Daily Wait for Vega Thread /w4v/

>>60213299

Not sure how to feel about these unverified leaks. Seems like it would make sense that AMD would release a weaker version of vega which leaked first.

Looks like they're directly competing with 1070, 1080 and 1080ti

File: amd2.jpg (52KB, 568x612px) Image search: [Google]

52KB, 568x612px

>I-It's going to bankrupt nvidia

>>60213324
>weaker
that score is higher than the 1080ti's lol

>>60213324
That's baby Vega 11 for you.
I don't think we've seen any leaks of Vega 10, or at least none of them line up with what the leaked advertised specs.

>>60213344

Yeah I meant the original benchmarks that leaked, show performance just above a 1070.

>>60213356

Yeah I think we've seen the lowest and highest powered so far. One above 1070 and one above 1080ti.

I hope Vega bombs, I already spent $1k on a gsync monitor

>>60213480

I've got a freesync monitor and willing to buy a 1080ti if vega has 10% less performance.

>>60213480

Bad move, that shit is so over-priced.

>>60213480
G-sync is already dead you dumb cuck, it's just a question of how soon it falls over.

The HDMI 2.1 spec has VESA Adaptive-sync style VRR.

>>60213480
>goysync in 2017

>>60213480
Man, Nvidia is gonna end up looking real stupid once Intel picks up adaptive sync

>>60213276
Let's all over hype it so when it comes out it can in no way meet the unrealistic expectations we've made in our heads

Still waiting

MAY 9TH
A
Y
.
9
T
H

>>60213535
how does this work then? is there some sort of setting that you enable that just causes "freesync/gsync" to kick in or does it just do it automatically?

File: 042d9a6f5fd42f2a9f8600880b948dad(2).jpg (282KB, 650x989px) Image search: [Google]

282KB, 650x989px

>>60214892
That's my birthday. Shiggity.

vega c3 is what i want and i want it at 350$ msrp if i can get it for that much i will
i want it to be roughly 1070 tier performance on average and better in vulkan and dx12

>>60214960
Dunno how HDMI 2.1 VRR will work, but VESA Adaptive-sync works by having the monitor send an extra flag at connection handshake time saying that it's able to accept any updates that occur in a certain frequency range, then the GPU or whatever just sends frames whenever it feels like, so long at the timing windows match.

Dead simple.

G-sync is a hack where the display scaler communicates back and forth over the DP link to tell the GPU when it's allowed to display a frame, which adds some tiny fractional millisecond of extra latency but mostly is just sloppy engineering.

It's currently broadly assumed that HDMI VRR will just build off the FreeSync over HDMI stuff that AMD already developed, which is again basically just the VAS protocol translated to HDMI.

File: speccy_muh.png (32KB, 657x524px) Image search: [Google]

32KB, 657x524px

So there gonna be a $300 vega? Will it be much better then the 4/580?

I'm a huge bang-for-buck style guy, usually purchasing something new every 2 years. However i stayed with the pic related set up for the past 3 years and am now feeling the the age.

Was thinking a whole new set up, with Ryzen 1600 that was suggested, but undecided on going for a 580 or wait and see if Vega has a lower end/bang for buck model.

>>60216357
so it'll just be plug and play for the display to synchronize its refresh rate with the amount of frames that the GPU is outputting?

>>60217833
pretty much.

>>60218118
awesome, last question but when will monitors start supporting HDMI 2.1 and DP 1.4?

If Vega releases at Computex, that lines up well with my college rebate.
Hopefully their 1070-equivalent isn't ridiculously expensive in comparison.
>unintentionally being a wait™ fag

>>60218197
DP 1.3/1.4 is (barely) already happening.

HDMI 2.1 is taking a while to finalize, so 2018 model TVs in the absolute best case with 2019 probably being more likely.

>>60218202
hbmeme are expensive

>>60218389
beats my gtx 760

>>60213276
how many TMUs did Hawaii, Fiji, and Polaris have?

>>60213356
>That's baby Vega 11 for you.

No, this is all vega 10. vega 11 is next year.

