AMD has confirmed that Vega utilises their new Infinity Fabric tech

[WYP]

AMD confirms that Vega utilises Infinity Fabric, opening up a lot of possibilities for multi-die GPUs.



Read more on AMD's Vega GPUs and the potential of Infinity Fabric.

 

[dwatterworth]

This article hits on something I've always wondered.
Why present a dual GPU card as dual GPU's to the system vs having the card manage it's own information tasking.

Maybe we will see a future GPU review with a chip looking like TTL's 100 fan AIO review.

'Enthusiasts will need to plan ahead with this GPU measuring in at 1500mm'.

In all seriousness though, fantastic idea! If the interconnects worked super well, maybe we could see a 'budget' monster GPU with larger process node sizes with a bunch of less expensive cores?

That multi-module approach sure would be more interesting for load distribution / power savings with GPU 'core parking/gating'?.

 

[NeverBackDown]

This was not surprising whatsoever

 

[Avet]

They had good idea with developing infinity fabric. With creating large dies becoming a problem multi die single units are probably the direction in which will future tech go.

 

[Tolemac]

Why has Buzz Lightyear suddenly entered my thoughts

 

[AngryGoldfish]

Navi = scaleable.

 

[Tripp]

Navi = scaleable.

Makes a lot if sense now that, exciting things in the future for gpus at amd anyway

 

[FluxTape]

This article hits on something I've always wondered.
Why present a dual GPU card as dual GPU's to the system vs having the card manage it's own information tasking.

Because it is not easy. The inherent latency of the pcie bus can lead to inefficiencys if you aren't careful. That said vulkan and dx12 make exactly this possible, showing only one unified gpu to the developer.

 

[NeverBackDown]

Because it is not easy. The inherent latency of the pcie bus can lead to inefficiencys if you aren't careful. That said vulkan and dx12 make exactly this possible, showing only one unified gpu to the developer.

It's less to due with the PC bus whereas it's more along the lines of latency that's added due to a PLX chip in addition to software latency in trying to get both GPUs to work. It would also hurt that old APIs and engines were using one core to feed two GPUs. There wasn't a standard in place which makes it harder. Even with modern software it's still difficult and now it's the issue of just trying to get it to work well than anything. A hardware solution like this should really help.