The Problem:
I have an old AMD Opteron 2000 system that is experiencing hyper-transport sync flood errors. I can't even install an OS without the system crashing/rebooting. At one point, the system crashed with sync flood errors while I was updating settings in the BIOS. I've tried installing Windows 10 2004 and Ubuntu 20.04.1 LTS with the same results: reboot + sync flood errors. I am hoping that someone will tell me what to do to fix these errors. I would rather not spend any money on replacement parts for such an old system.
The System:
Under normal circumstances, I would strip out the mobo, CPUs, and RAM, and take them all to Best Buy to be recycled. The gutted chassis + power supplies would be shoved into my rack for any future projects. However, I have promised this system to a neighbor who is stuck at home, having had two strokes, and is at a huge risk for a third which would certainly put him into the ground. This donation of equipment, my time, and my training to him is my attempt at helping his life to suck just a little bit less. We will try to get Proxmox or Windows 10 + VirtualBox installed so he can experiment with whatever catches his attention.
This is not a ploy to get you to donate parts or money to me to give to him. I won't take parts or money. All I want are repair ideas before I officially pronounce this system dead.
I have an old AMD Opteron 2000 system that is experiencing hyper-transport sync flood errors. I can't even install an OS without the system crashing/rebooting. At one point, the system crashed with sync flood errors while I was updating settings in the BIOS. I've tried installing Windows 10 2004 and Ubuntu 20.04.1 LTS with the same results: reboot + sync flood errors. I am hoping that someone will tell me what to do to fix these errors. I would rather not spend any money on replacement parts for such an old system.
The System:
- Dual AMD Opteron 2427
- 128 GB (8GB x12 Kingston 16GB 667MHz ECC Registered DDR2 PC2-5300 KTH-XW9400K2/16G)
- SR5690 chipset
- Supermicro H8DI3+-F motherboard
- Passive heat sinks
- 2U Supermicro chassis
- Redundant Power Supplies
- 2x SATA Intel SSD DC S3520
- 8x SATA 1TB HDD
- No add-in cards
- Verified BIOS at newest revision: v1.0b
- BIOS reset to factory defaults (optimal) while inside BIOS menu
- BIOS reset to factory defaults (fail-safe) while inside BIOS menu
- BIOS reset using the clear-CMOS procedure
- Exchanged memory for different modules
- Moved memory around to different slots
- Used one DIMM on CPU #1 and one DIMM on CPU #2
- Used only one DIMM on CPU #1 and zero DIMMs on CPU #2
- Used zero DIMMs on CPU #1 and only one DIMM on CPU #2
- Replaced BIOS battery
- Replaced heat sink paste on both CPUs
- Replaced heat sink paste on SR5690 chip
- Re-seated both CPUs
- Looked for bent CPU socket pins (none found)
- Looked for dirty/fried CPU chip pads (none found)
- Reduced hyper-transport link width from 16 bits to 4
- Disabled non-essential system peripherals via BIOS
- Swapped out Supermicro power supplies for other Supermicro power supplies
- Tried using only the power supply in top slot
- Tried using only the power supply in bottom slot
- Tried using a known good spare power distribution board
- Re-seated 24-pin power connector
- Re-seated dual EPS 12V power connectors
Under normal circumstances, I would strip out the mobo, CPUs, and RAM, and take them all to Best Buy to be recycled. The gutted chassis + power supplies would be shoved into my rack for any future projects. However, I have promised this system to a neighbor who is stuck at home, having had two strokes, and is at a huge risk for a third which would certainly put him into the ground. This donation of equipment, my time, and my training to him is my attempt at helping his life to suck just a little bit less. We will try to get Proxmox or Windows 10 + VirtualBox installed so he can experiment with whatever catches his attention.
This is not a ploy to get you to donate parts or money to me to give to him. I won't take parts or money. All I want are repair ideas before I officially pronounce this system dead.
Last edited: