10% faster than a 1080 on immature drivers, might see then match a 1080ti but are nowhere near a titan Xp and will be shortly outclassed again by volta

>>59900221
Good enough for competition.

>>59899915
Most estimates seem to put the top consumer entry as ~10% >1080 as far as power goes.

If that's all they can muster with HBM2 roll out and damn near a whole year of nvidia dominance in the high-range market, then that'll be pretty disappointing.

>>59899915
The Radeon group is seperate from AMD, and they don't share resources anymore, which was the issue before.

>>59900193

whats the point of having another Pascal? The 480x was already selling well and was cheaper then the 1060. I highly doubt that VEGA would come close to the 1080ti and Nvidia has gone full retard with releasing 2 titans card in 1 generation fucking with their high end market. If I were AMD i would just go ahead and release something to counter the 1070 at a much lower price.

>>59900287
Because they got access to better foundrys allowing them to remake the 480, but better. Now it really kicks the 1060 ass.
> If I were AMD i would just go ahead and release something to counter the 1070
That's what they're doing at least. AMD knows what's up.
They wont make a card that costs more than a 1070. If it does cost 20-30 more, its because it's taking swings at the titans.
They want the GPU market back.

>>59899915
>Realistically speaking is VEGA even going to be able to compete with Pascal
never

>>59900287
>whats the point of having another Pascal? The 480x was already selling well and was cheaper then the 1060
you mean another polaris?
AMD has been rebadging their GPUs since like 2011. R9 280x was a rebadge of the 7970. even nvidia is guilty of this with kepler, because maxwell was taking too long.

>>59900182
Well, you are retarded. Last ATI card was the x1950pro, hd4850 is AMD era.

>>59899915
Ok time for summary:

Vega is the biggest fundamental architecture change since they switched from Terascale to GCN 1.0
The new NCU means a massive boost to efficiency in the shader core and a much easier ability to fully saturate the GPU with work.
The biggest problem with Fiji(furyX) is that its 4096 shaders didnt mean very much because of large bubbles in the pipeline and large ammounts of shaders idling and doing no work each cycle.
Even with a low Level API like Vulcan which is best case scenario the FuryX is still not running with all cylinders firing because Vulkan only solved the Driver overhead problem. There is still bubbles in the workload because of the fundamental design of the GCN architecture itself.

AMD is switching from standard 4x16-bit blocks of a normal GCN CU(compute unit) to This new design of a variable 8+4+2+2 block and either four of these or a mix of these and standard 16-bit blocks in this new NCU(next compute unit)

They could not Do this in the past because they did not have the technology to build a Hardware Scheduler and Memory Scheduler Complexed enough to run this kind of architecture.

We have test samples of Vega with NCU's disabled and with very low clockspeeds and its beating a 1900mhz overclocked gtx 1080 in DOOM. Now there is a gain from using Vulcan over DX11 and OpenGL for AMD but thats on the Driver Overhead side not on the fundamental shader operations side. IT just means the CPU is feeding draw calls to the GPU at peak efficiency.

Alot of people dismiss the test because its vulcan but they dont know what they are looking at. The Vega sample is running FIJI 1.2 drivers Which means the instructions are running through the Vega NCU's like they are standard fixed 16-bit blocks like you have on a FuryX card.

So you need to look at it from this perspective.

This New architecture is more advanced than Maxwel/Pascal. You are going to see unreal performance from Vega when its finally released. Poor Volta

>>59900403

nobody I know of ever bought a titan. Titans are just a cash grab from Nvidia so i dont think AMD has to worry about those.Besides the 1080ti has already reached the final throughput with this architecture so theres little more they can do to make it better. They should just create a high end gpu with newer technologies that would go well with their FineWineTM branding. HBM2 is a good place to start. They should also concentrate on getting more Vulkan exclusives support to counter nvidia works which is very harmful to consumers to say the least.

>everything is just a rebrand of the 8800

>>59900559
this copypasta is so gay why does it keep getting posted

>>59900409

I think the point that I was making is that VEGA was scheduled to be released sometime in 2017. I think its weird to release a re brand next to a new architecture.

>>59900403
>Now it really kicks the 1060 ass.

what is there like a 100mhz difference in clock speeds? 50 mhz?

Vega engineering samples @1200mhz with alpha drivers were neck and neck with 1080s.

We already know the Instinct MI 25 card is clocked at 1500mhz boost clocks.

So expect the gamer card to overclock beyond that.

Nvidia releasing a fully uncorked titan should give you a hint about how powerful Vega will be. Right between the 1080ti and TitanXP^2.

>>59900669
the gaymer card is the one matching the 1080. the chip is as big as gp100. it's going to be dogshit, literally AMD's fermi.

>>59900667
Lower power consumption, and much greater overclocking potential. Unlike the 1060 that crashes at 2151mhz.
>>59900648
Because it's accurate.
>>59900669
>Nvidia releasing a fully uncorked titan
That's a little concerning. I have a feeling nVidias line up is ridiculously gimped.
I.E. The 1060 fully unlocked on Ln2 hits 8000Mhz...

it will end up just like ryzen
>g-guys it was meant to be a workstation gpu, totally not a gaming gpu with all the games in the background in the trailer

File: 1490702465219.png (68KB, 500x500px) Image search: [Google]

68KB, 500x500px

>>59900732
Keep stuttering that line 7700 fag.

