http://www.pcworld.com/article/3043993/hardware/amd-stakes-a-claim-on-virtual-reality-hardware-with-1500-dual-gpu-radeon-pro-duo.html
>AMD COMING OUT WITH THE MOST POWERFUL GPU MARKET
>1500 jewbucks
>>54265677
>titan z: $1600
>titan x: $1200
Nice try Pajeet
>>54265653
AMD is fucking retarded. Nvidia could afford to make titan x and titan z because the hardware was targeted at the machine learning/big data sector, NOT at the gaming market. AMD, on the other hand, is squarely targeting at the gaming market. They'll never break even on that crap.
>>54265738
>titan x
>targeting scientific computing and enterprise
oh look somebody who has no fucking clue on /g/, never seen that before
>>54265653
The 295X2 was still the most powerful GPU on the market, they just took the crown from themselves
>>54265738
They were never targeted at that, it was just luck that the first Titan's were basically cheap Quadro's
That's the only reason why people use them in workstations
>>54265653
Anon already has his
>>54264228
>>54265738
>titan x targeted at deep learning sector
>can't do double precision 64 bit floating point compute
titan x is squarely a gaming video card
>4GB VRAM
Wew lad
And it's not the most powerful GPU, it's the most powerful single card solution. Even then Xfire/SLI sucks dick.
>Runs at 200°C
>>54265653
Still can't run CUDA and cuDNN.
>>54265705
980ti 600 bucks
how new
amd and nvidia do this shit all the time so they can fire off a "muh fastest video card in the world" press release
>>54265653
Wow that actually looks bretty gud
>>54266441
Its 8GB you fucklechuck
AMD has no drivers or examples for developers.
>>54266986
>AMD
>Drivers
meme
e
m
e
>>54267005
its true. Look at nvidia extensions for openGL vs AMD (none). Does AMD even have vulkan yet?
>>54266984
That's not how VRAM works in crossfire.
>>54267017
Yeah but only shitty beta drivers
>>54267084
Well the article said 8gb
>>54267088
8 total, 4 per each. In gaming you get 4GB usable. Outside of gaming you may be able to dedicate individual GPUs to separate tasks to utilize all of it but that's not why people buy this card. People that have workstation workloads get workstation cards.
>>54267107
Well thats fucking useless.
Maybe they layed the ram out differently?
>>54267116
It's HBM, 2 Fury GPUs. It's the exact same layout of the fury X, albeit on one single PCB. So yes, completely useless. 2 fury X's cost less and will perform better (though marginally).
>electrical/computer engineering student here
>currently running 2 Radeon HD 7970s
>running a 40" 4k (didn't fall for the high dpi meme) display
Besides game development, is the Radeon Pro Duo actually designed to function as (or similarly to) a workstation GPU? AMD has undoubtedly released an absolute beast of a graphics card.
Do I save up for the Radeon Pro Duo? Would this be useful in the future for electrical simulations or should I put that money towards dual Xeons instead?
>>54267402
>4GB of usable memory in games
Nuh
>>54267154
Can you fit 2 fury x cards in an itx build?
>>54265738
No shit, Nvidia has been selling chipsets that don't pass QA as Geforce cards since Kepler.
>>54267505
In most itx cases yes.
>50% faster than a stock 980 ti
>35% faster than an oc 980 ti
>980 ti costs $1000 less
Who thought it was a good idea to release a card that is only 25% faster than a 980 ti for $1600? Going by those facts the fucking gtx 1070 or 1080 will have similar or better performance to the pro duo but will sell at $400-$500. AMD have madmen working for them. No wonder their company has gone to shit.
>>54267591
35% faster *
>>54265738
retardo
>>54265653
>hbmeme
>no hbmeme2
nice try sandeep, o.1 shekels have been deposited to your new sandals funding
Actually the 4 GB crossfire thing is partly true.
On most games/applications, it will only utilise 4 GB. But for anything that uses DX12, the application can take Control and use the 8 GB of vram on both crads.
Also keep in mind, 4 GB is still quite good considering that it's not GDDR5 vram, but infact HBM
>>54267673
Oh true so its a API limitation then huh i did not know dx12 could do that.
