Where does this come from, and can I check it?

A cache hit costs a few nanoseconds while main memory takes about a hundred, so chips copy nearby data first. The figure is taken from Britannica, which is linked in full on this page. Every fact on Mondo Knows names its source rather than repeating an unsourced claim.

Is this really a technology fact?

Yes. It sits on the Technology shelf, one of 2,600 sourced facts grouped into space, animals, the human body, history, science and more. Category pages collect every fact in the same field.

What else can I read about it?

The links under the fact point to a quiz, game, puzzle or calculator on the same subject, and the shelf below shows related facts. The wording here is original, and quoted phrases are marked as quotations.

Mondo Knows Facts

Processors keep small caches to hide memory delays

A cache hit costs a few nanoseconds while main memory takes about a hundred, so chips copy nearby data first Caches work because programs tend to reuse recently touched data and to walk through memory in order. A modern processor checks level one caches in a handful of cycles, then level two and level three, and finally asks the memory controller. Wulf and McKee named the growing mismatch between processor and memory speed the memory wall in 1995. It sits in Mondo's technology collection, one of 200 facts that each trace back to the reference below.

Where this comes from

Britannica

Facts that can change over time are reviewed and updated. If something here looks wrong, Mondo would rather know than be confidently incorrect.

Go further

Technology trivia

More facts like this

x86 processors still run code written in 1978Intel kept the 8086 instruction set alive through decades of chips so old software would keep working A cosmic ray can flip a bit in memoryAn energetic particle striking silicon can invert a stored value, an effect called a single event upset Dynamic RAM DRAM was invented by Robert Dennard at IBM in 1966DRAM uses a single transistor and capacitor per memory bit to store digital data Magnetic core memory served as main RAM for computers from the 1950s to 1970sTiny ferrite rings threaded on wire grids stored magnetic bits non-volatilesly The first stored program ran on the Manchester BabyIn June 1948 a small experimental machine executed 52 instructions held in a cathode ray tube memory Virtual memory first worked on the Atlas computerThe Manchester Atlas, running in 1962, gave programs an address space larger than the machine's real memory

Tags: processors, memory, performance

Processors keep small caches to hide memory delays: the short answers

Where does this come from, and can I check it?: A cache hit costs a few nanoseconds while main memory takes about a hundred, so chips copy nearby data first. The figure is taken from Britannica, which is linked in full on this page. Every fact on Mondo Knows names its source rather than repeating an unsourced claim.
Is this really a technology fact?: Yes. It sits on the Technology shelf, one of 2,600 sourced facts grouped into space, animals, the human body, history, science and more. Category pages collect every fact in the same field.
What else can I read about it?: The links under the fact point to a quiz, game, puzzle or calculator on the same subject, and the shelf below shows related facts. The wording here is original, and quoted phrases are marked as quotations.