Radeon RX Vega barely beats GTX 1070 in leaked results

>some engineering sample with unknown capabilities
>LEAKED RESULTS

>>60159202
We have this thread every day.

>>60150417
>>60146186

>>60159235
Every repost is another shekel in his pocket.

>>60159202
Wait wait wait!

>>60159235
>new from yesterday

You are deseperate

>>60159235
>News Posted: 10 hours, 28 mins ago
>We have this thread every day

Literally what?

>>60159202
is there a trade show they are saving it for?

or Another Month Delayed?

>>60159202
Are you surprised?
>>60159339
Maybe thats why the 4xx are rebranded?

>>60160330
>Shitty clickbait thread down to page 10
>OMG BETTER SAMEFAG IT TO THE TOP I AM DESPERTAELY STARVED FOR ATTENTION
>Phew it's back on page one again, I'm saved!

This shit was debunked yesterday.

>>60148806
>>60148898

JUST WAIT
U
S
T

>>60160383
How?

>>60160365
>8 posters
>12 replies
>samefagging

You must be new

it mite be weaker now but after they work out the drivers it will be faster than titan xp... while nvdia will gimp the 1080 to gt630 levels by then... better to invest in AMD if youre serious about your pc

>>60159202
>the city of amd
Like I said:

1070 < vega 20 < 1080 < vega 10 < 1080ti < etc

Vega can't beat a full GP102.

>>60159226
>>60159235
>>60159247
>>60160383
Amdrones are worst than nvidiots

>>60160365
>14 posts
>10 posters
>MUH SAMFAGGING
Dump cunt

>>60160575
>wait

>>60159202
I honestly hope it isn't true

if it is then the 1080ti will be the last decent price:performance flagship ever released

>>60159202
This is good, because that means it will need to be cheap. Like, $300-$350 or so.

IT KEEPS HAPPENING

>>60160675
This but also it will need to be cold/silent. 200W housefires should proceed to >>>/v/

>>60159202
I HATE THIS WEBM REEEEEEEE

>>60159202
why don't they add more gcus :(

>be nvidea
>get furry x modify bios and make it look like vega
>oc it a little

HEHE GOIYM DONT GET VEGA ITS A FLOP GET THE 1080 TI

JUST

>>60159202
I. Fucking. Called. It.

Amd doesn't have a single card that comes even CLOSE to the 1070 and the pajeets here keep saying how they are suddenly gonna whip out a gpu out of their asses that is magically 1080ti tier.

The denial ITT reminds me of when the 390 was supposed to "destroy" Nvidia and turned to be yet another garbage rebrand that pulls over twice the electricity from Nvidia gpus. Or how the pajeets were saying the 480 was gonna be 980 tier. Or the "overclocker's dream". Or everything said about the flop known as fury.

Cant wait until confirmation comes out so the amd shills start coming out the woodwork to say how "n-nobody s-said it was gonna be good, it was meant to be shit" like they ALWAYS fucking do. Fucking retarded jehowah witness tier cultists.

>>60159202
old BS, shill

>>60148310

>>60160365
You can't bump your own thread retard. Welcome to 4chan, marketer. Lurk before you post.

Shekels are flowing in this thread.

I'ld say AMD on suicide shills on watch, but not sure many still alive at this point.

>>60161302
Ah the good old AMD pajeet mantra that says when anything goes wrong, it's always Nvidia's fault. I mean, it's not like AMD has a track history of blunder after blunder or anything, nah. AMD is literally the dindu of technology.

Enjoy your outdated housefire that is probably gonna be more expensive than Nvidia stuff just like the Fury.

It has reprogrammable shaders disabled.

shitposting general?

shitposting general

>>60161335
can the 1070 run doom 2016 at 4k with 60+fps? we know vega can. so far that's the only bar we got for comparison

>>60159202
Ahhhaahahaha good thing I bought that 1080ti 2 weeks ago ppl that were waiting for vega going to cry bitter tears someone post that pic where amd anon is crying plz I'm on mobile rite now

>>60161439
Oh you mean that presentation where AMD didn't let people check the graphics configuration so they couldn't see if something was disabled? And on a game that heavily favors AMD stuff and runs extremely well in anything? What about it?

File: 1433023146625.webm (475KB, 640x480px) Image search: [Google]

475KB, 640x480px

>who could be behind these threads?

>>60161481

https://en.m.wikipedia.org/wiki/List_of_multinational_companies_with_research_and_development_centres_in_Israel

Nvidia is not in Israel

>>60160654
> Vega 20 being below Vega 10
You're not a very bright person anon.

>>60159202
Real title should read
>25% underclocked cut-down unknown Vega chip running on unknown drivers barely beats 1070 in leaked results

>>60160383
Thread should have ended here.

>>60162237
Vega 10 is faster than 20

>>60162814
Vega 20 is a die shrink of Vega 10 with FP64 due out in 2018. You're thinking of Vega 11

So what's the best deal now? 1080ti with a Ryzen 7? Or is Ryzen shit now?

File: image.jpg (38KB, 640x480px)

38KB, 640x480px

>>60161335
but the 480 is 980 tier?

I mean you'd have to think the 1060 wasn't 980 tier either given how close the 480 and 1060 are.

