Worst CPUs ever, now with poll!

[DrMrLordX]

The Celeron 300a was one of the best consumer CPUs ever produced, at least in terms of bang/buck. No way I would take a k6-III over one of those things.

 

[crashtech]

So long ago, but I seem to remember having one of those Celeron 300a's running at like 450 MHz. It was pretty much the chip to have for the budget-conscious adventuresome PC fanatic.

 

[Panino Manino]

This list seems strange to me.
Was the 286 bad?
PowerPC 60x wasn't decent? And the 970 "popular"?
I really don't understand.

 

[dark zero]

Ok guys, I saw that it passed more than 2 years and I want to add some new additions...

1. Snapdragon 8 Gen 1. Heck, what the hell Samsung did with that chip? In paper it looks OK until you see how Samsung fab is so badly done with 4 nm. And it left me question if Intel 14 nm is better than it.

When switched with TSMC with the 8+ Gen 1 the performance uplift was insane.

2. Exynos series in the latest years, sorry, but lately near all the Exynos chips were a thermal mess. It shows how badly Samsung screwed up.

 

[DAPUNISHER]

Itanic is winning by a landslide.

 

[aigomorla]

Itanic is winning by a landslide.

I dont understand how people can say it was so bad when less then what 5% of the people had them.
And when they did have them, they were very specific needs / use cpu's.

And why isn't the Smithfield / Prescott (Pentium D) on this list.
It was truely the worst cpu ever without even a competition.

Also missing on this list is the Cyrix M2 + Via C2.
We can also say Duron's were also pretty much straight up dangerous, as they would start fires.

 

[moinmoin]

I dont understand how people can say it was so bad when less then what 5% of the people had them.

I would say that's exactly the thing: Itanium was Intel's next best thing at the time and was supposed to replace x86. People simply didn't want it, as a result it failed to be Intel's next best thing and didn't make a dent into x86.

 

[Thunder 57]

AMD Family 10h (better known as K10) has no business being on this list while the P4 Williamette and Prescott are suspiciously missing

The original version had some issues but the revised die shrink was quite competetive. And it did very well in servers as Intel until intel dropped the FSB. Intel also copied the cache structure in future designs using a small L2 and L3 as LLC.

 

[Markfw]

AMD Family 10h (better known as K10) has no business being on this list while the P4 Williamette and Prescott are suspiciously missing

The original version had some issues but the revised die shrink was quite competetive. And it did very well in servers as Intel until intel dropped the FSB. Intel also copied the cache structure in future designs using a small L2 and L# as LLC.

The K5 should not be there either IMO. My memory is a little frayed from that time, but I thought the Intel chips of the time that were as fast were a lot more expensive. And yes, The Pentium D was like the worst Intel.

The AMD 10h and 15h were not great compared to Intel at their time, but a lot cheaper. So again, should not be on the list. Or the 80286, it was great for its time.

This list is messed up.

 

[Mopetar]

Ok guys, I saw that it passed more than 2 years and I want to add some new additions...

1. Snapdragon 8 Gen 1. Heck, what the hell Samsung did with that chip?

Snapdragon was Qualcomm. They had some impressive chips under that branding, but I remember they eventually had a bad design where the performance was awful and they got out of the custom ARM core business for a while since the ARM design all the Chinese companies were using was performing better. The acquisition of the Nuvea team has put them back in the game though.

I dont understand how people can say it was so bad when less then what 5% of the people had them.
And when they did have them, they were very specific needs / use cpu's.

It mostly comes down to it being an ambitious design that had a lot of things that didn't pan out. A big issue was that relied heavily on the compiler to be able to produce VLIW instructions that could be run in parallel unlike the x86 designs that would try reordering instructions to extract greater parallelism regardless of what the compiler had spit out. Depending on magical compilers is probably historically a bigger tech blunder than chasing clock speeds, but less we'll know.

Although Itanium was very powerful on paper and theoretically very efficient because it didn't have to include all of the usual superscalar architectural bits do handle reordering instructions, in practice people struggled with writing code to leverage those capabilities. It was also originally hyped as an x86 replacement that would be able to emulate x86 code and run it faster than native x86 CPUs. This did not work out at all.

