Worst CPU's of all time?

Stugots

Supreme [H]ardness
Joined
Feb 25, 2004
Messages
7,234
Inspired by a thread of the opposite! I'll go first

Willamette Pentium 4 1.3ghz - The slowest version of the worst version of the Pentium 4!

I personally am really fond of the old Pentium 4's, but the only ones I ever touched were between Northwood to Ceder Mill. I thankfully skipped over Willamette because I stretched my old 440BX with a Tualatin Celeron [email protected] for way too long and upgraded into an i875 with a Pentium 4C [email protected].
 
Worst is a very subjective adjective. I would say that the fact that Bulldozer was slower than Phenom II has to put it in the top 5. Did it not work at all? No. But IMO a successor CPU should never be slower than its predecessor MHz for Mhz.
 
I really did not like my AMD K6-2, its been a while so i dont remember much other than i hated that pc ever since i bought it.
 
K6-2 was actually pretty good!
The K5 OTOH, had WORSE FPU than Cyrix PR series. But for office only stuff was an OK beater chip. Combined with no name EDO, Amptron board, et al you had a satisfactory office system. Just don't ask about those horrific JTS champ hard drives! More like JTS brown (in name of Cleveland Browns haha)...

Bulldozer, haha I remember how the AMD fanboys were saying they were going to *destroy* Nehalem, Sandy Bridge and Westmere. LOL
Never was a fan of Duron either but kind of like the K5, beater status.

Socket 7 had a bunch of duds too. Remember IDT Winchip? Rise MP6?

Let's not forget the Pentium overdrive! I swapped out over 300 486 DX/33s in Dell slim desktop formats for these. Ran at 83MHz. Quite a bump in performance particularly FP. About as fast as a native Pentium 66 machine.

What was surprising was the Pentium II overdrive for socket 8 Pentium Pro machines. Extended the life out of quite a few Tyan and Supermicro based workstations with those.

There were other competing products (Power Leap IIRC) that allowed you to plug a socket 7 CPU into a socket 3 i486 board and those were less than interesting as most things designed to save money and keep hardware going that really should be sent to the scrapyards. ;-)
 
Inspired by a thread of the opposite! I'll go first

Willamette Pentium 4 1.3ghz - The slowest version of the worst version of the Pentium 4!

I personally am really fond of the old Pentium 4's, but the only ones I ever touched were between Northwood to Ceder Mill. I thankfully skipped over Willamette because I stretched my old 440BX with a Tualatin Celeron [email protected] for way too long and upgraded into an i875 with a Pentium 4C [email protected].
P3 Tualatin was a SOLID processor.

Early P4's woooo bad news I agree there.. lol
 
The worst one I ever had was an AMD K6 I bought in summer 1997. It had a bug that would occasionally crash the system if you had more than 32MB of ram installed. I had 64MB, so splat. Thankfully I found out about it in time and was able to RMA it and get a Pentium 200MMX. I just had to change a few jumpers when I swapped the AMD proc for an Intel one.
 
Wasn't Prescott a goddamn disaster? and Intel having to market recall their 1ghz chip wasnt great either
 
Idk. Never had a bad cpu. Didn't have many, though. Q6600>FX6300>Ryzen 1600>3600>5600X>7900X

Someone might say "But that piledriver chip!" and I'd probably agree if it wasn't a $80-$90 budget champ at the time when quads were still well above $200.
The original Bulldozer was a total disappointment when it came out performance-wise, but it actually worked properly so I can't put it on a "worst CPUs" list. I had one at work for a while and for what I was doing at work it was actually a pretty good proc for what I was doing with it. Lousy for gaming, but 8 integer cores were better for work stuff than a quad core Intel chip at the time. Later variations on that architecture such as piledriver were merely as disappointing performace-wise as expected, so, again, not bad enough to go on a worst CPU's list by the standard of anyone who's ever had a buggy CPU that causes BSODs and/or kernel panics.
 
If we're not counting the 8086 line in general as bad (because it killed superior architecture chips from other manufacturers due to the sheer inertia of the IBM and clones) then the original Pentiums with the bad FP math units. RIP Motorola 68K, 6809, Zilog Z80 and MOS 6502 CPUs, the world is often an unfair place...

