Sun Ray History
This is a personal perspective on my time developing the Sun Ray thin client during my years at Sun Microsystems.
Sun Ray is an Ultra Thin Client or Zero Client. Where as traditional thin clients typically run end-user applications and contain a small operating system (such as a Windows Based Terminal), a Zero Client actually runs no applications, has no operating system, and most importantly is is considered to be completely stateless. That means that when the power goes out you data is safe and secure back on the server where it belongs.
Sun Ray is an Ultra Thin Client or Zero Client. Where as traditional thin clients typically run end-user applications and contain a small operating system (such as a Windows Based Terminal), a Zero Client actually runs no applications, has no operating system, and most importantly is is considered to be completely stateless. That means that when the power goes out you data is safe and secure back on the server where it belongs.
NeWT - The very first Sun Ray
The concept began in Sun Microsystems Laboratories back in 1997 as a project codenamed NeWT or NetWorkTerminal. As the lead hardware engineer on the project it was my responsibility to develop the very first client. There was no preconception of what the client's architecture should be except to be low cost, low power, and silent.
Being in the research labs we weren't restricted to just using Sparc, quite the opposite. In fact some of the early processor candidates for NeWT included the Intel StrongARM, Philips Semiconductors (now NXP) TriMedia, and National Semiconductors Geode x86 chipsets. The StrongARM was of interest from the team because of its relatively high integration level and potential for low power operation and cost, but Intel's long term commitment to the ARM architecture at the time was in question by others in the industry since it had been part of their acquisition of Digital Equipment. TriMedia was evaluated, but a software simulator used to test its performance showed that a SparcStation 5 outperformed it with a 110MHz MicroSparc II, so it was quickly abandon. For all its multimedia acceleration features, it still lacked in the ability to keep up with a general purpose processor like the MicroSparc II. Ironically it was while working with the MicroSparc II I discovered that a new integrated version (MicroSparc IIep) had just gone into production and it was the perfect fit for just what we needed.
For graphics I turned to ATI who Sun already had an established and good business relationship with for our server products. The Rage 128 was ideal (later we used the Rage 128 PRO to support LCD displays), it was low power, had powerful 2D rendering performance, and relatively inexpensive compared to anything Sun had internally. It also outperformed most of its commodity PC competitors and ATI was willing to give us driver development support from their Mac team. Because the client had no requirement for 3D and all the 2D graphics rendering occurred on the server there was no need to use the best performing GPU. Years later this would be the subject of another labs project to build a 3D capable client, though the product never was put into production several of the technologies developed for it were later used in Sun Ray products I am told.
Sun at the time had a tendency to designing and using much of the chip technology it put in its systems. With few exceptions, commodity parts were not highly used. Mainly this was for performance issues. In a $50K server, part costs are far less of a concern than in a $399 appliance. But for our needs I knew we needed to leverage the economies of scale provided by using off-the-shelf PC components. With the exception of the processor and the Ethernet controller we used commodity devices for everything else. Now your probably thinking, why the Ethernet controller? Simple. While there were many commodity parts we could have used, this was one area where Sun's own Ethernet controller design was far superior in performance and quite cost competitive with commodity devices.
NeWT used an off-the-self power supply and mini-ITX PC case. The manufacturer offered us an ability to have custom bezel plastics made for the enclosure. It was prototype quality plastics but they held up pretty well if you didn't remove them from the case to much. An interesting side effect of this enclosure choice was that the power supply was mounted inside the case and not outside like the brick type power supply found on a laptop. Ironically however, because this was the same mechanical design constraint given to the Sun Ray 1 it inevitably contributed to a later problem with power supply failures and resulted in an 85,000 unit recall after its first year in production. But more on that later...
From the beginning NeWT was designed to have as much feature parity with the PC in every way as possible. Right down to following Microsoft's own PC design standard recommendations for USB, PCI, SmartCard, power savings, and audio and video performance. In fact, we exceeded PC's at the time almost every way using the exact same technology PC giants Dell and HP had available to them at the time. One unknown feature NeWT had that didn't get rolled into (or supported) the Sun Ray 1 was TV out. While an unfortunate loss, it was a sound business decision to remove it since it wasn't that Sun's customers were ever likely to ever it.