>overclocked Fury X

Makes perfect sense they'd wait 2 years for that, even Polaris two 10 strapped together at lower clocks would be much faster than this and would have been out in September
Conclusion: Shitpost

>>60162876
>best deal
>any card above $300

>>60162993
Fury x 2.0 without de watercooling system?

>>60163849
Fury X is perfectly coolable with air, it'll just run hotter, Hawaii used more power than Fiji and it was aircooled.

File: 1486847277382.gif (47KB, 408x410px)

47KB, 408x410px

>>60159202
>>60159235

>>60162852
I hope this will be the case, with some enhanced clocks/performance tweaks too. I'm sitting on a Fury X and part of me really would love to upgrade to Vega, but another part of me realises that Vega 10 won't be enough to justify it; but Vega 20 ought to be.
I'm playing at 2560x1440 so extra horsepower would be nice, but at the same time, I don't like the idea of upgrading every generation (and I don't consider Polaris as a generation relative to Fiji, considering it isn't high end.)

>>60159202
Based on what I've read I'm kinda wondering if this is either underclocked or a low power model like the Fury Nano. The clocks are too low to make this a flagship.

>>60165207
Thats probably not the 'big' vega

>>60165571
read
>>60160383

>>60165571
>>60165613
It's an old ES but then again it might not be all that different from small Vega, Lisa Su just confirmed at the ER that Vega will launch with multiple SKUs this quarter, most probably some smaller variation of Vega to fight off 1070/1080

>be AMD
>hype for a new architecture that will kill nvidia
>it's actually a rebranded Fury X

amd status rekt

>>60165645
My understanding is that small Vega is supposed to replace Polaris next generation. Has that changed

>>60165645
No, it's ES of big vega, BUT it has reprogrammable shaders disabled.

>>60165651
>I can't read
Only thing I gleaned from this post

>>60165669
No.. Polaris 20 just launched like a week or something ago, Vega is replacing nothing, Vega is a lineup upwards of Polaris perf

>>60165696
That just doesn't make sense. Are they trying to wedge another performance tier between mid tier and enthusiast/flagship?

>>60165613
Read what? Can I have a quick feedback inb4 its fake. The 3dmark cant lie...

>>60165723
You know there's a pretty fucking huge difference between Polaris 10 and GP104, right? Smaller vegas go there.

File: screen.png (35KB, 933x402px)

35KB, 933x402px

>>60165738
It's a taped out engineering sample with hardware disabled and underclocked. This benchmark literally means nothing.

>>60165669
Then why they release the rx580/570 aka rx480/470? Thats make no sense

>>60165787
1200 MHz is underclock Vega?

>>60165832
MI25 already clocks at 1525mhz and it's ML card.

>>60165792
580 and 570 are just overclocked Polaris. I guess I haven't really been keeping up as well as I should, I had thought it was going to be Polaris/Vega 10 and then Vega 11/Vega 20 would be next generation but apparently that's wrong.
>>60165832
They cut the transistor size in half. Transistors get smaller, power requirements go down, clocks go up. It's just physics.

>>60159202
but how much does it cost?

>>60165849
But the MI25 draw 300w vs 225/250w for Vega. That's why I think its legit.

>>60165909
MI25 is rated <300W TDP. We can only guess it's final TDP rating.

>>60165909
MI25 is dual chip card.

>>60165919
No you fucking retard.

File: 1493033815796.jpg (69KB, 460x613px)

69KB, 460x613px

>>60165849
>MI25
This card is not real my son

>>60165938
As real as anything GP100-based.

>>60165919
No, it not also its not even released lol

https://www.techpowerup.com/gpudb/2929/radeon-instinct-mi25

>>60165929
Why is it named MI25 then? 25 tflops is lot for one chip.

>>60165983
It's 25tflops FP16.

>>60165974
>GCN 1.3
Sasuga as always, TPU

>also doesn't know tflops are rated by base clock not boost clock on AMD GPUs


Why is TPU such garbage?

>>60165992
Well that explains it. Thanks

>>60165992
how much is that in leauge of legends?

>>60165938
is this a turbine simulator room?

>>60165993
>Why is TPU such garbage?
NVIDIA shills, that's why.

>>60165949
you know gp100 cards are being sold for quite some time now
amd shit is woodscews level real

File: Radeon_Instinct_cards.jpg (99KB, 1236x534px)

99KB, 1236x534px

>>60165938
They're fakes made out of VHS tapes and woodscrews

>>60159202
where's the crash dammit?
i waited.

File: Quadro_GP100_Top.jpg (1MB, 4000x2746px) Image search: [Google]

1MB, 4000x2746px

>>60165949
You can bought a quadro gp100 even the new drivers are avalible...

http://www.pny.com/nvidia-quadro-gp100

>>60166022
>nvidiot projecting THIS hard
Oh wew.

>>60162876
>amd + nvidia
you're gonna get crippled drivers.

>>60166028
>semiacrutate

You cant bought these card because they are not ready

Is there a chance... and hear me out here... is there a chance that just like how Intel has dominated the CPU market for the last ten years or so, and so games are generally optimised to work best on their architecture, and now that Ryzen has released games companies will most likely ensure their games are optimised for Ryzen too.. isn't there a chance the same thing is happening here if these are legit results. AMD cards are probably an afterthought as they barely even appear at the highest tiers of GPU's, so couldn't it be the case that benchmarks and games will run like shit on the Vega until companies begin optimising for it and taking it into account when developing their softare?