>>59900442
>Well, you are retarded. Last ATI card was the x1950pro, hd4850 is AMD era.
considering length of development cycles, you are the retard

>>59900758
actually i want to buy a ryzen 5 but it's too expensive right now and i don't want a i5 (x264 encoding)

>>59900695

Dog shit at what? Offering quadro compute for 25% less? Or killing Nvidia market share as AMD will have their infinity fabric laden Vega and Ryzen combos running rings around Intel and Nvidia as their drivers choke any serious asynchronous compute?

And we haven't even seen the entire Vega line up. Multiple card variants like Fury with an air cooler, a nano, an HBM2 or a GDDR5 version, an AIO water cooled version. If the top end Vega hits 1080ti performance, what's a cut down six pin Vega nano going to do against a 1080? Volta won't save Nvidia either.

>>59900732
>it will end up just like ryzen
killing comptetion?

>>59900779
i meant dogshit at offering decent performance without melting your house into a ball of carbon plasma.

>>59900775
>actually i want to buy a ryzen 5 but it's too expensive right now and i don't want a i5
The mobo's are cheaper for Ryzen. It's the RAM you should be trying to be thrifty with. Unless you just flat out don't have the budget at this time.

>>59900823

The 1080ti has a 250w tdp. 50 less than Vega instinct MI25.

So what the fuck are you talking about?

When is Vega actually coming out? Why are they releasing RX500 series when Vega is around the corner?

>>59900963
Because the 5XX series id an updated 4XX card. Made on a better process allowing higher clock speeds.
It is in no way meant to compete with nvidias high end. They are mid range cards aimed at anyone that hasn't bought a 480, or aren't retarded enough to buy a 1060.
Vega is hopefully the 1080 killer.

>>59900790
it will be slower in games and faster in benchmarks

>>59900995
if slower you mean a difference of 2.3 FPS but an overall stutter-free smooth experience then i'll take it over nvidia any day.

>>59901026
you look like a biased man
i mean 10-20% slower but sometimes the same or even faster, but with older games mostly slower, just like ryzen

>>59900287
What's there not to like about having Pascal-tier DX11/OGL performance coupled with non-crippled DX12/Vk performance?

>>59901071
Ryzen has been proven to have lower frametimes unlike stutter-lake. Higher statistical average means nothing if the game hiccups every once in a while.

>>59901071
Yeah, at first.
The 1070/80's didn't preform at the FPS they do today.
If you really think day one benches are the be all and end all, you don't belong here.

>>59901082
>non-crippled DX12/Vk performance
ahem

>>59901096
>If you really think day one benches are the be all and end all, you don't belong here.
i do know that amd gpus get better in games with time

>>59901088
i really hope it will be like this in every game, not only in some games
and there is still the problem with the cpu not getting 100% utilized

>>59900250
Those estimates are only based on gcn Arch. In reality it may be 50 percent faster in Dx12/Vulkan at 4k in a game that is decently optimized for it's extra capabilities. Even the Radeon pro so beat the Titan X Pascal in 4k dx12

>>59900165
>and use GeForce drivers.
why would I want that? did you miss the news? nvidia drivers are botnet now

>>59900193
There is no way it's under a 1080, it has the same shader count as a Fury X which gives a 1070 decent competition, and it has double the memory and 50% higher clocks than the Fury X. Even if there are no improvements on utilization and IPC like they claim there is, then it'll be above a 1080. Absolute minimum is 15% below a 1080 Ti, maximum is 20% over. It all depends on how good they get resource utilization over Fury. If they can get it to scale to the same level as Polaris 10 then it'll be a good 10% over.

>>59901845
I looked at the other numbers from that link in a thread a few days go.
That >>59900193 comment was me being conservative based on that alone.
I personal believe that it will be 1080Ti tier on release. And make some surprising gains shortly after.
But at current stats there's a fair few points that it needs to come up in. But as said those scores are on shitty drivers and a low clocked sample.
I like to 'undercut' expectation because of shills around here. They will soon start trying to say that it's an Xp killer ect. To create false hopes, in order to smash them down.
Like how they were trying to say the 480 was a 1080 killer.
At this stage there is little price data. So our best modest guess is 1080 levels.
I'm looking forward to it. Even if it's at 1080 levels, because fuck nVidia.

>>59899915
for what its wroth, amd's marketing department has said "Vega performance compared to the Geforce GTX 1080 Ti and the Titan Xp looks really nice."

realistically i'm expecting halfway between 1080 and 1080ti

>>59900769
No, youre a retard anon. Not getting out of that one. Im not even the same person.

>>59902166

you the idiot here

The leaked bench, which should be taken with a truckload of salt, put it between the 1080 and 1080ti, but that was at 1200MHz, so who the fuck knows how the performance scales up to 1600MHz

>>59903147
poor volta

File: 1482915175613.png (832KB, 1024x768px) Image search: [Google]

832KB, 1024x768px

>>59903147
Dank as fuck.
mate if what you're saying is right, and amd's pricing at 1070 levels nVidia is done for. Mate, I'd buy three of them just in spite.
Sell the other two as "used" to people that would post a fuck ton about them and drive even more sales.
Really, nVidia does have something to be worried about. But not quite sure if to the level you state.