They are just making multiple SKUs, like how they had Fury, Fury Nano, Fury X, and Fury X2. We'll have Vega, Vega Nano, Vega X, and Lou Vega Mambo No. 5.

File: 1493742333660.png (1MB, 1440x2392px) Image search: [Google]

1MB, 1440x2392px

>>60214892
you mean 8th.

>>60218483
4 GB of HBM2
8 GB of HBM2
16 GB of HBM2

>>60218389
Nah, it costs the same as any decent GDDR5 setup. Especially considering it's only two stacks.

As good as the 1070 with twice the TDP bet on it

>>60219137
>twice the TDP
Anon, Vega uses tiled rasterisation.

>>60219137
>thermal design poweremission instead of actual wattage

you don't truly belong to /g/ anymore

just curious, what are you all even waiting for?

what amazing game is currently out that a GPU in a similar price range couldn't easily handle?

there are literally NO GOOD GAMES OUT that need a next gen GPU

what the FUCK do you even want vega for

just buy a 1080 or a 1080 ti and be done with it

>>60219211
>just buy nvidia goyim
Happy merchant pls go.

>>60219211
I'm waiting for a 1070-equivalent to compare the 2 choices, and grab whichever suits me best.
Then, I wouldn't have to upgrade for ages.

>>60218859
>Nah, it costs the same as any decent GDDR5 setup. Especially considering it's only two stacks.

HBM itself is cheap, but the problem is that it needs an interposer so it can be put next to the GPU, and this completely ruins whatever yields they may have for the GPU itself (which might be pretty low too, since it is a huge chip).

Vega will start at $500, minimum.

>>60219264
I'd say $550-650 for big Vega.

https://youtu.be/b_ZMOn0X6jw

>>60219269
4u

>>60219187
There was a time when amd used them interchangeably

>>60219211
the moment you buy some tech shit, it is already outdated. i'm waiting to the very last second to buy shit, and because this tards are throwing shit closer and closer, i don't know really.

i hope to buy some day a ryzen + vega, or a 8th gen intel with volta. i live in a third world shithole, and have to plan for US trips with months in advance. i wish they made clear the launch dates so it wasn't so unexpected for me. and that's assuming they will let my in in the first place. it is cheaper to fly to the US and buy computer shit there than bring it and pay absurd taxes for a corrupt government.

fuck

>>60219282
AMD sometimes uses TBP and sometimes TDP, anyone with basic hardware knowledge will know the difference.

File: 1382663872113.png (497KB, 973x2259px) Image search: [Google]

497KB, 973x2259px

I'd need an edit of that for Vega.

Hi guys Raja here. Soon very soon we will release powerful graphics cart, but need to wait. It will kick nvidia ass, but please wait. We need optimization for ryzen drivers, need to wait. Please vait guys, soon we release graphics card. nvidia is shit

Just wait™ for AMD™ Radeon™ (RTG™) brand new release of Vega™ guys. It'll be what everyone's talking about!

>>60218507
This ... isn't real, right?

File: slides-24.jpg (235KB, 1920x1080px) Image search: [Google]

235KB, 1920x1080px

Raja while you're still samefagging pls explain what's dat.

>>60219389
It's about as real as anything posted so far.

>>60219389
Well it's fake, but at least someone put some effort into it.

>>60219401
it means

R -> R

>>60219401
Geometry culling

Or it's not and the whole board will be full of asspained nvidiots on Vega launch day.

File: 1491070627541.gif (491KB, 500x290px) Image search: [Google]

491KB, 500x290px

>>60219401
>primitive shader

>>60219428
It would be a nice change in the market, but Nvidia would still sell more.

do we get some caspian spice and ice cream with vega too?

and that poo in the loo game too?

>>60219446
NVIDIA always sells more, people bought fucking Thermis instead of Evergreens.
>>60219449
No, only Quake: Overwatch edition.

>>60219401
not raja, but looking at the pic it looks like its hardware overdraw elimination. It is probably related to their tile based renderer - which works by drawing small chunks of the entire scenery, like 16x16 or 32x32, in the ultra fast L1 or L2 cache, then copying the output to vram in the end. It is still bruteforcing draws (ie. it still draws stuff that isn't visible in the end because it gets covered up), but by doing so in internal cache, it can be done orders of magnitude faster than in the framebuffer. And since it takes many tiles to make up the full framebuffer, it can also be parallelized, say, across 64 compute clusters.

... that, or they have some other magic that culls unused geometry even before rendering anywhere, tiles or framebuffer, but I don't know how the hell you'd do that.

>>60219462
And that was when ATi was a pretty powerful brand name, now the mass of the market are Generation Z underage who didn't even know ATI or AMD exist.

AMD needs to straight up lead for 3-4 gens before it can reach 50-50 marketshare again

>>60213535
Na they will be first to have 144hz 4k ips Q2. You can't find anything about freesync2 monitors.

t. Freesync monitor owner looking for a 144hz version

>>60213737
yup this, then blame amd

>>60219529
I wonder how many people had 970 as their first video card.

>>60213737
Except this time no one expects anything because AMD was incredibly vague about Vega.

>>60219610
This.

The only thing to expect is performance slightly worse than 1080ti, like every other god damn release

>>60219610
>incredibly vague about Vega
>literally detailed the arch 5 months ago

It's only vague to idiots.

>>60219617
290x raped Titan and 780ti.

>>60219619
It's idiots that fall for marketing tricks and buy most high-end video cards. I mean Vega is already awesome if you know what made Maxwell awesome.

>>60219637
I don't care about idiot's opinion, I care about interesting technology be it Nvidia or AMD, and Vega is interesting technology, just as you said, Maxwell was.

>>60219623
Not at launch

>>60219658
But you see, these idiots buy these cards, and AMD needs to sell them. Awesome tech does not sell, marketing sells.

>>60219681
I can't help AMD there, I'm more interested in what they produce than their standing as a company and if you're worried about lack of R&D it seems the CPU division will now have more than enough cash to feed the GPU one.

>>60219719
Instinct cards will sell like hotcakes bundled with MI25's so GPU division will have money too. But i wish AMD had more influence over gaymur market, since things like Mantle/Vulkan are awesome, and software and APIs they develop are open-sourced, compared to one faggy green company with insane boner for proprietary frameworks.

>>60219745
bundled with Naples*

>>60219756
No point in bundling it with Naples for servers that don't need an accelerator.

There could be some 2U+ combos with them but the ML market is still tiny compared to the traditional x86 one

>>60219797
>ML market is tiny
Yet it grows, surely and steadily. And Naples is still the king of traditional x86 market due to sheer amount of cores and I/O.

>>60219581
UHD@144Hz = chroma subsampling garbage

DP 1.3/1.4 peaks out at UHD@120Hz for 24bpp and 96Hz for 30bpp.

>>60213335
nice strawman faggot

>>60219816
In pure specs, you're right, but I'm worried about AMD's on-site support, RMA and software compatibility, but Lisa Su has been doing these things for IBM for years so she should know that pure specs aren't all there it to datacenter

>>60219820
DP 1.4 has lossless compression

>>60219844
Considering the entire focus of Zen lies in dominating mobile and server markets, my guess they were preparing for it. And Su is no retard.

File: Arcturus.jpg (821KB, 1000x750px) Image search: [Google]

821KB, 1000x750px

>>60213276
My 7950's are waiting!

>>60219882
>not flashing his Tahiti PROs to XT

>>60219852
VESA Display Stream Compression is "visually lossless" 2x-3x compression, but it's optional and nobody uses it yet, either in GPUs or displays.

The new UHD@144Hz displays do 4:2:2 subsampling at their max refresh rate.

>>60219932
Wasn't 4:2:2 normal? Isn't 4:4:4 just on expensive as shit 12bit+ displays?

File: eye cancer explained.png (111KB, 1182x1586px) Image search: [Google]

111KB, 1182x1586px

>>60219959
No, 4:4:4 is normal unless you're using a UHD@60Hz setup over HDMI 1.4 or similar garbage.

4:2:2 and 4:2:0 both make color text and linework look like complete shit, and I hope you'd notice it if you saw it.

>>60220003
So there are no 4k 144Hz 4:4:4 displays out or even coming until DP 1.5 or someone adopts VESA display stream comrpession?

File: chrome_2017-05-04_21-03-34.png (490KB, 1485x926px) Image search: [Google]

490KB, 1485x926px

>>60213299

wew lad

http://www.3dmark.com/spy/1595952

>>60220054
Right, excluding multiple DP links. The new models now should do UHD@120Hz/4:4:4 though, unless the manufacturers are completely retarded.

But the next DP is at least several years away, and the next opportunity for DSC adoption is probably Volta or Navi.

>>60220003
>>60220054

The new Nvidia Gsync 4k 144hz IPS are DP 1.4

Seems like 1.4 can handle 4x @ 120hz no problem:

https://www.vesa.org/featured-articles/vesa-publishes-displayport-standard-version-1-4/

>>60220119
There's no difference between 120 and 144Hz and the former is better for videos so that's fine.

File: No_drivers_eh.png (200KB, 1000x492px) Image search: [Google]

200KB, 1000x492px

>3 drivers with in one month
Raja finally hired some driver guys it seems.

>>60220185
It's been what, a fucking year already since AMD had better drivers than novideo.

>>60220168
120 Hz is still a little too low for people with flicker sensitivity with ULBM or whatever strobing is enabled.

144 Hz is slightly better in that regard, but honestly just getting things up to 240 Hz would be good enough for anybody.

>>60220243
I thought Gsync/AdaptiveSync + ULBM doesn't work together?

>>60220200
Crimson is definitely better than recent Nvidia drivers, but honestly AMD was close to parity in the stability functionality for maybe closer to 2 years.

The new GUI and ReLive were just the final nails in the coffin.

>>60220243
For me 120hz is the sweet spot since videos look good in it, and 60fps capped japanese games look good in it, all without needed to constantly switch my refresh rate.

I've tried ULBM in the form of lightboost, and to be honest I wasn't impressed. The screen gets way too dim, the input lag is annoying, and I feel no appreciable gain from nuking motion blur from things.

>>60219932
I was really disappointed when I read that those monitors won't be using DSC, I was really looking forward to them.

>>60220262
they don't yet, at least not at the same time.

it's a fundamentally hard problem timing strobes and setting their amplitudes in such a way that average brightness stays constant at <10 ms scales, accounting for variable frame times, taking into account that pixel transition times are highly nonlinear with subpixel start and target colors, and that pixel rows still have to be refreshed sequentially even if there's a compressed time window readout fed by a local framebuffer.

>>60220267
Honestly it's 2014 when NVIDIA drivers went to shit and never came back. Also fucking pointless botnet telemetry and gayforce experience that requires you to use fucking goybook account to use it. Fuck NVIDIA.

>>60220267
They had better drivers than NVidia ever since the 7k cards got stable.

5k and 6k were so-so, but had some nasty bugs (my favourite was that they hard locked down if you did opencl while using hardware video decoding).

>>60220357
Man, shadowplay was so much better before. Its replaced by the unfiltered garbage called "share" now. There is no way to capture videos without turning on the telemetry.

>>60220322
what kind of games do you think you even be able to hit 120+ fps at UHD in anyway?

>>60220357
Eh, I switched from AMD not even a year ago and haven't regretted it one bit. Sure, I need to unpack the driver and delete all the shitware before installing, but that's not all that bad considering the drivers actually fucking work properly with recent games and more often than not with SLI too.

>>60220388
OBS faggot

>>60220392
I get like 170-300FPS in Overwatch, so shit like that at the very least. Doesn't have to be 120FPS anyway, with GSync/FreeSync even getting in the 80-90+ FPS range would be nice. 120Hz however is a nice sweetspot in terms of refresh rate though, 24 and 30FPS videos both play fine without refresh rate switching and so should games which are locked to 60FPS, or even 30.

