>$1500
HAHAHAHAHAHAHAHAHAAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHHAHAHAAHAHhahhaH
>>54296711
They'll find some way to defend this, whether it be "it's not a gaming card" or start floodding the thread with images of bad nvidia frame times. Either way, it's just deflecting.
>>54296711
ahahahah holy fuck AMD done guffed
SEND HELP
>>54296786
it's a feature
AMD, NOT EVEN ONCE
More than 16.666ms = shit.
>>54296786
AMD is drawing more pixels. Looks good to me.
>>54297078
Oh my sweet summer child
>>54297089
Nice shitpost, retard.
>>54296711
I WANT /v/ TO FUCKING LEAVE REEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE
>>54296711
Amd cucked to death
>>54296711
Frametime?
Someone explain. Is this the latency between frame updates or something making the game run at like 25 fps or something?
>>54298197
i seconf this
>>54298197
frametime is how long it takes to render each frame
check out this article, anon
https://www.mvps.org/directx/articles/fps_versus_frame_time.htm
>>54298197
It's the time that every frame lasts. If you want the game to play at 30 FPS, then every frame should last 1/30th of a second. If you wanted the game to play at 60 FPS, every frame should last 1/60th of a second. So, the smaller the frametime, the bigger the framerate and, therefore, the better.
The Fury X seems to have higher frametime than the 980 TI, so it runs slower, but at least it's constant, a.k.a the FPS isn't constantly jumping all over the place. Now look at the Radeon Pro Duo, AMD's new graphics card. It's all over the fucking place. Framerates go from something like 20 to 120 FPS within a single second and that results in an irritating gameplay.
>fury duo
Who care
>>54296711
does frametiming matter when Free/adaptive/G-sync is in play?
>>54296711
I wanna know what CPU they were running. GTA V is known as a CPU intensive game and AMD drivers have more CPU overhead compared to Nvidia
>>54299375
i7-5960X
http://www.pcper.com/reviews/Graphics-Cards/AMD-Radeon-Pro-Duo-Review/Testing-Setup-and-Professional-Testing
>>54299375
>http://www.pcper.com/reviews/Graphics-Cards/AMD-Radeon-Pro-Duo-Review/Testing-Setup-and-Professional-Testing
looks like a i7-5960x
nothing on the clock speed though
I'd assume stock or a minor overclock
>>54299428
>>54299435
looking in the comments, they can't test DX12 game frametimes because their testing method doesn't work with it yet
>>54296711
what you arent realizing is its all about asynchronous computing now. Next gen nvidia wont have that capability and amd has already been implementing it since the 7000 series. this means nvidia will not be able to brute force better performance than async enabled gpus. Now if generation after next they develop the capability then they might have a chance.
>>54296711
>>54296786
Too bad they've only used 4GB AMD cards, throwing 2 8GB 390Xs in there to compare to that 295X2 would've been interesting. GTA V is definitely the sort of game that can benefit from more than 4GB VRAM, especially at 4K. I wonder how much of that is shit drivers and how much is insufficient VRAM.
>>54299542
op wont post source he's too scared.
>>54299591
whoops meant
>>54299428
>>54299435
Fury X is born dead. The 'only' 4GB HBM1 per chips is the issue for this very high end card. Overhyped meme card.
>>54299575
There's a gigantic massive fucking watermark on the images moron.
>>54299602
this benchmark is at the top.
>>54299622
I see it as a test for HBM memory management for AMD in the field, and the early adopters are footing the bill.
>>54299640
thats not a source you dummy.
>>54299648
scaling for dual GPU cards is always weird, and Nvidia hasn't done any since the Titan Z
I'd like to see Fury x2 CF or R9 390X x2 CF, as well.
>>54299651
it seems most people itt dont even realize the potential of asynch computing it really doesn't surprise me though.
>>54299710
my point was all that frame lag is from running out of vram.
>>54299718
with async compute, effective RAM just stacks, right? so 2 4GB cards would have 8GB instead of the effective 4GB we're used to
>>54298150
/v/ is cancer. Shitting their own board is not enough
>>54296711
Fuck back to your daycare board you fat smelly piece of shit
>>54299744
yes that as well but async is more about being able to compute graphic commands and render at the same time.