>>54267704
dx12 Presents all "similar" video cards to the application as nothing more than resources to be be used. ie. X number of compute unites, Y amount of memory, Z numbers of schedulers, ect.
Vulcan does the same, and some of the newer versions of OpenCL do it as well. There are rumblings that Vulcan will be used for some scientific computing because of the options available in it.
>>54267704
It doesn't magically do it. The devs have to spend extra time working with the games memory management to get it to work across two or more cards. It's similar to async. It can be massively beneficial for the consumer to have this dx12 feature implemented but just like async, devs won't waste time with it unless it's absolutely necessary or they actually get paid to do it. They already have a lot of work to do and optimizing a game for less than 0.5% of all discrete gpu owners who use sli/xfire is a waste of valuable time and resources.
>>54267770
Interesting.
I assume DX12 and Vulkan is the new OpenGL and Direct X 9 so to speak they will stick around alot longer than DX11 i hope.
>>54267791
Damn.
They should really make them implement all DX12 features like the console devs have to do.
>>54267801
Two things. Because AMD had the foresight to win current and next generation consoles devs will be programming for DX12 on the xbone (and the next one) and some version of Vulkan or other low level API on PS4k and the Nintendo NX. This means that they will be used to lower level apis because the console market is ~80% of the video game market.
Secondly yes, you are correct in that Vulkan is the sucecssor to OpenGL, and DX12 is clearly the next DirectX.
The most important thing here is that in the future scalability is going to be how GPU's are made and will largely function. This is for economic reasons. It is far easier (because of yeilds) and cheaper to make a bunch of smaller gpu dies than a few huge ones. Then because the groundwork has already been laid with DX12/Vulkan devs will be used to simply making it possible for huge numbers of "individual" gpus to be used. AMD is going to put multiple ~200mm dies on a single interposer with HBM for Navi in 2018. This is how AMD wrecks nVidia in the future. This is also how we all have the compute capability to run CFD at home without needing millions of dollars of computer equipment.
>>54267801
No you've got it all wrong. Consoles don't use dx12 in the literal sense. They might use a form of directx but it'll be specifically designed for that specific console (xbox). All these features that directx 12 are bringing won't be used on consoles since consoles are already running at near 100% efficiency. Things like dx11 or 12 on pc are made to make pc's target console like efficency and with every new api this efficiency gets better eg dx12 is way, way more efficenct compared to dx11. It will be impossible to get to console like efficency due to the amount of pc configurations there are out there though. I think that guy who made ashes of singularity said that consoles with the same specs as a pc will run at 2x the efficency.
go amd go
>>54267867
Please, stop talking. You clearly know nothing about how api's work across consoles and the pc space. Go back to watching adoredTV and stop shitting up the board with his nonsense.
>>54266471
CUDA
kek
>>54267888
Microsoft forced DX12 to the Xbone, and the current PS4 SDK is extremely close to the metal. There is no reason to assume that sony and nintendo wouldn't utilize Vulkan.
Also, wtf is adoredTV? Some of us work in fields that use GPGPU. I personally wont use CUDA because I don't feel like doing much more than python, which I can use for OpenGL. Where do you currently work where you know how shit works?
>>54267973
>w-wtf is a-adoredTV
>i somehow managed to say the exact things as some retard youtuber word for word
Shut the fuck up retard. Anyone with a fucking braincell can see you're blatantly shilling for AMD based on the shit you watched on adoredTV. I bet you worship the uneducated cunt. You follow a guy who not only proven he has absolutely no knowledge of how software works on relation to a GPU but also gets ALL of his tech knowledge from fucking Anandtech and WCCFtech.
Go back to bed. You have school in the morning.
>>54268027
You are correct, I do have "school" in the morning. My compute time is at 10am, but I can connect from my apartment.
You never did state how you know how shit works btw...
I'm not blatantly shilling for AMD, I'm simply stating that multiple gpu's are going to be the way things go. It's already how most of the compute gets done anyway, aside from people with fucking stacks of Xeons. There is no way in hell we are going to stay in a world of 1 cpu (we used to think that 1 core would be enough) and 1 gpu. This shit isn't going to happen, we are going more and more parallel because we have to, and to go more parallel you need more dies. This is basic economics.
