Yup, I was one of them.For historical perspective. A whole bunch of us bought them from the same seller a couple years ago: https://forums.servethehome.com/index.php?threads/supermicro-x10srh-cf.28009/.
Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
The SRH is a SRi with a SAS controller, The SRL has better PCI-E connectivity.
| X10SRL |
|
| X10SRi X10SRH |
|
You already see by it being a rj-45 and not sfp+ that its not gone have any interesting ports for more than management tbh
All of my 10gig-e at home is RJ45. Even my other supermicro X10 boards, though those are Xeon-D with onboard copper 10gige.
I got my first 8-port for $680, and picked up a 12-port for $300 last year. Really not too bad. Netgear XS708T and 712T switches.
Thank you for the link and great info.
Since I would just be moving into v3/v4, I think I should wait for it to bottom since I haven't even fully utilized my ddr2 and ddr3 based servers yet. Yeah, I'd take an SFP switch any day over the rj45 version. Just use the copper as a pull cable for the fibre.
I agree, we had to pick up a 10g rj45 switch recently just because of the new Ryzen systems wanting to be able to do 10g without adding in a card to all the systems. Typically we use Arista 40g and break those out into 4x10g works so much better. Sadly those prices went up 2x over the last two years.
I could not find 32GB 2400 RDIMMS for less than $45, best I could find was $44.99. I got them last night and today the bay offered me 5% bucks
I couldn't find them either for less than that. But it comes and goes in spurts--someone in August got a 16GB module for $4 shipped so the deals are on there if you're persistent.
If you got the lowend XeonD boards they will probably be rj-45 yeah.
Lowend/small compute edge units in locations you generaly dont want fibre tends to be its usecase.
It's actually x8-in-x16 on the SRH, x16-in-x16 on the SRi.
It's bonkers. The SRH consumes eight lanes for the SAS3008 HBA and four lanes (at Gen2 speed) for the i350 NIC, so you only get 28 usable vs 40 on the SRL. I guess it makes a bit more sense on the four-port CLN4F variant. If you don't need those devices—and/or don't need the gigabit NIC attached directly to the CPU—you're really shooting yourself in the foot with the SRH.
It makes you realize how quickly PCIe lanes get used up, and how earth-shattering it was for EPYC to hit the market with 128 lanes from a single socket.
Last edited:
Lowest price is actually $45 for a 32GB RDIMM, there were lower prices on 32GB "kits", but why waste precious space with 8GB RDIMMs?
I think it depends on what you plan to do. There's a slightly different focus between the two boards. For the small set of folks for which the X10SRH-CF ticks all the boxes, it will probably cost less in the end and may be the better choice.
For maximum flexibility in component selection & upgrades, plus available CPU lanes assigned to slots, X10SRL is the clear choice.
But if you are going to need a RAID/HBA card anyway, and SAS3008 is acceptable, then you really aren't "losing" 8 lanes with the X10SRH-CF. The i350 NICs in the X10SRH-CF support SR-IOV, and some other server/virtualization oriented features, that would require an add-in card on the X10SRL because that board's onboard i210s don't.
With the original enterprise customers, the lower cost of a motherboard that includes an appropriate RAID/HBA & NICs (vs a Motherboard + add-in cards) was probably a selling point. I'd bet a lot of customers using X10SRH-CF for pretty basic server/virtualization roles never even used the PCIe slots, so the lower lane count wasn't an issue. I don't use the PCIe slots in mine.
Oh, don't get me wrong, I love the heck out of my SRH. Until very recently I had it decked out with a ton of NVMe drives, two SAS expanders, a 10GbE NIC, a Quadro, etc. and I was doing SR-IOV on both the GbE and 10GbE NICs. Now it's just a fileserver, no more virt, but I was able to move the 10GbE NIC to a CPU-attached slot and side-grade to a cheaper processor with fewer cores and higher clocks. I don't plan to replace it any time soon!
This is exactly what we are doing with the SRH.
If you are on SAS/SATA drives then there is a ton of life still left in the SRH boards for file server duties.
If, however, you are moving to an all NVME workload/storage, I don't think the e5 v3/v4 CPUs will cut it (and consequently the SRH). There is just not enough pcie lanes !
If you are on SAS/SATA drives then there is a ton of life still left in the SRH boards for file server duties.
If, however, you are moving to an all NVME workload/storage, I don't think the e5 v3/v4 CPUs will cut it (and consequently the SRH). There is just not enough pcie lanes !
Agreed. All my data vdevs for ZFS are SAS/SATA, but I've got five NVMe drives in the CPU-attached expansion slots: two for SLOG, two for the special vdev, and one for L2ARC. If you're willing to forgo a 10+GbE NIC you can squeeze in two more, but an NVMe-augmented zpool sort of necessitates high-speed networking and vice-versa.
There are two PCH-attached expansion slots left, but they're fairly anemic (Gen2 x4-in-x8 and Gen2 x2-in-x4) and should be avoided for anything important.
An x4 PCI-E 2 should be enough for a single port 10GbE port card, no?