Worst CPUs ever, now with poll!

[Dresdenboy]

Different question:

What is the worst software making good CPUs run bad?

 

[DrMrLordX]

Gah! Start your own thread for that. And get some fire extinguishers, 'cuz it's gonna be a flame war.

 

[Genx87]

Gee, and no one wants to mention the cacheless celerons? shame on you nerds

though in the areas of "oops" which I think would make an honourable mention would be the Pentium D and it's hype of being a dual core when it was just two high powered cpus in the same package (sharing the FSB just like normal dual cpu units).

or the pentium bug that intel down played as being "pointless" to the masses.

Heh I was coming in to mention the first Celeron without L2 cache. It is how celeron got the de-celeron label. Of course Intel did follow it up with the 300 that overclocked to 450. That was a great processor, maybe one of the greatest of all time.

Edit: Oh son of a necro.

 

[Dresdenboy]

Gah! Start your own thread for that. And get some fire extinguishers, 'cuz it's gonna be a flame war.

Sorry, but you misunderstood this question. There's always a link between hardware and software. There might be good designs, which miss expectations due to bad compilers, OS or simply applications. Transmeta CPUs needed translation software. Itanium and Bulldozer also needed the right compilers or simply applications using the right optimization (not for some other or generic x86 target).

 

[DrMrLordX]

Yeah, I know what you're asking. It's just that if you bring up that topic and somebody mentions AMD, 99.9% of the time you're going to start a huge fight over FX, Intel compilers, and all kinds of other vaguely-related stuff. It gets ugly.

Ignoring AMD products, I'd say Intel's Itanium takes the cake since it probably represents one of the few products Intel has ever released for which software support was not (eventually) robust and adequate. Their compilers normally support their own hardware nicely and their drivers are usually stable and functional (certain graphics drivers notwithstanding). Itanium was hung out to dry. It never reached the desktop where it (allegedly) was supposed to wind up eventually, and the compiler/OS support it would have needed to get there never materialized.

Software support for it might have been adequate in the markets where it saw use, but it was pretty-well bottled-up to a niche market. IA64 got plowed under by x86-64.

 

[ElFenix]

Yeah, I know what you're asking. It's just that if you bring up that topic and somebody mentions AMD, 99.9% of the time you're going to start a huge fight over FX, Intel compilers, and all kinds of other vaguely-related stuff. It gets ugly.

Ignoring AMD products, I'd say Intel's Itanium takes the cake since it probably represents one of the few products Intel has ever released for which software support was not (eventually) robust and adequate. Their compilers normally support their own hardware nicely and their drivers are usually stable and functional (certain graphics drivers notwithstanding). Itanium was hung out to dry. It never reached the desktop where it (allegedly) was supposed to wind up eventually, and the compiler/OS support it would have needed to get there never materialized.

Software support for it might have been adequate in the markets where it saw use, but it was pretty-well bottled-up to a niche market. IA64 got plowed under by x86-64.

i860 maybe? intel put out a competitive risc solution about the same time as the 486, but was noncommittal about its future.

 

[svenge]

Ignoring AMD products, I'd say Intel's Itanium takes the cake since it probably represents one of the few products Intel has ever released for which software support was not (eventually) robust and adequate. Their compilers normally support their own hardware nicely and their drivers are usually stable and functional (certain graphics drivers notwithstanding). Itanium was hung out to dry. It never reached the desktop where it (allegedly) was supposed to wind up eventually, and the compiler/OS support it would have needed to get there never materialized.

Software support for it might have been adequate in the markets where it saw use, but it was pretty-well bottled-up to a niche market. IA64 got plowed under by x86-64.

In my mind, that makes Bulldozer even worse, because there was already a plethora of x86/x64 software already in existence when AMD was designing that micro-architecture.

It's one thing to consciously go into completely unknown territory and fail, it's another thing to fail when the territory is already well-understood.