>>54299823
I was saying that because people seem to think OP was claiming the Pro Duo was hitting a VRAM cap, but yes.
I still want my fantasy future of GPUs with no VRAM, and dedicated VRAM/memory controller PCI-e cards
>>54299851
With that much lag there was definitely some sort of cap going on. and the most likely one is vram also this test obviously wasnt done in dx12, dx11 doesnt use async.
>Crossfire/SLI has tons of microstutter
WOW WHO WOULD'VE THOUGHT! IT'S ALMOST LIKE CROSSFIRE AND SLI CARDS RELY ENTIRELY ON DRIVERS!
GTA V = GameWorks
>>54299900
the way it stutters, I'd think it's referring to the CPU for a certain frequent operation. what that operation is would take a lot of knowhow about the hardware, game engine, and drivers
>>54300117
no its offloading to ram causing stuttering.
>>54300101
does GTA V have gameworks though?
>>54300101
95% of current games = gameworks.
What's your point?
>>54300117
No. It's a VRAM bottleneck.
>nvidia cards getting shit on by AMD cards
>suddenly frame time matters
>>54300136
>>54300134
that's one hell of a hivemind
>>54300145
except I dont think gtaV has gameworks.
>>54300143
frame time matters the most. thats showing vram running out.
>>54300166
I know it uses both AMD and NVidia special flavors of Anti-Aliasing, so I'd think they'd have to use GameWorks to get access to it
>>54300206
gameworks is a name for a suite of technologies. HBOA+ being one of them and I agree whenever they implement any of them it seems to make the game run like shit for everyone. I didnt realize gta V had HBOA.
what program did they use to record frametimes?
>>54300246
to add to this HBOA+ seems to be the least caustic to performance.
>>54300260
I was hunting around for a while, I can't find anything about their testing methodology for frame timing
>>54296711
>Card aimed at devs is kinda shit at games
Granted Crossfire is shit (so is SLI really just slightly less shit).
>>54300276
>>54300260
https://www.youtube.com/watch?v=CsHuPxX8ZzQ
>>54300321
didn't think to look at youtube, thanks
>>54300300
I doubt this test is legit to be honest. and if it is I was just informed gtaV has gameworks and that would explain everything.
>>54296711
>using crossfire
I remember when people were saying that that beta driver about 4 years ago was going to fix this issue.
>>54300300
>Card aimed at devs
Why would devs want to develop on something that their customers were not using?
>>54300143
>Frame time doesn't matter because AMD is terrible at it
>>54300397
unoptimized early builds that need extra muscle, but you don't want to use a workstation card because they are running different drivers. the issue is that AMD is assuming async is being used for all current development.
>>54296711
I'm so glad the frame timing thing came out. I always felt like in certain games even if the frame rate was 60+ it felt janky.
>>54300343
watch
>>54300321
they seem to be pretty thorough in their testing methodology
>>54300440
welcome to multi-gpu gaming
that's sometimes noticeable even on NVidia setups
>>54300406
total bullshit.
>>54300397
because its for vr development.
>>54300440
with these numbers it would be unplayable not just janky.
>>54300454
>watch
wut?
>>54300460
sli is no better than crossfire.
>>54300460
will be interesting to see how this pans out with asynch computing.
>>54300460
I noticed back on my 580 SLI setup playing things like Metro and what not.
>>54300435
>unoptimized early builds that need extra muscle
yeah unoptimized early builds are really the place where I want to deal crossfire issues.
>>54300504
>because its for vr development.
What the fuck does that mean? I'd develop VR games with the stuff that people play VR games with.
>>54300529
then you would want to use ati because they support async.
>>54300529
VR games typically use two GPU's fine as they have to render two scenes that are only partially linked.
>>54300529
amd dual cards dont use xfire though.
>>54300556
also they are going to be developed with dx 12 support.
>>54300550
>then you would want to use ati because they support async
async? You mean DX12 async shaders or whatever the 980ti lacked? I'm not sure why that's necessary for VR development. Vulkan seems like a better choice anyway, seeing as many people are not going to Windows 10.