If nVidiots weren't blind to reality they would see this. I suspect that when nVidia starts running multi-gpu chips in mainstream you will be the first to stand up and scream about how "innovative" it is and how "AMD BTFO on suicide watch". Kind of like how ATI was first on tessellation and now nVidia uses it as a means to exploit their architecture.
>>54268099
>and to go more parallel you need more dies.
>doesn't know in the slightest about how a gpu works
Holt shit I actually took the b8. I thought you were serious there for a second. My bad, my bad.
>>54268148
I know exactly how a gpu works, but you can only cram so many transistors on a given area of silicon.
God damnit, I took the bait...
>>54267867
Vulkan is not and was never meant to be successor to OpenGL, idiot.
>>54267770
>>54267867
get a load of this guy
>>54267973
Dx12 on the xbox is nothing like the dx12 on pc. Console apis and pc apis are totally different. Jesus Christ at least learn about what you're talking about before spouting random bullshit.
>>54268190
Don't reply to that guy. He's a fucking idiot that worships the uneducated ubershill adoredtv.
>>54268170
Are you really this fucking retarded?
>>54268148
Here is what he's talking about:
Current:
>design GPU module
>assemble N modules in the design stage
>fab all together
Future:
>design GPU module
>fab them separately, fully cut apart, etc
>put N modules on an interposer
>Voltron GPU
It's like how Intel made Core 2 Quads out of two Core 2 Duo dies.
>>54267084
Actually that's exactly how HBM works.
In crossfire (sli in the future as well) HBM and HBM2 stack on each other for resources.
>>54267550
Spotted the person that doesn't own a fury x. The rad is huge, personal experience of owning one, and it would be a nightmare fitting two of them in mid tower, let alone mitx.
>>54268270
you can fit two in most itx cases without issues, the problem is not haivng two pcie slots.
>>54268260
No, you massive faggot
Crossfire mirrors the VRAM usage of the smallest memory card, which is why you'll only get 4GB total if you pair a 390X with a 290X in Crossfire. HBM is nothing more than a different way of putting memory on a graphics card. It does not fundamentally change the way memory is treated or accessed.
>>54267867
>consoles
>80% of gaming market
What is league of legends for 400 trebek.
League has more players in one month than consoles have in probably 6 months. Not even exaggerated.
>>54268286
Go buy two fury x cards, put them in crossfire and take a peek at your usable memory. Then get back to me with your salt.
>>54268282
The cards will fit, sure, if you remove the rads lol. No, they won't fit. I own one, trust me on this
>>54268309
You really are special
>>54268309
HAHAHAHAHAHAHAHAHAHAHA
Fuck this guy is hilarious. Great bantz bro
>>54268309
That's still not how HBM works regardless of what your shitty software reports you retard.
Also, memory stacking is supposedly a feature of DX12 not HBM, which will not be implemented since 1% of the market use two GPUs.
>>54268309
Usable total is 8gb, sure. The game you're playing still won't be able to allocate more than 4096MB total. Crossfire still has to mirror memory across both cards, it merely splits the load on the GPUs.
>>54268329
The fact that the fucking retard ignored my earlier post >>54267791 means he's deluded himself into believing what he thinks is true. He's probably the same guy whose been shilling adored tv all thread.
Fucking fanboys.
>>54268322
>>54268328
>>54268329
>>54268346
Spotted the kids.
Go through with my simple test and you will see.
>>54268358
>software says it has 8 gb so it must be right
My sli 980's show I have 8gb vram in fucking gta 5. That doesn't mean I can use 8gb vram you absolute fucking spastic.
>>54268378
>calls someone absolute fucking spastic
Irony in its final form
>>54265653
Useless overpriced shit, they can't sell it to gaymers because it's just heavily underclocked 4gb piece of shit, so they're spewing some nonsense about vr. They released it because they had to keep the promise.
>>54268358
The shown memory usage in those scenarios is the total being used by both cards. So if you play rise of tomb raider with a single fury X and use 3.5GB then your total is 3.5GB out of 4 GB.
Now put that scenario in crossfire. You will now be using 7GB out of 8GB. You're not magically loading an extra 3.5 GB of new assets. You're just mirroring the VRAM across both cards for a combined total of 7 GB, 2 copies of the same 3.5 GB memory.