>>60166071
Fuck off with your games and go ask someone on leddit.

>>60166033
>amdpoojeet has no rebuttal
kek

>the mi25fags is here

Where is the '''''MI25'''''?

>>60160575
https://www.youtube.com/watch?v=b_ZMOn0X6jw

>>60166046
They are ready.

>>60166042
I really need a pc now. And I want to make it Ryzen based build but 580 is too weak and I don't want this amd+nvidia issues. Vega is still at least a month away from release and +1 month to get to my stupid country. And then retail will ask crazy money for it.
What did I do to deserve this? Amd is forcing me to go intel+nvidia.

>>60164553
It'll probably be workstation only, FP64 takes too much die area to be viable on a GPU of that size on a 14nm process. It won't be any faster probably just have FP64.

>>60166080

You do realise that 90% of graphics cards on the market are specifically created with gaming in mind? Like it or not it's the biggest market for card manufacturers. You probably have 100,000 gamers to 1 rendering artist, and even then they usually use specialist cards like the Quadro, not GTX 1080's.

>>60166125
Yeah, and I don't care about Steam users, take your gaming back to /v/

File: poo.png (312KB, 595x654px)

312KB, 595x654px

>>60166087
They are ready.
Just wait

>>60166132

>so elite and woke

'm sure no one here cares about the Vega's gaming performance...

>>60166097
We have tests, benchmarcks and drivers for the Pascal Quadros but NOTHING for the MI25. Not even the price...

>>60166150
/v/ cares, go ask them.

>>60165993
Citation

>>60166183
On what?

>>60166087
I can't give any numbers, figures, or time because NDA a fun. I think I can at least say they have no display output capabilities or conventional drivers though.

>>60166197
>tflops on compute cards are rated by base clock not boost clock on AMD GPUs

Citation

>>60166155
The product isn't in the same field as Quadro, they're competitors to Tesla. They don't have display output or real drivers either.

>>60166199
Thanks for confirm my point. The MI25 is a AMDrone gimmick

>>60166226
https://youtu.be/m5EFbIhslKU?t=693

File: tp100feat.jpg (24KB, 414x257px)

24KB, 414x257px

>>60166231
>

>>60166267
>youtube

>>60166303
That scared of a few hundred MHz?

Adorable

>>60166253
Well they do exist and do what they're advertised so I don't think that's really a gimmick. If you want an AMD gimmick just point to that TrueSound or w/e it was called.

>>60166289
Yes, no display output capabilities, same category as Phi and Instinct.

>>60166324
TrueAudio.

File: uhohhhh.png (542KB, 925x534px)

542KB, 925x534px

I sold my video card while it's still worth something to put towards vega assuming something will at least get announced by the end of may.

should I just get a 1080ti?

>>60166335

I'm in the same boat, finally decided on my new AMD build which I'm ordering in a week or two and now don't know whether to go for a 1070 or wait for vega -_-

>>60166370
Considering Lisa stated there'll be more than one Vega SKU, i'd wait.

>>60166334
Yeah the thing nothing fuckin uses. I still can't get why they wasted die area on that.

The consoles have it too but I don't think devs even bother on it.

>>60161335
> got a 760
> need to upgrade
> look at the AMD 3- series
> consider getting a 970 instead, because I only play at 1080p anyway, so VRAM isn't an issue
> hey anon RX480 soon, it'll be like between 970 and 980 at prices way below a 970!
> wow, i better wait
> RX480 comes and goes
> frying PCIe slots (yeah I know that's fixed, still not a god first impression)
> nowhere near the overclocking headroom we were told to expect
> sort-of matched the 970
> still with the 760
> look into the 10- series
> hey anon RX580 and vega soon, it'll be like between 1080 and 1080Ti at prices way below a 1080!
> wow i better wait
> 580 is just a slightly faster 480 with less overclocking headroom
> vega now looks like it's going to crash and burn
I'm honestly just going to buy a standard MSI 1080 and call it quits.

>>60166394
It's just a tiny DSP, I'm wagering no bigger than 2mm^2

You know what's wasted die size? A disabled integrated voltage controller in Ryzen CPUs

>>60166412
>A disabled integrated voltage controller in Ryzen CPUs
And a fuckton of other stuff like fucking GMI links. Zeppelin is a server die, no a consumer one.

>>60166392

If I wait I'll be using my 7950 in my new build, think I'll at least wait a few weeks to see if any new information comes forward.

Also I don't understand why multiple SKU's would be a reason to wait.

>>60166412
Disabled for consumers.

>>60166412
It's a waste until it's enabled...which it probably will for Raven Ridge/Naples and Pinnacle Ridge, that would be a nice perf/watt boost without a dieshrink so the next gen looks even more impressive

>>60166408
Not sure if troll or just dense idiot.

>>60166429
It makes sense to use the same die for consumer and server CPUs the way AMD is doing it because they can just stick a bunch of CCX's together and make a 32-core Naples processor instead of having to do an entire new process.