>>54300556
that makes sense. Is that where the async shaders come in?
>>54300559
Can they not do alternate frame rendering then?
>>54300614
>as many people are not going to Windows 10.
what the fuck is this guy talking about?
>>54300614
they can and do because they have to, but they are just setting the framework for asynch. nvidia is going to have upgrade their architecture generation after next to take advantage of asych. pascal doesnt do asynch.
>>54300670
>what the fuck is this guy talking about?
DX12 is Windows 10 only. I wouldn't want to lose the entire market segment that's still on Windows 7.
>>54300670
and to add to this someone had to make a hack to allow nvidia cards to be dedicated physx cards. amd on the other hand is allowing nvidia to take full advantage of thier new technology. amd has and always will be an open source company. nvidia is cancer on the entire community.
>>54300727
you have no idea how many people have switched to 10 even though dx12 hasnt been fully implemented.
>>54300727
async is also in Vulkan
>>54300759
That doesn't change the fact that a significant portion of the market will not end up upgrading and you're wasting development time supporting both DX11 and DX12, rather than just going to Vulkan, which seems to implement the same features.
>>54300786
dx 12 is no big deal to implement dumb dumb. only indie devs will tell you its not worth it. Also vulkan.
>>54300786
an insignifigant portion you mean. Its just a free update for normies. only weirdo's think 10 isnt better in every way. shit it has less overhead and performs much better vista 7 or 8.
>nb4 botnet.
>>54300852
better than*
>>54300798
is DX12 the exact same as DX11 with features added?
going from DX9 to DX10 seemed like a pretty massive change, from what I've seen. I've only had much experience with OpenGL though.
>>54300852
It's a free upgrade that breaks a ton of legacy compatibility like windows upgrades always do. A lot of people don't like Windows 10's other features either, such as forced updates, uninstalling pirated programs, etc.
I would imagine a significant portion won't upgrade to Windows 10 or will downgrade later.
That's why they need freesync to kick off so bad
the GPU equivalent of retard-strength, unrefined
>>54300916
Why would freesync improve this? Your refresh rate would stutter then, in addition to your frame rate.
>>54300910
I am both of your quotes. DX12 is not just another api it enables asynch. just wait till june everyone will take the upgrade before they stop offering it. It runs better it has less overhead and implements asynch its like the new XP.
I'm glad I never fell for the multi-GPU meme.
>>54300931
>DX12 is not just another api it enables asynch
so what would you have to change to get it to run in DX12? Would you just have a different FBO render method or something like that and otherwise the code would be compatible with DX11, if you just wanted the async shaders?
>just wait till june everyone will take the upgrade before they stop offering it
There are people that can't upgrade due to software incompatibility.
My Windows update broke, so I can't upgrade to 10 or update Windows. I use Debian for everything except games anyway, so it's not a big deal.
>>54300938
I went for GTX 480 SLI back when a GTX 480 was $100. GTX 680 performance with only a slight amount of stutter. I live in Canada so the warmth was nice.
>>54300938
this.
might be worth it with explicit/implicit multigpu in the next generation or two tho.
>>54300991
multi-gpu will probably always be such a small portion of consumers, it will never work particularly well, IMO.
>>54300953
>so what would you have to change to get it to run in DX12?
not much. even indie devs are retro patching their games to work with dx12
>There are people that can't upgrade due to software incompatibility.
backwards compatibility is still working. This would be a niche problem.
>>54301014
you would be wrong asynch fixes all the problems previously experienced with multi gpu's. You niggers are living in 2015
>>54301021
>not much. even indie devs are retro patching their games to work with dx12
I'm literally asking what elements of the renderer you have to change. Is initialization of DX12 the same? Is the method of rendering DX's equivalent of VAOs the same? is HLSL the same?
>backwards compatibility is still working. This would be a niche problem.
I've never seen a Windows upgrade that didn't break something. Banks are still paying for Windows XP support because of it.
>>54301043
>fixes all the problems
How does it solve all the problems? Why would it be any better than split frame rendering, for a non-VR situation?
>>54301048
there is not much difference its nothing major. 10 breaks some things but its nothing like going from xp to vista.
