Tuesday, August 18, 2026
Matthew Fricke joined the channel
Dan Pitt joined the channel
David Ozog joined the channel
Subhadeep Bhattacharya joined the channel
Sayan Ghosh joined the channel
Ryan Scherbarth (nvidia) joined the channel
Ryan Scherbarth (nvidia) renamed the channel from "2026-d2-1300-keynote-the-future-of-ai-interconnects" to "2026-d2-1300-keynote-3-the-future-of-ai-interconnects"
Jiaqi Lou joined the channel
bsahu joined the channel
Fabian joined the channel
Francois Labonte joined the channel
Daud Arslan joined the channel
Shahbozbek Hakimov joined the channel
Tong Xu joined the channel
myoung-gyun.suh joined the channel
xi.chen joined the channel
Mahfuz joined the channel
Varun Chotalia joined the channel
pierre-louis.benard joined the channel
Advm joined the channel
ahmadrazarehman033 joined the channel
Thursday, August 20, 2026
L
What is the actual traffic sent over scale across? Can it be used for inference or KV cache transfer?
L
Isn't optical circuit switch more applicable to scale out rather than scale up?
C
What are the special or minimum requirements of the DSPs that handle these new optical interconnects? Any special filter or algorithm? Minimum buffer/memory that handle the traffic QoA
D
I am still confused about the difference between ACC and AEC.
2 replies
G
AEC adds more signal processing so u get longer reach.
H
when we use scale out interconnection to connect two data centers instead of using scale across
Y
On In-Module-Liquid, is OSFP MSA working for it? Or is it IPEC LCO?
G
What's do you use for the out-of-band management network if you need more than 10 Gbps?
1 reply
G
my email is <mailto:gwhelan@teradio.co|gwhelan@teradio.co> love to chat out next gen OOB mgmt nets..
A
How do you compare competiveness of players from Coherent-side with from IMDD-side?
R
I see all the focus regarding CPO is on XPUs (processors), maybe some on memory, but no focus on Storage. Is that something in future roadmap of discussions for the industry coalitions you mentioned?
H
what is the maximum number of racks in a scale up domain
1 reply
B
For this one, there is no universal number but essentially the way at least I look at it is that the boundary is define by whether the XPUs (Accelerators) can still behave as a tightly coupled unit.
A
Korean skhynix just announced CPO for memory
V
what are use cases for coherent to IMDD media converter?
1 reply
B
One use case is a boundary between scale across network and an IMDD based scale-up/scale out fabric
L
Which hyperscaler uses scale across for inference or KV cache transfer?
S
@Bilal Riaz [Ciena] thanks for the great talk!
1 reply
B
Thanks, much appreciated!
🙌 1