I see a lot of K6 series CPU hate - I really believe a lot of that came down to the huge amount of crappy motherboards out there at the time. I never had trouble with any K6-2 or K6-3 CPUs, but I was also using Tyan Super Socket 7 motherboards. The only thing I had to worry about was making sure I kept the VIA chipset drivers for that platform up to date. K6 and lower-end K6-2 CPUs were also extremely reliable on Intel chipset boards - the K6-2-266 in particular ran very cool.
 
I had good experiences with ever Celeron I ever ran. But granted the last Celeron I messed with was a Celeron D 320.

Celeron 433@590
Celeron 633@1000
Celeron 600@1035
Celeron 1100@1260 (Coppermine)
Celeron 1300@1600 (Tualatin)
Celeron D 320 [email protected]

With the exception of the 1.1ghz Coppermine Celeron I had, they all overclocked great!
 
It's all relative.
The original Bulldozer was a total disappointment when it came out performance-wise, but it actually worked properly so I can't put it on a "worst CPUs" list. I had one at work for a while and for what I was doing at work it was actually a pretty good proc for what I was doing with it. Lousy for gaming, but 8 integer cores were better for work stuff than a quad core Intel chip at the time. Later variations on that architecture such as piledriver were merely as disappointing performace-wise as expected, so, again, not bad enough to go on a worst CPU's list by the standard of anyone who's ever had a buggy CPU that causes BSODs and/or kernel panics.
This. It was not good compared to the competition, and arguably mediocre at best compared to its predecessor, but it worked ok (minus the crappy 990FX boards that were out there), and if you REALLY needed 8 cores for cheap - it did the job. I used one for almost 9 years for certain tasks. It was... functional. Can't complain about it. Got it for $80 (8350).

K6 chips were (as several pointed out) all based on the board you had. I used several good ones, and I had some BAD SiS boards too (and some that I eventually learned the tricks on).
 
This was my first pc ever. Pentium 166mhz and 16MB of rams(Yes it is MB not GB). My parents bought it for retail at $3k at the time too. Mechwarriors 2 that came with the PC bundle became my favorite game for the longest time.

SonyOld.png
 
Inspired by a thread of the opposite! I'll go first

Willamette Pentium 4 1.3ghz - The slowest version of the worst version of the Pentium 4!

I personally am really fond of the old Pentium 4's, but the only ones I ever touched were between Northwood to Ceder Mill. I thankfully skipped over Willamette because I stretched my old 440BX with a Tualatin Celeron [email protected] for way too long and upgraded into an i875 with a Pentium 4C [email protected].
I decided to try an AMD server several years ago in my datacenter. That poor thing was a dog, but I can't recall the chip :(
 
And the 200 was also great. I was building in that era. Good chips both.

There was nothing wrong with either the P166 or P200 chips, I just remember that they were both rather quickly invalidated by the MMX versions of the chips... which had more cache on them. Even back then, the cache on the CPU was a big deal for performance. The K6-3 CPUs were probably the top aces with that. You got amazing performance with those K6-3's IF you had a good motherboard and did not mind the hit in FP performance vs the Pentium II and III CPUs of the time, with Quake games being the big beneficiaries of that FP performance.
 
I'd argue that the x86 architecture in general wasn't even good until the 80486DX2, maybe 80386. 640 KB conventional RAM addressing limit and half-assed 80286 protected mode, anyone?

The rival Motorola 68k was generally superior (note its use in the Macintosh, Amiga, Atari ST, early UNIX workstations like NeXT, SGI IRIS or Sony NEWS), yet IBM chose not to use it for the PC, and as it's the IBM PC that got ruthlessly cloned, that was what Intel needed to get a leg up on everyone else despite a fundamentally crappy design early on, to the point that they could ramp up the 80486DX2's clock speeds well past what the 68040 could reliably accomplish.