NeWT didn't have a need for an operating system, instead it had what was called an exec. This was developed in the labs as part of another project to produce an experimental Network or IP Camera for Sun Security. It was light weight, real-time, and was completely license and royalty free since it was written by one of our own staff. NeWT and Sun Ray use ALP (Appliance Link Protocol) to communicate with the server over standard Ethernet. ALP was based on the SLIM (Stateless, Low-level Interface Machine) Protocol developed in conjunction. Because so little actually needed to run on the NeWT itself it required very little memory. It had 8 MBytes of EDO DRAM and a 4 MBytes NOR Flash for exec and program storage (by comparison most PC's had 512 MBytes of DRAM at the time). The very first NeWT was built using a MicroSparc IIep-based development board codenamed Sumatra made by Sun Microelectronics. It lacked many of the system components (sound, smartcard, USB, etc) but it proved to be a highly valuable software development platform. Sumatra (see pictures above) was a life saver for the team because we knew we had working code before the first NeWT motherboard was ever powered on in the lab.
Story time...
One day while in the engineering lab the Director of Sun Labs came in for a visit and to see how the project was coming along. During his chat with the group lead, he asked "So why not just make this a software application? While bother with the hardware client at all?" He had a valid question. Our answer of course was "we could, but it does nothing to reduce the TCO as much as a dedicated client could". Sure, you could run a client application and it would work (just ask VMware), but how does that help with software upkeep? disk & system failure? power consumption? replacement cost?, etc.?? He quickly got the point - to effectively provide a customer with the lowest Total Cost of Ownership you must eliminate the fat client.
Months later I gave a talk on the NeWT architecture as part of the Sun Lab open house I was asked how NeWT might impact the JavaStation? I responded that Newt was a complementary technology as there were obvious use cases where a traditional thin client like the JavaStation made better sense. Stand-alone Kiosk's were a very good example of this because there may be no, or very limited network connectivity available. At that time there was no Sun Ray over DSL capability and the client minimally needed a 10Mb Ethernet connection to work. 3D applications was another area it couldn't compete. NeWT couldn't do local 3D rendering, but the JavaStation could and while these differences may seem trivial today, in 1998 they were serious concerns. The JavaStation did eventually fail, but for reasons that had little to do with Sun Ray.
As to the exact number of NeWT's that got built as part of the scalability experiment, frankly it escapes me (it has been over 13 years after all), but I think there were about 60 of them. They were used for quite sometime after Sun Ray 1 development began. Today sadly few survive. I have two in my own personal collection and I donated a third unit to the Computer History Museum in Mountain View California where maybe one day you'll see it put on display for all to enjoy.
Being in the research labs we weren't restricted to just using Sparc, quite the opposite. In fact some of the early processor candidates for NeWT included the Intel StrongARM, Philips Semiconductors (now NXP) TriMedia, and National Semiconductors Geode x86 chipsets. The StrongARM was of interest from the team because of its relatively high integration level and potential for low power operation and cost, but Intel's long term commitment to the ARM architecture at the time was in question by others in the industry since it had been part of their acquisition of Digital Equipment. TriMedia was evaluated, but a software simulator used to test its performance showed that a SparcStation 5 outperformed it with a 110MHz MicroSparc II, so it was quickly abandon. For all its multimedia acceleration features, it still lacked in the ability to keep up with a general purpose processor like the MicroSparc II. Ironically it was while working with the MicroSparc II I discovered that a new integrated version (MicroSparc IIep) had just gone into production and it was the perfect fit for just what we needed.
For graphics I turned to ATI who Sun already had an established and good business relationship with for our server products. The Rage 128 was ideal (later we used the Rage 128 PRO to support LCD displays), it was low power, had powerful 2D rendering performance, and relatively inexpensive compared to anything Sun had internally. It also outperformed most of its commodity PC competitors and ATI was willing to give us driver development support from their Mac team. Because the client had no requirement for 3D and all the 2D graphics rendering occurred on the server there was no need to use the best performing GPU. Years later this would be the subject of another labs project to build a 3D capable client, though the product never was put into production several of the technologies developed for it were later used in Sun Ray products I am told.