>>59903301
Polaris scales pretty nicely with clockspeed(and memory), some 5% perf from a 6% increase in clock, Vega won't have to worry about bandwidth so the pure clock scaling should be higher than Polaris as Vega has higher IPC.

>>59901181
>CPU not getting 100% utilized
it is an 8 core 16 thread, there is insane amount of headroom.There is a video of some korean dude playing 5 games at once on a Ryzen.

>>59903434
Your Mom!

File: 1490038010895.jpg (47KB, 480x480px) Image search: [Google]

47KB, 480x480px

>>59903798
nani?

>>59901964
>I'm looking forward to it. Even if it's at 1080 levels, because fuck nVidia
this desu, match the 1080 performance and i'm sold

will be perfect for my new work pc build that will likely run some flavour of linux for vm hosting

>>59903852
if it manages to match 1080 at launch then in six months or so it will become between 1080 and 1080Ti due to better drivers. Just look at the 290x at launch and now.

>>59899915

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

>>59904073
That guy is clueless.

>>59904099
Welcome to shittube hardware speculah, enjoy your stay

>>59903901
>become between 1080 and 1080Ti due to better drivers.
Try titan Xp.
Seriously. it'd a major shift.

>>59904073

U see this video posted one more time and I'm going to file a copyright complaint.

>>59903901
FineWine™

>>59903901
Just goes to show how much more competent Nvidia engineers are at squeezing the most out of their cards early.
As far as hardware is concerned though Vega looks promising, Nvidia releasing a second Titan is disconcerting.

Better luck with next year's 690X equivalent if it releases with HBM2, I have a feeling it will be hard to get a vega because they will fail to produce enough volume just like fury before it. As if announcing the 500 series refresh isn't enough of an indicator, they pulled the same shit with 300 series.
They can't keep getting away with this.

>>59905606
No it just goes to show how fucking ancient DX11 and the Intel/Nvidia platform is. Nvidia didn't have to change shit. Meanwhile DX12/Vulkan/Ryzen/Vega is a new incoming set of hardware and software for games developers to optimize for so will take time to mature.

>>59900221
but when are we getting volta

you know what's really sad about /g/?

that they are so much of a consumer whore that they don't see the motives behind nvidia and amd

>>59900559
>We have test samples of Vega with NCU's disabled and with very low clockspeeds and its beating a 1900mhz overclocked gtx 1080 in DOOM. Now there is a gain from using Vulcan over DX11 and OpenGL for AMD but thats on the Driver Overhead side not on the fundamental shader operations side. IT just means the CPU is feeding draw calls to the GPU at peak efficiency.
link

Damn too late.

Everybody bought already the 1070 and 1080.

Vega will be only a huge success if it costs 400 USD and gives more than 1080ti.

Nobody is going to spend time doing a sidegrade.

>>59905677
By the time DX12/Vulkan is relevant Nvidia would have come out with a new architecture better utilizing it, or rev up their underhanded gimping bullshit.

>>59905708
Q2 2018

>>59905742
why upgrading for games which are barely worth it

>>59905748
kek, better sell your 1080ti now when it's worth something

>>59905866
I am still on a 980, if I can squeeze close to 1080Ti performance with vega at 4K and VR for good price/performance and be able to buy it off the shelf then I would consider it.
They don't seem to have support for asynchronous re-projection yet, or equivalent optimization technique to Multi resolution shading implemented in current releases.

Don't know what the state of liquidVR is at the moment, but I know I already have access to these features on my current card which are handy. Async reprojection is just a safety net during performance dips, but Multi res shading does give a good boost to performance. Hopefully developers would implement a hardware agnostic solution, for now I go with what is currently accessible. Not in a rush to upgrade.

>>59905753
when are we getting zen+

>>59905931
Q1 2018

Navi? Probably Q3-Q4 2018

>>59904099

RX 480 8gb - 232 mm2
1080 - 314 mm2
1080Ti - 471 mm2
Titan Xp - 471 mm2
Vega 10 - 530-560 mm2

He might be right

>>59905922
>AMD announced back in February 2017 that they were working on adding asynchronous reprojection support to their GPUs, and the April 2017 ReLive 17.4.1 driver released for Radeon RX 470 and RX 480 cards now delivers this promise

???

https://www.vrheads.com/what-asynchronous-reprojection-and-how-do-i-use-it

why everyone is putting Vega on pat with gtx 1080? Vega is way too big chip to have such low performance.

>>59906080

this >>59905997

>>59905997
No, he's clueless, he didn't even mention the die sizes.

First of all, we already have a bench of a 1200MHz Vega against a 1080 and 1080Ti, even if it is a compute benchmark there are 3D benchmarks included for the manchildren.

Secondly, we knew Vega's core setup and clockspeed in DECEMBER, after that it's simple napkin math to gauge its performance without any architectural improvements(and there's too much of those to get a proper gauge on performance, so without them is worst case and a good baseline)


tl;dr/tl;dw, some fag coming to a flawed performance conclusion on hardware without any technical background or investigation.