Beyond that, I can't think of anything that stands out as particularly bad for x86, besides Intel's later NetBurst offerings (mainly Willamette and Prescott) and the Itanic - I mean Itanium/IA-64. AMD dropped the Hammer on them with Opteron/Athlon 64 and probably would've overtaken the market if not for Intel's anticompetitive practices with the major PC OEMs. (If anything, EPYC is the current-day Opteron for its devastating effect on Intel's big iron market share.)

Too bad AMD had their own NetBurst moment in the form of Bulldozer/FX - and you know it's bad when the outgoing Phenom II had better performance in certain workloads and Intel imposed 4C/8T stagnation for so long until they finally got Zen 1/Ryzen/Threadripper/EPYC out the door. Just look at AMD's stock prices from the FX era to the Ryzen era, and the relative market share - that sums it up right there.

As for non-x86, I'd like a word with the now-defunct Sun Microsystems as to why the hell a dual-1.015 GHz UltraSPARC III Cu workstation like the Blade 1000/2000 manages to feel like a total slog compared to a single 400 MHz MIPS R12000 SGI Octane (which can go up to dual 600 MHz R14000s), let alone a Pentium 4/Athlon XP-era PC or a dual G4 Power Mac while costing several times the price when new. Maybe it's Solaris 10 earning that Slowaris reputation.

IBM also NetBursted themselves with the PowerPC 970 G5 - a major reason why the AIM alliance collapsed and Apple went to Intel until recently. Big iron POWER apparently still survives today, but I don't know of anyone using it over x86-64 or ARM.

If we're not counting the 8086 line in general as bad (because it killed superior architecture chips from other manufacturers due to the sheer inertia of the IBM and clones) then the original Pentiums with the bad FP math units. RIP Motorola 68K, 6809, Zilog Z80 and MOS 6502 CPUs, the world is often an unfair place...

I see a lot of K6 series CPU hate - I really believe a lot of that came down to the huge amount of crappy motherboards out there at the time. I never had trouble with any K6-2 or K6-3 CPUs, but I was also using Tyan Super Socket 7 motherboards. The only thing I had to worry about was making sure I kept the VIA chipset drivers for that platform up to date. K6 and lower-end K6-2 CPUs were also extremely reliable on Intel chipset boards - the K6-2-266 in particular ran very cool.
Oh god, don't remind me about the cheap-ass K6-2 350 build my father cobbled together for me decades ago, with its cheap-assedness being apparent in the AT case and PSU (right as ATX was taking over), and moreso in his choice of PC-Chips M598 mobo with SiS 530 chipset.

That is the single worst motherboard I have ever had the displeasure of using. Ridiculous layout where some of the rear I/O headers sit between the PCI/ISA slots and thus making card installation/removal more of a pain than needed, a CPU socket sitting behind two of the PCI slots such that you can't install Voodoo2 SLI due to clearance issues, the apparent inability to run 100 MHz FSB stably despite having PC-133 SDRAM installed (I thought for years that I had a K6-2 366 MHz because of this!), a built-in CMI8338 that never worked (but everyone in their right mind ran an SB Live! or Aureal Vortex2 card then), prominently mentioning AGP support despite not having an AGP slot because that was for the craptastic SiS 530 graphics that lose DirectDraw acceleration the moment you update DirectX too far and make your UI even more of a slog...

Unsurprisingly, I haven't had it for years - it went to recycling during a computer decluttering run long ago. The sudden inflation of Super Socket 7 hardware values on eBay over the past few years has me wondering why anyone would put up with that when I'm not usually seeing the good motherboards for sale.

That really is something we all take for granted nowadays - overall motherboard quality is much better than it used to be in the late '90s, perhaps as a consequence of the CPUs having their own integrated northbridge functions (memory controllers, I/O) for the past decade and a half.
 
…Big iron POWER apparently still survives today, but I don't know of anyone using it over x86-64 or ARM.

I’ve been an IBM Power / AIX engineer since the mid 2000’s. I’ll take AIX over Linux any day of the week for enterprise servers. I’ve used every generation of IBM PowerPC since Power 4, and from AIX 4.3 to 7.3.

Nice to see some non-x86 input here!
 
I know this isn’t a CPU, but back in the Pentium 2/3 days I attempted to upgrade to an Intel i810/815 motherboard just to discover how much Intel castrated them. Ended up switching back to a 440BX when I figured out you could modify a slotket adapter to run Tualatin chips on them.
 