>>60166433
Because you want a 1070 or it's competitor. Flagship Vega will cost more.

>>60166465
They are sticking Zeppelins, not core complexes.

if youre dumb enough to believe ES benches as final you should just be a good nvidiot and overpay for your midrange card

>>60166492
>ES bench

keep making excuses and just WAIT

>>60166465
If AMD had money they'd have separate dies for consumer and server, more dies is a pure cash investment, which would be offsetted by higher margins due to die size and specialization.

However AMD doesn't have that much cash on hand, so for a while they'll stick to one die that can handle 60W-200W

>>60166468

Ah right I see what you mean. Yeah I'd look to spend around the same amount as a 1070, so would be good to see what AMD offers in that range. But still, will be a long wait using a 7950.

>>60166438
RR is going to be a different die anyways with different uncore, and Naples is just MCM of the 8 core die. Pinnacle Ridge is anyones guess, but there has been rumbling on 14 HP.

>>60166508
>specialization.
How would you specialize consumer dies?

>>60166508
These are already small dies and yields are good. In fact, these dies are actually smaller than Intel's Skylake dies. AMD isn't worried about adding a little unused die space to save a bunch of money.

>>60166458
Full blown retard.

>>60166539
Just bad wording, I meant 'one die that can do one thing well'
Which is pretty much what Intel does and AMD has done for many years until Zen

>>60166555
They're small, but they can be better, a few % on margins on $300+ CPUs selling in millions is no joke.

>>60166585
Spending billions to retool machinery and set up a new fab process is no joke either.

>>60166600
Yeah, but that's a one time investment, this one looks much better on the next year or two earnings report.

>>60166625
My point is financially it'll probably end up being a wash.

Is there a reason you're trying so hard to criticize this decision? What the fuck is the point?

>>60166650
We're talking pure profits here, I'm not criticizing AMD's decision for a one-size-fits-all die, but if they had money and knew they could sell chips in high volume they'd have dedicated dies for consumer and server.

>>60166671
There are no dedicated server dies, Naples is 4-die MCM of Zeppelins.

>>60166676
?
That's what I said

>>60166685
There's no point of making dedicated consumer dies.

File: 1475461066506.gif (912KB, 240x176px)

912KB, 240x176px

>STILL stuck with a gtx 670 after all these years
FUCKING POOS RELEASE SOME INFO
SHOULD I GET A 1080 FOR $590 OR WAIT FOR YOUR SHITTY VEGA CRAP? ANSWER ME, PAJEET

>>60166693
Tell that to Intel and AMD's every lineup before Zen

>>60166693
Except mobile, RR is probably all consumer

>>60166408
You should, you sound like a retard so get the retard purchase.

>>60165696
small vega is 1070 tier

>>60166748
And that means upwards of Polaris perf, you don't seem to realize the rather big gulf between the Polaris/GP106 and GP104(cut)

>>60166757
no shit retard

>>60166408
1080 is the worst price performance, you'd get a decent card but not for the asking price.

If you are going to go Nvidia either go balls to the wall with a 1080Ti or grab a 1070. Personally I am going to wait on Vega and see if it is worth upgrading to from 980 for 4K and VR.
If it disappoints then I will grab the 1080Ti once it is discounted because 1000USD equivalent for an after market version here is fucking retarded.
Next year might be a better time for me to upgrade.

>>60166702
I upgraded from GTX 670 -> RX 480 last summer and have been pleasantly surprised by AMD.

I'm going to upgrade to Vega if it's 1080-tier or above and remotely competitive in pricing, out of spite for Nvidia's garbage policies more than anything.

ReLive >> ShadowPlay, Crimson >> Post-2014 GeForce drivers (especially with bullshit social media tie-ins and all the "telemetry" functionality).

>>60159226
Fucking retards like you literally have no idea what the world revolves around.

Every single time one of AMD's engineering samples came out, they were somewhere between 2% and 10% more powerful than the final product.

If a crappy engineering sample from AMD barely beat a 1070, the final product will either have the same performance as the 1070 or lower. Get your shit together you idiot.

>>60166833
You must be this ------------------------ SPECIAL to believe Vega is a overclocked Fury X

Have a nice day.

>>60162169
and AMD is
makes your synapses fire electric pulses

>>60166788
>1080 is the worst price performance
is this still true after the price drop?

>>60166833
People look at Zen, where ESs were clocked at a mere 2.8 GHz (to many cries of "AMD is finished and bankrupt") compared to final retail base clocks up to 3.6 along with roughly Haswell IPC.

It's also known that the fanless enterprise Vega variant (MI25) is 12.5 TFLOPS @ ~1525 MHz and ~225W, so the consumer version will probably be some 13 TFLOPS/1600 MHz/275W factory overclocked housefire deal again.

Perf/W will probably disappoint as usual, but there is literally 0% chance that the uncut Vega 10 will come in below GTX 1080.

File: 1467254443174.jpg (59KB, 513x472px)

59KB, 513x472px

>>60166833

>Every single time one of AMD's engineering samples came out, they were somewhere between 2% and 10% more powerful than the final product.

how retarded are you?

>>60166947
>Perf/W will probably disappoint as usual
Why?

