board
OCuLink Riser for the Supermicro AIOM Slot
Public Unreviewedby John Lauer
Passive OCP NIC 3.0 SFF card that breaks a Supermicro AIOM slot's PCIe x16 into four OCuLink 4i ports (4 x4, Gen4): 6-layer, 85 ohm, length-matched, DRC-clean KiCad project with Gerbers, BOM and CPL.
← Commit history
Highlight v2: what the card is for, captions and chapters in the adom-video player
6 files changed
+275−3
clips/README.md+5−3@@ -2,11 +2,13 @@ The whole build, filmed off the screen while Claude worked in Adom Hydrogen, cut to 16x speed with a voice-over for each part. New clips land here as the build goes on. -## The highlight (draft, 4 min 53 s)+## The highlight (draft, 6 min 22 s) -<video controls preload="metadata" width="1080" poster="clips/highlight-poster.jpg" src="clips/highlight.mp4"></video>+::adom-video{src="/john/oculink-aiom-riser/render/clips/highlight.video.json"} -[Download the highlight](clips/highlight.mp4). The whole build in five minutes: the request, the spec research, the 85 ohm pairs and length tuning, the board landing live in KiCad, the field-solved current check, the export and the finish line.+What the card is for (the server, the AIOM slot, the drives and GPU docks it opens up), then the whole build in five minutes: the spec research, the 85 ohm pairs and length tuning, the board landing live in KiCad, the field-solved current check, the export and the finish line. Captions on (press C) and chapters on the timeline. [Download the highlight](clips/highlight.mp4).++Product photos: Supermicro, ServeTheHome, Delock, ChargerLAB (see the credits in the video). | Clip | Real time | What happens | |---|---|---|
clips/highlight-poster.jpg⋯ 1 unchanged line ⋯
clips/highlight.chapters.jsonadded+34@@ -0,0 +1,34 @@+[+ {+ "t": 0,+ "title": "The request"+ },+ {+ "t": 20,+ "title": "What this board does"+ },+ {+ "t": 108.97,+ "title": "Spec research and impedance"+ },+ {+ "t": 148.97,+ "title": "Routing the high-speed lanes"+ },+ {+ "t": 220.97,+ "title": "The board in KiCad"+ },+ {+ "t": 278.97,+ "title": "Current and thermal checks"+ },+ {+ "t": 313.97,+ "title": "Silkscreen and 3D"+ },+ {+ "t": 331.97,+ "title": "Export and the finish line"+ }+]
clips/highlight.mp4⋯ 1 unchanged line ⋯
clips/highlight.video.jsonadded+39@@ -0,0 +1,39 @@+{+ "video": "/john/oculink-aiom-riser/render/clips/highlight.mp4",+ "poster": "/john/oculink-aiom-riser/render/clips/highlight-poster.jpg",+ "chapters": [+ {+ "t": 0,+ "title": "The request"+ },+ {+ "t": 20,+ "title": "What this board does"+ },+ {+ "t": 108.97,+ "title": "Spec research and impedance"+ },+ {+ "t": 148.97,+ "title": "Routing the high-speed lanes"+ },+ {+ "t": 220.97,+ "title": "The board in KiCad"+ },+ {+ "t": 278.97,+ "title": "Current and thermal checks"+ },+ {+ "t": 313.97,+ "title": "Silkscreen and 3D"+ },+ {+ "t": 331.97,+ "title": "Export and the finish line"+ }+ ],+ "captions": "WEBVTT\n\n00:00:00.400 --> 00:00:05.096\nA YouTube viewer asked for an OCuLink\nriser for the AIOM slot on their\n\n00:00:05.096 --> 00:00:06.304\nSupermicro server.\n\n00:00:06.304 --> 00:00:09.256\nBefore we build it, here is what that\nmeans.\n\n00:00:20.500 --> 00:00:23.854\nThis is the kind of server the viewer has:\n\n00:00:23.854 --> 00:00:25.850\na Supermicro rack server.\n\n00:00:25.850 --> 00:00:32.318\nInside, besides the normal P C I e slots,\nnewer Supermicro boards have one or two\n\n00:00:32.318 --> 00:00:33.196\nAIOM slots.\n\n00:00:34.600 --> 00:00:37.976\nFrom the back, Supermicro labels them:\n\n00:00:37.976 --> 00:00:40.997\nP C I e by sixteen, AIOM, OCP 3.0.\n\n00:00:40.997 --> 00:00:47.128\nA full sixteen lane slot, the same\nbandwidth as a graphics card slot.\n\n00:00:48.520 --> 00:00:50.584\nBut only one kind of card fits:\n\n00:00:50.584 --> 00:00:52.914\nan AIOM network card like this one.\n\n00:00:52.914 --> 00:00:57.640\nIf you don't need another network card,\nthose sixteen lanes sit unused.\n\n00:00:59.040 --> 00:01:00.799\nThis board changes that.