Thanks for checking! I'm glad we may have found a real tray that fits but that cost is pretty nuts. So far the 3d printed caddies are working great but I agree that we should upgrade them with steel pins (where the hell would we find those?) or screws.
48 Drive 4U Chassis: IBM Slicestor 3448 aka Chenbro RM43348
- Thread starter llowrey
- Start date
Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
I was thinking along the lines of something like this:
Stainless Steel Select Ø2mm Ø3mm Ø4mm Ø5mm Ø6mm Ø8mm Dowel Pin Rod With Knurl | eBay
2.5x8mm stainless pin with knurled base.
Makes sense, the MCP-220-94601-0N is the same carrier that goes in the 90drive SuperMicro Chassis - SC946ED-R2KJBOD | 4U | Chassis | Products | Super Micro Computer, Inc. looking inside mine it looks nearly identical to the drive area that @llowrey had in the first post.
yea.. all u need to do is get a small metal rod and tac weld it or maby cold weld with something like jbweld or other strong epoxy to the case and u have a rod for the latch on those trays and they would fit perfectly... when i get the cpu and i can verify that the light rod does line up i am going to decide if i want to do that or buy a 3d printer...
Ordered a set so I'll let you know how it works out!
Man I'd be super curious if the rails for this chassis would work for us. I'm having a hell of a time locating rails!
Any update on this? I picked up one of the chassis and am also debating whether to buy these SuperMicro units or just get a 3D printer for half the price.
yea sorry, thought the post was dead... i purchased a single E5-2630LV4 and paired it with memory HMA42GR7MFR4N-TF Hynix 16GB PC4-17000 2133MHz ECC REG CL15 and it works just fine.
--also still debating on the caddys lol!!! i have not run out of space yet on my msa70 arrays so there is no rush on getting this thing into my rack.. 500 dollars on caddys alone or 200 dollars on a 3d printer and some work... /shrug.. i keep checking ebay.. hoping i can get a lot of those cheap dell plastic things for a real good deal..
This is about the best deal I have found for the Dell plastic caddies. You do get a hundred, plenty for spares just in case modding goes wrong.
Lot 100 Dell Optiplex Hard Drive Caddy YJ221 U6436 GX620 745 755 760 780 Tower | eBay
Lot 100 Dell Optiplex Hard Drive Caddy YJ221 U6436 GX620 745 755 760 780 Tower | eBay
I've actually picked up a Monoprice MP Select Mini Pro for less than the cost of any caddies, so I'll be putting together a complete replacement for the SuperMicro style trays now that I can start prototyping them. Looks like I'll have to print either simple side rails, or do a whole caddy/tray in two pieces, given how small my print volume is.
I'll report back, and provide a file for my results.
I'll report back, and provide a file for my results.
Update for whomever continues using these for builds in case you're bad at the wiring side like me courtesy of the post author.
The front panel connector is labeled "MBI Side Band Signal Connector" on page 26 of the Chenbro 43348 User Manual, take that wiring out and replace it with custom cables with the appropriate headers for your system and it's actually not bad to swap out. Also recommend popping out the 2 x 2.5" slot backplane and running U.2 Cables there for newer drives using Molex -> SATA power for the U.2 drives.
Can anyone recommend a source for actual rack rails? I ordered a set of Chenbro 4U rails off newegg but I'm not confident they'll be correctly sized.
Edit: sourcing on the parts: Those connectors are called "Crimp Connector Headers" and they're on a 2 x 6 .1" grid on the PDU. The wires easiest to use are "Wires with Pre-Crimped Terminals" and I got 24" wires. My motherboard had a 2 x 12 .1" grid that will also use a crimp connector of the same variety all from Pololu.
Looks straightforward to do after you map the pins across. On my X10DRi-T system I ran into a problem where no NVMe boot was supported, but used a 983DCT 960GB that was plug and play bootable to Windows 10/CentOS7 despite that when I set the slot the adapter was in to EFI mode, and the boot to UEFI. Waiting on a more expensive ($190 M.2 RAID Card from StarTech). If that works I should be able to get 2 x U.2 drives off one slot with one bootable, the other serving as BeeGFS metadata location.
The headers on the MBI Side Band Signal Connector are also in different pinouts on the X10DRi-T motherboard, not all the pins are in the "Front Panel" group, but most of the pins were available on the motherboard in other pinouts. On the MZ32-AR0, all the pins are in the "Front Panel" group.