 

[DrMrLordX]

i860 maybe? intel put out a competitive risc solution about the same time as the 486, but was noncommittal about its future.

I had forgotten about that (and the i960). Hmm! I guess it's really a matter of which CPU consumed more R&D resources (relative to what was available) and which was imagined to be more important to Intel's future product lineup.

IA64 was supposed to replace x86 pretty much everywhere and was Intel's original plan for moving from 32-bit to 64-bit. I'm not really sure what the i860 represented in term's of Intel's long-term plans back in the late 80s.

 

[jhu]

Yeah, I know what you're asking. It's just that if you bring up that topic and somebody mentions AMD, 99.9% of the time you're going to start a huge fight over FX, Intel compilers, and all kinds of other vaguely-related stuff. It gets ugly.

Ignoring AMD products, I'd say Intel's Itanium takes the cake since it probably represents one of the few products Intel has ever released for which software support was not (eventually) robust and adequate. Their compilers normally support their own hardware nicely and their drivers are usually stable and functional (certain graphics drivers notwithstanding). Itanium was hung out to dry. It never reached the desktop where it (allegedly) was supposed to wind up eventually, and the compiler/OS support it would have needed to get there never materialized.

Software support for it might have been adequate in the markets where it saw use, but it was pretty-well bottled-up to a niche market. IA64 got plowed under by x86-64.

While true, what if the Itanium family was instead used to remove competition instead? When Itanium was announced, the end-of-life roadmap for PA-RISC and Alpha were assured. In that respect, it was one of the best processors, but not due to its processing capabilities.

 

[Dresdenboy]

In my mind, that makes Bulldozer even worse, because there was already a plethora of x86/x64 software already in existence when AMD was designing that micro-architecture.

It's one thing to consciously go into completely unknown territory and fail, it's another thing to fail when the territory is already well-understood.

Of course, but ...

Designing architectures to work best with legacy code somehow prevents them from improving significantly. It's always a trade off.

 

[StinkyPinky]

I think I've posted in this thread before, but IMO it's the early pentium 4's. They weren't any better than p3's, used weird ram, and weren't really up to running the software of the day. Until a few years ago we had some at work and those machines were just sloooow. Unusable almost.

Maybe the reason people look at the Core 2's so fondly is because the P4's were so bad (even the later gen versions weren't that good).

 

[Centauri]

I think I've posted in this thread before, but IMO it's the early pentium 4's. They weren't any better than p3's

They were actually worse, and at significantly higher clock speeds (think 50%+). The people who complain about how much of a step back Bulldozer was on IPC weren't around for the Pentium 4 launch, I imagine.

 

[DrMrLordX]

While true, what if the Itanium family was instead used to remove competition instead? When Itanium was announced, the end-of-life roadmap for PA-RISC and Alpha were assured. In that respect, it was one of the best processors, but not due to its processing capabilities.

Ha ha! Ouch. Yeah. Poor Alpha.

I think someone, somewhere within Intel really wanted IA64 to be the ISA "of the future", regardless. It doesn't seem that Intel is much worse for its failure.

Of course, but ...

Designing architectures to work best with legacy code somehow prevents them from improving significantly. It's always a trade off.

Well, also, consider the position of any design firm that is #2 on the market (or lower). If you're constantly designing against legacy code, you are probably going to be reverse-engineering the market leader's processors from one or more generations ago (assuming you can do so legally). That's a fine niche for those that can occupy it, but it isn't like the mid 90s or earlier when x86 clone companies could crank out socket-compatible knockoffs. Things are different now.

 

[Sheep221]

VIA C3 Nehemiah - so far the worst CPU I've ever worked with, everything in poll is actually much much better than this one

 

[dark zero]

VIA C3 Nehemiah - so far the worst CPU I've ever worked with, everything in poll is actually much much better than this one

Actually the Atom N450 and the AMD E1-2100 are by far, worse than the C3.

Adding a new example I am starting to think that Braswell ended into a massive flop from Intel after advancing great with Bay Trail.

