[Discussion] Backplane

After much research I’d like to compile what I believe is the current data on the HL15 and its specs. Starting with the backplane.

  1. Data

According to a 2021-06-23 post. 45Drives moved from a previous configuration of 3x 5-Bay backplanes (per row) to a single 15-Bay backplane. In the aforementioned link they mention their reasons which include speed.

  • Data for 5 drives share 1 SATA cable (3Gb in the original design), limits performance to 60MB per second per drive (mechanical drive limit approximately 150MB/s)

    So, direct wired became our thing.

My own “napkin math” suggest a 3Gb/s ÷ 5 drives = 600Mbps ÷ 8B/b drives = 75MB/s per drive instead of the 60MB/s quoted above.

A modernized SAS/SATA 6Gbps “refresh” of the old design could* now limit performance to 150MB/s per drive.

A modernized SAS 12Gbps “refresh” of the old design could* now limit performance to 300MB/s per drive.

Notably, this “modernized refreshed” backplane could* be fed by a single SFF port since they would use 3/4 lanes. Many motherboards have a SlimSAS or equivalent port, 8i cards are very abundant, economic, and consume less power. This could potentially be a light version of the HL15 in the future (?).

(Modern mechanical hard drive limits are commonly reported as “up to 250MB/s”)

Feel free to discuss below the differences between hard drive technologies, discrepancies between advertisements and real life speeds, sustained vs peak, etc

If you take a closer look at the above picture you can see the where the data and power connections are made. From right to left, between the 2nd and 3rd slot, between the 6th and 7th slot, , between the 10th and 11th slot, and between the 14th and 15th slot. Labeled “0”, “1”, “2”, and “3” for a total of 4.

Similarly, each “group” of 4 ports can be seen with a white line printed on the PCB as a visual divider from the next group. Except the last group which is a 3 port group presumably because either 16 drives would not fit in a standard network rack or the added drive would create issued with ventilation, cooling, power, or any other combination of reasons.

  1. Power

Here you can see a close-up of the ports when seen from below.
Taking the image from this post.

We can see all 4 of the connectors, referenced simply as “Molex power connectors”, daisy chained one after another and seem to come from a single source cable, closer to the #0 side and consequently the PSU side of the case.

These “Molex power connectors” seems to be commonly fed by a single 6-pin power port (unsure if this port is labeled as “Peripheral/PCIe” or something else in PSUs).

It is as-of-yet uncertain if the HL15 will feed power to the backplane via a single cable or multiple. Could be 1 cable, could be 4. Stay tuned.

It has been now confirmed the HL15 needs 4 independent connections from the PSU to the “power distribution daughter board”. I will create a separate topic for this PDDB component later.

“There is a separate power management board for powering the drives, which accepts 4x 4-pin Molex connections.”


The required power for the backplane can’t be found, yet, but published guides for other products (AV15) seem to suggest a 300W “Maximum”, 200W “Standing”.

  1. Connectors

Having discovered 45Drive’s “direct wire” implementation. It seems unlikely that connecting a single cable to the backplane would operate ports other than those in the group (up to 4). So if you were to connect only the #0 conector, you could ONLY Access drive bays 0–3.

With that being said…

“There are four MiniSASHD (SFF 8643) data connectors on the backplane.”

Some motherboards have SATA 7pin connectors, others may have SlimSAS or any 3-4 different variations of these connectors. make sure you inspect your MB and its manual to convert your available ports into the four 8643 necessary to drive all your bays.
SAS/ SATA Cables Guide – SFF-8087, 8088, 8470, 8482, 8484, and single device connectors.

Keep in mind that if you use certain cables, for example, 4xSATA3 from the motherboard to 1xSFF-8643 in the backplane you will need “reverse cables”.

Gbps x1 lane x4 lanes x8 lanes x16 lanes
SATA3/SAS2 6 24 48 96
SAS3 12 48 96 192
PCIe 3.0 8 32 64 128
PCIe 4.0 16 64 128 256
PCIe 5.0 32 128 256 512
DMI 3.0 8 32 64 128
DMI 4.0 16 64 128 256
GB/s x1 lane x4 lanes x8 lanes x16 lanes
SATA3/SAS2 0.75 3 6 12
SAS3 1.5 6 12 24
PCIe 3.0 1 4 8 16
PCIe 4.0 2 8 16 32
PCIe 5.0 4 16 32 64
DMI 3.0 1 4 8 16
DMI 4.0 2 8 16 32

It is definitely possible to use an HBA card or a SAS expander to populate all four 8643 ports of the backplane.