Sun at the time had a tendency to designing and using much of the chip technology it put in its systems. With few exceptions, commodity parts were not highly used. Mainly this was for performance issues. In a $50K server, part costs are far less of a concern than in a $399 appliance. But for our needs I knew we needed to leverage the economies of scale provided by using off-the-shelf PC components. With the exception of the processor and the Ethernet controller we used commodity devices for everything else. Now your probably thinking, why the Ethernet controller? Simple. While there were many commodity parts we could have used, this was one area where Sun's own Ethernet controller design was far superior in performance and quite cost competitive with commodity devices.
NeWT used an off-the-self power supply and mini-ITX PC case. The manufacturer offered us an ability to have custom bezel plastics made for the enclosure. It was prototype quality plastics but they held up pretty well if you didn't remove them from the case to much. An interesting side effect of this enclosure choice was that the power supply was mounted inside the case and not outside like the brick type power supply found on a laptop. Ironically however, because this was the same mechanical design constraint given to the Sun Ray 1 it inevitably contributed to a later problem with power supply failures and resulted in an 85,000 unit recall after its first year in production. But more on that later...
From the beginning NeWT was designed to have as much feature parity with the PC in every way as possible. Right down to following Microsoft's own PC design standard recommendations for USB, PCI, SmartCard, power savings, and audio and video performance. In fact, we exceeded PC's at the time almost every way using the exact same technology PC giants Dell and HP had available to them at the time. One unknown feature NeWT had that didn't get rolled into (or supported) the Sun Ray 1 was TV out. While an unfortunate loss, it was a sound business decision to remove it since it wasn't that Sun's customers were ever likely to ever it.
NeWT didn't have a need for an operating system, instead it had what was called an exec. This was developed in the labs as part of another project to produce an experimental Network or IP Camera for Sun Security. It was light weight, real-time, and was completely license and royalty free since it was written by one of our own staff. NeWT and Sun Ray use ALP (Appliance Link Protocol) to communicate with the server over standard Ethernet. ALP was based on the SLIM (Stateless, Low-level Interface Machine) Protocol developed in conjunction. Because so little actually needed to run on the NeWT itself it required very little memory. It had 8 MBytes of EDO DRAM and a 4 MBytes NOR Flash for exec and program storage (by comparison most PC's had 512 MBytes of DRAM at the time). The very first NeWT was built using a MicroSparc IIep-based development board codenamed Sumatra made by Sun Microelectronics. It lacked many of the system components (sound, smartcard, USB, etc) but it proved to be a highly valuable software development platform. Sumatra (see pictures above) was a life saver for the team because we knew we had working code before the first NeWT motherboard was ever powered on in the lab.
Story time...
One day while in the engineering lab the Director of Sun Labs came in for a visit and to see how the project was coming along. During his chat with the group lead, he asked "So why not just make this a software application? While bother with the hardware client at all?" He had a valid question. Our answer of course was "we could, but it does nothing to reduce the TCO as much as a dedicated client could". Sure, you could run a client application and it would work (just ask VMware), but how does that help with software upkeep? disk & system failure? power consumption? replacement cost?, etc.?? He quickly got the point - to effectively provide a customer with the lowest Total Cost of Ownership you must eliminate the fat client.
Months later I gave a talk on the NeWT architecture as part of the Sun Lab open house I was asked how NeWT might impact the JavaStation? I responded that Newt was a complementary technology as there were obvious use cases where a traditional thin client like the JavaStation made better sense. Stand-alone Kiosk's were a very good example of this because there may be no, or very limited network connectivity available. At that time there was no Sun Ray over DSL capability and the client minimally needed a 10Mb Ethernet connection to work. 3D applications was another area it couldn't compete. NeWT couldn't do local 3D rendering, but the JavaStation could and while these differences may seem trivial today, in 1998 they were serious concerns. The JavaStation did eventually fail, but for reasons that had little to do with Sun Ray.
