Worst CPUs ever, now with poll!

[positivedoppler]

I agree in hindsight Bulldozer was far better than most reviewers initially thought, but AMD stuck with the same FX lineup for almost 5 years and it ended up being compared to Intel's 7th gen on 14nm by 2016 and artificially made bulldozer look worse. AMD decided to invest in apus at the time, and while I thought what they were able to hammer put at 28nm and HSA was impressive against Intel's 14nm, it just wasn't what the market demanded at the time.

So Itanium is hands down the worst cpu, i think 2nd place depends on the definition of worst.

Worse engineered cpu at the time of release? Can't be Bulldozer as I rather have a 8350 fx than a i5-3570 right now.

 

[Insert_Nickname]

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.

Can't say I'm surprised. Having only 2 physical cores is just not going to cut it, even with HT. I think the reason 2C/4T i3/i5 did well at the time was their fairly high frequency as DX11 is fairly dependent on a single rendering thread (faster core = better). Now that games have become multithreaded, FX is going to have the advantage simply by having (far) more physical execution resources.

Edit; error 484; Cat on keyboard...

 

[dark zero]

Can't say I'm surprised. Having only 2 physical cores is just not going to cut it, even with HT. I think the reason 2C/4T i3/i5 did well at the time was their fairly high frequency as DX11 is fairly dependent on a single rendering thread (faster core = better). Now that games have become multithreaded, FX is going to have the advantage simply by having (far) more physical execution resources.

Edit; error 484; Cat on keyboard...

That last part loled me hard.

But you are right. Even by having very strong single core performance, now most things are multithread and I see that having 8 core / 16 thread might become the next standard for the following years.
And even in ARM it is shown that going multi core is now the standard.

Meanwhile after seeing the mess Samsung did with their 5 and 4 nm process it made me think... what might happen if AMD released their FX in said process? How hot it might be?
It would defeat Itanic or Presshot in how hot can be the chip?

Also, the Transmeta chips might count? or is an underrated one?

 

[DAPUNISHER]

Can't say I'm surprised. Having only 2 physical cores is just not going to cut it, even with HT. I think the reason 2C/4T i3/i5 did well at the time was their fairly high frequency as DX11 is fairly dependent on a single rendering thread (faster core = better). Now that games have become multithreaded, FX is going to have the advantage simply by having (far) more physical execution resources.

Edit; error 484; Cat on keyboard...

One correction, I said it was the i3 that couldn't handle Tomb Raider, but it was instead, the i5 in the video I posted.

I agree in hindsight Bulldozer was far better than most reviewers initially thought, but AMD stuck with the same FX lineup for almost 5 years and it ended up being compared to Intel's 7th gen on 14nm by 2016 and artificially made bulldozer look worse. AMD decided to invest in apus at the time, and while I thought what they were able to hammer put at 28nm and HSA was impressive against Intel's 14nm, it just wasn't what the market demanded at the time.

So Itanium is hands down the worst cpu, i think 2nd place depends on the definition of worst.

Worse engineered cpu at the time of release? Can't be Bulldozer as I rather have a 8350 fx than a i5-3570 right now.

I don't know which CPU can truly be considered the worst, but Itanic was still infamous when this thread started over a decade ago. Being fairly fresh in people's minds was likely why was/is winning or is that losing? the poll so easily.

Also, don't forget the 8350 is Vishera. From Anand's conclusion of his review -

Vishera is a step in the right direction for AMD, it manages to deliver tangibly better performance than last year's disappointing FX processor without increasing power consumption. Thanks to architectural and frequency improvements, AMD delivers up to 20% better performance than last year's FX-8150 for a lower launch price, while remaining within the same thermal envelope.

Bulldozer was indeed a really bad CPU with a bad MSRP. Definitely hall of shame worthy.