>AYYMDPOORFAGS THINK AYYMD WILL SELL A 500mm2 GPU AND EXPENSIVE HBM2 BELOW $600

TOPPEST KEK

>>59906148
>he seriously thinks 11GB GD5X and its accompanying traces are cheap

>>59906122

>vega 530 mm2 with hbm2 performs on par with 1080 314 mm2

how can one be this autistic?

>>59906194
Where did I even state that you retard?

>>59906179
Still way cheaper than interposer and doing all the through silicon vias for HBM2, SK Hynix can't even deliver HBM2 higher than 1.4GHz on their website, meaning 400GB/s bandwidth only with 2 HBM2 stacks, slower than a 1080 Ti's 484GB/s bandwidth

Dumb moron like you don't know anything

>>59906207

that's the context of what you're disagreeing with, check the replies timeline, retard.

>>59900193
>flogging
>nipping
>does in
reddit

File: 1491630247557.png (190KB, 1000x1408px) Image search: [Google]

190KB, 1000x1408px

>>59906236
It's the context that your head came up with, if anything I just stated that Vega is doing rather fucking well against a 1080ti with much lower clocks than retail and infant drivers.

>>59906019
That is good news, which leaves MRS.

Only released game I own so far that makes of use of MRS is poolNation VR.
I'd have to do some more digging at least with regards to performance comparison between 390X/Fury and 980 in pool nation VR or equivalent to see if that feature is worth it or not. All it does is reduce the amount of redundant pixels being rendered that are lost or not easily noticeable when the image is warped to properly match the optics.

>>59906080
>>59906091
/thread
Even if they just produced a 560mm^2 Polaris it would be almost on par with a 1080Ti.

>>59906220
Way cheaper? Prove it, you're acting as if that much GDDR5X costs fucking nothing, apparently the 384bit memory controller using it costs nothing too.
Is 2 stacks of HBM2 on a interposer more expensive? Probably, but to a notable difference? Very fucking unlikely, Fiji had more HBM stacks and the Nano was still sold for like $500

>>59906282

so then what the fuck are you whining about here >>59906122

in the video they say that vega is going to beat the 1080 and match/outperform the titan XbigP, the pic you show >>59906282 shows that as well

you're basically agreeing with him except you're showing other evidence of this

stop crying faggot

>>59906380
I'm crying because he came to a conclusion based on fucking nothing..
"I think the Vega will be near the Titan X.." based on fucking what? I can claim that Navi will be either 30% slower to 50% faster than Volta, great huh?

>>59906411

>speculated
>was right

who gives a shit anon, he was right even if he came to the conclusion out of thin air or didn't provide the evidence of how he came to that conclusion

>>59906468
But that's nothing but a lucky guess.

As said I can claim in 10 different videos how X hardware will perform in 10% increments compared to the competition, one of them will be true at least!

>>59906492

Holy shit at how triggered you are that somebody made a lucky guess, step back from the keyboard and get some fresh air.

>>59906511
Suck my cock, dude

>>59903147
considering AMD has scaled almost perfectly with clock speeds in the past, I'd say they're going to be at or a little above the 1080ti if the leaks are true.

>>59906606
You can see the drivers are simply not there yet, there are tests where it significantly loses to the 480 even.

>>59906520

MI25 is already confirmed... which means minimum clock of 1525mhz is confirmed for professional cards.

This is likely an ES like Ryzen was at around 2.8-3.2Ghz

>tfw becoming a major waitcuck

Is Vega/Volta really gonna be worth it lads?

>>59906769

vega = yes, you can build a 4k pc for ~$1000

volta = we don't know but since it's nvgreedia, don't hold your breath for that.

File: amd-deck-03-840x472.jpg (78KB, 840x472px) Image search: [Google]

78KB, 840x472px

>your face when a few of the April Fools slides end up being true

Hiding the truth among lies.

>>59906769
maybe, maybe not. I'm not in a hurry so if vega disappoints I'll just get a 480/580.

>>59906874
Is this fake? I want more info

>>59906926
http://wccftech.com/vega-teaser-slides-leak-nda/

It's clearly fake, but not all of them, was an April 1st joke.
For example the clockspeed is not a lie.

>/g/ is utterly convinced AMD would be bankrupt in 2017
>shares increase nearly tenfold in 6 months
>/g/ is utterly convinced ryzen couldn't hang with Intel's mightiest
>get proven wrong
>/g/ is convinced vega can't hang with Nvidia's mightiest consumer cards
>????

Remember, /g/ is always wrong.

>>59906908
So you buy card based purely on fanboysm?
both 480 and 580 are poointheloo shit for price/perf
compared to nvidiot2017 mid range cards.

>>59907169
There's no 6144 ALU version of Vega, period.
Not only is there no fucking way to fit that on a single die(multi die packages are 2 years away), the top end Vega cards for enterprise are 4096 ALUs

>>59907169
Why is it written by an Indian

>>59907011
Forgetting that /g is just bunch of edgy teenagers
that care only about muh FPS in games that need max 2 cores. = they were right

>>59906953
Augmented reality to make it blievable for the normies

>>59907169
What a fuck up it was by nvidia to sponsor DOOM

I mean it's cool for the devs they got to be the first real game to get out there with Vulkan but Nvidia basically paid them to advertise for their competitor

>>59907191

>that care only about muh FPS in games that need max 2 cores. = they were right

Then you buy the 7350k and pray you don't run anything in the background and enjoy 1% lows lower than a whores drawers.