>>60166947
>>60166863

Why are you arguing with ___THAT___ shitposter? This is his work hour and he started shitposting in multiple threads at once some 10 minutes ago, it's very transparent.

Think again about arguing with a macroing wall

>>60166828
I run with linux as host and play games every once in a while on a windows VM passing my dedicated GPU back and forth between guest and host, so the telemetry doesn't affect me as much since I can just use the shitty open source drivers on linux while the windows nvidia drivers along with windows 10 telemetry can play in their little sandbox.
I hear AMD's open sourcce drivers are improving so I might grab that if it means less headaches on linux.
Not sure how their GPUs fair when it comes to hardware decoding for video.

I am hoping Vega would be in a sweetspot for 4K 60fps without AA, and VR. So perhaps within 5~10% 1080Ti performance for a good price, otherwise I'd be better off grabbing 1080Ti on discount, or waiting on next year's GPUs. The VRM issue on some 1080Ti models has me worried however.

>>60166940
well not sure, it just might be that out prices in fakin straya are retarded.

File: Screenshot_2017-04-19_at_11.34.23_AM.png (115KB, 344x171px) Image search: [Google]

115KB, 344x171px

>>60166947
Ryzen IPC is higher in everything except 256 bit AVX. Agner already ran tests and updated his guides.

>>60167030
>living in straya
You poor thing.

>>60167012
AMD's open sores drivers are actually becoming downright solid, but it's been a slog for them to get all their stuff mainlined in the kernel since it was initially build off a fairly heavy HAL that the maintainers balked at.

>>60166947
I'd argue that perf/watt will actually be phenomenal if AMD doesn't have to overclock the chip to the moon due to bottlenecks like they did with Polaris and Fiji,

>>60167034
> that lying nigger
> I literally checked Agner's stuff yesterday and saw nothing
> no threads on /g/ appeared to talk about it either
> (checks just in case)

wow, today anon was not a fag:
> http://www.agner.org/optimize/
> http://www.agner.org/optimize/microarchitecture.pdf

>>60167094
I'm inclined to agree, they aren't adding in moar corz, Vega seems like a pure Fiji architectural upgrade + dieshrink

>>60167049
That is good to hear, I might look into it further especially if it means that I can swap GPU between VM and host OS without Xserver restart or logging out.

File: Screenshot from 02-05-2017 01:48:50.png (253KB, 1920x1080px)

253KB, 1920x1080px

>>60167114
Oh I'm gonna have so much fun reading this, Kanter has been slacking off on these things.

>>60167134
Kanter is highly overrated in general.
His primary merit was that nobody else was writing anything at all in that rough region of technical detail.

Fog actually has real street cred.

>>60167114
RWT thread started, Linus is already posting.

Should be chalk full tomorrow

http://www.realworldtech.com/forum/?threadid=167900&curpostid=167900

>>60167202
We plebians have no choice but to take breadcrumbs, I can't afford a $10000 subscription for these things so Kanter was my only choice.

File: 1493252773106.gif (1MB, 320x180px)

1MB, 320x180px

>>60166029

>>60167034
Ryzen's scheduling/execution/retire throughput is higher than any prior x86 processor's but that by itself isn't a guarantee of higher sustainable IPC.

In particular, post-Haswell Intel chips have much fatter datapaths (64B/cycle for both I and D), which are necessary for 256b AVX and certainly beneficial for even a lot of int workloads.

Agner's stuff so far is good as usual, but I'm a bit disappointed his didn't do detailed timing analysis on a lot of "uncore" stuff, like the timing anisotropy of L3 slices in CCXs, inter-CCX latencies, or even inter-core or inter-SMT thread latencies.

>>60167540
Nothing's stopping him from updating it latter, I'm also personally curious how AMD got those relatively low L3 latencies and high bandwidth(at least compared to BDW-E)

Last I checked, Intel's SRAM was the highest performing

>>60166940
I mean, why are you asking him about value of something? Just go check tests of 1070 vs 1080 and ask yourself if you want to pay extra for that boost. He might be poor neet and for him it is not worth it.

>>60167563
The conventional wisdom is that per area or per Watt, Intel's SRAM is the best on the planet, and that's very likely still true.

However, AMD can get slightly higher densities and the resulting modest latency gains from that by cutting corners on datapath widths and who knows what else.

Honestly though it's probably mostly just like it looks with the L3 slices being faster and less bottlenecked on a crossbar than on some huge crusty ring.

> sort-of matched the 970

And you know what's most funny about it? It actually doesn't beat the 970 in the vast majority of games because the 970 overclocks like a monster. You can easily get one to stock 980 levels and even if you OC the 480, the 970 will still mop the floor with its carcass.

Anyone falling for this waiting meme from amd pajeets is an idiot. They always say this crap, then the thing comes out and doesn't come even close to meeting what they promised, so they start damage controlling and pretending they weren't promising great stuff. Then the next amd trash gets announced and the cycle repeats itself. It never fucking ends, but I feel like people are wisening up to this fact by now due to the ryzen blunder.

I like how he's not even trying anymore.

>>60167666
I rate this post 3.5/4.