>>54301088
educate yourself on async. its a good read. It allows all your cores from your cpu to your gpu to be utilized fully. something that has been needed for at least 8 years.
>>54301094
>there is not much difference its nothing major
have you ever actually developed for DX12?
>>54301116
have you?
I didn't notice another GPU thread so I'll just ask here.
Are the H.264 encoders on GCN1.1/1.2 and Maxwell superior to GCN1.0? And I don't just mean spec wise, but actual quality. Even if I crank the bitrate up to absurd levels (like 20Mbps) while recording with my 7870, it still looks like utter garbage compared to x264 recording via the CPU.
>>54301124
of course not. I asked about the specifics of the DX12 API, which should indicate that.
If you're not able to provide such basic specifics, I assume you've not used the API at all and are just making stuff up.
>>54301129
youre supposed to record at lossless then convert.
>>54301145
I will tell you this. Every single game that comes out in 2016 will have dx 12 support. This is fast tracked api unlike the others it has a huge performance boost because of its asych capabilities and is the only reason 10 is free.
>>54301208
The 7870 can't seem to do lossless, the output is full of dropped frames if I set QP values to 0.
Lossless via x264 looks great though.
>>54301208
I'll take that as a no then.
Anyway I'm hoping people will move to Vulkan. I'm also hoping the wine people implement DX12 using Vulkan and it works well, but that's just a pipe dream.
>>54301259
>I am an indie dev and can't be asked to look into it.
>>54301247
unfortunately video capture needs some overhead when using the gpu. Stick with what works for you. Theres no advantage to using the gpu if your cpu can handle it.
>>54301341
>I am an indie dev and can't be asked to look into it.
1. I'm not an indie dev
2. Looking into it means learning enough of the DX11 and DX12 API to build a renderer from scrach. That's a pretty long way to go to verify the claims of someone who's clearly never worked in the field.
>>54301259
>>54301341
Linux users will always be btfo when it comes to games. dx is the standard.
>>54301393
you assume too much especially since you dont dev games.
>>54296711
>dem lag spikes
Why is AMD such a garbage.
>>54301408
read the thread before you shitpost.
>>54301396
Yeah that makes me sad. OpenGL seems to be a perfectly good API. If only Wine worked better, gaming on Linux would be good.
>>54301405
I don't develop games, but I understand enough to know the changed required when moving from one rendering API to another. You don't develop games either, at least DX ones.
>>54301426
dx 12 is not a big change from previous implimentations. Its designed that way. youre obviously a neck beard from your first post just mark my words dx12 will not take years for devs to implement like the others. this is the first api that actually has performance improvements over the others. Considering consoles use this to their advantage and basically all ports are shit because of this difference its a huge benefit to pc gaming. think what you want about be but this is a truth.
>>54301489
>dx 12 is not a big change from previous implimentations. Its designed that way.
You have made this assertion several times. Please provide evidence for this assertion by explaining what elements of a DX11 renderer would need to be changed to accommodate DX12, as well as what would need to be changed to add in async shaders. If you're not making shit up, it should be trivial to explain this.
>youre obviously a neck beard from your first post
cool story bro
>this is the first api that actually has performance improvements over the others
That's definitely not true. Programmable shaders had a performance impact.
>>54301417
>look at the damage control not the graphs
ok lmao
>>54301655
its been posted it was a vram limitation. If you had any clue on how shit works it would be obvious.
>>54301542
>You have made this assertion several times.
linuxbeards btfo.
heres a pre-emptive I told you so.
>>54301709
are you 12?
>>54301763
you'll see. I would recommend investing whatever you have in amd. you wont be sorry.
Whoever didn't see this coming was an idiot. 390 owner here, you have to wait a long time for Crossfire or dual GPU profiles to catch up.
>>54301777
you are a premium fuckboy
>>54301789
dx 12 benefits users of amd all the way back to the 7000 series will see significant performance improvements nvidia cant into asynch until gen after next.
4GB was a big gimp to Fury and this card.
>>54301792
you just havent updated your database kiddo.
>>54301819
hopefully memory stacking soon
>>54301813
Async itself isn't that important for most use caases, IIRC only Ashes of the Singularity is the only DX12 game to implement it precisely because it's an RTS with lots of individual units on screen.