>>59906874
What kind of ports are those I don't recognize them.

File: spooky.png (657KB, 1600x1200px) Image search: [Google]

657KB, 1600x1200px

How many generations do I have to wait before upgrading from hawaii GPU? It performances on par or better with poolaris in games and Fiji(fury) isn't much of a improvement over it to buy on the spot again.

>>59907290
DP and HDMI?

>>59907204
It's only a fuckup because someone in the engineering department neglected to tell someone in the marketing department that their hardware actually can't do async compute and dx12 / vulkan while their competitor can.

File: 1795t4.jpg (81KB, 500x616px) Image search: [Google]

81KB, 500x616px

>>59907301

Vega is seemingly aimed at getting hawaii owners to shift. My 290x won't cut it for 4K and i'm hoping vega does so I can go full AMD this upgrade cycle (ryzen yo).

>>59907343

See pic.

>>59906874
its fake because die on has 2hbm stacks and support max of 4gb per stack. so there is noway in hell it will have 16gb hbm vram.

File: 50916_02_sk-hynix-hits-mass-production-4gb-hbm2-q3-8gb-q4-2016.png (149KB, 620x464px) Image search: [Google]

149KB, 620x464px

>>59907832
Eh... There are clearly 8GB stacks in the works.
4GB is way too small for enterprise and HPC, even at 4 stacks.

File: wpshrine_Lucky_Star_91_1680x1050.jpg (435KB, 1680x1050px) Image search: [Google]

435KB, 1680x1050px

>>59899915
Well, you said realistically.. nobody will know until benchmarks come out. benchmarks are REAL. all the big talk is just a bunch of shit designed to make the hype go up so that when the card is relaesed ppl start buying the vega $$ i doubt there will be a significant difference in performances with NVIDIA's top tier GP102 pascal 1080ti. they will probably be roughly the same. with AMD getting more bang for buck then NVIDIA like they always have. Buy a 1080ti now and you won't get cucked by Vega i guarantee this. you might as well too if you already have an expensive G-sync monitor. or you could just wait till 2018 for volta. if by some miracle AMD has a huge gain over 1080ti for sligtly cheaper (which i highly doubt) then volta will probably come faster than 2018 forcing NVIDIA to compete harder with AMD.

File: b0dc5fa189d9a52116889fd04ae3106c_XL.jpg (107KB, 1200x600px) Image search: [Google]

107KB, 1200x600px

>>59907940
amd is using 4gb stacks, 8gb arent available yet, its the same reason nvidia GP100 uses 4 4gb stack...

>>59907011
What are you even talking about? /g/ sucks on AMDs dick like they're being paid to

>>59908107
But they'll be available? You were claiming that 4GB is max of HBM2

>>59908136
If they're not available yet then how the fuck are we supposed to see Vega in Q2 of this year? They simply won't have the time to produce the cards

I belive that when vega comes out and it is given real fair benchmarks amongst a wide variety of new games that it will be just as good as a 1080ti or slightly less since it's bound to be cheaper than NVIDIA. the DOOM test they did with vega/vulkan/4k was perfect for testing vega cause they have the advantage of manipulating or possibly using tom-foolery for Vulkan api and the graphics settings to be the best for DOOM. i doubt that card will be able to do that for ALL triple a titles.

File: 1456292519685.jpg (104KB, 600x800px) Image search: [Google]

104KB, 600x800px

So is it a bad time right now to buy a 1070 or 1080?

That we should wait to see Vega first and thats definitely getting shown in May?

>>59908117

Only since march 2nd 2017.

>>59908157
Nobody said there will be Vega with 16GB memory, that was an april fool's joke, even AMD states Vega goes up to 8GB

But 16GB down the line isn't impossible, especially for enterprise crap

>>59908136
right now max is 4gb, im sorry i triggered you by not elaborating on my answer some more...

>>59908166
No? Do you not remember back during the 290x vs 970 days? It's been, at the very least, a couple of years.

>>59908161
Getting a non-Ti 1080 is full retard

Getting a Ti is acceptable though here probably will be an 1100-series half step this year before the Voltas next year.

>>59908178

> Do you not remember back during the 290x vs 970 days?

Do you? When the 970 dropped /g/ spent weeks declaring AMD dead and buried with a few anons insisting on finewine technology. It wasn't until the 970's vram partitioning become apparent did /g/ swing the other way (along with the usual damage control).

>>59908269
/g/ always talked about how AMD was the way to go, but nice revisionist history tho

>>59908161
Yea, Vega will have good equivalents for the 1070/1080 that are going to be cheaper bang for buck. Remember there is no DEFINITE until the real benchmarks start coming out. AMD testing DOOM in 4k with vega/vulkan and getting 75/76 FPS on ultra doesn't count cause they have the advantage over using little tricks here and there to achieve that FPS since vulkan is AMD's baby. watch when the real benchmarks come out on alot of variety of titles i bet it will be different and equal out with the 1080ti for their best vega card. don't get all hyped by all the talk of new architecture. just wait it out and see the real benchmarks and laugh.

>>59908224
Yea an 1100 half-step. If AMD vega by some miracle outperforms their TI and then NVDIDIA will come in and be like heres the 1100 so hopefully aMD doesn't get any real edge or else we'll get cucked by the GTX1100

Question: does MSI Afterburner work well with GTX 1080ti no matter the brand? like for example the STRIX oc one.

>>59908421
Yes.

>>59902977
no u

>>59905748

>new arch

More like they need to remake everything to support async compute at the hardware level with a massive scheduler to handle it dynamically, the way AMD does it, and the way DX12 and Vulkan expects you to design your arch for.

Of course, if Nvidia does that, the entire card will burn to the fucking ground--as you'll have to dedicate a significant amount of die space to said scheduler. Further, it will suck up a lot of power and generate a lot of heat. That's the main reason why AMD cards have run hotter. They've had a hardware scheduler in them since GCN was first introduced in the 7xxx series.

Nvidia had one too, that's why the 4xx series is synonymous with house fires, and why the 5xx series though improved, still ran very hot. The 6xx series was the last generation where the hardware scheduler continued to exist on die, and thereafter Nvidia gutted it. As a result, it allowed them to massively pull ahead in GPU core clocks. They then invested heavily into handling the concept of asynchronous compute at the software level via command lists to maximize workload splitting to concurrent CPU threads.

It's for that very reason that CPU utilization on Intel CPUs and AMD CPUs in DX11 in any game, is higher than with AMD cards.

AMD's model: CPU feeds the GPU, the GPU decides how to best split the workload

Nvidia's model: CPU feeds the GPU, but the driver intervenes and decides for the GPU how best to split the workload.

To use a car analogy, since this entire board loves those:

>AMD
>driverless car for the last four generations
>each new model has improved on previous generations
>The ability to navigate roads/highways and deliver passengers safely has improved greatly over time, but its not at its best yet

>Nvidia
>traditional taxi service
>the driver knows the best shortcuts, and has all the tricks to get you from point A to B efficiently and quickly
>but if you replaced him with an AI suite tomorrow, the vehicle will probably stall

>>59909155
>no food analogy

Didn't understand a word desu

File: 1464988884962.jpg (87KB, 986x553px) Image search: [Google]

87KB, 986x553px

>>59909187

>>59908341
>we'll get cucked by the GTX1100
on the contrary, technology improving at a faster pace is better.
Poorfags get better bang for their cuck and emofags get their self harm fix by riding the bleeding edge roller coaster of fun times on a regular basis.

And everyone lives happily ever after, except you because you're a faggot.

https://forum.beyond3d.com/threads/nvidia-pascal-reviews-1080xp-1080ti-1080-1070-1060-and-1050.57930/page-38#post-1974379

>Just another useless youtube video whose author does not have a clue what he's talking about... Can I have 20 minutes of my life back?
It's hard to even comment on this with people not having a clue about what a scheduler does and which scheduler does what (Yes there's more then one). Fermi vs. Kepler scheduling has been discussed around here quite a bit already. It has to do with how warps are scheduled after kernel dispatch or draw call is issued. That is to say after a shader instruction is executed for a group (warp) of 32 pixels/vertices/threads when can an SM pickup next instruction from this same warp. This was done in hardware in case of Fermi and has been done in software since. The way this software scheduling works is that a shader compiler will emit control codes/hints within shader code. You can read more about it here. This is compile time job. Once shader is compiled CPU doesn't do anything anymore.
How this works against his fantasies further on is anybody's guess.

AYYMDPOORFAGS STILL POSTING LIES ABOUT ASYNC, THINGS THAT THEY DON'T UNDERSTAND

TOP KEK

File: 1325270032277818150.jpg (25KB, 320x304px) Image search: [Google]

25KB, 320x304px

>>59909155
>Of course, if Nvidia does that, the entire card will burn to the fucking ground
What's their plan?

>>59907011
/g/ is not one person anon

>>59908063
The hyping is so people with stocks and pump and dump. Didn't you learn anything from the last couple of launches?

>>59904099
obviously not, he was completely right.

>>59909545
> except your because you're a faggot.

Explain. Do you know who i am? No. how can you judge someone bades on just my comment about getting cucked. You wanna argue? i can go all day long lets go around in circles arguing and get more and more pissed.. Lets! :)

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

42KB, 653x726px

>>59909637
DELET

>>59909637
>>59911366
samefag

get out with your shitty bait/shilling

File: tidus.jpg (30KB, 475x343px) Image search: [Google]

30KB, 475x343px

>>59911395
>poojeet has no rebuttal

>>59900669
this is another one of those 'on paper' scenerios.

nvidia likely got wind of what amd cards can clock like when not passively cooled, but remember, we are going to get shit binning again so expect no oc headroom.

what nvidia likely knows is

1) whet vega clocks to
2) probably has a few engineers pulling out their hair trying to reverse engineer amd cards from marketing speak and tech events where they talk about it and what efficiency gains they may b e able to expect from it.