\n\n00:01:00.799 --> 00:01:06.662\nIt plugs into the AIOM slot in place of\nthe network card, and splits the sixteen\n\n00:01:06.662 --> 00:01:11.352\nlanes into four OCuLink ports, each a full\nP C I e by four link.\n\n00:01:12.760 --> 00:01:18.420\nRun an OCuLink cable from each port to a U\ndot two NVMe drive, and the server gains\n\n00:01:18.420 --> 00:01:23.808\nfour more fast drives, each running at\nfull Gen four by four speed, about eight\n\n00:01:23.808 --> 00:01:25.240\ngigabytes per second.\n\n00:01:26.640 --> 00:01:31.888\nOr cable a port to an OCuLink dock like\nthis one, and hang a graphics card or\n\n00:01:31.888 --> 00:01:35.568\nanother P C I e card off the server,\noutside the case.\n\n00:01:36.960 --> 00:01:38.269\nThere is one catch:\n\n00:01:38.269 --> 00:01:42.607\nthe server's BIOS has to support splitting\nthat slot four ways.\n\n00:01:42.607 --> 00:01:47.635\nWhen it does, an unused network slot\nbecomes four general purpose P C I e\n\n00:01:47.635 --> 00:01:48.048\nports.\n\n00:01:49.370 --> 00:01:50.427\nNow the build.\n\n00:01:50.427 --> 00:01:54.653\nWe typed that one request into Claude, in\nAdom Hydrogen.\n\n00:01:54.653 --> 00:02:00.314\nIt reads the OCP 3.0 spec and the OCuLink\nstandards, pinout tables and all.\n\n00:02:09.370 --> 00:02:14.713\nIt writes a small field solver and finds\ntrace widths for 85 ohm differential pairs\n\n00:02:14.713 --> 00:02:16.258\non a six layer stack-up.\n\n00:02:29.370 --> 00:02:34.171\nThirty six high speed pairs are routed\nfrom the gold fingers to four OCuLink\n\n00:02:34.171 --> 00:02:36.066\nconnectors, with no crossings.\n\n00:02:49.370 --> 00:02:54.283\nEvery lane in a port is tuned to arrive\nwithin a hundredth of a picosecond of its\n\n00:02:54.283 --> 00:02:54.890\nneighbors.\n\n00:03:04.370 --> 00:03:05.548\nThe design record:\n\n00:03:05.548 --> 00:03:10.850\nimpedance per layer, skew inside every\npair under two hundredths of a millimeter.\n\n00:03:24.370 --> 00:03:29.822\nOn the adom-aiflow timeline, the run moves\nthrough design rules:\n\n00:03:29.822 --> 00:03:32.122\nJLCPCB's six layer process.\n\n00:03:41.370 --> 00:03:43.801\nThen the real board opens in KiCad:\n\n00:03:43.801 --> 00:03:47.274\ngold fingers at the bottom, connectors at\nthe top.\n\n00:03:59.370 --> 00:04:04.818\nForty seven parts land live, one at a\ntime, each its own undo step.\n\n00:04:19.370 --> 00:04:24.268\nThen one hundred sixty nine traces land\nthe same way, every one checked against\n\n00:04:24.268 --> 00:04:25.322\nthe design rules.\n\n00:04:39.370 --> 00:04:42.141\nThe current check needs a real field\nsolve.\n\n00:04:42.141 --> 00:04:43.816\nClaude finds why it fails:\n\n00:04:43.816 --> 00:04:47.554\na tenth of a millimeter via ring the\nsolver could not see.\n\n00:04:59.370 --> 00:05:02.970\nWith a finer grid, current and thermal\nboth pass.\n\n00:05:14.370 --> 00:05:18.617\nClean silkscreen, and the finished board\nin KiCad's 3D viewer:\n\n00:05:18.617 --> 00:05:20.946\nevery test point and port labeled.\n\n00:05:32.370 --> 00:05:37.329\nGerbers for all six copper layers, drill\nfiles, pick and place and the bill of\n\n00:05:37.329 --> 00:05:40.890\nmaterials, each read back and checked\nagainst the board.\n\n00:05:57.370 --> 00:05:58.413\nFinish passes:\n\n00:05:58.413 --> 00:06:01.543\nzero design rule errors, zero unconnected.\n\n00:06:01.543 --> 00:06:07.280\nA manufacturable PCI Express Gen four\ncard, from a one line comment, designed\n\n00:06:07.280 --> 00:06:10.186\nand checked by Claude in Adom Hydrogen.\n\n"+}