>>60213276

actually tho, if the drivers can pick out if a calculation needs to be 16 or 32 bit precision, and take advantage of that potential 25TFLOP number it could be an easy contendor with the 1080ti

but this is AMD so the drivers won't be able to do anything that cool for at least a year after launch.

Nvidia cards suck on any CPU with more then 4 cores so vega or nothing till Volta (witch may also have the same issue)

>>60220444
Nah, using FP16 requires some game code optimizations.

I have my 1080 so long as I stay sub 4k it will last at least 5+ years

>>60220185

Half the time the drivers have broken Freesync.

I do not upgrade the drivers because almost always, the new drivers break Freesync. There are only a few good drivers which have Freesync not broken and actually working. Its really quite pathetic, I dont know how they even manage to do it.

A 1070 is $350 just fucking buy that

Waiting for tech is always dumb

>>60220444
This thing is clocked like 50% higher than Fury X and the switch to tile-based rendering improves efficiency and performance quite significantly. Those two things together mean that the NCU architecture is going to be MUCH faster than Fiji. That's ignoring all other improvements to the architecture. I think it's safe to say that Vega will compete with the 1080 Ti. The real questions, to me, are whether or not AMD will be able to satisfy demand with such a fuckhuge chip, and how well will it scale down for the midrange market, which is where most of the money is

>>60220627
I have freesync on my monitors and I wanna use it. I refuse to pay a G-sync tax, I would rather use this money to buy a better GPU.

>>60220647
This is the first time I'll planning to buy a GPU on release so I'll finally have to deal with shortages.

Will these be used for mining stuff? Might get fucked then.

>>60220664
A 1070 should work with free sync

>>60220647
You forgot MUCH better geometry throughout.

>>60220673
That's not how it works, unless there's been a cheeky exploit to do so.

>>60220672
Yes, it's fucking 12.5tflops card. Prepare for 290x 2.0: electric boogaloo.

Yes yes pls buy wega it wery good yes

>>60220673
"Should" but it wont. Nvidia actively blocks its usage on their cards.

>>60220672
>>60220690
My understanding is that it's no longer viable to GPU mine and you need an ASIC designed for mining now.

>>60220872
Buttcoins? Yes. But there's other cryptocurrency.

>>60219401
>pls explain what's dat.

nobody really knows yet, but since it has goofy throughput (~11 tris per cycle with 4 shader engine blocks), people are speculating that it has something to do with using the general shader ALU array for geometry and maybe tesselation shading in addition to vertex shading.

>>60220093
fake? is original creator of >>60213299
just messing with us by using 1080Ti score and photoshop?
I'd be all over it if they can even reach within 5% of 1080Ti performance just to escape the forced telemetry and get to make use of freesync.

>>60222446
dunno the graphics and cpu scores are slightly different, think it might be legit

>>60222478
I sure hope so, synthetic benchmarks are one thing though and actual results in game are another.
Wonder how it will fare on titles using DX11 since that is what the majority of recent games still use until Vulkan/DX12 becomes the norm.

>>60213276
>le wait meme

lmao i've been sitting on this solid 1440p60 performance for like 6 months now with my $320 1070. why would i wait?

>>60220627
Why?

My currently in use GTX-780Ti is still chugging along fine at medium-high settings with my i7-3850k.

But with how things are looking, I'll be getting an AM4 based guts for my tower next. Probably a R7-2800X + Big Vega, depending on how long I feel my current setup will last to my satisfaction.

>>60222540
not everyone is on the same hardware for some there is benefit to waiting if release date is close.
My 980 is barely coping with 4K and VR, and a 1080Ti aftermarket version is far too expensive here compared to reference model price/perf, could get one but I'd be wasting my money if AMD could offer acceptable performance at a better price point. They usually sell for $100 cheaper around these parts.

Banging my head in a wall for not buying that last 290X sapphire card in stock for 400AUD compared to the $768 I payed and falling for Nvidia marketing with "full DX12 support" and VR featues at the time. To be fair they did manage to support asynchronous reprojection for a good while before AMD and DX12 did not matter in the least.

>>60222795
i also had a 980. it was a solid card till it broke. i had mine oc to 1518 mhz and it smoked my best friends 290x 8gb edition.