Possible setups:
4i HBA with SAS expander:
1(port)x4(lanes)x6Gbps ÷ 15 drives = 1.6Gbps ÷ 8B/b = 200MB/s per drive
1(port)x4(lanes)x12Gbps ÷ 15 drives = 3.2Gbps ÷ 8B/b = 400MB/s per drive

8i HBA with SAS expander:
2(ports)x4(lanes)x6Gbps ÷ 15 drives = 3.2Gbps ÷ 8B/b = 400MB/s per drive
2(port)x4(lanes)x12Gbps ÷ 15 drives = 6.4Gbps ÷ 8B/b = 800MB/s per drive

16i HBA
4(ports)x4(lanes)x6Gbps ÷ 15 drives = 6.4Gbps ÷ 8B/b = 800MB/s per drive
4(port)x4(lanes)x12Gbps ÷ 15 drives = 12.8Gbps ÷ 8B/b = 1.6GB/s per drive

These numbers assume you have the chipset, PCIe, HBA, SAS expander, cables, backplane and drive to NOT bottleneck connections.*


All of the calculations made above assume both a completely populated chassis and that all drives are being accessed together as one, using drives in smaller groups may increase theoretical or actual performance.

Feel free to discuss in the comments your thoughts about this write-up. What was done right and wrong, and what can be improved.

I’m just a customer like most of you and want to hit the ground running once I get my HL15.


Hi @ZVeguillaCotto and welcome to the Forum. first off thank you for putting all of this info into 1 place.

I’m not going to cover everything you posted but I will go over it and confirm/correct some things

  1. Data

You are correct in saying the backplane had 4 sets or 4 drives connected with 1 Molex and 1 HBA cable for data. the reason we switched to this design was for the performance. our older designs where it had a multiplexer on the backplane which allowed 5 drives to connect to 1 HBA cable.

The newer design gives a direct connection from the HBA to the drive without this multiplexer which increases the performance as you mentioned

  1. Power

As you stated each backplane has 4 Molex power connectors. each connector powers 4 drives. In the HL15 this harness of Molex connectors feeds back to the power distribution daughter board by the PSU and is a larger 24-pin connector that the daughter board breaks out into 4 single Molex connectors from the PSU. It is not 1 cable strung together and instead 2 sets of 2 Molex connectors coming from the 24-pin connector. So the first set of 8 drives is powered from 1 run of wires and the second set of 7 is powered with a second run of wires all coming from the 24pin connector.

This backplane has been tested to withstand 1.07A on the 5V and 1.52A on the 12V for each drive

  1. Connectors
    Yes 1 HBA cable will connect 4 drives to the backplane. If you only connect cable 0 you will only get the first 4 drives to be connected.

Edits made.

Some corrections were done to the post, primarily a miscalculation on my part for SAS lanes, a single SFF-8643 port has 4 lanes, it’s NOT a single 6Gbps/12Gbps connection as originally stated.

Added bits and bytes table for easy reference when sourcing your HBA or SAS expander.

For a real life (theoretical bandwidth) calculation example:
I’ll be using ASUS Pro WS W680-ACE motherboard.
It has 2 x PCIe 5.0 x16 (x16/x0 or x8/x8) giving me a 64GB/s connection straight to the CPU.
It has 8 x DMI 4.0 giving me a 16GB/s throughput through the W680 chipset.
It has 2 x PCIe 3.0 x16 (x4 mode) giving me a 4GB/s connection on each trough the W680 chipset.

I’ll be using a 9300-16i which has a PCIe 3.0 x8 in a PCIe 3.0 x4 slot giving me a 4GB/s connection to the backplane trough the W680 chipset.
I’ll be using a 9300-16i which has a PCIe 3.0 x8 in a PCIe 5.0 x8 slot giving me a 8GB/s connection to the SSD cache (x8 870 EVOs) straight to the CPU.

4GB/s ÷ 15 drives ≈ 250MB/s per drive
8GB/s ÷ 8 drives ≈ 1GB/s per drive

If in the future I were to install faster drives I would need to move my backplane HBA to another slot like the PCIe 5.0 x8 to achieve PCIe 3.0 x8 speeds of 8GB/s or ≈500MB/s per drive. Even faster drives would need a different HBA altogether to achieve their max speeds.

Being as the backplane is 15 x SAS3 ports for 1.5GB/s each, we would need a ~24GB/s HBA with a PCIe 4.0 x16 or PCIe 5.0 x8 in a matching PCIe slot to completely saturate each drive.

After a lot of research I tried to make the basic concepts of the math involved easy to understand. Since there are so many possible configurations you need to understand your specific case and use the math in your own formula to calculate your own numbers.

Feel free to mention your own hardware and I’ll do my best to run the numbers for you. (hopefully after a couple of examples it will become even easier to understand).