Why Don't They Make BIGGER CPUs?

Would larger CPUs help performance?

AFFILIATES & REFERRALS
---------------------------------------------------

FOLLOW US ELSEWHERE
---------------------------------------------------

FOLLOW OUR OTHER CHANNELS
---------------------------------------------------

[Drza]

You heard the man: Bigger isn't always better

[Bajicoy]

"bigger isn't always better"

Threadripper owner: is this some kind of peasant joke I am too rich to understand?

[christiandominiclangreo]

"Bigger isn't always better"

This gives me hope.

[fish]

I’m sorry, I just can’t imagine being named “Ben Benson”

[movedaccount]

"... and Ben Benson..."
*dear god Ben Benson, it's like my friend John Johnson*

[lonzoformvp]

Dammit Linus, you gave intel new ideas to continue using 14nm

[FaceyDuck]

This explains why the squares in waffles are so small! Thanks, Linus!

[qwijbo]

It's nice that after 30 something years you're finally able to grow a beard

[wheelybigcheez]

“Bigger isn’t always better”

Yeah, Linus.

[ascendria]

Even if Linus ends up ditching the beard, he should keep that haircut

[brianm]

When he gave the car engine metaphor all i heard was "bring the V10s back to F1!"

[suryanarayanan]

Last time I was this early, the only folding phone was the iPhone 6 Plus

[twylanaythias]

Also keep in mind that, with a 4.2 GHz CPU, electricity can only travel ±3.5cm during a single cycle. Factor for the time demanded by capacitance and latency, and you're lucky to traverse a single cm during a complete processor cycle. The AMD Zen 3 8-core die is 1.29 x 0.964 cm, with the shared 0.7cm² cache flanked by 0.3cm-wide CPU cores - keeping most operations within 0.6cm in order to obtain reliable operations at the rated 4.7GHz boost speed.

Prior to ±2000, CPU clock speeds were doubling roughly every 2-3 years:
~ 1982 : (1.5 μm) 80286 @ 6 MHz; 68000 @ 8 MHz
~ 1985 : (1.5 μm) 80386 @ 12 MHz; 68020 @ 16 MHz
~ 1989 : (1.0 μm) 80486 @ 25 MHz; 68040 @ 30 MHz
~ 1993 : (800 nm) Pentium @ 60 MHz; 68060 @ 55 MHz
~ 1995 : (500 nm) Pentium Pro @ 120 MHz; PowerPC 603/604 @ 120 MHz
~ 1997 : (350 nm) Pentium II @ 260 MHz; PowerPC 750 @ 260 MHz
~ 1999 : (250 nm) Pentium III, Athlon, PowerPC 7400 @ 600 MHz
~ 2001 : (180 nm) Pentium 4, Athlon XP @ 1.2 GHz
~ 2005 : (90 nm) Pentium D, Athlon 64 X2 @ 2.4 GHz


By 2005, CPU manufacturers were already approaching the physical limits of semiconductor technology and it took roughly a decade to refine manufacturing sufficiently (down to 32-10 nm; 7 nm for Zen2/3) to reliably deliver ±4.8 GHz on consumer-grade CPUs.

[nathannguyen]

0:03, no Linus, not arguably. We all know RGB is the most important

[SIGSEGV]

Do a fast as possible on the FINFET manufacturing process

[madil]

Title should have been: "Does size matter?"

[MrFreakzoidrj]

People: “americans should use metric”

Americans: “a cpu measures 3 post it stamps”

[dalec]

IMO a decent mini-series for this channel would be the the "Top 5 advancements in X" series.

Like the Top 5 advancements in videocard history.

The Top 5 advancements in display technology.

The Top 5 advancements in PC case design.

The Top 5 advancements in water bottle technology... You get the idea.

[haystackdmilith]

You forgot to mention that's also about path length between elements. Longer paths mean more time electrons have to travel… which uses more power and takes precious pico-seconds (that later adds together).

[jezuconz]

I was actually wondering about this exact topic a few days ago. Didn't even google it or say it loud (you know Google's love for their users' microphones) and here it is. Thanks guys!