Kek

http://hexus.net/gaming/news/industry/103957-amd-vega-powered-liquidsky-streaming-servers-go-live/

>>60167683
Was gonna review yours but sorry, had to get to my phone because my 480 fried my motherboard so I lost what I had typed.

>>60167734
Well I guess we know where those Mi25's went

>>60166947
i literally dont care if i need a 800+wat psu for a single video card as long as its good.

>>60167734
>>60167817
I have never and will never give two shits about cloud gaming, but I would really like competent same-box virtualization to keep my W7/W10 cancer in a sandbox that doesn't perform like shit in gaming.

I'll never buy Nvidia again. I hope Vega is great, but even if it's just average, I still wont have to pay to sync up my monitor and gpu when theres a great free alternative.

>>60166102
Then go for it.

File: 1493603052352.png (9KB, 160x160px)

9KB, 160x160px

AMD Radeon RX Vega beats GTX 1080 in leaked tests

Read more: http://www.tweaktown.com/news/56645/amd-radeon-rx-vega-beats-gtx-1080-leaked-tests/index.html

>>60166408
Obvious bait, literally nobody said the 580 would be that good, not even close to a 1070

>>60162169
>using wikipedia as a source

>>60166161
/v/ is for PS4 fans, /g/ is for shitposting COMPANY X BTFO gaming hardware threads

>>60168562
Nvidia's solution though marked up and more expensive has more stringent certification. They will likely adopt freesync the way things are going since it is better value for money.

File: Vega.gif (221KB, 470x372px)

221KB, 470x372px

>>60166702
you look Fury-ious.

>>60166370
I would say buy the r9 fury nitro but that was when it was 250 dollars

you missed the boat my friend, now you're better off getting a 1070 or waiting for vega. If you're going ryzen it's possible that nvidia still hasn't gotten their shit together with drivers but they might eventually.

>>60159202
Vega's going to be 20% faster than the 1080TI, screencap this

File: Screenshot_2017-05-02_14-37-21.png (9KB, 451x86px)

9KB, 451x86px

>>60170327
sent :)

>>60167737
Man, same here, but my house burnt down due to my EVGA branded Fermi card.

>>60170368
/g/ does better work than mexicans.

File: 1481012579153.jpg (264KB, 595x801px)

264KB, 595x801px

lets get this shit started
>amd poojeets

>>60168440
What do you mean by samebox virtualization? as in have windows run in windowed mode within host OS both using the same single dedicated GPU?
That would be very convenient, for now easiest and cheapest way to do it is using an iGPU and dedicated GPU on a single monitor with multiple inputs.

File: via_quad_core.jpg (139KB, 500x260px)

139KB, 500x260px

Hmmm...
If AMD is to Pajeets, and Intel is to Jews, then what is VIA to?

>>60170438
>Hmmm...
>If AMD is to Pajeets, and Intel is to Jews, then what is VIA to?
CIA (Chink Intelligence Agency)

>>60170074
Nvidia support the anti israel sjw

>>60170429
I mean I want:
- a HEDT CPU
- a nice 1080-tier or better GPU
- some Linux flavor as primary OS and gui/shell
- ghetto Windows sandbox
- GPU can run gaymen on Windows sandbox at speeds comparable to PCIe pass-through
- can run primary system on same GPU as guest OS
- maybe on different display outputs, maybe even in a goddamn window. how hard should this shit be?

>>60170695
It sounds tricky to do, with PCIe passthrough the GPU is being utilized as though it were bare metal.
Emulating a screen in a window means passing through the output of the GPU in software back to the host OS with minimal latency constantly which could add a lot of pressure on the CPU.
But that is only my humble opinion, I don't know much about virtualization. Hopefully devs finds an elegant solution.

>>60170753
Maybe if some new hardware feature came out on the GPU that allows it accept instructions from a VM directly and the Host OS simultaneously then automatically draw the VM's output in a window with minimal latency and CPU intervention besides whatever the VM requests to be calculated/drawn. So no piping output via CPU between host and guest.
There already is monitor output emulation but the 3D acceleration is weak for some reason.

>>60159202
WhyCry literally has his hands on 20 different 3DMark scores. Have you ever wondered why they don't go up on VideoCardz?

>>60159202
>t. desperate nvidia shill

Any news about the release date/prices?

Saging myself because this thread sucks.

>>60165993
they are nvidia shills now.
they openly trashes amd.
not sure why amd is giving them cards.

>>60166833
>Every single time one of AMD's engineering samples came out, they were somewhere between 2% and 10% more powerful than the final product.

That is literally the exact opposite of what has happened with all of AMD's previous engineering sample leaks.

>>60160654
vega could beat gp 102

its simple, at amds worst,
if vega meants 1080 at 1500mhz, then it meets the 1080ti at 1700 and gp102 at 1900
If vega is polaris efficiency, it now meets the 1080ti at 1600 and gp102 at 1800


Now, I consider baseline being polaris efficiently because why in the fuck would vega go backwards? However, if amd, and they seemingly did, spend all their fucking time alleviating bottlenecks, sky is the fucking limit, as at 1525, the passively cooled pro line, they match gp102 in flops, If amd got their shit together on the bottlenecks, its going to be close. BUT this is a passively cooled 1525mhz gpu, add 2 fans and an after market cooler and how high would that clock to? nvidia is at 2000mhz, what is stopping amd, who spent nearly 2 product cycles sense fury, optimizing the card hardware wise, from not alleviating bottlenecks at the same time allowing higher clocks? even polaris can get nearly 1500 on the more recent gpus and that's not a clock rate maximizing design.


