The History Of Intel CPUs: Updated!

About the author
Michael Justin Allen Sexton

Michael Justin Allen Sexton is a Contributing Writer for Tom's Hardware US. He covers hardware component news, specializing in CPUs and motherboards.

Read more
Create a new thread in the US Photo reports comments forum about this subject
11 comments
Comment from the forums
    Your comment
  • mitch074
    Strange that Itanium is missing, the Celeron 300A/333 is gone, the original Pentium bug disappeared, no mention is made that the 487 was actually a fully active 486DX, and that AMD led the desktop for "a short time" while it led from the moment Netburst came out (2000) to the moment Core replaced it (2006). On another note, a 64-bit CPU doesn't run 64-bit software faster : it is required to have one to run some. But since AMD came up with x86-64, I guess some approximation is allowed...
  • lunyone
    And where is the AMD CPU history?
  • Tom Griffin
    Remember the ABIT BP6a motherboard; I was running dual Celery (Celeron) 300mhz processors overclocked to 533mhz. What a flashback.
  • TheDane
    @Tom Griffin: Yeah - those were the days :)
  • AndrewJacksonZA
    1) Where can we see what was updated please?
    2) Where are the Phi CPUs please?
    3) Where can we view this as a one-pager please?
  • ta152h
    There are quite a few mistakes here.

    For one, the 8086 was not available in higher clock speeds than the 8088.

    The 8086 could NOT run 8080 code. Source code compatible does not mean that. It means it was very easy to recompile the code so it would work with the newer processor, not that the compiled code would work.

    The 286 section is oddly very limited. It was an enormous improvement over the 8086, as it added more memory, much more performance, and also virtual memory and hardware assisted multi-tasking.

    The 80386 was not significantly faster than the 286 running 16-bit code, despite what the author says. Clock for clock, they were very close, although 386 based systems tended to get SRAM caches, whereas the 16 and 20 Mhz 286s rarely did.

    The 386SX not only cut down the data bus to 16 bits, it also cut down the address bus to 24-bits.

    The remarks on the i860 are bizarre. " ... it was nearly impossible to correctly list every instruction from beginning to end when compiling the program. " This is wrong, it was just very difficult to order the instructions very efficiently. Of course it had no problem listing them correctly, or the program wouldn't run. Intel tried it again with Itanium, and depended on the compilers to order instructions very efficiently, and also had difficulties.

    The author oddly left out the most significant part of the 486; it was the first pipelined x86 CPU, and that was a large part of the performance improvement.

    The Pentium's FPU was not 10x faster than the 486, but it was the biggest improvement. Unless you're comparing a very low clocked 486 to a high clocked Pentium. Clock per clock, it was not nearly 10x faster.

    Also MMX instructions were not related to the FPU, but were actually integer based.

    The first Pentiums actually ran at 66 MHz, but ran really hot, and they had yield problems, so sold 60 MHz Pentiums along with them, at a significant discount. Most people bought the 60 MHz because they were so much cheaper initially, but 66 MHz was out there.

    The Pentium MMX only reached 233 MHz as sold by Intel, not 300 MHz.

    The Pentium III (Katmai) was a Pentium II with SSE instructions added, nothing more. The nonsense about fewer pipelines and IPC improvements (outside of SSE code) is fabricated with regards to Katmai.

    The Celeron originally had no cache, but there was another version (not a Coppermine based) that had 128K cache on the processor. In some ways, it was faster than the Pentium III it was based on, because the cache was faster.

    Coppermine's cache was wider, and generally superior, but Celeron's were very competitive with Katmai based products, and were the favorites of overclockers.
  • mitch074
    Anonymous said:
    The Celeron originally had no cache, but there was another version (not a Coppermine based) that had 128K cache on the processor. In some ways, it was faster than the Pentium III it was based on, because the cache was faster..


    The Celeron 300A/333 you're referring to is the Mendocino core, wasn't based on Pentium III - it didn't have support for SSE. It can be seen as a precursor for Pentium III inasmuch as it had 128 Kb of level 2 non-inclusive low-latency cache - and it was the first Intel P6-family CPU to have that.
    It was soon replaced with the Coppermine core, which was indeed based on the Pentium III core (SSE) as it was a smaller die (smaller engraving) thus much cheaper to produce.
  • milkod2001
    Still on Haswell 4770k. Don't see any reasons to upgrade apart from m.2 SSDs maybe but have regular SATA SSD and there might be no difference in actual performance. Right?
  • Dymension
    OK, I believe the article says a history of INTEL chips, not AMD.
  • mitch074
    Anonymous said:
    OK, I believe the article says a history of INTEL chips, not AMD.


    Indeed - but it's strange how the jumps from 8 to 16-bit and then 16-bit to 32-bit got a whole lot of 'splaining, and then the jump to 64-bit is actually glossed over. That, and "AMD led the market for a short while" which lasted for several years and Intel got its collective butt handed to it:
    • faster than GHz frequency (K7)
    • DDR RAM support (K7)
    • fast double precision FPU (K7)
    • exclusive L2 cache (K7)
    • high-speed serial bus instead of FSB (K7: northbridge to southbridge, K8)
    • integrated memory controller (K8)
    • 64-bit (K8)
    • native dual core design (K8)


    Of course, if this appears on a "History of AMD CPUs", which does need an update since Ryzen came out, us AMD users won't complain as much :)