As to the exact number of NeWT's that got built as part of the scalability experiment, frankly it escapes me (it has been over 13 years after all), but I think there were about 60 of them. They were used for quite sometime after Sun Ray 1 development began. Today sadly few survive. I have two in my own personal collection and I donated a third unit to the Computer History Museum in Mountain View California where maybe one day you'll see it put on display for all to enjoy.
Sun Ray 1 - Out of the lab, into the fire!
Somewhere durring the middle of developing NeWT the effort began to find an eventual home in the product group for our technology. As a research team our goal was to produce new technology that would eventually make it in to customers hands. Sun Labs had an excellent track record of transferring technology and we were no exception. Seven of us made the move to Menlo Park and joined the Javastation group (later renamed). An organization of around 200 people at the time.
Work began immediately on Sun Ray. The project was codenamed "Corona" (as in the Sun, not the beer) and eventually involved ruffly forty people. Our manufacturing partner was Mitac International in Taipei Taiwan. Mitac was chosen by Sun operations because it was already building the Ultra 10 workstation for Sun and they had the line capacity, global distribution, and service network we needed. You see, when your send in your Sun Ray in for repair it actually went to a Mitac repair depot, not to Sun. The Mitac team was fantastic. I've never seen so many people working so hard 12 hours a day, 6 days a week!
In Sun Labs we were pretty self-sufficient. We did everything ourselves and contracted out only what we needed. I drew the schematics, I dealt with the suppliers and vendors, I did the enclosure bezel design, and yes... chose the colors. There was no SQA staff, compliance staff, product boss, project manager, supply chain manager, or buyers. We did it all ourselves. But that was about to change. At Mitac I had a full engineering team assigned to me and it was my job to lead them. First order of business was the schematics. NeWT was designed using Cadence Concept which was Sun's schematic capture tool. Mitac used OrCad. Since there was no translation tool, Mitac had to start from scratch with paper copies and rebuild the schematics and libraries from scratch.
Choices also had to be made. What part of the NeWT design would be kept, and what would be tossed? This was an uneasy time for me since I had so much invested into the design and I really didn't want to give anything up. I was also nervous what features I might be asked to add. Fortunately I had gotten much of the design right and little needed to change, except video out. The choice to remove it was a logical one since ATI who made the encoder refused to allow Sun to know about its register configurations, so we couldn't write a driver for it. It also cost $4 on the BOM (Bill Of Materials). So away it went.
One great feature that stayed was the smartcard interface. Most customers at the time didn't even know what one was even though Sun had made some inroads with the Javacard. Sun Ray changed all that. For the first time, users and their desktops were no longer anchored to a single location, nor did they need laptops. As long as they had a smartcard, they had session mobility. By inserting a smartcard into any Sun Ray a customer could securely authenticate themselves and tell to the host server where we were and where to route thier desktop. Anywhere in the world. Hot Desking (as it would be later called) was Sun Ray's killer app. Its what made the client so popular with customers.
Story time...
We built the Sun Ray 1 in record time. Approximately 8 months, which was lightning fast by Sun standards. Most products at the time took anywhere from 2-5 years, usually because of chip development delays. Customers always ask me why the Sun Ray Type 5 keyboard had such a long USB cable? Well the reason was as simple as it was foolish. Sun had a history of introducing new keyboards but very rarely. A new "type" would emerge as a result of a new product family being introduced like the Ultra series and that would be the new keyboard for basically everything that came after. At the time Sun Ray was in development an effort had already began in the server group to design first USB keyboard - the type 5. It would go with a whole new class of servers. Because the server group had bought and paid for its development we could only use the keyboard, not influence any of it's design. Therefore since the server needed a very long cable, everyone got a long cable. Whether you needed it or not.Surprise!