 

[Shivansps]

Ive already mentioned this, but the worse cpus, ever, were AMD small cores, starting from Bobcat, Bobcat itself was not considered bad at their time because they were way better than the Atoms of that time, but it was really bad regardless, their cpu performance was below of the previous Neo budget cpus like the L335 and and igp perf was gimped due to single channel memory and with CPU cores not helping, perf was generally below Intel HD2000/3000 that launched that same year with the 2nd gen core.

But the worse came after that, AMD started to really milk the small cores and allowed OEMs to put them anywhere, the number of DOA 100% unusable e-Waste from Day 1 notebooks that i saw with the C-30 to C-70 APU was increible, the damage that caused to AMD brand is increible, and they never stopped!!! they keep this up for a few years until the small cores were dead. The last few ones, like the Beemas quad cores were not terrible, but they were still bad and the dual core Beemas were really slow regardless. And use considerable more power than a Gemini Lake.
They are still around, im sitting next to a box of Asrock QC6000Ms.... im actually suprised that after Ryzen AMD didnt re-buy all those chips that were still around and burn them.

 

[Insert_Nickname]

One correction, I said it was the i3 that couldn't handle Tomb Raider, but it was instead, the i5 in the video I posted.

Since the 3470 is a physical quadcore, I don't think the i3 would have a particularly easy time of it. Even with HT. There is only so much that can do compared to two extra physical cores.

Remember the Celeron Anniversary Edition (G3258)? Someone should really do a performance revisit on that, as at the time it was touted as almost the be-all, end-all cheap gaming solution. I don't think many modern titles will even start on it. One thing it -did- do well was emulation on a budget if you could OC it far enough. Or if you had a particular piece of single threaded software.

 

[Insert_Nickname]

Ive already mentioned this, but the worse cpus, ever, were AMD small cores, starting from Bobcat, Bobcat itself was not considered bad at their time because they were way better than the Atoms of that time, but it was really bad regardless, their cpu performance was below of the previous Neo budget cpus like the L335 and and igp perf was gimped due to single channel memory and with CPU cores not helping, perf was generally below Intel HD2000/3000 that launched that same year with the 2nd gen core.

But the worse came after that, AMD started to really milk the small cores and allowed OEMs to put them anywhere, the number of DOA 100% unusable e-Waste from Day 1 notebooks that i saw with the C-30 to C-70 APU was increible, the damage that caused to AMD brand is increible, and they never stopped!!! they keep this up for a few years until the small cores were dead. The last few ones, like the Beemas quad cores were not terrible, but they were still bad and the dual core Beemas were really slow regardless. And use considerable more power than a Gemini Lake.
They are still around, im sitting next to a box of Asrock QC6000Ms.... im actually suprised that after Ryzen AMD didnt re-buy all those chips that were still around and burn them.

Agreed. I've cursed Bobcat more times then I can count on that front. Bobcat + 5400RPM HDD with too little RAM makes for looong waits. Don't even think about doing Windows Updates on them. Best strategy was to let them simmer overnight.

 

[DAPUNISHER]

Since the 3470 is a physical quadcore, I don't think the i3 would have a particularly easy time of it. Even with HT. There is only so much that can do compared to two extra physical cores.

Remember the Celeron Anniversary Edition (G3258)? Someone should really do a performance revisit on that, as at the time it was touted as almost the be-all, end-all cheap gaming solution. I don't think many modern titles will even start on it. One thing it -did- do well was emulation on a budget if you could OC it far enough. Or if you had a particular piece of single threaded software.

Yes, my poorly made point, was that the i5 got its cheeks clapped. That means the i3, HT was fairly weak compared to real cores, was going to have a horrible time.

Your timing on the 3258 could not be better. One of my favorite newer tech tubers just did a revisit. It ain't pretty.

 

[DAPUNISHER]

Agreed. I've cursed Bobcat more times then I can count on that front. Bobcat + 5400RPM HDD with too little RAM makes for looong waits. Don't even think about doing Windows Updates on them. Best strategy was to let them simmer overnight.

I will have to check if that is what I have soldered on a m-ITX board in my parts bin. If it is, it can't even handle the modern internet with 16GB of DDR3 1600 and SATA SSD. It is torture to even mess with it using windows. It can handle firefox on Mint IIRC if I keep youtube to 720p. I will have to dig it out and confirm; I have free time.

 

[sdifox]

Holy necro

 

[Insert_Nickname]

It can handle firefox on Mint IIRC if I keep youtube to 720p.

Mint? Xfce edition works well for this kind of system. Within the hardware limitations of course.

Brazos video acceleration hardware is limited to h.264@720p, unfortunately. It's far too slow to software decode anything newer. So don't waste your time.

 

[Tuna-Fish]

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.

That was not what it was for. It was intended to eventually replace x86 in every market. Only, it turned out to be so bad that it could only compete in small niches, and it was only competitive in those niches because Intel was intentionally segmenting the market for it (that is, they were leaving features off from x86 products to make room for Itanium). Once AMD got 64-bit working, the only remaining market for Itanium was the one where it's RAS features shined, and all those features could have just as well been on Xeons (and would eventually make their way into them).

Or the 80286, it was great for its time.

80286 was a significant upgrade from 8088, but I think calling it "great for it's time" is stretching it. IIRC IBM XT 286 launched on the same year as the ARM2, and ARM2 was ~4 times as fast at the same clock, and clocked almost as high on an inferior process. Of course, ARM was only available for weirdo Europeans, and was not compatible with anything, so Intel won.

The story of x86 from 8086 to Pentium 1 is that when one vendor is shipping vastly more chips than any of it's competitors (because they are targeting the lowest-margin, highest-volume segment of the market), they can afford to upgrade their process faster than those competitors, allowing them to compete against superior designs with inferior designs built with better manufacturing. In those days, people who could afford them bought much more expensive workstation computers from vendors specializing in that market, that performed much better than anything available from Intel. Then Intel came out with Pentium Pro, which was their first design that was legitimately competitive design-wise, on iso-process metrics, and suddely all the RISC vendors were hosed, when the best process in the business got mated with a good design.

 

[bba-tcg]

The biggest issue with the 80286 is that it had no builtin mechanism to go from protected mode to real mode, thus requiring a workaround that performed a physical reset of the processor in order to accomplish this transition. A very time consuming endeavor. At the time, real mode (AKA DOS mode), was where people spent most of the time. This was fixed with the 80386. This limitation is what led to the "break up" of IBM and Microsoft and derailed OS development by years.

Personally, I consider the 80286 one of the worst CPUs in history from a usage standpoint. It did, however, lead to the changes in the 80386 so it wasn't all bad.

 

[dark zero]

Holy necro

Funny story it can be necro untill you realize that there were chips that can compete in the worst CPUs ever like the Snapdragon 8 Gen 1 made by Samsung or the Exynos series since 9810.

And that is not including some unknown ones like the Zhaoxin ones which performs very badly for the current standards

 

[sdifox]

Funny story it can be necro untill you realize that there were chips that can compete in the worst CPUs ever like the Snapdragon 8 Gen 1 made by Samsung or the Exynos series since 9810.

And that is not including some unknown ones like the Zhaoxin ones which performs very badly for the current standards

Thread is more than a decade old, of course someone is going to make a bad chip.

 

[LightningZ71]

I'm still kind of surprised that no one has recently brought up any of Russia's abortive attempts at x86 processors... Witness the MCST Elbrus-8C in all it's glory...
Elbrus-8C falling on it's face...

The Elbrus situation is so bad that they've actually moved to trying to source chips from China... like the KaiXian KX-6640MA, which isn't exactly very performant either...
KX-6640MA not doing very well against an i3 and an R5...

We may complain about some chips of the past, but, we certainly are spoiled for quality...

 

[Insert_Nickname]

The Elbrus situation is so bad that they've actually moved to trying to source chips from China... like the KaiXian KX-6640MA, which isn't exactly very performant either...
KX-6640MA not doing very well against an i3 and an R5...