The front panel connector is labeled "MBI Side Band Signal Connector" on page 26 of the Chenbro 43348 User Manual, take that wiring out and replace it with custom cables with the appropriate headers for your system and it's actually not bad to swap out. Also recommend popping out the 2 x 2.5" slot backplane and running U.2 Cables there for newer drives using Molex -> SATA power for the U.2 drives.
Can anyone recommend a source for actual rack rails? I ordered a set of Chenbro 4U rails off newegg but I'm not confident they'll be correctly sized.
Edit: sourcing on the parts: Those connectors are called "Crimp Connector Headers" and they're on a 2 x 6 .1" grid on the PDU. The wires easiest to use are "Wires with Pre-Crimped Terminals" and I got 24" wires. My motherboard had a 2 x 12 .1" grid that will also use a crimp connector of the same variety all from Pololu.
Looks straightforward to do after you map the pins across. On my X10DRi-T system I ran into a problem where no NVMe boot was supported, but used a 983DCT 960GB that was plug and play bootable to Windows 10/CentOS7 despite that when I set the slot the adapter was in to EFI mode, and the boot to UEFI. Waiting on a more expensive ($190 M.2 RAID Card from StarTech). If that works I should be able to get 2 x U.2 drives off one slot with one bootable, the other serving as BeeGFS metadata location.
The headers on the MBI Side Band Signal Connector are also in different pinouts on the X10DRi-T motherboard, not all the pins are in the "Front Panel" group, but most of the pins were available on the motherboard in other pinouts. On the MZ32-AR0, all the pins are in the "Front Panel" group.
Last edited:
Additional info for anyone building with this. The midplane fans respond only to HARD DRIVE temps, not to the motherboard. The proper alignment with the pins for the fan headers is to have the 2 blank pins aligned with the side of the PWM headers with the clip, and the final tach wire not in contact to the motherboard PWM header. If you swap the motherboard, I would recommend an active 2U heat sink at least for the back CPU in a dual socket config, as there's just a hair over 110mm clearance off the motherboard to the top of the case, I'm not planning to try a Noctua 3U active cooler. With HDDs it's entirely possible that the amount of airflow through the case would be sufficient with passive cooling, however I did have thermal throttling in testing with no HDDs. Still waiting on those caddies to finish printing.
Instead of rails, I've opted to try 1U star tech shelves capable of holding 200lbs. All in all, this seems like an excellent budget chassis. Replacement parts may be non-trivial to come by (Fans/PSUs) so keep that in mind if you're buying to be cheap for a small business/low budget operation like me.
Instead of rails, I've opted to try 1U star tech shelves capable of holding 200lbs. All in all, this seems like an excellent budget chassis. Replacement parts may be non-trivial to come by (Fans/PSUs) so keep that in mind if you're buying to be cheap for a small business/low budget operation like me.
Final Thoughts:
Custom cables worked great, as did the Epyc motherboard in the chassis. Ended up needing to clip and modify some PWM fan extension cables in order to get the in-chassis fans all the way back to where the PWM headers were in the case. 2U cooler comes tomorrow, and should be a great setup. Whole system is working as intended, and the 3D printed drive rails were easy to use. I ended up having to file the inserts down a bit, but that was easy. Can't wait to build the second one next week, and then set up the storage cluster.
Custom cables worked great, as did the Epyc motherboard in the chassis. Ended up needing to clip and modify some PWM fan extension cables in order to get the in-chassis fans all the way back to where the PWM headers were in the case. 2U cooler comes tomorrow, and should be a great setup. Whole system is working as intended, and the 3D printed drive rails were easy to use. I ended up having to file the inserts down a bit, but that was easy. Can't wait to build the second one next week, and then set up the storage cluster.
So were you able to quiet the fans down at all? Mine are all seated and running but are at 100% and you are right.... It moves a ton of air and is loud!
Has anyone swapped out the fans with say noctua fans or vardar fans?
@mciver I would like to know more about your setup and what you did with the fans in particular. Do these fans work off the MB headers? I read they draw 3.3 amps where most MB headers only do ~1.3 max. Would that cause any problems?
Has anyone swapped out the fans with say noctua fans or vardar fans?
@mciver I would like to know more about your setup and what you did with the fans in particular. Do these fans work off the MB headers? I read they draw 3.3 amps where most MB headers only do ~1.3 max. Would that cause any problems?
