Chapter 281 Dual-Core and 64-Bit
Su Yuanshan led Claude toward his office.
Although Yuanxin's desktop CPU design department had already moved entirely to Shanghai, Su Yuanshan's office still held the most complete set of reference materials and the latest technical updates.
"We've been thinking very seriously," Claude said as he sat down across from Su Yuanshan, staring thoughtfully at the rising steam from the coffee pot, "and we believe that in the future, floating-point performance will be a critical indicator of a CPU's capabilities. It's heavily influenced by frequency in a very intuitive way."
"One reason Cyrix never managed to confront Intel head-on," he continued, "was not just because we lacked resources, but also because we couldn't break free from architectural limitations."
Though it sounded like Claude was explaining Cyrix's history of imitating Intel, in reality... he was paying Su Yuanshan a subtle compliment.
Because now, the Sol architecture provided by Su Yuanshan had given them something completely different — and when Claude first laid eyes on it, he instinctively knew this new architecture could surpass Intel's Pentium.
Su Yuanshan smiled slightly, poured a cup of coffee for Claude, and said softly:
"Instruction Per Clock."
Claude froze for a moment.
Su Yuanshan thought for a second and added, "Let's abbreviate it — IPC.
It measures the efficiency of instructions executed per clock cycle.
That's the real fundamental metric; frequency is just a surface indicator — a means to an end."
"Take brick-carrying as an example.
If you want to move bricks faster, you can either run faster — that's raising the frequency — or you can carry two or three bricks at once. Even if you don't move faster, your efficiency doubles or triples."
Claude fell into deep thought.
Su Yuanshan wasn't in a hurry. He sipped his coffee leisurely.
There were many ways to increase IPC — superscalar designs, super pipelining, dynamic branch prediction, and in the future, even Very Long Instruction Word (VLIW) architectures.
"You're saying we should focus on superscalar and super-pipelining architectures?" Claude asked a few seconds later, lifting his head.
"Not focus on one direction — but don't neglect any of them either," Su Yuanshan replied calmly. "If we focus only on frequency, we'll end up down the wrong path."
He smiled faintly, half-jokingly adding, "For example, right now, Intel and AMD are both walking that crooked road."
Claude couldn't help but feel a bit speechless.
Only Su Yuanshan dared say things like that so bluntly.
He himself wouldn't dare.
Over the past year, Claude had exchanged over a hundred emails with Su Yuanshan.
He already knew this young man wasn't just a genius architect — every time they discussed technical issues, Claude almost forgot he was talking to someone so young.
It felt like talking to an old master who had spent decades immersed in CPU design.
"Of course, we can't abandon frequency entirely either," Su Yuanshan said with a chuckle.
"1GHz is too tempting a milestone.
The performance boost from frequency is too intuitive."
"Come on, let me show you the ideas for our next-generation architecture."
...
That afternoon, Li Mingliu and Gao Xiaodi arrived from Shanghai.
As the leaders of the CPU team, they were the ones truly in charge of frontline technical work.
Together with Claude, the four gathered in Su Yuanshan's office to brainstorm the direction for their next-generation architecture.
Right now, Yuanxin and Xinghai's technical foundations were still shallow.
Even with Su Yuanshan "cheating" with his future knowledge, at best they could just about catch up to Intel.
Real overtaking would only be possible once they achieved "one generation in production, one in reserve, one in development."
Since both CPU teams had completed their current projects, it was natural to start early research for the next generation.
"32-bit processors will inevitably be replaced by 64-bit processors," Su Yuanshan said softly, twirling a pencil in his fingers.
"It doesn't matter what outsiders think — we must clearly recognize this ourselves."
"But if we move too fast, won't we hurt ourselves?" Li Mingliu said seriously, flicking cigarette ash onto the floor.
Seeing Claude's confusion, he quickly repeated it in English.
Claude chuckled. "We all agree that 64-bit processors are the future.
But it requires operating system support...
Right now, even today's operating systems aren't fully 32-bit yet."
"Moving too fast could turn us into cannon fodder," Gao Xiaodi nodded.
"We could end up building CPUs so advanced that no one can use them."
Seeing the three of them so cautious, Su Yuanshan smiled — and he was genuinely pleased.
What he feared most was that because he was "too good," his colleagues would blindly follow him, turning into mindless yes-men.
That would be meaningless.
"You're right," Su Yuanshan said, acknowledging their concerns, "but we can't wait for the need to arise before rushing to develop — that's what mediocre companies do."
"I'll make a bold prediction:
Even without us, within five years, 64-bit processors will emerge."
The three of them froze.
Especially Li Mingliu, who stared at him with utmost seriousness.
He knew — whenever Su Yuanshan made a prediction, it was no joke.
Was he getting insider information?
Impossible.
Xinghai was a purely commercial enterprise — not an espionage outfit.
"Moreover, developing 64-bit processors would mean one other thing —
We could finally break free from Intel's x86 dominance," Su Yuanshan said calmly.
Whether it was multi-core or 64-bit, these technologies had already been proposed.
Intel and AMD were both thinking about it — but at this stage, only thinking.
They had spotted the trend but hadn't yet taken action.
Meanwhile, Yuanxin and Xinghai hadn't even "seen" it yet — or if they had, they thought it was still a distant fantasy.
First, focus on making CPUs profitable.
That mindset made sense for small companies like Cyrix — survival was the priority.
But for big companies, it was deadly.
When Cyrix was sold to National Semiconductor and later VIA,
they produced the MII and M3 chips — and then gave up CPU development altogether.
Why?
Because they couldn't keep up.
When your competitors are already implementing next-generation technology and you're still struggling to replicate their old designs, how can you compete?
"That's why I propose," Su Yuanshan said, smiling,
"Xinghai's team should start preliminary research on multi-core architectures, while the Shanghai team explores 64-bit instruction sets."
He paused, then added, "Of course, I'll participate in both."
"How much manpower will we need?" Claude asked seriously.
Thank you for the support, friends. If you want to read more chapters in advance, go to my Patreon.
Read 30 Chapters In Advance: patreon.com/Albino1