>>54301826
the future is now. dx12 allows mem stacking with amd and nvidia. you see the difference now?
>>54301827
BULLSHIT
>>54301834
>you see the difference now?
don't games have to support it to work
>>54301834
>dx12 allows mem stacking
what the fuck does that mean?
>>54301838
Quantum Break at least does not use async compute and generally favors AMD nonetheless. But isn't that game still basically a broken port?
>>54301839
all games in 2016 will support it unless they are incompetent.
>>54301841
it means it will allow nvidia cards to stack with amd. you do know that dx12 is an amd technology dont you
>>54301853
>all games in 2016 will support it unless they are incompetent.
oh shit, i can see it now
>>54301841
Take a 970 SLI on dx11 as an example.
3.5GB VRAM + 3.5GB VRAM = still 3.5GB regardless of how many 970s you have in SLI. The VRAM is mirrored, not stacked.
In dx12 3.5GB + 3.5GB = 7GB to use. That's stacked.
>>54301860
You can already test gains with dx12 using the dolphin wii emulator, of all things. Pretty high FPS increase.
>>54301853
>>54301864
Oh, that makes sense.
>you do know that dx12 is an amd technology dont you
I thought it was Microsoft...
>>54301827
async only makes a difference when you have multiple GPUs, right?
>>54301877
asynch is the future boys.
>>54301880
it is encapsulated within dx12 they basically adopted it because they didnt want to lose out to vulkan.
>>54301880
dx12 is an M$ proprietary fork of mantle, and mantle is AMD. Vulkan is the open source version.
>>54301880
>async only makes a difference when you have multiple GPUs, right?
fuck no bruh.
>>54301898
where does it improve performance?
>>54301916
read the thread. I already explained it.
So if Polaris cant catch up with the Fury and Duo pro hopefully everyone dumps crossfire its been stuttery shit since it came out
>>54296711
>Crossfire
You know what you are getting into when you purchase a dual gpu.
>>54301938
https://www.youtube.com/watch?v=4qJj1ViyyPY
Pajeet does want to move away from crossfire.
>>54301938
thats the thing dx12 fixes all of the problems. instead of the cards rendering half the screen they offload every frame. do you get it now? brought to you by amd.
>>54301963
every other frame*
>>54301963
this is the difference between amd and nvidia. amd is all about performance and optimization and open source, nvidia is all about gimpworks.
>>54302037
this was meant for>>54301938
>>54301963
>instead of the cards rendering half the screen they offload every frame
Those two things describe split frame rendering and alternate frame rendering, which both exist in SLI and probably xfire
>>54301682
>I-It's vram limitation
No shit faggot, that doesn't exempt AMD from making a shit product and having shit frametimes
>>54301963
>DX12 magically fixes xfire issues
Well if it fucking supported it i imagine it would...
>>54302037
AMD has Hairwerks and made for amd as well they both do it mate dont deny it.
i remember the Nvidia performance in BF4 has always been fucking terrible compared to the same card in BF3
Ive owned both a 290 housefire and my 770 for ages so i know what they do
>>54301961
I know its great, besides now we have the tech to put 2+ gpu's on a single card, leave the xfire to nvidia and workstations
The current solution is terrible for gaming because it introduces microstutter due to sync issues i assume.
>>54301954
Pretty much this.
Until they sort out the underlying issues hopefully going to Vulkan and DX12+ will solve alot of them.
DX11 was good but the lack of low level hardware access is still a massive issue.
i saw a 150% increase in performance on mantle so i hope there still room to sqeeze out more performance now that we're hitting out head on the ceiling especially with CPU tech
8 cores @ 5.0ghz is the best we can do, saving AMD's ZEN which might catch up to intel's older i5's
>>54296711
>>54296786
is that DX12?
>>54302073
no they dont but nice try.
>>54302075
>2016
buying nvidia
>>54302078
>AMD has Hairwerks and made for amd as well they both do it mate dont deny it.
that was in one game dont deny it.
everyone with a brain realizes amd is the open source alternative. and most everyone will tell you gimpworks is a shit,
>>54302111
>no they dont but nice try.
http://docs.nvidia.com/gameworks/content/technologies/desktop/sli.htm
the way they were described, yes they do
>>54302110
are you kidding?
