Could the 1060 ever have bridgeless SLI support? AMD cards do

>>57885792
I think dx12 can do it because ashes of singularities did it

>>57885823
If I'm not mistaken that wasn't SLI you're referring to. MDA and LDA E

http://www.gamersnexus.net/guides/2519-gtx-1060-sli-benchmark-in-ashes-multi-gpu

>>57885792

bridgeless sli is a thing and you can already do it with any nvidia card, but it has the same problems that AMD's xdma meme has and that is crazy stuttering caused by high frametime variance

>>57885861
Crossfire is different than MDA, that's the extent of my knowledge. The Ashes of Singularity was not SLI. Look at the article I posted, they even say it.

Crossfire is more stable than MDA or LDAE, so whatever Nvidia comes out with (given they do) would hopefully be more stable than Crossfire, which is currently the best bridgeless solution.

>>57885857
Yeah that's what i was referring but a 1060 won't have sli because 2x1060=1080, and nvidia don't want this to happen

>>57885908
Right, I know that is what's stopping them, but I just want to know if it's possible for them to implement bridgeless SLI support with just a driver update.

What's even the benefit of it lmao if bridges work why change it

>>57885923
If AMD did it, nvidia can also do the same thing, but they would released this new tweak but the release of next gpu. nvidia doesn't give a lot gifts.

>>57885953
I'm not saying bridges are bad. Have you ever seen a GTX 1060? They don't have SLI support. I was asking if it is possible to add SLI support without a physical update (drivers).

>>57885923

it already is implemented... bridgeless sli has been a feature for literally 12 years when nvidia released the 6000 series, and its been used by OEMs that ship those super low end cards for a while too. though knowing nvidia they've probably disabled intentionally for cards that 'dont support' sli like the 750ti and 1060.

>>57886001
http://www.gamersnexus.net/guides/2519-gtx-1060-sli-benchmark-in-ashes-multi-gpuhttp://www.gamersnexus.net/guides/2519-gtx-1060-sli-benchmark-in-ashes-multi-gpuhttp://www.gamersnexus.net/guides/2519-gtx-1060-sli-benchmark-in-ashes-multi-gpuhttp://www.gamersnexus.net/guides/2519-gtx-1060-sli-benchmark-in-ashes-multi-gpu

fuck im so sick of

>>57886010
people "knowing" things.

dont know why my posts keep getting cut off.

>>57885792
>Could the 1060 ever have bridgeless SLI support?
Almost certainly not. DX12/Vulkan explicit multi-GPU rendering will become a bigger thing in the future though hopefully.

>AMD cards do not require a connection, so would it be possible for the 1060 to be run in SLI with driver updates?
That's because GCN cards have beefed up programmable DMA engines that can arbitrarily push data to other cards or pull data into their own buffers.

>Would mobos need updates as well if this is possible?
No, PCIe is an inherently switched fabric, and it's just a matter of having peripherals flexible enough to handle peer-to-peer (instead of just device-to-host) messaging.

>>57885861
CFX's XDMA jitter (now actually on par with SLI) had nothing to do with the DMA engine or PCIe bandwidth, which absolutely dwarfs that of SLI or old bridged crossfire. It was another case of shitty AMD drivers not pacing frames effectively at the software level.

>>57885953
>What's even the benefit of it lmao if bridges work why change it
There's literally not enough bandwidth available for higher resolutions and frame rates, even with the double wide 2-way connectors.
Nvidia disallowed 3-way SLI on 1080s because there's simply not enough bandwidth to push e.g., UHD@>60Hz between three cards.
IIRC each link is 1 GB/s unidirectional vs. 16 GB/s bidirectional for x16 PCIe 3.0.

Get a GTX 1080 and call it good. SLI is a meme and modern game developers still don't support it as well as they should. I ran a crossfire 970 setup, and my brother had a crossfire 280x setup. They are more hassle than they were worth.

>>57886125
>CFX's XDMA jitter (now actually on par with SLI) had nothing to do with the DMA engine or PCIe bandwidth,

i never claimed it was, because i know that the bad frametimes are caused by latency over the pcie bus, which the sli bridge is designed to workaround.

>>57885792
It already does. What the fuck are you talking about?

>>57886391
http://www.gamersnexus.net/guides/2519-gtx-1060-sli-benchmark-in-ashes-multi-gpu
http://www.gamersnexus.net/guides/2519-gtx-1060-sli-benchmark-in-ashes-multi-gpu
http://www.gamersnexus.net/guides/2519-gtx-1060-sli-benchmark-in-ashes-multi-gpu

fuck off god dammit

>>57886217
Bridges were designed ages ago to not choke old narrower PCI/PCIe buses, and peer-to-peer PCIe latency is on the order of a microsecond or less.

CF's old jitter issues were on the multi-ms timeframe regardless of whether vsync was used, on both bridged CF and bridgeless CFX, so it had nothing to do with PCIe.

>>57886436
>Bridges were designed ages ago to not choke old narrower PCI/PCIe buses,

just because that was its original purpose doesn't mean it's the same today, pcie 2.0 and 3.0 gave enough bandwidth that it doesnt really matter anymore

> and peer-to-peer PCIe latency is on the order of a microsecond or less.

yeah and it adds up when you need to send 100s of pcie packets between each card before a frame is ready to be presented

>CF's old jitter issues were on the multi-ms timeframe regardless of whether vsync was used, on both bridged CF and bridgeless CFX, so it had nothing to do with PCIe.

it's not an old issue, it was never 'fixed' because the problem is inherent to the hardware and amd's shoddy design.

>>57886436
>>57886362
>>57886217
You guys clearly know more than I do about this, so let me just get clarification.

Nvidia does not officially support two 1060's like they would two 970's or two 1080's. (SLI)

DX11, DX12. and Vulkan all allow the use of multiple GPU's in their own way. (MDA, etc.)

DX11 support is currently shit and DX12 and Vulkan are good and getting better.

Official Nvidia support is unlikely due to bandwidth issues? (Other than 2x1060 is op)

Is that generally right?

>>57886515
on what basis do you assert that PCIe CFX designs use store-and-forward frame buffer behavior (and that bridged designs are cut-through)?

(PCIe is a queued messaging protocol with strict receive buffer token tracking btw, so it's not like you need to wait for RTTs or something.)

if you can't provide sauce, everyone here will just have to assume you are an ignorant toddler.

>>57886531
To be honest I would be absolutely fine with it if it wasn't for obvious fact that they only did it to 1060 because there wouldn't be a point in buying a 1080, well at least on paper. SLI/Xfire was trash for years now with more and more titles not even supporting it or supporting it like a month after release. I know some of you are probably American and you think about this issue in the same way you think it's okay to burn American flag because it's a freedom of expression but in the uncucked part of the world putting money in xfire/sli is absolutely retarded.

>>57885792
it doesn't matter since most games have shitty sli support