clips/highlight.vttadded+197@@ -0,0 +1,197 @@+WEBVTT++00:00:00.400 --> 00:00:05.096+A YouTube viewer asked for an OCuLink+riser for the AIOM slot on their++00:00:05.096 --> 00:00:06.304+Supermicro server.++00:00:06.304 --> 00:00:09.256+Before we build it, here is what that+means.++00:00:20.500 --> 00:00:23.854+This is the kind of server the viewer has:++00:00:23.854 --> 00:00:25.850+a Supermicro rack server.++00:00:25.850 --> 00:00:32.318+Inside, besides the normal P C I e slots,+newer Supermicro boards have one or two++00:00:32.318 --> 00:00:33.196+AIOM slots.++00:00:34.600 --> 00:00:37.976+From the back, Supermicro labels them:++00:00:37.976 --> 00:00:40.997+P C I e by sixteen, AIOM, OCP 3.0.++00:00:40.997 --> 00:00:47.128+A full sixteen lane slot, the same+bandwidth as a graphics card slot.++00:00:48.520 --> 00:00:50.584+But only one kind of card fits:++00:00:50.584 --> 00:00:52.914+an AIOM network card like this one.++00:00:52.914 --> 00:00:57.640+If you don't need another network card,+those sixteen lanes sit unused.++00:00:59.040 --> 00:01:00.799+This board changes that.++00:01:00.799 --> 00:01:06.662+It plugs into the AIOM slot in place of+the network card, and splits the sixteen++00:01:06.662 --> 00:01:11.352+lanes into four OCuLink ports, each a full+P C I e by four link.++00:01:12.760 --> 00:01:18.420+Run an OCuLink cable from each port to a U+dot two NVMe drive, and the server gains++00:01:18.420 --> 00:01:23.808+four more fast drives, each running at+full Gen four by four speed, about eight++00:01:23.808 --> 00:01:25.240+gigabytes per second.++00:01:26.640 --> 00:01:31.888+Or cable a port to an OCuLink dock like+this one, and hang a graphics card or++00:01:31.888 --> 00:01:35.568+another P C I e card off the server,+outside the case.++00:01:36.960 --> 00:01:38.269+There is one catch:++00:01:38.269 --> 00:01:42.607+the server's BIOS has to support splitting+that slot four ways.++00:01:42.607 --> 00:01:47.635+When it does, an unused network slot+becomes four general purpose P C I e++00:01:47.635 --> 00:01:48.048+ports.++00:01:49.370 --> 00:01:50.427+Now the build.++00:01:50.427 --> 00:01:54.653+We typed that one request into Claude, in+Adom Hydrogen.++00:01:54.653 --> 00:02:00.314+It reads the OCP 3.0 spec and the OCuLink+standards, pinout tables and all.++00:02:09.370 --> 00:02:14.713+It writes a small field solver and finds+trace widths for 85 ohm differential pairs++00:02:14.713 --> 00:02:16.258+on a six layer stack-up.++00:02:29.370 --> 00:02:34.171+Thirty six high speed pairs are routed+from the gold fingers to four OCuLink++00:02:34.171 --> 00:02:36.066+connectors, with no crossings.++00:02:49.370 --> 00:02:54.283+Every lane in a port is tuned to arrive+within a hundredth of a picosecond of its++00:02:54.283 --> 00:02:54.890+neighbors.++00:03:04.370 --> 00:03:05.548+The design record:++00:03:05.548 --> 00:03:10.850+impedance per layer, skew inside every+pair under two hundredths of a millimeter.++00:03:24.370 --> 00:03:29.822+On the adom-aiflow timeline, the run moves+through design rules:++00:03:29.822 --> 00:03:32.122+JLCPCB's six layer process.++00:03:41.370 --> 00:03:43.801+Then the real board opens in KiCad:++00:03:43.801 --> 00:03:47.274+gold fingers at the bottom, connectors at+the top.++00:03:59.370 --> 00:04:04.818+Forty seven parts land live, one at a+time, each its own undo step.++00:04:19.370 --> 00:04:24.268+Then one hundred sixty nine traces land+the same way, every one checked against++00:04:24.268 --> 00:04:25.322+the design rules.++00:04:39.370 --> 00:04:42.141+The current check needs a real field+solve.++00:04:42.141 --> 00:04:43.816+Claude finds why it fails:++00:04:43.816 --> 00:04:47.554+a tenth of a millimeter via ring the+solver could not see.++00:04:59.370 --> 00:05:02.970+With a finer grid, current and thermal+both pass.++00:05:14.370 --> 00:05:18.617+Clean silkscreen, and the finished board+in KiCad's 3D viewer:++00:05:18.617 --> 00:05:20.946+every test point and port labeled.++00:05:32.370 --> 00:05:37.329+Gerbers for all six copper layers, drill+files, pick and place and the bill of++00:05:37.329 --> 00:05:40.890+materials, each read back and checked+against the board.++00:05:57.370 --> 00:05:58.413+Finish passes:++00:05:58.413 --> 00:06:01.543+zero design rule errors, zero unconnected.++00:06:01.543 --> 00:06:07.280+A manufacturable PCI Express Gen four+card, from a one line comment, designed++00:06:07.280 --> 00:06:10.186+and checked by Claude in Adom Hydrogen.+