Another question many have asked about was how the name Sun Ray came into being. This is a very funny story and I suspect slightly embarrassing to a certain branding agency Sun used at the time. As I had mentioned before, Corona had been the working codename of the product. Many of us on the product team actually liked it a lot and thought it should be the real product name. However marketing had a different idea. They hired an outside agency to come up with a product name, and they did. Corona was to be called "Clickster". No, I am not kidding.
The representative of the agency described to us proudly how the name had been derived. The "Click" meant "Snappy" and "ster" was from "Boxster" as in the popular Porsche sports car. So the take-away meaning was "snappy performance". Cute. Only one slight little problem. The phonetic pronunciation of "Clickster" in certain European countries sounded exactly like that of the word "Klyster". In English its phonetically different, however in French speaking regions and even in the UK, the meaning was resoundingly different - it meant enema bag. Once proof of this oversight surfaced the name was dropped and a new name was suggested by the team. Sun Ray became the winner by a simple majority vote.
The product had been slated for release in the early part of the summer of 1999, however a senior Sun VP nixed the idea in favor of announcing later in September at a Design Automation Conference in New York City. While most were disappointed by the launch delay, it gave us and Mitac much needed time for additional product testing. It also meant that the team was going to New York City for the launch, something we all liked!
Work began immediately on Sun Ray. The project was codenamed "Corona" (as in the Sun, not the beer) and eventually involved ruffly forty people. Our manufacturing partner was Mitac International in Taipei Taiwan. Mitac was chosen by Sun operations because it was already building the Ultra 10 workstation for Sun and they had the line capacity, global distribution, and service network we needed. You see, when your send in your Sun Ray in for repair it actually went to a Mitac repair depot, not to Sun. The Mitac team was fantastic. I've never seen so many people working so hard 12 hours a day, 6 days a week!
In Sun Labs we were pretty self-sufficient. We did everything ourselves and contracted out only what we needed. I drew the schematics, I dealt with the suppliers and vendors, I did the enclosure bezel design, and yes... chose the colors. There was no SQA staff, compliance staff, product boss, project manager, supply chain manager, or buyers. We did it all ourselves. But that was about to change. At Mitac I had a full engineering team assigned to me and it was my job to lead them. First order of business was the schematics. NeWT was designed using Cadence Concept which was Sun's schematic capture tool. Mitac used OrCad. Since there was no translation tool, Mitac had to start from scratch with paper copies and rebuild the schematics and libraries from scratch.
Choices also had to be made. What part of the NeWT design would be kept, and what would be tossed? This was an uneasy time for me since I had so much invested into the design and I really didn't want to give anything up. I was also nervous what features I might be asked to add. Fortunately I had gotten much of the design right and little needed to change, except video out. The choice to remove it was a logical one since ATI who made the encoder refused to allow Sun to know about its register configurations, so we couldn't write a driver for it. It also cost $4 on the BOM (Bill Of Materials). So away it went.
One great feature that stayed was the smartcard interface. Most customers at the time didn't even know what one was even though Sun had made some inroads with the Javacard. Sun Ray changed all that. For the first time, users and their desktops were no longer anchored to a single location, nor did they need laptops. As long as they had a smartcard, they had session mobility. By inserting a smartcard into any Sun Ray a customer could securely authenticate themselves and tell to the host server where we were and where to route thier desktop. Anywhere in the world. Hot Desking (as it would be later called) was Sun Ray's killer app. Its what made the client so popular with customers.
Story time...
We built the Sun Ray 1 in record time. Approximately 8 months, which was lightning fast by Sun standards. Most products at the time took anywhere from 2-5 years, usually because of chip development delays. Customers always ask me why the Sun Ray Type 5 keyboard had such a long USB cable? Well the reason was as simple as it was foolish. Sun had a history of introducing new keyboards but very rarely. A new "type" would emerge as a result of a new product family being introduced like the Ultra series and that would be the new keyboard for basically everything that came after. At the time Sun Ray was in development an effort had already began in the server group to design first USB keyboard - the type 5. It would go with a whole new class of servers. Because the server group had bought and paid for its development we could only use the keyboard, not influence any of it's design. Therefore since the server needed a very long cable, everyone got a long cable. Whether you needed it or not.Surprise!