If it's all just clock speed, for amd to match a 1080ti, they need to clock to 1750 on polaris levels of efficiency
to beat a titan Xp, 1925mhz

if amd is able to match a 1080ti at pro level of clocking, they need 1750 mhz to get into titan Xp range,

but speculation is speculation, what we know for a fact is the gpu clocks to 1525, and amd would not put a gpu with worse efficiency out, so assume polaris level of efficiency.

it puts them half way between a 1080 and 1080ti.

>>59911311
Calm ya tits, Its just bantz m80

>>59903147
>>59906606
Probably slightly behind or slightly above in 1080p.
But completely stomping the 1080Ti's shit in 4K.

>>59906874
>>59906926
They're fake, but based on reasonable assumptions.

We know there are 11 Vega 10 SKUs and that some of these are for servers like the Radeon Instinct MI25.
But there must be more than one of those that are for consumers, probably 3-5 of them.

So the reasonable assumption is that there will be a 4GB HBM2 model clocked around 1400mhz.
Maybe mITX form factor, and probably around $350-$450.
Then there will probably be the 8gb HBM2 model that's higher clocked than the server one that has been seen at 1536mhz iirc. So probably 1600mhz+. It's typical for consumer models to be higher clocked outside of the usually optimal perf/watt ranges.

>>59907940
Hm. So they could have 4GB without reduced bandwidth compared to 8GB?

>>59908161
You could probably still sell the 1080Ti and not lose much thanks to Nvidia fanboys.
I think 1070 and 1080 value is going to tank, though.

>>59909187
The first sentance that anon gave sums it up:
Nvidia hardware is designed for DX11 and OpenGL.
DX12 and Vulkan, with will always outperform the best possible OpenGL/DX11, requires hardware like AMD's.
AMD's architecture is simply better. Plain and simple. But OpenGL and DX11 hold it back.

The reason Nvidia didn't release a driver for DX12/Vulkan on Fermi is because Fermi's architecture is more like AMD's since GCN. The older Fermi card would have crushed their newer Maxwell and Keplar cards in many cases if they could take advantage of DX12/Vulkan.
And instead of going back to that type of architecture, Nvidia bet on paying devs to hold back their games with DX11/OpenGL, and hold back progress in general, since they had a DX11/OpenGL advantage against superior hardware.
(which worked)

>>59911767
>nvidia likely got wind of
I don't know how so many people don't understand that corporate espionage is a thing.
a huge thing.

All big companies roughly know what other big competing companies are doing.

>>59899915
I'd say there is high chance that it is ~5 % slower than 1080 Ti on launch, and during one year it will catch up to be 1% faster (that is how it more or less went with 980 ti and Fury X).

In games that will get FP16 optimizations, it will likely beat both 1080ti and the Xpppp Titan.

I think there is some non-zero chance for it to be actually fater than top nvidia on launch, but that chance is smaller, so count on Radeon RX Vega to be between 1080 and 1080ti (closer to ti).

File: Vega Final Presentation-27_575px.png (349KB, 678x381px) Image search: [Google]

349KB, 678x381px

>>59911767

We know for a fact it's not just clock speed improvements. It's massive IPC gains and data throughput. It's primitive shader culling, it's load balancing.

Fury was a great card that no developer ever took advantage of simply because half of it's functionality was waiting to be utilized as the APIs still stuck with single core supremacy. Now we know how AMD intends to let developers use Vega fully, by having hardware schedulers not just break down graphics and compute, but schedule threads and CPU access so that half of it's shaders no longer sit unused.

>>59912431
Personally I'm more interested in the tile based rasterizer, the 2x throughput per clock geometry units (great for gimpworks and its ilk) and L2 to ROP association.
The first one is a massive perf/watt win while the other two are pure pixel throughput wins

File: GPU Compared.jpg (56KB, 512x512px) Image search: [Google]

56KB, 512x512px

>>59899915
In high-end? No.

Vega will be Ryzen, GPU edition. Maybe a little bit less performant still, but not bad either.

File: fury 4l.png (31KB, 500x690px) Image search: [Google]

31KB, 500x690px

>>59912224
The Fury X is still slower than the 980Ti at 1080p.
But it's faster at 4K.

File: AMD-Vega-Slides-Architecture-Diagram-500x281.jpg (26KB, 500x281px) Image search: [Google]

26KB, 500x281px

>>59912516

It's amazing that people don't understand the impacts those different "engines" are going to have in comparison to Fiji.

If the tile based rasterizer is half as good as Nvidia, then efficiency is going to be extremely competitive.

>>59912570
And they're equal at 1440p

>>59912576
It's going to depend heavily on whether "rapid packed math" and if the detection and discarding of polygons that aren't visible will be automatic, or if they'll require developers to optimize for.

If it does it automatically, or in the driver, than it'll be fucking huge and we could see Vega stomping the 1080Ti.
If it's game by game optimizations which is often the case.. then maybe we'll see Prey do 30% better at 4K on the Vega card over the 1080Ti, but then slightly less at everything else.
That's the problem.

>>59912580
It's a bit less at 1440p unless you lower tessellation and other GimpWorks stuff.

