Cerb
Elite Member
- Aug 26, 2000
- 17,484
- 33
- 86
Pentium could be beat, not running recompiled/specialized code, by 486 variants and tweaks (AMD and IBM, not sure if Cyrix' were unique, or 486-derived).
Pentium division bug.
The PPro had poor performance outside of a 32-bit protected environment (NT4 flew on those things, though!).
Several cases of Pentium II and III cores coming late to market. Luckily, most scaled in clocks well enough to not be too big of a deal, and AMD was even later and slower with competition.
The i740. Like Itanium, it was a good implementation of a bad idea.
Rambus. Even discounting the Intel hubris issues of Netburst, anything without an i850E ($$$ RDRAM) wasn't much faster than a P3, and you were consistently better off with AMD. If you could afford a 760 board, Athlon XP was a no-brainer, once they were released. It took far too long for Intel to come out with good DDR chipsets, entirely due to contract issues.
Prescott. Not bad, but even lower IPC, and higher power per clock, when people clearly wanted less power, and greater IPC.
SB chipset SATAs.
In the opposite corner:
The entire series of K6 can be summed up by being too late. Hot, too, but it seems every one of them was late enough to market that they were just competitive by the time they got out in volume. Great value (I fondly and bitterly remember my K6-2 350.Voodoo2 rig), but very much stuck in the low-cost rut, like the current Athlon II CPUs are. Not only that, but thinking about the K6 ALi chipset mobos, right now, gets my stomach in a knot. I'd love to find a few and bash them in, Office Space printer style.
Athlons had issues, not only for delays and such, which weren't nearly the problem of the k6, as they performed well, and Intel had their own delays, but due to chipsets. You couldn't always find an AMD 750, and there were quirks to worry about with others. As time went on, 60mm fans cooling 50-70W CPUs proved to be a liability, as well.
Athlon XPs ended up with the same problem, while competitive. The SiS 735-748 were nice bright spots, but as the platform got old, they got harder to find. The same later became true until well into socket AM2's lifetime.
The Phenom bug hurt AMD's server share, then Intel came out with very nice Xeons. I've read rumors that there have also been anti-competitive practices involved in keeping the Phenom II gen Opterons from gaining a good foothold, but I'm not sure how true it is, given that Intel had no problems at all selling cheap 2-socket Xeons.
While not a late-to-market problem, the Cyrix-based AMD Geodes are real POSes. Throw that special random number generator on something better (like a Bobcat-based SoC?), and move on, please!
...and that's just off the top of my head, without getting into VIA and nVidia AMD or Intel chipsets (AMD/nV was often a lesser evil, but Intel/nV was just people being masochists for SLI), and I still wonder why SiS couldn't get the quirks out of their P4 chipsets, when the Athlon ones generally were flawless. So, I'm going to wait for launch, and see what pricing and performance are like. I have high expectations for server use, but not so sure about desktop and notebook BDs. There's no good reason they shouldn't be 30-50% faster than Phenom II, but that only barely gets them into Intel's current midrange, and it will be highly-tweaked productivity apps where the lower peak IPC could be really detrimental, as it will necessarily mean taking more clock cycles to execute code that can take advantage of bursts of >2 IPC (OTOH, high clocks and an excellent cache system could be enough to offset that, so you never know).
Pentium division bug.
The PPro had poor performance outside of a 32-bit protected environment (NT4 flew on those things, though!).
Several cases of Pentium II and III cores coming late to market. Luckily, most scaled in clocks well enough to not be too big of a deal, and AMD was even later and slower with competition.
The i740. Like Itanium, it was a good implementation of a bad idea.
Rambus. Even discounting the Intel hubris issues of Netburst, anything without an i850E ($$$ RDRAM) wasn't much faster than a P3, and you were consistently better off with AMD. If you could afford a 760 board, Athlon XP was a no-brainer, once they were released. It took far too long for Intel to come out with good DDR chipsets, entirely due to contract issues.
Prescott. Not bad, but even lower IPC, and higher power per clock, when people clearly wanted less power, and greater IPC.
SB chipset SATAs.
In the opposite corner:
The entire series of K6 can be summed up by being too late. Hot, too, but it seems every one of them was late enough to market that they were just competitive by the time they got out in volume. Great value (I fondly and bitterly remember my K6-2 350.Voodoo2 rig), but very much stuck in the low-cost rut, like the current Athlon II CPUs are. Not only that, but thinking about the K6 ALi chipset mobos, right now, gets my stomach in a knot. I'd love to find a few and bash them in, Office Space printer style.
Athlons had issues, not only for delays and such, which weren't nearly the problem of the k6, as they performed well, and Intel had their own delays, but due to chipsets. You couldn't always find an AMD 750, and there were quirks to worry about with others. As time went on, 60mm fans cooling 50-70W CPUs proved to be a liability, as well.
Athlon XPs ended up with the same problem, while competitive. The SiS 735-748 were nice bright spots, but as the platform got old, they got harder to find. The same later became true until well into socket AM2's lifetime.
The Phenom bug hurt AMD's server share, then Intel came out with very nice Xeons. I've read rumors that there have also been anti-competitive practices involved in keeping the Phenom II gen Opterons from gaining a good foothold, but I'm not sure how true it is, given that Intel had no problems at all selling cheap 2-socket Xeons.
While not a late-to-market problem, the Cyrix-based AMD Geodes are real POSes. Throw that special random number generator on something better (like a Bobcat-based SoC?), and move on, please!
...and that's just off the top of my head, without getting into VIA and nVidia AMD or Intel chipsets (AMD/nV was often a lesser evil, but Intel/nV was just people being masochists for SLI), and I still wonder why SiS couldn't get the quirks out of their P4 chipsets, when the Athlon ones generally were flawless. So, I'm going to wait for launch, and see what pricing and performance are like. I have high expectations for server use, but not so sure about desktop and notebook BDs. There's no good reason they shouldn't be 30-50% faster than Phenom II, but that only barely gets them into Intel's current midrange, and it will be highly-tweaked productivity apps where the lower peak IPC could be really detrimental, as it will necessarily mean taking more clock cycles to execute code that can take advantage of bursts of >2 IPC (OTOH, high clocks and an excellent cache system could be enough to offset that, so you never know).
Last edited:
).
Facebook
Twitter