>>54302136
they may exist but they suck dicks.
>>54296786
>>54296711
the spike is literally one single frame
on that particular graph it looks dramatic, but only one frame will spake beyond 70ms (still not noticeable to the human eye) out of 1 second. @ 60 fps this is only 3% of frames, which is not frequent enough to be noticed ontop of the fact the actual latency spike itself is not noticeable
nvidia is in full damage control to take pride in such data
the test is entirely worthless, funny how a worthless test is the only thing nvidia truly excels in
sombody needs to keep the hype real
niggers need to believe
>>54298150
I thought you were saying that as a joke but then I realized I actually wasn't on /v/.
>>54302165
its an obvious vram issue no reason to take it beyond that.
>>54298268
this is wrong
>>54296711
>shit card with no target market performs shit
I don't know why AMD even bothered. No one asked for this.
>>54302078
>TressFX
>an MIT License Open Source Lib
>comparable to Gameworks
memes
https://github.com/GPUOpen-Effects/TressFX
>>54302458
/thread
>>54302165
>on that particular graph it looks dramatic, but only one frame will spake beyond 70ms (still not noticeable to the human eye)
you clearly don't own a fury card, the stuttering is unbearable and nearly unplayable in some games. with two fury x's, most games become completely unplayable. the problem is just as pronounced in older amd multigpu setups too, can't wait to dump these pieces of shit on some dumb goyim and upgrade to pascal when it comes out.
>>54301682
it's not a vram limitation, when i had a 970 it played gta5 just fine at 1440p without hitting the 3.5gb wall of stutter. this is just AMD putting out shit hardware and then hiring paid shills to post bullshit.
>>54302705
you clearly dont understand what dx12 is.
>>54302742
you better update your database kiddo.
>>54302756
>you clearly dont understand what dx12 is.
kek, do you think dx12 is magically going to fix the problem of there not being enough bandwidth (and tons of latency) between the two GPUs?
>>54302773
its exactly what it does among other things.
>>54302783
so dx12 is so revolutionary that it's going to make the 295x2 and fury x2 communicate over some sort of imaginary interconnect instead of the pcie bus? wow
>>54302789
yes. you are retarded sir.
>>54302806
>yes. you are retarded sir.
i know this is b8, but all multigpu cards AMD have released since the 7990 haven't been interconnected. they're limited by the bandwidth of the pcie bus, which is 1/50th the bandwidth of HBM1. dx12 isn't going to magically fix this hardware limitation. what is most likely going to happen (due to vulkan and dx12 forcing devs to do much more work, especially with multiple gpus and synchronization and other bullshit) are lazy implementations where only 1 gpu is ever used, or the second only does post processing stuff that can be offloaded without significant bandwidth cost.
>>54302832
look up dx12 dumb dumb. amd is all about making the platform better for everyone yes that includes nvidiots.
>>54302843
>amd is all about making the platform better for everyone yes that includes nvidiots.
except for their own customers, it would seem
>>54302843
you're retarded, for real
>I can't describe anything technical about dx12, but I know for a fact that it's going to make everything better because I read it on the internet!
>>54302850
you're gay and thats ok with me.
>>54302855
that goes for you to.
nvidiots everyone.
>>54302864
>that goes for you to.
he went into far more technical detail than you did
>>54302873
this whole thread is about dx12 goober. gimpworks is the mind killer.
>>54302226
Actually, no it isn't.
>>54302165
The human eye is sensitive to events with a duration as low as 1/300 of a second. This was discovered in experiments done by the US Navy, where they'd flash images of aircraft in front of pilots for tiny fractions of a second, and the pilots would still be able to identify the aircraft.
>>54302938
once you let it flow through you only amd will remain.
>>54302965
Don't get me wrong, I'm a committed AMD fanboy through and through, but I think we now know why the dual Fiji card was retargeted away from gaming, because this performance is retarded.