Now don't get me wrong, I don't believe vega at stock will beat out a 1080ti, but I do believe it will beat out a 1080 and if vulcan/dx12 beat out a 1080ti, but thinking amd wont at least beat or at least play with a 1080 even on their cut cards is a bit stupid.

>>60161473
they did, and it was set to nightmare on a gpu that reported fiji due to half baked drivers.

>>60161335
>Or the "overclocker's dream"
Mark it as shit never said, give me a video where they call the gpu an overclockers dream.

>Or everything said about the flop known as fury.
You mean the amazeing gpu that contends with a 1080 when properly utilized, that sold for 650, a good 1-200$ cheaper then everyone was saying it would sell for, that only lost its rightful place as being the best gpu because nvidia held off on the 980ti till amd put their card out, shoved it out before the fury came out, and then price matched the fury, because nvidia was so scared of losing the gpu crown?

The gpu that was designed and built for, along with the 390 line, for 20nm, a process that fell through, and is why amd took a file and ended their contract with tsmc over?

yea its a bit fun reading shit people who want to be willfully ignorant, or are just fanboying for their own team write.

>>60147488
They didn't, they offered up a very viable card to the masses, and forced nvidia to play ball with pricing, but then again the mid range founders edition for 450$ and high end founder's edition for 700 were pretty funny, not going to deny that.

>>60148925
Fury, a card made for 20nm, got fucked over and had to have a last minute redesign for 28nm, still played ball with a 980ti back in the day, and with a 1080 today when properly utilized, when it went up against a desperation move of a gpu, the 980ti because nvidia were shitting themselves at raw specs. Now that the 1080ti is out, I wouldn't be shocked if amd does not start showing off the hardware running at full spec and teasing the hardware costing more then a 1080ti just so nvidia can't preemptively price drop. If amd was smart they would let reviewers have it two weeks in advance, and allow them to review it, it comes out largely the same as the 1080ti in performance or slightly above depending on the game, without knowing the final price.

then 0 day, 30 minutes before you are able to buy it, they drop a hammer, cheaper than the 1080ti, significantly so, 550 range.

>>60161459
at best, people waited 1-2 months to see if amd had anything ball park the same, or better fit their price range. they will also likely get a cheaper card then you too because they aren't paying the shiney new premium.

at worst, vega beats you and you will have to cry 'newer hardware beats older hardware, who would have guessed' incessantly for the next 3 years.

>>60166833
Yeah just like ryzen

Oh wait

>>60166563
the issues is scalability, intel hit a fucking wall hard with theirs, amd found/designed a way to scale better, this also allowed them to have smaller dies and mcm them instead of being monolithic, so overall cheaper production.
doing one thing well is good and all, but if a gpu is better suited for a purpose then a cpu, use the gpu, intel only invested in its fpu on the cpu because they hit a wall with integer, something amd crushes intel in right now.
Now if you can use the same server die for consumer dies, awesome, it allows you to go full production and be very discriminatory with what you put in the server dies and pass the crap off to consumers.
then the final piece, does consumer use really need specialised cpus for it? I mean look at intel, they were 7% ahead raw ipc, newest bios increased amd by 2% so now they are 5% faster raw ipc, and 17% faster clocks (saying 4.8 is an average people hit before delid is necessary) so around 20~% more powerful at absolute most when thing suse less than 4 cores, and this gap decreases even more when amd has faster memory,

Now in my case on my stock 1700 I have somewhere around 3-10% idle use, so on intel that would be 6-20% use our the window right there. Yea, its very hard to justify intel knowing that when its used 100%, the user experience is shit, but when my 1700 is used 50%, I don't even feel it.

>>60159202

9 mins in.....how long will this take...

>>60166625
>>60166650
actually, just because you can save a few mm^2 doesn't mean you make the die smaller,
lets say you have a 200mm^s die and you can save 5mm^2
to save die space the die would need to be 5x40

And lets be honest, once all is said done and designed, removing something from the die in materials alone is sub pennies, probably sub .001 penny per die. even if you sold 1 billion of them it would cost more to redesign and validate then to just disable

>>60166671
I honestly doubt they would ever do that again, I mean look at intel's server chips, how high they can boost, so on so forth.

If you can make a small chip, lets say 8 cores, that can be mcm'ed and are more then viable as a single chip as they are multi dies, why in the fuck would you ever design a consumer AND server chip? The only applications where intel shines are single core and games, everything else its either a wash or heavily in amd's favor when compared to amds 1700, and when you talk server parts, amd's 1700 is on par with shit intel sells for 1200$, and these are the worst of amds 8 core chips, these aren't the high as fuck binned shit.

>>60166704
Intel does it because they are a greedy as fuck company, they make many skus as small as possible to maximize profits, cost of production be damned.

amd has always made fewer skus and just culled them, see the athlon and phenom II 4 cores where every die was made off of a 4 core, and later on all the phenom II's were made off of 6 core dies.