Another question many have asked about was how the name Sun Ray came into being. This is a very funny story and I suspect slightly embarrassing to a certain branding agency Sun used at the time. As I had mentioned before, Corona had been the working codename of the product. Many of us on the product team actually liked it a lot and thought it should be the real product name. However marketing had a different idea. They hired an outside agency to come up with a product name, and they did. Corona was to be called "Clickster". No, I am not kidding.
The representative of the agency described to us proudly how the name had been derived. The "Click" meant "Snappy" and "ster" was from "Boxster" as in the popular Porsche sports car. So the take-away meaning was "snappy performance". Cute. Only one slight little problem. The phonetic pronunciation of "Clickster" in certain European countries sounded exactly like that of the word "Klyster". In English its phonetically different, however in French speaking regions and even in the UK, the meaning was resoundingly different - it meant enema bag. Once proof of this oversight surfaced the name was dropped and a new name was suggested by the team. Sun Ray became the winner by a simple majority vote.
The product had been slated for release in the early part of the summer of 1999, however a senior Sun VP nixed the idea in favor of announcing later in September at a Design Automation Conference in New York City. While most were disappointed by the launch delay, it gave us and Mitac much needed time for additional product testing. It also meant that the team was going to New York City for the launch, something we all liked!
Getting the cost out - Product diversification
Almost immediately after launching Sun Ray 1, the group began work on cost reductions and new models. We were under pressure to reduce product cost. After all, it was a $399 appliance so there wasn't much in margin to speak about compared to what Sun made on a workstation or server. The cost reductions did have their silver linings. One was integration, but the other was improved functionality which helped bring about some wonderful products. With the changes we were able to create the Sun Ray 100, Sun Ray 150, and Sun Ray 1 G.
The first motherboard cost reduction spin was codenamed Andromeda. This project was lead by a co-worker of mine so I could focus on the development of the 100 and 150. The motherboard introduced a new combined Ethernet controller and USB interface from the folks at SME along with the ATI Rage 128 PRO which allowed direct interface support LCD panels needed in the 150. Andromeda was used in the Sun Ray 1 and 100, but because of mechanical constraints Samsung had to re-layout the design for the 150. The Sun Ray 1 had been designed and manufactured by Mitac International, however with the advent of the 100 and 150 we needed a manufacture who could provide additional elements that Mitac couldn't. Sun operations recommended Samsung's display division, so the development moved to Korea. Did I mentioned we were on airplanes a lot?
The Sun Ray 100
The Sun Ray 100 was an Andromeda based client that was mounted in the base of a 17" CRT display. This was a low cost all-in-one product intended to target cost sensitive customers in the education and call center markets. It utilized the same motherboard the Sun Ray 1 used but with a few small modifications, mainly an internal video cable to connect to the display board. The motherboard was still manufactured by Mitac and then flown to Korea where they were assembled and tested by Samsung.
The Sun Ray 150.
The Sun Ray 150 was my favorite. Of all the Sun Ray's it was the one I wanted to build more than anything. That's not to say I didn't love all of designs, but this one by far was my personal favorite because of the tight integration with the LCD. To me it truly represented a new direction of desktop computing. The 150 was originally intended to be a 17" diagonal display like its 100 counterpart. However due to supply constraints and the high cost a more economical 15" panel was chosen. The RAGE 128 PRO's TMDS LCD interface was also a problematic as TMDS LCD panels were expensive, but fortunately there was a good supply of 15" TTL panels. So we used external translation chips on the power board to make the conversion.
One feature I added to the 150 was the VGA connector. This was done to provide customers a way to attach and external projector if they wished, or scan converter to get NTSC/PAL video. It was a hit. Conference rooms through out Sun used 150's with their projectors. People really loved the 150, just not its price. Sadly it was the one thing we couldn't control very much as the price of the LCD panel dictated pricing the product at a premium well above what we had hoped. Even so, the product line continued with the Sun Ray 170 and now with the Sun Ray 3i.
