The Sun Ultra 5, introduced in February 1998, was Sun’s answer to the growing demand for affordable 64-bit workstations. It marked a significant departure from previous Sun designs, adopting industry-standard components like IDE hard drives and PCI expansion slots while maintaining the UltraSPARC architecture that made Sun workstations so capable.
History
The Ultra 5 represented a strategic shift for Sun Microsystems. By the late 1990s, the workstation market was facing increasing pressure from high-end PCs running Windows NT and Linux. Sun responded by creating a workstation that could compete on price while offering the reliability and performance of Solaris and the SPARC architecture.
The machine used the UltraSPARC IIi processor, which integrated the CPU, cache controller, and memory controller onto a single chip. This integration reduced costs and simplified the system design. Clock speeds ranged from 233 MHz to 400 MHz depending on the configuration. This box is the 400 MHz variety.
Perhaps the most controversial aspect of the Ultra 5 was its use of IDE hard drives instead of the SCSI drives that had been standard in Sun workstations. While this reduced costs significantly, it also meant lower disk performance compared to higher-end models. The PCI bus replaced SBus, making it easier to use commodity graphics cards and network adapters.
The Ultra 5 found its niche in educational institutions, software development shops, and as a low-cost Solaris development platform. It was often paired with its sibling, the Ultra 10, which offered more expansion options and better performance for users who needed more than the entry-level configuration could provide.
Specifications
| Attribute | Value |
|---|---|
| Release Date | February 1998 |
| Model | Ultra 5 |
| CPU | UltraSPARC IIi @ 400 MHz |
| Architecture | Sun4u |
| RAM | TBD |
| OS | Solaris 9 |
| Graphics | TBD |
| Environment | CDE |
| Disk | Spinning IDE (boot) + ZuluSCSI (user data) |
Timekeeper Replacement
Like all Sun workstations, the Ultra 5 uses a timekeeper chip that holds the machine’s host ID, Ethernet address, and real-time clock backed by an internal battery. After a decade or so the battery dies and the machine loses its identity on every power cycle.
The Ultra 5 uses the MK48T59 timekeeper. Unlike many of the older Sun timekeeper models that have gone out of production and require surgery with a Dremel tool to attach a new coin cell, the MK48T59 was still available as a new part. I purchased a replacement, installed it, and programmed the host ID and Ethernet address through the OpenBoot PROM.
For details on the IDPROM programming procedure, see Resetting the IDPROM on Sun Workstations.
Graphics
The Ultra 5 ships with an onboard PGX24 framebuffer based on the ATI Mach64. It is notoriously slow, even by late-1990s standards, limited to basic 2D acceleration with poor performance. Upgrading to a better card like the PGX64 or a Creator3D is possible, but you run into the same FCode ROM problem that plagues the PCI SCSI cards. Only Sun-branded versions of these graphics cards carry the OpenBoot-compatible firmware needed for the system to recognize them. Generic PC versions of the same ATI or 3DLabs silicon will not work.The Flash Storage Problem
My goal with every machine in this collection is to avoid spinning rust wherever possible. Flash storage is silent, reliable, and easier to back up. For most systems that is a straightforward swap. The Ultra 5 turned it into an odyssey.
The problem starts with the on-board CMD646 IDE controller. It is what Sun chose when they decided to cut costs and go commodity, and it has been making restorers miserable ever since.
CompactFlash Attempts
My first attempt was a standard CF card in an IDE adapter. That failed immediately. Consumer CF cards assert the removable media bit in their IDENTIFY response, and Solaris sees that and refuses to treat the device as a fixed disk. No fixed disk, no VTOC disk label, no boot.
So I went looking for industrial CF cards, the
kind that suppress the removable bit by design.
Found some, tried them. Got further, but still
hit a wall. The CMD646 and Solaris have a geometry
negotiation problem. Even with the right card,
format would either refuse to write a label or
write one that did not survive a reboot. I was
close enough to taste it and still could not get
there.
Disk On Module
DOMs, Disk On Module, the 44-pin IDE flash modules, were the next attempt. They identify as fixed disks by design, so you do not get the removable bit problem. But DOM technology is old now, the good ones are hard to find, and used units are expensive. Trading one headache for another.The PCI SCSI Card Dead End
At some point I figured if I could not beat the IDE controller, I would go around it. The Ultra 5 has PCI slots, so I hunted for a SCSI card I could boot from. This is where I learned an expensive lesson.
Sun’s OpenBoot firmware can only boot from cards with OpenBoot-compatible FCode option ROMs. The generic market versions of those cards, same silicon, same everything, have x86-only option ROMs that OBP simply cannot interpret. From the firmware’s perspective the card does not exist as a boot device. I accidentally bought the non-Sun version first. That was a fail.
The Sun-branded cards with proper FCode ROMs exist, but tracking down the right Sun part number and finding one at a sane price is its own adventure. I eventually decided that was a dead end too.
The ZuluSCSI Solution
The solution I landed on was the ZuluSCSI. The Ultra 5 has an on-board narrow SCSI controller on the same bus as the CD-ROM drive. OBP knows how to talk to that natively, no third-party firmware, no option ROM issues. I put a ZuluSCSI on that bus and use it for user data and home directories.
The machine still boots from a spinning IDE disk, but everything I actually care about lives on the ZuluSCSI image where it is easy to back up. It is a hybrid, but it works cleanly and reliably.
Conclusion
The end result is a fast machine. The 400 MHz UltraSPARC IIi has real horsepower and Solaris 9 runs beautifully on it. But the Ultra 5 is a frustrating system to work on. Its cost-cutting decisions created a machine that sits awkwardly between the commodity PC world and the proper Sun ecosystem without fully belonging to either.
The IDE controller in particular feels like a mistake Sun knew they were making. Everything else about this box is quintessentially Sun. The disk subsystem is an afterthought that restorers are still paying for twenty-five years later.