Please go back and read up on the meme magic that SLI/crossfire use to split GPU loads.
>>54268389
>he doesn't deny he's a fucking spastic
Atleast you know what you actually are. Now end yourself.
>>54267550
Yes if you cram it anywhere it doesn't belong, it'll fit.
>>54267770
>>54267867
>>54267973
>>54268099
>>54268170
>>54268260
>>54268309
>>54268358
>>54268389
Stop replying to this fuckbrain amd shill.
>>54266499
>shilling for Gimpvidia
There is a reason its cheaper m8.
I have a Fury X and a 4k Freesync IPS
Nvidia pisses do not know the true 4k single gpu experience
>>54268520
U wot?
>>54268496
>Gimpvidia
Holy shit I swear you're the same fuckwitt I saw on youtube using that exact phrase.
>>54268520
enjoying that silky smooth 30 fps?
>>54268548
>780ti slower than 960
T-They are not gimping Cards I-I swear
>>54268547
>using outdated graphs
Here's one with more updated drivers you ignorant jewish fuck.
>>54268686
200fps counter strike 45fps arma 3(runs shit anyway) and 45-55fps gta v goy which is fine
>>54268746
If this was true other than for maybe really unoptimized games then the 960 would right now be faster than a 780 ti.
This is the most recent game (dx11) that is out. If what you were saying is true then the 960 would be ahead, except it isn't. The 780 ti is nearly 100% faster.
>>54268808
So with a small +200mhz oc to a 980 ti you can gain 10% extra performance whilst also putting the fury x in the ground. Damn imagine how much of a performance boost an average +400mhz oc will give.
Thanks for showing. You just sold me a 980 ti.
>>54268881
>>54268808
Oh shit it's actually 20%! I read it wrong.
Fuck me I need this card in my life.
Costs more then two Fury X's or two 980 Ti's, but performs worse then both. And costs way more then two Fury Nanos but barely compete in power efficiency.
There's literally no reason to buy it unless you're going to do some Linus Shit tips Level retartedness like building a high-end gaming rig with a mini-Itx board.
>>54268828
See this is what confuses me. You guys claim that nvidia shit isn't gimped, but you say that amd cards are re-badges. Now I'm willing to admit AMD cards are pretty much re-badges, but I also think Nvidia cards are too, but you fuckers are too ignorant to believe that your beloved would gimp your cards.
>>54265653
Indians won once again
Fastest 4k card on the market
>Costs a shitton
>Only 4GB per GPU
>Dual GPU, so possibly compatibility issues
>Fuckton of Watt
Seems good for what it is, but you'd have to be retarded or loaded or in a very niche market to buy this.
what can you do with it?
>>54270822
Same thing you can do with other cards but faster.
Has anyone actually confirmed that it can use both gpu cores in VR?
Last I heard most VR games didnt support 2+ gpus.
>>54270822
None, AMD Crossfire despite having almost 100% scaling is more useless than Nvidiai poor 70% scaling because AMD drivers sucks, The Division which was released months ago still has issues on Crossfire setups.
The 290X2 which was another dual GPU by AMD and the "strongest GPU" is only the fastesr CPU when it works, when it doesn't which is almost always it behaves slower than a 290.
>>54265912
>deep learning
>using double precision
>ever
It benefits so little from it that the hot thing to do is to use half-precision floats (16bits).
Fuck off, retard.
>>54265753
>performance showcases were mostly on machine learning and computer vision tasks
>n-no u-ur dum!1
Fuck off, retard.
>>54265779
See above. Additionally, nvidia gave away titan x's and z's to ML labs and made presentations at large ML conferences promoting these cards (e.g. nips and iclr). Fuck off, retard.
>>54267612
Fuck off, retard.
>>54270967
Lol this is retarded if true.
>>54271067
You sound like one of those retards that says shit like
"Nvidia surround, 5760 x 1080, isn't supported by many games."
>>54270743
>fuckton of watt
>is the lowest drawing duel gpu ever made
>>54271067
>le amd drivers sux
>meme game actually fucking matters
>nivida cuck
>>54271497
You sound like a retard.
>3 8pin connectors
AMD HOUSEFIRES
AMD HOUSEFIRES EVERYWHERE
>>54269842
Please tell me how that picture shows a 780 ti being gimped so a 960 is faster
>>54271628
That's a pretty stupid thing to say. It's like saying a tank doesn't use a fuckton of fuel just because it's the most fuel efficient tank around, it'll still use a fuckton of fuel. But you could say this (and other points) are more gripes against dual GPU in general, than gripes against the Pro Duo specifically.
The fuck did you do to that image though? Shit's all blurry.
>>54271725
>fuckton of fuel
>is duel gpu card doing no worst than 980ti
>>54271939
>293 isn't more than 272
You're fucking retarded. Not only because of what you're saying by the way.
>>54270697
still can't handle 4k 60fps
>>54272061
>muh fps meme
AMD sets out to create a virtual reality in which they are the leading GPU manufacturer
once they're done, it happens for real
P O T E R Y
O
T
E
R
Y
>>54272061
likely beast at 1440p though
>>54265653
why does it have to be >NVIDIA SHILLS ON SUICIDE WATCH
why can't it be >LOOK AT THIS COOL THING AMD IS DOING
god this board is trash
(btw i own no nvidia hardware :^) )
>>54268190
It was literally called glNext. It's not a successor to OGL because OGL is garbage, it's a replacement while they keep OGL alive for compatibility reasons
>>54271125
>h-hurr durr
Your a retard.
>>54265738
>Nvidia could afford to make titan x and titan z because the hardware was targeted at the machine learning/big data sector, NOT at the gaming market
This was the case with the original Titan. As >>54265912 mentioned, the Titan X is 100% a gaming card and nothing more, just using the Titan brand to justify the cost.
>>54272173
openGL is definetly not garbage, especially with the command_list and token buffers giving near vulkan performance
>>54272299
the titan was just as "gaming" as the x it just happened to have floating point since it was based on a professional architecture
>>54272299
>gaming card
you fell for a meme. Its just an excellent, high perf card.
>>54272318
You clearly missed NVIDIA's entire marketing campaign behind the release of the original Titan then.
>>54272319
Fuck off retard.
>>54272330
how about you stop replying in memes, kid.
>>54272364
How about you crawl on over back to /v/?
I can't tell if this thread is mostly trolls or people are actually this stupid (on both sides of the debate). Either way it's cancer.
>>54272371
Why are you so upset? You said something really stupid, and now you can't even explain your position
>>54267867
What is cfd?
>>54272403
The Titan was brought to market as a consumer-friendly workstation card. I know you were too young to remember that but consider not posting because you'll only embarrass yourself further, retard.
>>54272434
Define workstation
>>54272299
You don't need specialized compute capabilities to run CUDA. The Titan X is a gaming and a general card for the CUDA dominated enterprise/professional market.
There's really no reason for the Furyx2 to exist. AMD multi gpu game support is genuinely trash and they abandoned their firepro support back in 2012. It's an awful consumer card for the price and it's an awful product for professional use.
>>54272440
>>54272459
https://blogs.nvidia.com/blog/2013/03/13/geforce-gtx-titan-cuda/
>>54272467
thanks for the link but answer the question in your own words
>>54272475
You have been spoonfed enough already as it is, you need to learn these things on your own.
>>54272494
I've worked in graphics since 2013, went to gnomon from 2011-2013. Im trying to enter your mindset, but you dont seem to have any opinion on your own other than memes
>>54272494
Not him but cut the bullshit and tell people what you think or don't participate in the first place. Nobody got time for your "guess what I'm thinking and feeling based on this url" bullshit.
>>54272518
>>>/reddit/
You don't need a Titan to use Cuda, plenty of people who work with Cuda actually buy the cheaper cards like the GTX970/GTX980, source the official cuda forums.
>>54272518
>>54272536
>NVIDIA shills desperately trying to find something to nitpick and attack because they can't stop meme spouting
ok
>>54272561
>work with Cuda
Why would anybody sane do this?
>>54272562
>I don't like it when people disagree with me
>why can you all agree with me
>/g/ is my safe space
>pls be my echo chamber
>I don't want to talk, I want you all to agree with me
Why are you even here
>>54272543
>>54272562
>>54272591
>literally describing yourself
I'm not the one trying imply the Titan X is anything but a gaming card.
