Skip to main content

max / everycycle

hardware: settle the CPU on merit, not price The EPYC 4565P and the Ryzen 9950X are the same silicon: 16c/32t, 4.30 base, 5.70 boost, 170 W, 64 MB L3, two DDR5-5600 channels, 28 Gen 5 lanes. AMD's own datasheet against the 9950X's spec, line for line. So there was never a performance argument, and the earlier "higher clocks" framing was wrong. It comes down to officially validated ECC UDIMM. Ryzen 9000 ECC works where a board wires it up and ASRock Rack does, but board-vendor support is not CPU-vendor validation, and this machine's whole job is being a trustworthy reference whose gate requires readable, correct ECC counters. The 9950X's only real edge is an iGPU the AST2600 already covers. Fallback recorded: it drops in with zero performance change if the niche part is hard to buy. Two SKUs considered and rejected in writing, since the datasheet raised them. The 4545P is the same 16 cores at 65 W, tempting for an always-on box, but a mostly-idle host spends its hours at idle where both parts draw the same — the big part only draws more while building and finishes sooner, so energy per build is a wash and wall-clock is better. The 4585PX has TBD clocks and X3D trades all-core clock for cache that compilation barely uses.
Author: Max Johnson <me@maxj.phd> · 2026-07-25 15:40 UTC
Signed with PGP, not checked
Commit: 87dc7c00b3fb39bec84e441d9bbe3b72d3162a88
Parent: 6fac3bf
1 file changed, +42 insertions, -1 deletion
M docs/hardware/bench-v1-bom.md +42 -1
@@ -35,7 +35,7 @@
35 35
36 36 | Component | Choice | Approx | Notes |
37 37 |---|---|---|---|
38 - | CPU | **AMD EPYC 4565P** (16c/32t "Zen 5", AM5) | ~$700–800 | 16 Zen 5 cores at high all-core clocks is competitive throughput-per-dollar for clean builds, container builds and the ISO's zstd pass. 2× DDR5-5600, 28 PCIe Gen 5 lanes, 170 W class. **Note the ceiling: EPYC 4005 stops at 16 cores**, so if build throughput ever binds, the answer is a new machine, not a new CPU. Acceptable for a declared stepping stone. Also live: a Ryzen 9950X is ~$150 cheaper with the same core count and higher clocks — the EPYC premium buys official ECC UDIMM validation and server firmware support, which is worth confirming against this board's Ryzen ECC story before paying it |
38 + | CPU | **AMD EPYC 4565P** (16c/32t "Zen 5", AM5) | ~$700–800 | 4.30 base / 5.70 boost, 64 MB L3, 170 W, 2× DDR5-5600, 28 PCIe Gen 5 lanes. Sized for throughput: clean builds, container builds, the ISO's zstd pass. See the CPU note below for why this and not a Ryzen 9950X. **Ceiling: EPYC 4005 stops at 16 cores**, so if throughput ever binds the answer is a new machine, not a new CPU — acceptable for a declared stepping stone |
39 39 | Board | **ASRock Rack B650D4U** (base variant) | ~$400–450 | mATX server board. **AST2600 BMC** with IPMI 2.0, iKVM, vMedia, dedicated management LAN. Base variant chosen over -2L2T/-2L2T/BCM specifically because **it is the only one with two M.2 slots** |
40 40 | RAM | 2× 48 GB DDR5 ECC UDIMM (96 GB) | ~$450 | 1 DPC runs 5600; 2 DPC drops to 3600. Two slots stay empty on purpose |
41 41 | Storage 1 | 2 TB NVMe on M2_1 (PCIe 5.0 x4) | ~$180 | **xfs** root. Not ZFS — see the note in `mm-v1-bom.md` |
@@ -49,6 +49,47 @@
49 49
50 50 Vendor manuals for the decided parts are in `_private/docs/hardware/bench-v1/`.
51 51
52 + ## CPU: why the EPYC and not a Ryzen 9950X
53 +
54 + Decided on merit rather than price (Max: "$150 is marginal, whichever is the better part").
55 + From AMD's EPYC 4005 datasheet, held locally, against the 9950X:
56 +
57 + | | Cores | Base | Boost | TDP | L3 | Channels | Gen 5 lanes |
58 + |---|---|---|---|---|---|---|---|
59 + | EPYC 4565P | 16/32 | 4.30 | 5.70 | 170 W | 64 MB | 2× DDR5-5600 | 28 |
60 + | Ryzen 9950X | 16/32 | 4.30 | 5.70 | 170 W | 64 MB | 2× DDR5-5600 | 28 |
61 +
62 + Same silicon, same numbers. **There is no performance argument either way**, so this comes
63 + down to everything else, and the deciding factor is **officially validated ECC UDIMM**.
64 +
65 + Ryzen 9000 ECC works on boards that wire it up, and ASRock Rack does — but board-vendor
66 + support is not CPU-vendor validation. This machine exists to be a trustworthy reference; its
67 + characterization gate below requires ECC counters that are both correct and readable; and a
68 + silent memory error corrupting a build artifact is exactly the failure ECC is bought to
69 + prevent. Validated beats de-facto for that job.
70 +
71 + Secondary, both real: the embedded/server line carries a longer supply and firmware-support
72 + window, and ASRock Rack's BIOS and QVL work on this board targets EPYC 4004/4005 first. And
73 + since this file doubles as a TM product template, "server CPU with validated ECC on a server
74 + board with a BMC" is coherent to sell under a Care Policy in a way that a consumer part with
75 + unofficial ECC is not.
76 +
77 + The 9950X's only genuine edge is its iGPU, which the AST2600's own VGA already covers.
78 + **Fallback:** the EPYC 4005 is a niche part, and if availability is bad the 9950X drops into
79 + this board with zero performance change. Preference with a documented alternative, not a hard
80 + requirement.
81 +
82 + **Considered and rejected: the 4545P.** Also 16c/32t Zen 5, but 65 W instead of 170 W
83 + (3.0 base / 5.4 boost) and cheaper — tempting for an always-on box in a living space. Rejected
84 + because a mostly-idle CI host spends most of its hours at **idle**, where both parts draw
85 + about the same. The 170 W part only draws more while actually building, and it finishes
86 + sooner, so energy per build is roughly a wash while wall-clock is materially better. If heat
87 + or fan noise later becomes the actual complaint, the 4545P is the drop-in answer.
88 +
89 + **Considered and rejected: the 4585PX** (16c, 128 MB via 3D V-Cache). Clocks are listed TBD in
90 + the datasheet, so it may not be shipping; compilation gains from doubled L3 are modest, and
91 + X3D parts usually trade all-core clock for the cache.
92 +
52 93 ## Enclosure: open frame, and airflow solved at the card
53 94
54 95 The earlier draft of this file argued for a 4U rackmount because a passive card "only cools