The Sun Ray 1G
The story of the Sun Ray 1G is an interesting one as it almost didn't happen.The original Sun Ray was designed with a 1 Megapixel framebuffer. No one really knew if a 100Mbit Ethernet connection had the bandwidth or if the server was capable of rendering good enough to make it work. Marketing said that due to this resolution limitation we weren't getting any interest from our ECAD customers. So into the lab I went. The change was simple enough, take out the old 8MB SGRAM and replace it with 16MB parts doubling the size of the framebuffer. After a few firmware changes we turned it on. Much to our pleasant surprise it worked. Having done this I setup a small demo to marketing and management and they were surprised by what they saw. Marketing seeded some key accounts with the 2 Megapixel prototypes. Needless to say the Sun Ray 1G got the green light to eventually go into production. This result inspired me to also adapt the Sun Ray 150 motherboard to try and support larger LCD panels. We successfully got a 21" LCD panel (1600x1200) to run and it looked really good, but the effort was shelved because of the absurdly high cost of the display.
Copernicus
Copernicus wasn't a Sun Ray client, but a system-on-chip designed to be used in all models of Sun Ray. Its roots go back to Sun Labs where it was envisioned as the next logical step to integrate the client into a single chip solution. In many ways it was ahead of its time.The idea of merging a processor, memory controller, USB, smartcard interface, I2C, and 4 Mbytes of DRAM onto one die was simply amazing at the time. Few had ever tried combining a DRAM process with a processor, and for obvious reasons. It was risky and had limited market appeal. After all, most processor implementations tend to grow their memory foot prints over time, not shrink it. By holding the memory size to 4 MBytes it limited the device's appeal in other uses.
Eventually the DRAM process used to make Copernicus had run its course and the semiconductor manufacturer wanted to halt its production. Though it had a good run, Copernicus life was a short one. With the introduction of the Sun Ray 2 the design had moved to the AMD Alchemy 1550. I once evaluated using the Alchemy 1550 as a possible alternate while Copernicus was being proposed. But the high cost of porting everything to the MIPS platform was just to much to bare at the time, so the idea got shelved. In retrospect I am happy Copernicus had a good run and was implemented. I am even named on the patent as one of its contributors. But it was a slave to the fluctuations of an unforgiving DRAM market that eventually did it in. A market that doesn't stand still for anyone, no matter who they may be... .
The first motherboard cost reduction spin was codenamed Andromeda. This project was lead by a co-worker of mine so I could focus on the development of the 100 and 150. The motherboard introduced a new combined Ethernet controller and USB interface from the folks at SME along with the ATI Rage 128 PRO which allowed direct interface support LCD panels needed in the 150. Andromeda was used in the Sun Ray 1 and 100, but because of mechanical constraints Samsung had to re-layout the design for the 150. The Sun Ray 1 had been designed and manufactured by Mitac International, however with the advent of the 100 and 150 we needed a manufacture who could provide additional elements that Mitac couldn't. Sun operations recommended Samsung's display division, so the development moved to Korea. Did I mentioned we were on airplanes a lot?
The Sun Ray 100
The Sun Ray 100 was an Andromeda based client that was mounted in the base of a 17" CRT display. This was a low cost all-in-one product intended to target cost sensitive customers in the education and call center markets. It utilized the same motherboard the Sun Ray 1 used but with a few small modifications, mainly an internal video cable to connect to the display board. The motherboard was still manufactured by Mitac and then flown to Korea where they were assembled and tested by Samsung.
The Sun Ray 150.
The Sun Ray 150 was my favorite. Of all the Sun Ray's it was the one I wanted to build more than anything. That's not to say I didn't love all of designs, but this one by far was my personal favorite because of the tight integration with the LCD. To me it truly represented a new direction of desktop computing. The 150 was originally intended to be a 17" diagonal display like its 100 counterpart. However due to supply constraints and the high cost a more economical 15" panel was chosen. The RAGE 128 PRO's TMDS LCD interface was also a problematic as TMDS LCD panels were expensive, but fortunately there was a good supply of 15" TTL panels. So we used external translation chips on the power board to make the conversion.