>>54302938
I remember watching a miniseries hosted by Richard Hammond about speed where they tried to get some point across about how we can miss things that are only visible for a fraction of a second by claiming to have flashed something on screen during playback but then I immediately realized that it is getting broadcast in 60i at best which means that there are only 30 full frames worth per second so anything "faster" than the length of one frame literally wouldn't be shown at all, unless you blended it with a frame and that would be cheating and not even emulate the effect.
>>54302982
you keep ignoring dx12 and vulkan son. its all ati from here on out.
>>54303000
barely anything is running on dx12 and vulkan, and half the things that do are worse on it
>>54302998
Heh. The magic of television, indeed.
>>54303006
dx 12 will be compatible with every game that comes out in 2016 unless they are incompetent devs.
>>54303027
The whole point of DX12/Vulkan is that it gives game devs more control over what the hardware does, which does make multi-GPU systems easier to exploit. But the sacrifice is that you actually do need to do the work of controlling the hardware manually. If I were writing a Vulkan renderer today, I would not bother adding multi-GPU support, becuase so few people have those high-end cards, it's not worth the development effort. Don't expect a Vulkan or DX12 game to magically exploit multi-GPU systems correctly just because it's using a next-gen API-- the API gives devs the opportunity to do so, but not the obligation.
>>54303068
your wrong. It makes computers work more closely to the architecture of consoles as they have enjoyed asynchronous computing since the beginning. it makes porting that much easier for them in other words.
>>54303088
I'm not wrong. Next-gen APIs will be more similar to current consoles, true. But there is no current game console that has a topography anything like a dual fiji, or any kind of multi-GPU setup for that matter, so game devs will not be implementing that functionality to take advantage of consoles. Also, I don't think you know what "asynchronous" means. Desktop graphics APIs have been asynchronous for decades now (i.e. you make a bunch of draw calls but nothing actually blocks until you call glSwapBuffers.)
Crossfire/SLI on DX12 is DOA, developers use Nvidia/AMD tools to automatize the code path for those to work and even if they don't Nvidia/AMD can almost always hack their way via drivers, the same can't be done on DX12 so it is all completely up to the developer to implement it which they won't even bother to touch.
>>54303133
you keep saying multi gpu. dx 12 has benefits for multi gpu setups sure but its goal is asynchronous computing,
>>54303068
>If I were writing a Vulkan renderer today, I would not bother adding multi-GPU support,
to make matters worse you need to design from the ground up for multiple gpus, since even consoles don't have them we're going to see a slew of driver hacks to find ways to make it work without forcing all devs to waste money implementing it.
>>54300559
Yes they do, the 290X2 behaves like a normal 290(actually slower) when the game has no CF profile.
>>54303166
sure thats why every game coming out in 2016 unless its an indie game will be dx12 compatible right? this isnt some bullshit dx update this is something that actually uses all the cores something htat has never been done before. if you ask me its about time we have had 8 core+ gpu's for more than 5 years now.
>>54303195
they are setting up the framework for their designed vulkan/dx12. buy nvidia see if I care.
>>54303196
Every game could be DX12 dumbass that doesn't means every game will be using all features sets from DX12.
Look at that newly released MICROSOFT game, it don't even has CF/SLI not to mention it runs like crap, and guess who is trying to push DX12?
I like AMD more than nVidia but when will you guys understand that high-end is always a meme and a ripoff?
>tfw always buying one year old used gpu for 150, get best bang for the money all the time
>>54303168
As I said before, graphics APIs have been asynchronous for decades now. The benefits of the next-gen APIs are reduced driver overhead, better concurrency in the renderer, and more ability to optimize for the GPU directly. Those benefits are very, very real, and I'm psyched about them. They should also favor AMD, because NVIDIA has historically had more resources to plow into driver hacks and game-specific optimizations to get more performance out of their GPUs than they really deserve, and these lower-level APIs should make all that moot. Also, with respect to AMD, the drivers for these types of APIs are pretty idiot-proof, and that's good for them too.
However, it still probably won't save the Radeon Pro Duo as a gaming card.
>>54303214
by default it will be using the multicore optimization dub dumb.
>>54303226
this whole statement is retarded though considering it just uses the cores out of the box there is no extra anything in the code. It literally is defaulted to work.
