With dx12 and vulkan multiadapter improving on all the flaws multigpu had before, and IGPUs improving in performance, why doesn't Intel release a standalone series of pcie GPUs like the iris pro and HD530?
You could get a $99 i3 6100 + $40 HD530 for example and enjoy 1080p high settings.
On a side note, how well does the IGPU in an APU work in dx12?
How good would it be for multiadapter?
Other urls found in this thread:
gamedev.net
twitter.com
That actually so sounds pretty sweet. Intels newer iGPUs are much better and dx12 compliant, and drivers would be up to devs
because it's just outdated roughly ripped from nvidia tech tacked onto their babby-targetted mainstream cpu's as an afterthought.
But it will sell and you know it
intel are innovators, they have no interest in low hanging fruit like that. If they thought they had anything new or interesting to offer the GPU market im sure they would participate, but for them it's just low hanging fruit. Not a priority, and certainly not worth the r&d or fabs.
Doesn't asrock/asus already do that? Without the need for dev support or dx12 even. It works well in low res, but craps itself in high res.
How?
Considering the majority of people have an IGPU sitting idle in their machine if they have a dedicated graphics card it would be kinda nice if dx12 could leverage it in some way.
...
i consider mine as insurance, my computer isn't hobbled if my housefire gpu dies
Women are not technology!
R u k?
Not a bad idea I'd love for some other company to enter the mainstream GPU market, but I question how much you would realistically save compared to just buying a regular Intel CPU.
Nvidia and AMD already offer low-end -100$ GPUs if you need that kind of expansion for your desktop.
I don't think Intel has a market that needs to be fulfilled, and I wouldn't be so sure that the top of the line Intel GPU would cost 100$ if it was being sold by itself, me thinks it would probably cost more than that. And if it costed something like 150$ you would probably already want to look at some higher-end older GPUs.
The more I think about it the less sense it makes. But I'd definitely love to see Intel try to come up with a good line of dedicated GPUs if they had any money to make there.
That's almost nothing.
It's free performance
Surely it requires more power, I wouldn't say that's free.
I don't know how much extra the IGPU uses, it seems to be pretty low because it doesn't majorly affect CPU package temperatures.
So when exactly are we going to see this "Multiadapter" thingy in gaymes?
Is there anything that already uses it? Will it be a setting that can be turned on/off?
When DX12 games come out.
I think the new Tomb Raider has multi GPU implemented in DX12
No, that thing has to be specifically optimized for. If devs are lazy the integrated won't be used.
Intel cannot into gpus. Their hardware is fairly impressive given the die size but a standalone card which die a painful death as Intel is absolutely not equipped to provide the software support needed for gaming.
Literally just the igpu on a pcie card.
show me her dick
Firaxis said they'd implement the multiGPU thing in Civilization VI. For what purpose is beyond me because it's a turn based game with mobile graphics and probably CPU bottlenecked anyway.
You still need drivers and Intels are biblically shit - there is absolutely no market for intel in making a discrete gpu.
Funny thing about low level APIs..
If someone came along and wrote some nice and clean code forthis hypothetical gpu sure, but anything DX11 or older will basically not work as Intel lacks the software team to fix all the fucking shit and hacks most games use.
gamedev.net
Yeah but aren't Intel GPUs the most dx12 compliant right now?
They don't need drivers for dx11 anymore, they can skip that since pretty much everything new is dx12
>Yeah but aren't Intel GPUs the most dx12 compliant right now?
They may support more features than GCN or pascal do (and overlooking the potato nature of igpu vs discrete card) but Intel's architecture still has its own nuances that might not match up well against what AMD and/or Nvidia are doing.
>They don't need drivers for dx11 anymore, they can skip that since pretty much everything new is dx12
For gaming on pc breaking backwards compatability is heresy. I'm not well versed enough in Intel's ipgu to say anything gospel but DX11 support is still very, very important and IHV's still need their drivers to fix game dev fuckups.
The point of DX12/vulkan is to reduce the need for the likes of AMD and Nvidia stepping in and hand tuning shit to make it work.
Intel makes most of its money from mobile
Why the fuck would you make a desktop exclusive when your own like 90% of the laptop market if not more
They can make mobile versions you plug into mpcie?
Why would they instead of just doing more soc chips?
They laid off people because portable CPUns didn't sell.
Recent Atoms are good
>Intel making GPU
No.
Crystalwell was a complete disaster.
>With dx12 and vulkan multiadapter improving on all the flaws multigpu had before
There are 2 multi-GPU DX12 games I know of: AotS and RotTR. AotS scaling isn't great at all and and RotTR runs worse than DX11 SLI and uses more RAM. The current release of Vulkan doesn't actually support multi-GPU. RotTR not only has straight-up lower FPS compared to SLI but the frame time also seems fucked up, not to mention that when I tested it on my own system one of the cards only hit like 80% load.
So no, theory and future development aside, DX12 and Vulkan at this very moment do not improve multi-GPU in any way.
Vulkan doesn't support multiadapter right now. I guess tha's an improvement.
MS is trying but failing to make useful. Almost no DX12 game supports multi adapter
Intel would never waste fab time on shit-tier graphics chips that wouldn't even compete with AMD's APU graphics. Even if Intel intended to do something so monumentally stupid, they'd have to price the cards at 120USD minimum to justify the fab time against CPU chips they can sell at $300+.
I don't expect software multiadapter to make sense prior to the VR market becoming much larger. It's simpler to get improvements when you have two separate images to display.
and yet only AMD pushes for multiadapter in VR, nVidia is more comfortable with a single GPU handlign all the work and their software tech, like simultaneous Multi-Projection, handling the hard part.
When VR gets to be big, it's a pretty safe bet that dual GPU solutions will become far more common since it's much cheaper to produce two smaller chips than a single large die.
and then do 10 time more work trying to adjust latency.