One feature I added to the 150 was the VGA connector. This was done to provide customers a way to attach and external projector if they wished, or scan converter to get NTSC/PAL video. It was a hit. Conference rooms through out Sun used 150's with their projectors. People really loved the 150, just not its price. Sadly it was the one thing we couldn't control very much as the price of the LCD panel dictated pricing the product at a premium well above what we had hoped. Even so, the product line continued with the Sun Ray 170 and now with the Sun Ray 3i.
The Sun Ray 1G
The story of the Sun Ray 1G is an interesting one as it almost didn't happen.The original Sun Ray was designed with a 1 Megapixel framebuffer. No one really knew if a 100Mbit Ethernet connection had the bandwidth or if the server was capable of rendering good enough to make it work. Marketing said that due to this resolution limitation we weren't getting any interest from our ECAD customers. So into the lab I went. The change was simple enough, take out the old 8MB SGRAM and replace it with 16MB parts doubling the size of the framebuffer. After a few firmware changes we turned it on. Much to our pleasant surprise it worked. Having done this I setup a small demo to marketing and management and they were surprised by what they saw. Marketing seeded some key accounts with the 2 Megapixel prototypes. Needless to say the Sun Ray 1G got the green light to eventually go into production. This result inspired me to also adapt the Sun Ray 150 motherboard to try and support larger LCD panels. We successfully got a 21" LCD panel (1600x1200) to run and it looked really good, but the effort was shelved because of the absurdly high cost of the display.
Copernicus
Copernicus wasn't a Sun Ray client, but a system-on-chip designed to be used in all models of Sun Ray. Its roots go back to Sun Labs where it was envisioned as the next logical step to integrate the client into a single chip solution. In many ways it was ahead of its time.The idea of merging a processor, memory controller, USB, smartcard interface, I2C, and 4 Mbytes of DRAM onto one die was simply amazing at the time. Few had ever tried combining a DRAM process with a processor, and for obvious reasons. It was risky and had limited market appeal. After all, most processor implementations tend to grow their memory foot prints over time, not shrink it. By holding the memory size to 4 MBytes it limited the device's appeal in other uses.
Eventually the DRAM process used to make Copernicus had run its course and the semiconductor manufacturer wanted to halt its production. Though it had a good run, Copernicus life was a short one. With the introduction of the Sun Ray 2 the design had moved to the AMD Alchemy 1550. I once evaluated using the Alchemy 1550 as a possible alternate while Copernicus was being proposed. But the high cost of porting everything to the MIPS platform was just to much to bare at the time, so the idea got shelved. In retrospect I am happy Copernicus had a good run and was implemented. I am even named on the patent as one of its contributors. But it was a slave to the fluctuations of an unforgiving DRAM market that eventually did it in. A market that doesn't stand still for anyone, no matter who they may be... .
Postmortem
There is more of the Sun Ray story to be told, but I must stop here because I wasn't a part of it. By the time I was getting the Copernicus Sun Ray's ready
for their debut, word came down that there was going to be a redeployment. It was the summer of 2001 and the dot com meltdown was well
underway. Sun was losing customers and the stock was sliding It was the beginning of a slow, long drawn out phase in Sun's history that eventually led to it being acquire by Oracle.
Its possible I have left out some details in this historical account. I may have gotten a fact or two even wrong, and for that I apologize. This all happen over a decade ago and nobody's memory is perfect. I am very thankful for my time on this projectand it inspired me to return to college and finish my undergraduate degree. I stayed on for three more months at Sun before leaving for good, then took a nice long much needed holiday in Europe.
Its possible I have left out some details in this historical account. I may have gotten a fact or two even wrong, and for that I apologize. This all happen over a decade ago and nobody's memory is perfect. I am very thankful for my time on this projectand it inspired me to return to college and finish my undergraduate degree. I stayed on for three more months at Sun before leaving for good, then took a nice long much needed holiday in Europe.