Last edited:
So the fans in the midplane are completely powered off the SAS3 backplanes. The 5-pin "PWM-like" headers are just putting out a tach reading. only 1 of them is in the proper alignment to be read, the second signal should be off the 4 pin headers and not in contact with a male pin on the motherboard. The fans are probably not at 100% if they are fully seated unless you have it stuffed with drives and it is sensing hot drive temps. I'd say to make sure you checked the pin out diagrams off the chassis and put them into the right headers on your motherboard according to your best guess. I think it's possible the drive temps are only read on the backplane or in UEFI mode, since my BIOS is in legacy mode due to the software stack I'm currently running I can't see the HDDs. If I recall properly, when I was in UEFI mode I could see the hard drives. I should be back in UEFI mode ~August or so. Just depends on when OpenHPC releases their CentOS 8 guide.
For the people that may get this you will have to do a continuity test to confirm yours but I found my wires to be on the MB-J42 cable connected to the MBI :
| PIN | COLOR | |
| 1 | ORANGE | FP LED STATUS R N |
| 2 | RED | FP PWR LED + |
| 3 | ORANGE | FP PWR LED - |
| 4 | BLACK | FP ID LED + |
| 5 | BLACK | FP ID LED - |
| 6 | BLUE | FP ID BTN - |
| 7 | BLACK | FP PWR BTN - |
| 8 | BLACK | FP RST BTN - |
| 9 | N/A | GND |
| 10 | N/A | N/C |
| 11 | N/A | SMB IPMI CLK |
| 12 | N/A | SMB IPMI DATA |
@mciver Where did you find the pin-out for the 5-pin "PWM like" fan connectors? Per your statement I will have to rewire mine, which is no problem, but I am not finding the pin-out for them. Or I am overlooking it which is probably the case. When I traced the cables back to the plug in the backplane they split and go to the Fan clock input pin 6 on the JF1-JF4. Where should this hookup on the MB? Following your input of one of them being correct it would either be the Ground or case fan pwr (being the opposite side of the PWM).
I apologize I don't follow what you mean when you said " the second signal should be off the 4 pin headers and not in contact with a male pin on the motherboard ". Currently there are no drives in the unit as I am trying to get everything running properly before I start populating it.Also I confirmed the fans are completely seated and running what seems to be 100%.
I apologize I don't follow what you mean when you said " the second signal should be off the 4 pin headers and not in contact with a male pin on the motherboard ". Currently there are no drives in the unit as I am trying to get everything running properly before I start populating it.Also I confirmed the fans are completely seated and running what seems to be 100%.
Last edited:
I think it was earlier in the thread, otherwise I just plugged it in and looked at what happened. It appears there is literally no use in plugging in the fans using the 5 pin cable. It seems to do absolutely nothing other than allow you to read the speed of the fan not set it.
It does not give input control on the fan speed. The right end of the cable aligns with the PWM right end as seen in that picture, and the second wire hangs off the edge of the PWM connector. I doubt they're running at 100% personally, unplug one and see if they get louder would be my recommendation.
It does not give input control on the fan speed. The right end of the cable aligns with the PWM right end as seen in that picture, and the second wire hangs off the edge of the PWM connector. I doubt they're running at 100% personally, unplug one and see if they get louder would be my recommendation.
Hi,
I just started putting one of these together, and I had an old Supermicro CSE-PTJBOD-CB2 that i was hoping to get working here so I could use it as an expansion chassis. However, from reading the chenbro manuals and the comments here there appears to only be a PWR Button N connector. I thought power buttons worked by just briefly completing the circuit to turn the machine on (which is why its position didn't really matter), but with only the one connector, i'm not sure how this is supposed to work here? Or if it even can work here. Sorry if this is a stupid question, but what am I missing about how this chassis works?
I just started putting one of these together, and I had an old Supermicro CSE-PTJBOD-CB2 that i was hoping to get working here so I could use it as an expansion chassis. However, from reading the chenbro manuals and the comments here there appears to only be a PWR Button N connector. I thought power buttons worked by just briefly completing the circuit to turn the machine on (which is why its position didn't really matter), but with only the one connector, i'm not sure how this is supposed to work here? Or if it even can work here. Sorry if this is a stupid question, but what am I missing about how this chassis works?