>>60222907
I only OC mine to around 1450 Mhz, it is a bit hit a miss, If I go any higher it could crash a few minutes in to a few hours depending on the game I play.
COH2 for some funny reason would crash even with the slightest OC above factory default But for other games it is golden.
Not a bad card, but price/perf wise I could have done far better even with a 980Ti but I would have had to upgrade for 4K anyways.

>>60222585
If you want an upgrade now

But the other Anon made a good point about how async not working on Nvidia cards is dumb

So what's the current state of ryzen+geforce performance issues? Is it worth replacing 960 if I only need 1080@60?

>>60225003
it's fine. the only people who are complaining are either radeon fanboys who only want you to buy radeon or high fps asswipes who cry about not getting 200 fps. for 1080p60 it's perfectly fine. i'm gonna get a 1600 for 1440p60 with my 1070.

File: ume.jpg (96KB, 1280x720px) Image search: [Google]

96KB, 1280x720px

>>60213276
>ayymdpoojeets still waiting

3 SKUs, one supppsedly better than 1080ti, one between 1080 and 1080ti and one between 1070 and 1080?
Sounds fine to me. Hopefully we get them before october, that would complete my 1440p Ryzen build.

>>60226237
I think it will be 2 SKUs, one big vega and its cut-down version. I only started to hear stuff about that 3rd SKU since april. So, unless AMD had managed to keep it under wraps for that long or had prepared a SKU in the last 2-3 months its unlikely to exist.

when is the next jew convention where they will Reveal Vega?

how long will the wait™ be?

>>60227099
Computex

>>60225003
At this time it would be better to wait for ether Volta or vega if your not having problems

>>60213480

Why? If Vega doesn't flop, will it die?

>>60227099
computex, end of this month.

there's a possibly fake slide saying NDA of may 8th though.

>>60227244
Buyers remorse is a shit feeling.

>>60213276
I hope to get a Vega card that can play completely new modern games consistently above 100fps on ultra at 1080p. Freesync will smooth out the variation here and there. If it can get 144hz on ultra for older games like GTA and the witcher I would be ecstatic.

>>60227244
It would be extremely painful

All I want VEGA to be is a 1070 performance card, 8GB HBM2, low power draw like nano, be under 200mm length, and cost $250 or less

>>60213480
Dude why? I feel bad for you, because I fell for this meme too, but that was like 2 years ago. G-Sync is legit inferior in every way to FreeSync.

>>60227963
Pick 2 out of those.

>>60227963
>1070 performance card, 8GB HBM2
>$250 or less
AHAHAHAHA GET A LOAD OF THIS DELUSIONAL PAJEET
AMD AREN'T A CHARITY YOU FUCKING RETRAD

>>60219211
Fuck off, a lot us do rendering and editing too, so this shit is important.

>>60227886

You're a big monitor

>>60227054
Imo should give us 3 SKUs, otherwise they will leave out a lot of potential customers.
Releasing cards that competd with 1080(ti) is all fine and good, but without a third SKU AMD can't compete against the 1070 which has a huge market share of the current gen.

>>60227963
If it's comparable to the 1070 it's going to cost at least $320.

>>60228012
They kind of have to be right now though to reestablish themselves as a competing company. I honestly think AMD is going to cut prices much lower than they have to this time around to really get a new market.

>>60228809
not even amd is stupid enough to release a new architecture thats closed in price/greater in performance to the recently release 500 series. At the cheapest small vega will be $300

>>60218507
>Xp
Basicly proof that 1080tI will btfo vega, even amd was afraid to compare against it.

>>60219211
Please don't wait
Buy ou... I mean NVIDIA (tm) it will run your games! :)

>>60219637
The difference between a 1000 buck titan and a 500 vega is nothing if your spending that much

>>60213521
This

All the otehr brand loyalty is like listening to nine year olds talk about fucking soccer

>>60218483
>Vega X

I just hope the top-tier card doesn't have a dumb water cooler.

>>60213276
It would be nice to have a powerful GPU that also renders things correctly, yes. >>60230318