I’ve been an IBM Power / AIX engineer since the mid 2000’s. I’ll take AIX over Linux any day of the week for enterprise servers. I’ve used every generation of IBM PowerPC since Power 4, and from AIX 4.3 to 7.3.

Nice to see some non-x86 input here!
I just wish it was more affordable - I love Power, but it can't compete with x86 really on price for anything but HANA workloads, so it's steadily dying. And I wish the CPUs were cheaper for home use. Want me a Raptor Talos.
 
I just wish it was more affordable - I love Power, but it can't compete with x86 really on price for anything but HANA workloads, so it's steadily dying. And I wish the CPUs were cheaper for home use. Want me a Raptor Talos.
RISC was a good idea at the time but it's time has passed. TBH Power isn't even RISC anymore. Originally the RISC idea was simplify the CPU to make room for more cache etc. and focus on what really mattered. Now it's all about power consumption and not melting the chip and you can afford a bunch of extra inactive logic while still having a nice cache, more registers, etc. Maybe someday we'll get a new architecture. Intel tried with Itanium but messed it up and tried to offload too much to the compiler. Of course AMD also torpedoed that attempt with the Opteron/AMD64 architecture. But still, we could gain a lot from a clean slate design tailored to modern chip making reality if anyone ever has the balls and the billions to do it. IMHO it's more likely to happen if ARM starts taking big chunks of the server space from x86. It'll be easier for an upstart to gain market share if 2+ incumbents are engaged in real competition.
 
Last edited:
It's gonna be hard to find something worse than a Cyrix Cx486SLC. I mean, they were glorified 16-bit 386SX style processors but with worse compatibility. Not only were they objectively bad, but marketing deception on top of that.
 
It's gonna be hard to find something worse than a Cyrix Cx486SLC. I mean, they were glorified 16-bit 386SX style processors but with worse compatibility. Not only were they objectively bad, but marketing deception on top of that.
I had a SLC, roommate had the DLC. They both sucked, but they were stupid cheap. Back when a DX2-66 ran like $2-300, the DLC/SLC chips were like $30. And they ran on old 386 boards. Actually now that I think about it, the SLC I had was soldered to the board. You could play Doom on them, but try to run Wing Commander or X-wing and it was a slide show.
 
I can't think of a CPU I've ever worked with that was inherently "bad." Never touched a Cyrix (of any sort) and even the poor performers did the job appropriately for how much they cost. I've definitely had some trash motherboards, though. Models that would make a good CPU look bad and would have compatibility issues that no others would.
 
Every time I see one of these threads, I give the same answer.
I once was the unhappy owner of a Fujitsu Lifebook P series with a Transmeta Crusoe 800Mhz CPU.
It made the overheating, 900MHz Duron in my Compaq Presario, feel like a rocket ship in comparison. With no SSE, running XP was a brutally painful affair. It once took me 3hrs to reformat it with XP, and over 2 days just to get SP3 back on it.

I can't remember what I did with it, but I'm pretty sure I gave it away for free. The sad part is they still got ripped off. Fuck you Transmeta.
 
I had a P150+ Cyrix chip. My first PC. Ya the FP sucked but it was cheap enough for me to afford and got me online 56K baud and at the time that was amazing.
I don't know how we realize "worst cpu", as that is a tuff one. I think the award should go to a chip with apparent defects that hurt the customer, or something massively over priced that did not deliver in an extreme way.
Myself, when I buy budget I expect budget and don't complain. Doesn't equal a bad part IMO.
Wanting the thread to reply with some POS now!
 
I had a P150+ Cyrix chip. My first PC. Ya the FP sucked but it was cheap enough for me to afford and got me online 56K baud and at the time that was amazing.
I don't know how we realize "worst cpu", as that is a tuff one. I think the award should go to a chip with apparent defects that hurt the customer, or something massively over priced that did not deliver in an extreme way.
Myself, when I buy budget I expect budget and don't complain. Doesn't equal a bad part IMO.
Wanting the thread to reply with some POS now!
I think Cyrix CPUs are far game. A lot of us were scammed being told they were just as good if not better then the compensation. Benchmarks and stuff was not as big and in-depth like it is now either
 
Didn't we have one of these threads? Anyway, my vote is easy: the Western Electric WE32000, in its 3B2-300 incarnation. Horrible git of a thing, dog slow, full of errata (compilers had to sprinkle NOPs around liberally because interrupts wouldn't get the PC right), and other faults that I've managed to forget by now.