Honestly thinking of it now, amd has done this for a long LONG time, but intel hasn't. They would rather segment everything then bin higher core shit lower.
Don't get me wrong, intel should not only make 22 core dies and bin i3's from it, but still.

File: 1476920317729.png (360KB, 1230x1341px)

360KB, 1230x1341px

lol

Pft wait for this

It still doesnt touch the 1080

What is wrong with peoples brains?

>>60160383
/thread

>>60173145
Mi25's TFLOP rating is at base clock, it's probably well over 1630MHz turbo

>>60173570
Difference is in AMD's case the GMI links take a significantly bigger part of the die than 5mm2

>>60174283
This is about as useful as the C1 "leaks" since we know nothing about this card, how old it is.

>>60174422

Its quite sad that we know more info about Nvidia Volta than RX Vega which supposedly is coming soon (tm)

and I want Vega since I have a freesync 4k and Fury x is gettin old

>>60174443
We know ZERO about Volta.

We know plenty about Vega, precise clocks, features, core config, etc.
Which is more than enough to give calculate a baseline performance, and that's close to a 1080ti without architecture advances

>>60174283
F-f-f-fFAKEEEEEEEEEEEEEEEEE

>>60174534

So we get to see RX Vega demod for 5 minute at Computex, which is June 1 and then its in stores in July if we are lucky

As long as it's lower price, which I could imagine it won't be heavily, if at all, it won't be so bad.
Also, I'm hoping to see if it's going to be because pre-release drivers, as well as engineering sample. I'm really hoping, anons.

>have enough money to buy a decent 1070 right now
>waiting for Computex, or at least closer to the date for more "leaks" to get a general idea of the price
>sticking with i7 5775r iGPU
Fuck my life up tbqh guys

>>60174812
That's all you need to play your visual novels anyway my bubbla dubbla

>>60174839
H-h-how did y-you know a-about m-my visual nov... uh I m-m-mean uh I don't uh p-p-p-play visual n-n-novels!!

>>60174584
some announcement announced in shareholders meeting on May ~15

>>60162169
>AMD is in israel

hmmm...

>>60161420
Remember 5000 series was the last amd video card. Had a 5970 behemot. Hot as hell nice of a card. Sold it to a bicoin miner, got a 7970. This one was good too. In the end got a 980. then 2x 1080. Now I have the noney, but amd don't have the products. Sad.

>>60159202
AHAHAHAHAHAHAHA

>>60166788
>1080 is the worst price performance

The higher end is always going to be bad price-performance

But there is no card that compares, there's literally no choice

>>60176807
>Had a 5970 behemot. Hot as hell nice of a card.

Not sure about that, I had a 5950 and I could run it with an Accelero S1 and a 12cm fan zip tied on. It was extremely cool. But the 5970 might have been different.

>>60166788
>>60176922
this desu
please into economics
but seriously 1080 right now is a bad deal
1080ti and titanxp is where to go

Just wait™ for the upcoming™ AMD™ Vega™ line-up of graphics cards. Radeon™ Technology Group (RTG™) would be proud, anons™

>>60176953
5970 was dual-gpu card, but I don't think 5950 was... I could be wrong, can't quite remember, and I'm too lazy to google it proper.

So when is vega coming out and pricing options?

Im a bang for buck type of guy, so was disappointed in the 500 series rebrand. Still, looking at a 580 possibly, but l wondering for vega and ryzen build.

Currently using 760 and a fx 6350 cpu. Built 3 years ago for very cheap but feeling the age.

>>60177045
good chance is it'll be announced early June, and released maybe a month or so later. keep in mind I have no knowledge of Radeon' plans for the release, so take this with a grain of salt

>>60177045
Vega is gonna be over $350 for the cheapest one

I'm waiting for Vega for the sake of seeing NVidia prices drop :^)

>>60177082
>>60177085
Thanks for heads up. Wondee how big of a difference between 580 and cheapest vega, performance wise.

>>60177112
Rumours point at a Vega card that has similar performance to a GTX 1070, so probably the $350 that the other anon mentioned

>>60177112
has to be at least 50% or it's not worth the money

>>60177106
>Nvidia
>giving goys a discount

>>60177183
all I gotta do is make a simple phone call to my boys at envidya and they'll hook me up tbqh :^)

>>60166693
The most bizarre thing is Intel i5 revealed to be on x299 lineup.

>>60176953
You probably had 5850. 5970 was two dies on a PCB two 5870. 350w if I remember correctly

>>60174283

Fake!

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

This is a 1080 Ti score.

>>60159202
So a CERTAIN vega model beats 1070?
Lets hope its the model thats competing with 1070.

When Vega launches Nvidia will launch the Titan xP

>>60159202
Source on webm pls, need closure.

>>60159202
>CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT = 1
It's running at literally half speed, you fucktard.

this thread is prime example of why people think /g/ is a joke

>>60183142
/o/ /v/ /tv/ and /pol/ here we R all equally succc

>>60160383
Nvm this guy beat me to it. Didn't notice.
Yet the post continued for 200 replies when it should have ended there.

>>60159202
Fake news decoy leak sent by AMD to sandbag.

AMD is playing N-Dimensional Parcheesi.

>>60182170
What's ''CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT = 1'' mean?