Sad Walk Down Memory Lane
-- NapMaster (aka slaphappy)
This is my theory showing what happened with the Project Dream development effort and what went wrong. Based on the available facts, this is what I believe to be true, but I am always open to contradictory opinions and information pointing out any flaws in my analysis.
Let's Begin. In 1993, Nintendo began development of Ultra 64.
In establishing the hardware specifications, the decision was made by Nintendo to incorporate the Rambus Unified Memory Architecture into the Ultra 64, later renamed to the Nintendo 64. When the N64 was released in 1996, it used 36Mbits of RAMBUS 9bit DRAM, equating to 4 megabytes (expandable to 8 via the expansion slot). This was a huge win for Rambus in gaining mainstream acceptance for RDRAM and RIMM. Unfortunately, the first generation RDRAM, now referred to as Base RDRAM, had a severe initial latency flaw. Not so problematic for Nintendo, but it had quite a negative impact on Rambus's battle with SDRAM for leadership in the PC market.
The Rambus memory incorporated in the N64 architecture was state of the art and by it's very design created an interesting but potentially troublesome situation. It became apparent early on that the RDRAM would not be immediately cleared when the system was powered down. This was due to the RDRAM's unique power dissipation modes, primarily the nap and powerdown modes.
SDRAM Power Management
This situation was quickly discovered during development efforts by Nintendo, and was documented explicitly in the N64 Programming Guide. The guide states in the Data Cache & RDRAM Consistency section that program data in the N64 Game Pak ROM is moved to RDRAM overwriting whatever is there already.
N64 Programming Guide
Overwriting what is already there implies the RDRAM does not immediatley clear when the system is powered down. Nintendo wanted to make sure all games initialized properly to handle this memory holdover without causing error conditions; endless looping or program hangs as they stated. Rare didn't see this as a drawback, but instead as an amazing opportunity, and Project Dream was born. The possibility existed that game data could be stored in RAM and if handled properly and not overwritten, be accessed by a new game ROM. I have been unable to find supporting documentation, but viablity of this memory has been spectulated to be anywhere in the 10 to 30 second range.
So Rare began it's work on Project Dream, later to be renamed Banjo Kazooie. An interesting proposition, but one that would be considered an amazing accomplishment in the gaming community. The transfer of data Proof of Concept was coded and successfully demonstrated by transferring the Ice Key data between beta versions of BK and DK64 (see the DK64 SNS Special for more details). Not via a hot-swap of ROMs, but by turning off the console first, removing one ROM and inserting the other. There was more than enough time to stop the N64, and swap the game ROMS, hence Stop 'n Swop. If the user took to long, nothing bad would happen, just no data transfer.
Rambus, meanwhile, had been working on a fix for the initial latency problem with the Base RDRAM. This issue was beginning to cost them marketshare in the SDRAM battle, due to RDRAM's comparable to slower access times but much higher cost. Rambus was able to successfully resolve the latency issue and the 2nd generation, Concurrent RDRAM, was announced. The initial latency problem was occurring during the powerdown to active mode transition, so modifications to the power dissipation modes must be assumed. By late 1997, Concurrent RDRAM began shipping and would now be used in the fabrication of new Nintendo 64 units.
Concurrent RDRAM
In 1998, Banjo Kazooie was released to the public, with the Stop 'N Swop feature fully intact. Rare got to work on Banjo Tooie, showcasing it at E3 in 1999, with some great unexpected features, including Bottles Revenge. They also put out the infamous BT Tepid Seat interview in 1999, so to assume they were unaware of a devasting upcoming problem would seem fair. Several question in that interview regarding the BK secret items and Blackeye supports the belief the SNS concept was alive and well and was nearing completion. A note regarding Q25, I don't recall ever seeing anyone suggest the BK pak concept.
BT tepid Seat Interview:
Q14: In Mad Monster Mansion, there is a room with two pictures. One of Grunilda, and one of this evil pirate bloak. What I want to know is: Will he be the next villian to face BK? Or is he somehow in league with Grunty? Is he mates with BK? Or is including his picture just arseing with my mind? (Chris Todd)
A: Well spotted sir! The evil pirate bloak has a mission and purpose all of his own, but this is neither the time nor place to reveal even the merest hint. You might learn more from a certain tavern in the future...
Q25: Have you ever been a bit surprised at how close and/or unclose some of the rumors were? (DPrice@aol)
A: Some of the rumours were fairly close, bust most of the suggestions on how to find the ice key, Sharkfood Island, secret eggs, etc, were so amusingly inaccurate it kept us entertained for quite a while.
Q31: Obviously aside from the ice key, Sharkfood Island, and desert door, are there other places in B-K that are locked until B-T comes out? I'm thinking, maybe the closed barrel in the wine cellar in Mad Monster Mansion, or the two doors in the room with Grunty's transformation machine. (Josh Gelman)
A: heh... heh... heh... keep on guessing... ho... ho... ho... all will be revealed... hee... hee... heee... Sorry about that, the simple answer is yes.
It is unknown what quantity of Base RDRAM Nintendo still had in inventory, or how quickly the Concurrent RDRAM N64's where built and shipped to market. Unfortunately for Rare, change was about to come and it was going to hit like a ton of bricks. The newly shipping N64s now had the 2nd generation RDRAM, which adversely affected the memory holdover time. Be it by fixing the initial latency problem, or due to some other change, the memory retention was shortened immensly. The holdover period was spectulated to now be in the 2 second range following power down, clearly way too fast to have someone switch ROMs. Project Dream was officially dead.
It was way too far in the development cycle for Rare to change course with the SNS concept. Switching to a controller pak or a lock-on cartridge were impossibilites, since Banjo Kazooie had already been released with no contingency coding in place. How could anyone at Rare have anticipated a hardware change on a gaming console? PC Game developers must deal with this situation, but not console development shops. Rare quickly had to scramble to come up with a viable alternative, one that would give the BK players the items they were promised. Hence, the bouncing BK packs were conceived. All things considered, not a bad idea with the incredibly short timeframe they now had to work. Unfortunately, all development resources had to be thrown at this, leaving Bottles Revenge to forever remain incomplete. The poster artwork for Tooie had already been completed, so removing Devil Bottles from that could not be done.
Banjo Tooie was released to the public with the BK paks and we were left scratching our heads. Rare was so close to pulling off the unprecedented SNS, but at the end of the day, got the shaft. Blackeye apparently got the shaft too. No more purpose all his own, such as giving us the sandcastle codes? Makes me wonder if Rare was somehow referring to Rambus as the Bear in the "a bear stole my glory" statement. Rambus as the bear, hmmm. I'll have to work on that one.
In my mind, one final question remains, though. Did Rare actually have enough time to remove the SNS code from BT? There was the infamous statement from the Rare employee claiming the connection between the games was still there. Rare claims he was misquoted. Maybe, maybe not. The clean up effort appears to have been a little sloppy. Blackeye still has the Pink Egg portraits all around him. Remnants of Bottles Revenge remain. The Base Egg being left in the NTSC version of Tooie.
So, I am holding on to the final remaining glimmer of hope. Assuming SNS does still exist, one would need a pre-1999 N64, the NTSC version of BK, and the NTSC version BT to find it. If this, however unlikely, is true, it creates a tough position for Rare. Is SNS still there? No, well yes, well kind of. I wouldn't comment on the subject either. I also wouldn't have taunted everyone with the chalkboard in GbtG either, but hey, Rare will be Rare.
---