For x86 I'd probably have to go with the 80286 for the segmented architecture and ridiculous lack of a way to revert to real mode without doing a reset. (I think that was it...) Never had my hands on any of the x86-alikes (other than AMD).
 
Worst is a very subjective adjective. I would say that the fact that Bulldozer was slower than Phenom II has to put it in the top 5. Did it not work at all? No. But IMO a successor CPU should never be slower than its predecessor MHz for Mhz.
Bulldozer was a pretty bad performer in games at the time. Interestingly, it's aged fairly well with modern games taking advantage of more parallel threads than they used to. Even so, back in the day an Intel i3 chip was more than enough to beat AMD in most games. I think the worst ones were really during the early Pentium days. That being the NexGen NX586 and Cyrix 6x86 series CPU's. The NX586 didn't even have an FPU and while the Cyrix 6x86 did, it's FPU's design limited its ability to perform in games. In a lot of games at the time this was fine but this is when games like Quake were starting to really leverage the FPU and Intel CPU's were considerably faster. In fact, it was a horrible miscalculation on Cyrix's part as the Quake engine (most notably Quake III) took off like crazy with a crap ton of popular games utilizing it.
I really did not like my AMD K6-2, its been a while so i dont remember much other than i hated that pc ever since i bought it.
To be fair, while AMD did overstate the K6-2 as being a "Pentium Pro Killer" (which was a hilarious lie), the main thing that really impacted the perception of the K6-2 was the Super 7 motherboards of the day. Most of them were based on SIS and VIA chipsets. These boards were all basically trash with a host of hardware compatibility problems. This included but was not limited to AGP compatibility issues. Furthermore, one of the biggest issues we saw at the time was that like Cyrix and other alternative CPU's of the time, there were application compatibility issues with the CPU itself. These were fewer and further in between than in previous generations but these problems still existed. It wouldn't be until the Athlon era that this ultimately stopped being an issue.
 
To be fair, while AMD did overstate the K6-2 as being a "Pentium Pro Killer" (which was a hilarious lie), the main thing that really impacted the perception of the K6-2 was the Super 7 motherboards of the day. Most of them were based on SIS and VIA chipsets. These boards were all basically trash with a host of hardware compatibility problems. This included but was not limited to AGP compatibility issues. Furthermore, one of the biggest issues we saw at the time was that like Cyrix and other alternative CPU's of the time, there were application compatibility issues with the CPU itself. These were fewer and further in between than in previous generations but these problems still existed. It wouldn't be until the Athlon era that this ultimately stopped being an issue.

Good info there. Like I said I couldnt remember much, think that would have put me at bout 16 or 17 years old. The chipset and agp compatibility you mention is starting to ring some bells though. I had bought a pre-build by a company called Midwest Micro lol.
 
I think Cyrix CPUs are far game. A lot of us were scammed being told they were just as good if not better then the compensation. Benchmarks and stuff was not as big and in-depth like it is now either
The problem is that at the time the 6x86 launched, it had been in design for awhile. Even in its early days its weak FPU was only an issue in a handful of titles. Most notably, Quake. In most other games it performed similarly, although it had write back cache instead of write through which did negatively impact it from time to time. However, it wasn't as big a deal as the FPU issues were. For the money you saved the performance hit using write back cache (which boards were not built for, nor were most applications) wasn't a deal breaker. There were even titles where the Cyrix chips were faster than Intel CPU's. Redneck Rampage is one such title that did better on Cyrix chips.
 
I had Cyrix 486 chips, and those pretty much ruled. I see someone mentioned the Cyrix SLC chips as a negative, and in fairness, they were ultra-low power draw CPUs that were really meant for portables.
 
Back
Top