 

[waltchan]

Actually the Atom N450 and the AMD E1-2100 are by far, worse than the C3.

Adding a new example I am starting to think that Braswell ended into a massive flop from Intel after advancing great with Bay Trail.

AMD C-50 is 50% worse than E1-2100. Try it...

Celeron J1800 and Pentium J2900 are the fastest single-thread speed (2.41GHz) sub-10W processors we'll see for a while.

 

[dark zero]

They were actually worse, and at significantly higher clock speeds (think 50%+). The people who complain about how much of a step back Bulldozer was on IPC weren't around for the Pentium 4 launch, I imagine.

BD was a dissaster, but Vishera practically was a Real evolution, Sadly Vishera seems that was figthing Nehalem instead of Sandy.
Meanwhile the BD APU IS a real dissaster. Llano APU wasn't as bad as many thought. AMD shouldn't went to the CMT, even without SMT, the moar cores were effective.

 

[soccerballtux]

While true, what if the Itanium family was instead used to remove competition instead? When Itanium was announced, the end-of-life roadmap for PA-RISC and Alpha were assured. In that respect, it was one of the best processors, but not due to its processing capabilities.

that's what made it interesting to me. I think eventually we would have figured out, granted after insane compiler development, and it could have turned out to be awesome. But IDK.

 

[dark zero]

that's what made it interesting to me. I think eventually we would have figured out, granted after insane compiler development, and it could have turned out to be awesome. But IDK.

Or maybe Itanium wasn't well built at the beginning.
Poor Alpha. It wasn't so bad...

Another competitor... Any common MIPS processors.. So slow that makes you enter on despair

 

[skipsneeky2]

I think I've posted in this thread before, but IMO it's the early pentium 4's. They weren't any better than p3's, used weird ram, and weren't really up to running the software of the day. Until a few years ago we had some at work and those machines were just sloooow. Unusable almost.

.

Rambus socket 423 Pentium 4+478 adapter=winning:awe: I think some Northwood chips as fast as 2.6ghz ended up being compatible?

 

[pyjujiop]

They were actually worse, and at significantly higher clock speeds (think 50%+). The people who complain about how much of a step back Bulldozer was on IPC weren't around for the Pentium 4 launch, I imagine.

Worse is an understatement. A Willamette P4-1400 was about as fast as a P3-900. The first P4's were dreadful. Once Intel beefed up the cache, better compliers were written, and the clock speeds went past 2 GHz, it was a competitive part, but still was terribly inefficient. I voted for the Itanic in the poll, but I would have voted for Netburst as the worst CPU architecture. Even as bad as AMD K15h was when it first came out, it looks like masterful execution compared to early Netburst. Intel even sold Willamette systems with SDRAM memory for a while, because of the Rambus fiasco, which took away the only redeeming quality of the first P4's, their enormous memory bandwidth.

 

[mizzou]

My old cyrix 686 was such a let down. I remember it performed quite well, but the day I switched to the slot-style AMD Athlon I felt for the first time what a real upgrade was like.

It was interesting tech and had good performance, it was the FPU I think that really did it in.

 

[dark zero]

I revive this topic that I can add one competitor: Snapdragon 810.

Despite being sold as a high end processor, it performed and overheated to badly that performed like a mid tier one.

 

[superstition]

BTW why is PowerPC 970 on the list?

Because the list is poor-quality. That this thread has received so much attention is unfortunate.

 

[superstition]

Ok, Intel is a monster now but I think if it wasn't for the huge revenue they got from IBM's pretty poor decision to chose 8088 (to think the 68000 was around back then and had 32-bit data and address registers) and they would likely be much much smaller today.

1. 8086. For all those who have struggling with 64KB segmented memory, DOS memory managers, 640KB limits (and millions did) there can be no other #1

The 8086 was just outdated when IBM picked it. It's not Intel's fault that IBM chose an outdated CPU instead of the 68000.