DirectX 12 Will Allow Multi-GPU Between GeForce And Radeon

[P4-630]

http://www.tomshardware.com/news/microsoft-directx12-amd-nvidia,28606.html

 

[Ja.KooLit]

This is interesting.......

M$ maybe allow but I dont know about both camps particularly the green team will do some drivers lock. This has happen before when some gamers wants to use radeon as main and nvidia for dedicated physx.

Its good to know that the DX12 would want to "unite" both camps for consumers benefits

 

[Ebo]

I think both will lock their drivers from being able to do that. Otherwise some smart people out there can make a hack just like they have done before.

 

[xfia]

this is dreaming for sure and some of the comments are living the dream..

 

[CAPSLOCKSTUCK]

the fact it is on Toms gives it credence.



How well it works............ ?

 

[HumanSmoke]

So all that's required is for Nvidia and AMD to get on the same release schedule for their drivers to maintain continuity of experience. Does this mean the Nvidia guys only go into the office 2 days a week, or the Catalyst guys have to go on double shift?

I guess all that's needed to get the monthly driver install up to 1GB/month would be to include an Intel CPU (for Hybrid CrosSLI) on a board using a Lucid Logix Virtu setup. Troubleshooting level up!

the fact it is on Toms gives it credence.

Dunno about that. Tom's was a major cheerleader of Larrabee for over four years, forecasting the end of both Nvidia and AMD in the face of Intel's gaming prowess ( remember Tom's shilling for Project Offset?). Their pronouncements are as prone to hit or miss as many others.

 

[Easo]

I am betting on driver lock, most likely from the Green team.
Though it is kinda understandable, i foresee so much issues trying to make them work together.

 

[erixx]

at least let people use 2 different cards from the same brand!

 

[MustSeeMelons]

at least let people use 2 different cards from the same brand!

If I'm not mistaken, AMD does allow that.

 

[RCoon]

It won't happen. NVidia are too proud (no dedicated PhysX card when using an AMD card), and AMD are too slow (where's that Mantle SDK that was supposed to release publicly last year?).

 

[Daemmon]

Do you guys really think it will be driver locked? I mean, yeah it will be driver locked but not intentionally, the guys that write drivers for both camps are too bad to be able to actually pull this off.

 

[Ferrum Master]

Won't happen, due to arch difference and the micro latency due to different data execution in data pipes.

 

[qubit]

Sounds like a great idea.

I can just see nvidia wanting to play well with amd though...

 

[Fluffmeister]

It won't happen. NVidia are too proud (no dedicated PhysX card when using an AMD card), and AMD are too slow (where's that Mantle SDK that was supposed to release publicly last year?).

I wonder if Mantle is ever going to be open? Seems like it was just lip service to appease the open is awesome crowd.

 

[Easy Rhino]

Won't happen, due to arch difference and the micro latency due to different data execution in data pipes.

That's what software is for...

 

[Ferrum Master]

That's what software is for...

Upper layer languages does not count... not bare metal Assembly code and data execution. We are whining about micro stutter in between one ASIC as it doesn't properly chew in parallelism the data stream from the game engine thus it needs more cycles to process it...

And VLIW synchronization when the hardware has different pipe length? I don't believe. Even the L2 Cache size and speed will tear everything apart.

Simply there is not enough budget time, otherwise you will have to introduce lag caused by frame/scene buildup, and the controls and sound will delay, it may work for 24FPS not proper PC"ish frametimes.

 

[GhostRyder]

Yea I am going to say BS to this ever working well even if it does happen. Neither camp will support this almost at all and they might just lock it out period to save from issues down the line or alternative ways to get performance. Would love the idea honestly, but it seems like its a dream that is not going to happen or if it does its not going to work for very long.

If I'm not mistaken, AMD does allow that.

They allow multi-GPU configurations on any card in the same series of GPU's. For instance an R9 280X can CFX with a R9 280, an HD 7970 can CFX with an R9 280, R9 290X can CFX with R9 290, etc. The requirement is they have to be the same GPU even if one is cut down a bit but its something not many people do much as it is.

Sounds like a great idea.

I can just nvidia wanting to play well with amd though...

Yea, I think I would believe NVidia dropped out of the GPU market before they would work together with AMD (Mostly speaking about the ATI part) on something like this.

 

[Easy Rhino]

Upper layer languages does not count... not bare metal Assembly code and data execution. We are whining about micro stutter in between one ASIC as it doesn't properly chew in parallelism the data stream from the game engine thus it needs more cycles to process it...

And VLIW synchronization when the hardware has different pipe length? I don't believe. Even the L2 Cache size and speed will tear everything apart.

Simply there is not enough budget time, otherwise you will have to introduce lag caused by frame/scene buildup, and the controls and sound will delay, it may work for 24FPS not proper PC"ish frametimes.

hrm, it is almost as if an entire field of science is dedicated to solving the exact problems you describe...

 

[FordGT90Concept]

I think it is a function of close-to-the-metal programming that DX12 allows and the uniformity of compute unit architecture DX10 and newer has demanded. I don't think AMD nor NVIDIA would block this because it doesn't directly impact the sales of either. For example, if you have a GTX 970, you're not going to go out and buy an R9 290X to boost your performance. You're most likely going to buy another GTX 970. I'm sure the technology DirectX is using is similar to Intel's on higher end chipsets but software instead of hardware. It is long past time for NVIDIA to stop pretending like SLI is worth protecting.

 

[Disparia]

Not being hung up on the headline I read through the entire article. Good stuff. Nice quick adoption of DX12 too, with those free Windows 10 upgrades.

 

[Tallencor]

If I buy a card and I can continue to use my old card in my x8 slot for crossfire or added vram or added horses or whatever, I'm happy. saves me from giving away my card. Do I want Shadowplay and true audio to work together along with mantle and G-Sync. Nope. I want to play Witcher 3 @ 60fps with no bloody stutter. That's it.

 

[natr0n]

History always repeats.

 

[Ferrum Master]

The only thing that could work is if a game developer... It actually could be done even now via openCL...

Uses game effect data , such as FXAA(MSAA not possible) and some physics(HAVOK CPU does fine IMHO useless) or lighting reflection processing, fluids... that is actually blocked feature for nvidias department, then using the other card as processing cluster... but it ain't something new really... But again... accessing it to other card's driver kernel ain't a really fast solution, it still introduces lag(we can see it when enabling hybrid Physx), no matter how there will be two drivers that have their own capricious characters. The overall goodness in this is questionable really...

 

[phanbuey]

Yeah I think "technically possible" and "practical/beneficial" are two different things here... I doubt this type of solution will ever really make sense for most people.

 

[TRWOV]

If this lets me use Physx without hooping through loops I'll be a happy camper. W10 can't come out fast enough.