2 of 17

SERVER

OCP NIC Community Update�December 2022

��Sub-group Project wiki: https://www.opencompute.org/wiki/Server/Mezz�

Mailing List: https://ocp-all.groups.io/g/OCP-NIC

NIC3.0

3 of 17

Agenda

Thermal Test Fixture Update
SI Test Fixture Update
TSFF Thermal Test Fixture challenges
I3C updates
Rev1.21 specification feedback
SFF-TA-1034

4 of 17

TTF was designed to support speeds up to PCIe® 5.0 32GT/s

Uses very low-loss material

Cable assembly must be updated for 32GT/s signaling transmission

Received cable assemblies from two vendors
Lab experimentation requires NIC 3.0 PCIE Gen5 unit
NVIDIA samples received end of Nov
Dell arranging test with two cable assemblies

PCIe® 5.0 Thermal Test Fixture (TTF)

OCP NIC 3.0 TTF in use

Example cable assembly from Samtec

Jason

5 of 17

SI Test Fixtures

CLB 5.0 32GT/s status:

Protos arrived Sept 1^st

Dell: impedance mismatch; prior rev was correct
Tek: out of spec channel loss
Material analysis ongoing
Ordered material from different fabricator; samples to arrive end of December.

CBB 5.0 32GT/s status:

Based on 4.0 design; updates to align with PCI-SIG^®
Netlist reviewed; ECAD parts placed; routes tuned
Currently in Dell SI review
Will gerber with ‘known risk’ once review completes

OCP NIC 3.0 SFF CLB 5.0 prototype

Calibration

CLB board 1

Impedance Coupon

CLB board 2

Jason

OCP NIC 3.0 SFF CBB 5.0 block diagram

6 of 17

TSFF thermal test fixture challenge

Presentation by Anna of Nvidia

7 of 17

I3C: Out-of-band Interface Speed Comparison

No RMII

No I3C

No RMII

No USB2.0

No I3C

Mb/s

Do we see use-case with I3C?
Request feedback about I3C use-case & roadmap from 4 NIC vendors within our CLA group

8 of 17

I3C: NIC vendors feedback

I3C:

Some NIC vendor feedback did not see significant use-case of I3C.

Most NIC vendor think we will get to I3C sooner or later but slowly

Some NIC vendor see I3C as performance improvements over I2C but also cited challenges that I3C needs to over-come and cost adder

USB:

All NIC vendors mention USB as sideband

Replacement of NCSI as additional MCTP management interface

9 of 17

Revision3 Version1.21 OCP NIC clarifications

CLA members review feedback received

Table 42 note from “may” to “optional”
Section 3.12 as “informative reference”
Suggestion to remove IO current max/min from table 45
Question around RBT_Isolate don’t care & high state

10 of 17

SNIA new SFF-TA-1034 project

A project by Intel, supported by Cisco & HPE
New engagement to look for solution space around 32 PCIe lanes + NIC sidebands + High Power
Targeting higher power than “150W” LFF that OCP NIC 3.0 defined
Meeting at every Tuesday 11am-12pm Pacific
Suggest OCP NIC members that is also SNIA members to influence the direction of the project

12 of 17

Gather feedback throughout the year

Provide your feedback at:�https://tinyurl.com/feedback2ocpnic

13 of 17

Rev1.21 Minor spec clarifications (1/5)

Date	Items to fix in next release	Resolution	Date Resolved	Resolved in..
Apr-22	PCIE PRSNTB[3:0] descriptions should explicitly state what a '0' and '1' state of the PRSNTB pins are.	This should be taken care of with the 0b1111 definition clarification, below	9/27/2022	Version 1.2.1 draft
Apr-22	[3:0] = 0b1111 == no card installed. Pullups on host	Added additional text for 0b1111 condition in Section 3.4.2. This state is already discussed in Section 3.5	9/27/2022	Version 1.2.1 draft
May-22	Minor nit I found in the 1.2.0 spec (and maybe prior revs). Naming of Pin OCP_A14 is different between Table 21 and Table 26. It looks like both terms are used in different places. REF-TA-1012 uses RBT_CLK_IN so I would prefer to use that term but let me know if it should change.	Resolved. RBT_REF_CLK (and other permutations) have been changed to RBT_CLK_IN.	9/27/2022	Version 1.2.1 draft

14 of 17

Rev1.21 Minor spec clarifications(2/5)

Date	Items to fix in next release	Resolution	Date Resolved	Resolved in..
May-22	Add text for Activity LED that it ONLY blinks when link is up and there is activity present. Otherwise, this LED remains OFF. "* The Activity LED will not blink when the link state is down."	Resolved. Added text that the Activity LED shall blink only when link is up and there is activity	9/27/2022	Version 1.2.1 draft
9/27/2022	From E-mail: ��I found 6 "PRSNTB" being mistyped as "PRSTNB" while Reading "OCP NIC 3.0 Design Specification Version 1.2.0" (P. 131, 135, 136, 137, 138, 139).�I don't know where to report this typo problem. So email you directly.	Corrected "PRSTNB" typos	9/27/2022	Version 1.2.1 draft

15 of 17

Rev1.21 Minor spec clarifications(3/5)

Date	Items to fix in next release	Resolution	Date Resolved	Resolved in..
8/9/2022	Section 3.4.4 - Clarify that once RBT REF_CLK is enabled with the first rising edge of AUX_PWR_EN == 1 and NIC_PWR_GOOD ==1, it's probably better to use the latched version of NIC_PWR_GOOD (RBT_ISOLATE_N) to prevent the clock from getting disabled while MAIN_PWR_EN transition is occuring.	Reworked text in Section 3.4.4 as a coherant description for the RBT pins and RBT_CLK_IN isolation. Terminology with RBT_ISOLATE# has been clarified and a truth table has been included.	9/27/2022	Version 1.2.1 draft

No change to existing RBT_ISOLATE definition in timing diagram

16 of 17

Rev1.21 Minor spec clarifications (4/5)

Date	Items to fix in next release	Resolution	Date Resolved	Resolved in..
9/28/2022	What is the appropriate VIHmin value for each DC I/O?	Added additional entries to the DC I/O specification in Section 3.12. Broke out additional rows for RBT/PCIe/SMBUs/Scan Chain and OCP NIC 3.0 signals not covered by other specifications	9/28/2022	Version 1.2.1 draft

17 of 17

Rev1.21 Minor spec clarifications (5/5)

Date	Items to fix in next release	Resolution	Date Resolved	Resolved in..
10/19/2022	Check the RBT TSKEW >= or <= symbol… I think it's currently wrong	Tskew max timing inequality changed to TSKEW(Max) >= (equation). Max skew of board side (RHS) should be equal or less than Tskew(max) (LHS).	10/29/2022	Version 1.2.1 draft

1 of 17