Because Intel was convinced that Itanium was the future, they decided not to bother creating a 64-bit instruction set for x86, which is why AMD beat Intel to the punch there and basically threw egg on their face by making Intel adopt AMD's specification. Intel also killed off the DEC Alpha because they were believing their own hype about everything else being pointless. HP, who was working with Intel on the project, also killed off their own PA-RISC line because they bought into the hype. MIPS faded away around this time due to all of the FUD being thrown around.

Itanium wasn't just a bad chip, but a bad chip that had a lot of negative impacts for the market as a whole as well as for Intel itself. There are other chips that have been awful for various reasons, but most of them failed quietly in a corner or limited the misery caused to the company that produced it and a small number of consumers unfortunate enough to buy them. The disaster that became known as the Itanic was felt by so many different companies across the industry that it truly is the worst CPU of all time.

 

[gdansk]

Snapdragon was Qualcomm. They had some impressive chips under that branding, but I remember they eventually had a bad design where the performance was awful and they got out of the custom ARM core business for a while since the ARM design all the Chinese companies were using was performing better. The acquisition of the Nuvea team has put them back in the game though.

No doubt they mean what did Samsung foundry do to mess up the chip. The TSMC refresh was much better.
But I don't think it was an exceptionally bad, personally. Just worse than it should have been. It wasn't a doomed architecture like Itanium/iAPX432 and didn't nearly kill Qualcomm like Bulldozer nearly killed AMD.

 

[Mopetar]

Didn't Qualcomm axe their in-house development team though? They were building their own custom core for a while and were generally pretty competitive with Apple for a time, but at some point they had hot garbage for product and killed their custom design and went back to the ARM designed core.

I don't think there was much risk of killing the company because their foray into SoCs was a side business and having the cellular modem integrated was a good deal for many vendors, but running a chip development team isn't cheap and if your offerings are inferior to the design ARM is selling, it makes your integrated product less attractive by comparison.

It was still a financial loss though. Maybe not as dire as Bulldozer was for AMD, but AMD didn't have nearly as much of a fallback so any failure was always going to be more perilous for them.

 

[positivedoppler]

I dont understand how people can say it was so bad when less then what 5% of the people had them.
And when they did have them, they were very specific needs / use cpu's.

And why isn't the Smithfield / Prescott (Pentium D) on this list.
It was truely the worst cpu ever without even a competition.

Also missing on this list is the Cyrix M2 + Via C2.
We can also say Duron's were also pretty much straight up dangerous, as they would start fires.

Cyrix M2 was awesome....cheap and got me on the internet. It was like 1/4 price of a pentium if you didn't care about games.

Cyrix M-II 300

www.anandtech.com

"For the price, the M-II 300 can't be beat as a processor for business applications. As a gaming platform, you're better off spending the additional $65 and getting a K6-2 as the M-II will put you through more frustration during gameplay than you can imagine. Cyrix has come a long way since the original 6x86 PR-120+ hit the streets, running at a meager 100MHz clock speed, that was the processor that started the 6x86 generation and it is the reason the M-II is here today."

 

[Storm-Chaser]

Why have you not listed the FX as one of the worst processors? That one is right near the top of the list.

 

[Thunder 57]

Why have you not listed the FX as one of the worst processors? That one is right near the top of the list.

AMD Family 15h is FX. The poll would be easier for most people if they had used K10 instead of AMD Family 10h and FX/Bulldozer instead of AMD Family 15h.

 

[Insert_Nickname]

Cyrix M2 was awesome....cheap and got me on the internet. It was like 1/4 price of a pentium if you didn't care about games.

It wasn't a bad chip at all. It just got compared to superior chips because of the somewhat unfortunate naming scheme. Compared to Pentium/K6 at similar speeds, it wasn't half bad.

The whole Quake thing was way overblown IMO, and had more to do with optimization then performance.