>>54265653
>this means anything
>amd good for anything
the best all purpose choice for your money is the gtx970
theres nothing it wont do
and its very fairly priced
amd stopped being good some time ago
>>54272571
It's a nice thing for speeding up video rendering on basically every video cutting software.
>>54272606
Could've just said that from the start you fucking drama queen.
And no you're wrong.
>>54272571
For ML/big data stuff, cuda is significantly faster (4-10x easily) than opencl, nevermind the fact there's no cudnn equivalent for opencl.
>>54272646
>And no you're wrong.
xD
>>54272622
3.5 memebytes
>>54272606
>I'm not the one who cares about reality
OK kid
>>54272663
>Titan X is anything other than a gaming card
>"reality"
Fuck off retard.
>>54272651
just putting in equal amount of effort m8
you need to pull your own weight if you want people to take you seriously in a discussion
>>54272723
There's nothing to discuss outside your stupidity.
>>54272729
lol k
Buy Nvidia and support zionism!
>>54272670
OK retard. Whatever helps you sleep at night.
>>54272793
We don't want that now do we Pajeet?
>>54268828
Am I the only one not taking people serious who talk about "optimized games"?
Some real developer-gamer perspective cringe.
>>54272925
Yes. Now go back to >>>/reddit/ from whence you came.
>>54265653
stick your discussion wherever you want but
>nvidia market share 90% or so I don't care google it yourself
>amd the small rest
>there are 10000x more amd shills on /g/ than nvidia shills
how is that even possible
someone explain that to me
>>54271628
>duel
>>54272844
>not that guy
>no ECC VRAM
>1/32 DP rate
>workstation
My sides
>>54268828
What I see from this graph is that a 280x has the same performance as a gtx 780
Why? The 780 was so much beefier when it came out
>>54273076
Found the inbred, everybody!
>>54270697
$1500 and still cant get over 40-50fps at 4k
Either those games are hitting their vram limit because 4GB or ati need to work on those fucking drivers.
Tomb raider especially look at that shit.
That or the benchmark pc is struggling to keep up.
Digital Foundary found that even with a 8 core i7 the Crossfired Fury X's beat the shit out of the cpu even when overclocked
Fuck me thats brutal i guess 4k is still a ways off.
if a 4.5ghz~ 8 core cant keep up in DX12 games we gonna need a big leap in cpu tech
>>54272952
>edgy /v/-tard being edgy and ignorant
Never change
>>54273138
OK r.eddit
>>54271695
>comparing pears with apples
Benchmark the game after some months, only then you /could/ say that /maybe/ it's getting gimped or not.
it's over. Microsoft won. seriously canot stang GPU faggots.
>>54273273
e.g. fallout 4 1.3 am i rite?
>>54273326
>Canot stang
linuxfags everyone
>>54273000
A lot of people on /g/ are not the same people that buy products from companies like nvidia, Microsoft, adobe or apple.
They avoid them either for ethical reasons or bang/bucks ratio.
i see many biznes opportunities in suicide watch
>>54268520
>gaming on a ips
>>54273552
>using a noun as a verb
>using any kind of monitor that's not IPS
>>54273602
I already did the ghosting on it was fucking worst thing in the world even the text on the browser ghosted.also tweaking was of no help.
>>54273663
Lol'd
>>>/reddit/
>>54273725
I am not from reddit you faggot
>>54273085
>can't come up with a reply
>lel inbred cuck xDD
It's not my fault that facts are against you but I don't expect a room temp IQ to be able to understand this.
>>54273108
My two 980tis don't sweat in 4K, never below 40 in Gta 5 maxed
ENOUGH.
People out there are ACTUALLY falling for the AMD meme.
They actually pay for non functional garbage and ruin their lives. They spiral into drug abuse and crime. It isn't funny anymore.
>oy vey
>>54272409
Computational Fluid Dynamics.
>2016
>Buying anything other than Nvidia
>>54277260
>nvidia
>buying anything other than 2016
>>54275383
i've been using an amd exclusive build for like 2 years straight it's fine