>>59906148
580 is $400 canadaroos

1070 is $600
1080 is $800
1080ti is $1100

How many variants of Vega are there? 2? 3?

It's honestly hard to figure where AMD will price everything because Nvidia's skewed the market so much with their overpriced GPU's.

I'd be willing to bet the top end Vega is cheaper than the 1080ti, and the low end Vega is cheaper than the 1070.

I'm a little worried about my PC case clearance for the ASUS strix 1080ti. I have a NZXT source 220 case and have measured the differences of my old and new GPU and here's what i need expressed in inches:

Difference: Length: 0.7457 need this much more.
Width/Height: MSI GTX 980 has more
Thickness: 0.65268 need this much more.

think i'll be O-kay?

i just bought a Zotac GTX 1080 AMP edition for 400usd, how is my Price/Performance compared to Vega?

>>59912632
There's no way in hell there's that much of a price variance from a single die with 2-3 bins, tops.

>>59912676
Sounds like a top of the line cooler (amp edition) you did well check those vega benchmarks when they come next month you'll be neck and neck with their high end vega most likely.

>>59912624

Here's the thing that developers are finally catching on to when it comes to optimizations.

Optimizing for DX12/Vulkan and by extension AMD hardware capabilities benefits everyone. Across the board performance gains which means no more brute force approach to PC gaming.

Optimizing for Nvidia only gives a modest increase to Nvidia hardware and completely fucks over AMD cards.

>>59912690
Yep.
The AMD recommended way to thread a game and to use Vulkan improves performance for anyone with 4 threads, which includes people with G4560s.

It's just the superior way to fucking program, but most game programmers are awful and/or bribed.

>>59912676
Around $400 is where I'm expecting them to drop after Vega launch, but with more choice instead of just getting what's on sale some given day.
If it's a good 1080, that's still a nice price.

>>59912742
NVIDIA being the competitor will surely do somehting simillar with how DX12 works later and work with game developers for optimizations.

>>59912765

They already announced Gameworks DX12.

And it was almost universally eyerolled.

>>59912765
It seems likely that Nvidia claiming such a massive performance increase with Volta is that it's going to be that performance increase in DX12.

Currently their DX12 is far worse than DX11. Volta might just be 10% better in DX11, but 30-40% better in DX12 *~compared to their own shitty DX11 cards~*.
So they might be going back to hardware scheduling to make those performance improvement claims true.

I think I'm gonna buy a RADEON R9 FURY 4GB HBM NITRO TRI-X is it a good choice?
I can get it for as low as 260Euro

Where as a gtx 1070 is at best 400Euro and a rx 480 8gb is 255Euro, I wanted an AMD because freesync

>>59912799

They will HAVE to improve compute for it to be taken seriously at the very least.

>>59901243
He is suggesting that you use GeForce, aka, the consumer video game brand of drivers for a professional non-GeForce drivers that come naturally with a Quadro.

>>59912799
A hardware scheduler is a massive change, I don't think you'll see it in Volta.

>>59912845
I think that's why it'll come with Volta.
Volta was planned for 2016 and pushed back to 2018.
It's what Nvidia has been working toward for a long time, like how AMD has been working toward Vega since before Fiji.

I see no other way that Nvidia can reach the performance they claim with Volta without them having meant DX12 performance and it being a significant architecture change.

>>59912829
It's more than just compute.

>>59912765
>>59912781
Nvidia does so well with DX11 because their driver is a marvel of software engineering. It's able to break up work being sent to the main render thread from a game so that multiple cores can work on it, then reassemble the render thread so the GPU can receive it and it does that on the fly with minimal performance penalty. However, I doubt their software solution will be able to scale the same way a hardware scheduler can. For example, AMD's hardware scheduler can receive instructions from up to 64 threads. To try to split a render thread into more than three pieces is probably not something that's technically feasible, which is probably why we see Nvidia cards perform so poorly with Ryzen. DX12 exposes hardware schedulers to game engines in a way that wasn't really done before, so it kind of renders Nvidia's software solution obsolete. Gameworks is just a set of tools that helps create graphics assets that are optimized for Nvidia; it's nowhere near as low-level as DX12 or Nvidia's DX11 driver.

>>59912909
>I doubt their software solution will be able to scale the same way a hardware scheduler can
It absolutely won't, really.
It's the application (game) itself that needs to break up instructions across multiple cores. Trying to do it after the game is already sending instructions won't work.

It works with Nvidia because it has a lot of work to be done that more cores can leverage. There is much more overhead on the CPU with its software scheduler. With a hardware scheduler, there is almost nothing that needs to be done to those instructions.

>>59913268
>It absolutely won't, really.
I'm sure there were people at Nvidia that thought removing the hardware scheduler and handling threading in software was impossible, so I'm not going to discount the possibility that Nvidia rewrites the driver to work with more cores, but again, I doubt they will. At this point, I imagine the cost of doing it in software hugely outweigh the benefits of having an arch with no hardware scheduler, especially since AMD has paved the way for widespread HW scheduler support.

>>59913313
They did.
And it works for DX11/OpenGL.
Does not work for DX12/Vulkan where there are many threads sending instructions directly without needing much driver interference.

t. decent enough graphics programmer