 

[positivedoppler]

It wasn't a bad chip at all. It just got compared to superior chips because of the somewhat unfortunate naming scheme. Compared to Pentium/K6 at similar speeds, it wasn't half bad.

The whole Quake thing was way overblown IMO, and had more to do with optimization then performance.

back in the days, some complained the M2 ran too hot...i think that was like a under 20 watt processor? ultra mnobile by today's standard

 

[igor_kavinski]

I personally was AFRAID to have Itanium succeed coz articles I read about it suggested that if Intel abandoned x86 in favor of Itanium's ISA, we would all end up running our x86 software in emulation mode. That was a nightmare scenario for me at the time. My entire game collection was at risk of going obsolete!

 

[Abwx]

Why have you not listed the FX as one of the worst processors? That one is right near the top of the list.

Certainly not the worst, and by far...

The last reviews from Computerbase including this CPU are telling, it ended being faster than a i5 2500 and matching a i7 2600K in games while outperforming both in most MT benches, and if multitasking was added in the balance the i5/i7 were vastly outperformed.

 

[DAPUNISHER]

Certainly not the worst, and by far...

The last reviews from Computerbase including this CPU are telling, it ended being faster than a i5 2500 and matching a i7 2600K in games while outperforming both in most MT benches, and if multitasking was added in the balance the i5/i7 were vastly outperformed.

Honestly I think FX is an example of what we are currently witnessing. That is, the OS needed to catch up with the CPU tech. Both AMD and Intel are having issues because Windows can't fully leverage their new tech. It was the same with FX. Win7 was holding it back as much as most games of the time.

The downside, and why it deserves at least a dishonorable mention for worst CPUs, is iterations of it were all we got for 5yrs. I dig my FX 8350, but the overclocked 4770K that I had, would take it behind the woodshed. Of course it was over $200 more than the 8350 when it launched, so there's that.

 

[positivedoppler]

Honestly I think FX is an example of what we are currently witnessing. That is, the OS needed to catch up with the CPU tech. Both AMD and Intel are having issues because Windows can't fully leverage their new tech. It was the same with FX. Win7 was holding it back as much as most games of the time.

The downside, and why it deserves at least a dishonorable mention for worst CPUs, is iterations of it were all we got for 5yrs. I dig my FX 8350, but the overclocked 4770K that I had, would take it behind the woodshed. Of course it was over $200 more than the 8350 when it launched, so there's that.

I wonder if there's any recent benchmark that revisits FX 8350 vs Ivy Bridge i3 to i7. I recall when both came out, the entire ivy lineup was faster in gaming. Gaming is not just about raw FPS though. I had FX 6100 which I finally give up on when playing Path of Exile, game notorious for throwing hundreds of units and excessive effects on screen. The FX 6100 could not keep up and the game constantly freezes for long periods when the screen floods with objects. When I finally upgraded to my 5900X using the same video card, game became smooth even if I ran it with using 4 cores or less. So my perception of the entire FX series is that it's pretty bad...but I do wonder if more modern OS helps.

 

[dark zero]

No doubt they mean what did Samsung foundry do to mess up the chip. The TSMC refresh was much better.
But I don't think it was an exceptionally bad, personally. Just worse than it should have been. It wasn't a doomed architecture like Itanium/iAPX432 and didn't nearly kill Qualcomm like Bulldozer nearly killed AMD.

Funny story despite Snapdragon 888 was a mess, it was the 8 Gen 1 the one who let it lose some terrain and allowed Mediatek to pull some impressive chips like the Dimensity 8100 (still waiting for a phone that fully uses the chip with Desktop mode) and 9000.

 

[DAPUNISHER]

I wonder if there's any recent benchmark that revisits FX 8350 vs Ivy Bridge i3 to i7. I recall when both came out, the entire ivy lineup was faster in gaming. Gaming is not just about raw FPS though. I had FX 6100 which I finally give up on when playing Path of Exile, game notorious for throwing hundreds of units and excessive effects on screen. The FX 6100 could not keep up and the game constantly freezes for long periods when the screen floods with objects. When I finally upgraded to my 5900X using the same video card, game became smooth even if I ran it with using 4 cores or less. So my perception of the entire FX series is that it's pretty bad...but I do wonder if more modern OS helps.