>>54303196
Do you even know anything about computing at all?
>>54303249
this whole statement is retarded though considering the poster clearly has no idea what is happening on computers/consoles
>>54303227
That doesn't fixed the problem, it just makes the CPU more efficient, OpenGL already can do near zero driver overhead and so does Mantle, yet BF4 was a mess on Mantle for years.
>>54303256
do you? my wifes daughter wants to use the computer and I'm bored with this conversation anyways. just google it. I'll be here when you realize your folly once Polaris comes out.
>>54303271
google it dumb dumb. amd is kicking nvidias ass in the dx 12 market nvidia even shit talked the game for not being nvidia friendly.
>>54303290
Bf4 runs wonderful on Nvidia hardware and this is a Dice game, everyone knows Dice is in bed with AMD.
>>54303280
>my wifes daughter
I don't want this meme to die.
>>54303313
my stepdad said to ignore you.
>>54303196
8 core plus GPUs? Do you even understand how GPUs work? My Radeon HD 4850 from 2011 had 800 cores.
If you're talking about multi-GPU setups like the Radeon Pro Duo-- there is an argument that this might be a good way to go in the future, because smaller silicon dies have higher yields. But it will require game developers to change their habits and add muli-GPU support to their games. Next-gen APIs like DX12 and Mantle will help with that effort, but they won't do it for you.
>>54303249
Literally the opposite of true.
You show promise, kid, consider taking a CS class at your local community college during your summer vacation. You might learn something.
>>54303181
I hope we can avoid that nightmare scenario of driver hacks. That's exactly what crossfire and SLI were, and it didn't work that time either.
>ITT: /v/
>>54303328
my stepdad said to ignore you. stop replying. the blinking tabs are distracting.
>>54300185
Not in the case of dual cards. It more likely shows incorrectly delivered frames through the crossfire pipeline.
Something that can mostly be fixed in driver updates. Though I won't deny that GTAV bench.
>>54303320
>>54303353
Your stepdad is having sex with your mom right now and she's faking every minute of it.
>>54303367
what are you talking about dude.
>>54303376
shut up she loves him.
>>54303392
Also I should point out that your typing style is exactly the same as your supposed stepdads which is probably the most retarded thing you could do when pretending to switch person on an Anonymous board.
>>54303435
stop making my tabs blink idiot.
>>54303453
If you're going to throw a tantrum on the Internet, why not threaten to hold your breath instead?
>>54303468
I will, I'm dead go away.
/v/ pls go and take your games and children with you
>CTRL+F
>GIMPVIDIA
>NO RESULTS
You're not memeing hard enough /g/.
>>54301129
>>54301247
Recording/encoding over a GPU is always going to look worse than via the CPU due to the nature of bit streams and how work gets dispatched over some cores versus many smaller cores.
Maybe in the future some magic compiler, format, or GPU encoding program will come along and resolve most of the issue, but it's highly doubted to ever match CPU quality.
3.5
Woodscrews
3.5
;_; kill me
>>54301129
Maxwell cards can record 4K at 130mbps, but if you want quality, to this day recording using FRAPS still gives the best quality.
>>54296711
B-BUT IT'S A CONSPIRACY!
IT'S NVIDIA'S FAULT!
IT'S INTEL'S FAULT!
AMD DINDU NUFFIN!
>>54296748
this graph would be fair if the card used consumer drivers specified for games, it uses another set of drivers for another set of tasks
>>54303791
D A M A G E
A
M
A
G
E
C O N T R O L
O
N
T
R
O
L
>>54304436
Not him, but /v/tards need to go. And no damage control as Intel master race. No need for useless vidyacards.
Well it's a multi-GPU card, isn't it?
AMD have never been able to into multi GPU
>>54304517
5970
6990
7990
295x2
All bad cards. nVidia's Titan Z and 690 were also terrible. The 590 exploded.
Bad card is bad. Why are people surprised that a dual GPU card is bad? When was the last time we have a good dual GPU card?
>>54302938
Those flashes were done in complete darkness and were done with trained plane spotters who were asked to identify planes based on those flashes. If I remember right they did a test with the room lit and it was nowhere near 1/300th of a second.