The modern OS did indeed help. RA Tech has a number of FX videos. I will add one at the end of this post, of the FX 8350 v i5 3470. I also have a long thread here on the FX 8350 if you want to wade through or skim some of it. Includes Philscomputerlab revisit of it.

I have the FX 6100 and it aged better than the i3s of the era, that's for certain. Not too mention it overclocks a bit. Mine only does 4.2Ghz in a 990FX board. Still, 6 threads turned out better than a 2 core with HT for modern usage. RA Tech has vid on it. As I get into in that long thread, it was better even for some more contemporary games, but how reviewers test would never show it. Rich from Digital Foundry did a vid on Witcher 3 showing how even the i5 of the time, Haswell I think? was having frame pacing issues in Novigrad where the 8350 was doing much better. He pointed out how the FX did better in CPU demanding scenes of Assassin's Creed Syndicate than the i5 too.

I tested my FX8350@4.6GHz with DDR3 2133 in Witcher 3 vs Ryzen 3200G and the 8350 was a better experience in Novigrad and Toussaint. Anytime there is a bunch of A.I. collision detection, etc. the 3200g would max out at times, which meant frame pacing issues. The 8350 was a much smoother experience.

There a situations where the poor IPC of the FX cannot be overcome, but again, vs the CPUs that were in its price range, it has held up really well. Another thing RA Tech did well is show how bad the benchmark data from GN and HUB is. They don't sit and watch and listen to the bench runs. They take the logs and compile and post the data. A good example is the i3 they claimed was better than FX, during the Shadow of the Tomb Raider bench, had texture pop in and rendering problems. Completely failed to render NPCs, and had audio issues. None of which show up on the final bench results. You have to watch and listen to observe it.

So yeah, I agree 100% it ain't all about raw FPS. Because in my experience, and some others, that is where the FX was better than the bar charts sometimes made it look. While at times, making the i3 and i5 look better than they actually were. i5 craps the bed way more often than the 8350 in BF5 64 player, as RA shows. You weren't going to see that pointed out years back by reviewers.

A few more comments: I bought the FX 8350 for $90 in 2015, that is less than the cheapest i3 cost at the time. The Ivy and Sandy i7 aged magnificently. Truly the best bang for buck mainstream CPUs ever. Nothing else was ever as relevant for as long as they have been. You can still play everything out there with them, with compromises. Where as the i3 and i5 lack enough threads for some newer games running a modern OS. Hell they were already struggling in some games as I pointed out above.

 

[Mopetar]

Another thing RA Tech did well is show how bad the benchmark data from GN and HUB is. They don't sit and watch and listen to the bench runs. They take the logs and compile and post the data. A good example is the i3 they claimed was better than FX, during the Shadow of the Tomb Raider bench, had texture pop in and rendering problems. Completely failed to render NPCs, and had audio issues. None of which show up on the final bench results. You have to watch and listen to observe it.

No one with a large test suite and an automated process is going to sit and watch every single CPU and every single game.

There's also no good objective way to measure phenomena like that either. Maybe you can report the observation, but what's the reference point and how do you ensure it's being reported systematically.

Bulldozer was certainly disappointing, but I don't know if I'd consider it any worse than Prescott. AMD at least priced it competitively for what it could do and perhaps the biggest complaint against it would be the power draw for anyone trying to hit the obscene clock speeds the architecture was designed for. The extra cores did help it age a bit more gracefully than the Pentium 4 did at least.

 

[DAPUNISHER]

No one with a large test suite and an automated process is going to sit and watch every single CPU and every single game.

That's the point. Hence, their numbers/bar graphs mean diddly crap with CPUs that have become janky due to OS and gamer habits. It's lazy quick and dirty testing for clicks. That it plays to their confirmation bias is the perfecta. Stating a Pentium 5500 is better than an FX 8350 is fake news, and the result of the aforementioned lazy testing.

And forget aging more gracefully than Presshot. It aged more gracefully than similarly priced Intel CPUs.