Ouch. That is beyond brutal.

Unfortunately, I think the point is these will run x64 software 1:1. Just not very fast. It's somewhere around C2D, C2Q performance looking at it. Perhaps a bit better. So around ~15 year old performance.

As a side note; anything that slows down the Russians must be a good thing. So if you view it in that perspective, it's actually a positive thing. Lets not get political though. P&N is that way -->

 

[LightningZ71]

Elbrus, for their process tech, isn't abysmal. They are on what appears to be a .28 class node, so c2d performance isn't exactly awful. Interestingly enough, if memory serves, they were trying the transmeta/itanium VLIW path for a while, though I don't think they were ever competitive with their contemporaries.



As for the chinese parts, I think they are performing in the Sandy/Ivy brige level overall on a per core basis at the top end. There's a separate thread here for that though.

 

[blckgrffn]

LMAO.

"One of the surprising things about the Elbrus-8C server was that it is a real product," said Zhbankov. "It was a real server that we were given. […] It is an actual product that has its disadvantages, loads of disadvantages, but we can work with them."

Tom's article is IT gold. Not in a P&N way, but an IT nightmare fuel way.

"Every server begins from its chassis and some general features such as remote management, which Sber evaluates under its Functional Testing procedure. Apparently, an MCST Elbrus-8C machine failed 84% of Sber's Functional Testing as it could not be easily removed from the rack, lacked proper LED indicators, and came without remote management, which to a large degree made it unusable for usage in commercial datacenters. There is some hope, though."

If you completely remove all of your requirements for what a server is, you can can use these as "servers".

 

[dark zero]

I tought that Itanic was bad... but Elbrus managed to defeat Itanium on failure level!
So we have a new king on this thread?

Meanwhile how the indian Shakti processor performs? I heard that was made in 180 nm process.

Ouch. That is beyond brutal.

Unfortunately, I think the point is these will run x64 software 1:1. Just not very fast. It's somewhere around C2D, C2Q performance looking at it. Perhaps a bit better. So around ~15 year old performance.

As a side note; anything that slows down the Russians must be a good thing. So if you view it in that perspective, it's actually a positive thing. Lets not get political though. P&N is that way -->

Meanwhile... Zhaoxin is going back to Enhanced Pentium M era? That is brutal.

 

[Mopetar]

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's taking your point a bit to far. No testing approach will be perfect simply because of time requirements. Someone can do some really detailed testing on a single game by spending hours looking for cases where performance problems occur, but how do you distinguish it from cherry picking.

There was some sock puppet in another thread recently extolling the virtues of this highly focused testing methodology that just so happened to coincide with their own preconceived notions. You can ha e depth, but realize it's being traded for breadth.

I agree in hindsight Bulldozer was far better than most reviewers initially thought

Not really. It was bad and the only saving grace is that software grew to leverage more cores so it tends to outperform the low-core Intel CPUs that could be had for the same price around the same time.

Bulldozer basically saw AMD's withdrawal from the server market. If it weren't for their ATI acquisition giving them a better APU in those regards, I don't think they would have been able to survive in mobile either.

I'm still kind of surprised that no one has recently brought up any of Russia's abortive attempts at x86 processors... Witness the MCST Elbrus-8C in all it's glory...
Elbrus-8C falling on it's face...

The Elbrus situation is so bad that they've actually moved to trying to source chips from China... like the KaiXian KX-6640MA, which isn't exactly very performant either...
KX-6640MA not doing very well against an i3 and an R5...

We may complain about some chips of the past, but, we certainly are spoiled for quality...

I don't think it's surprising that communist countries make worse products. Go watch some of the old Top Gear episodes where they drive Soviet cars. Anyone from Russia or China that could design a great CPU got the hell out a long time ago.

Every year there's some disappointing draft picks, but they still got to play in the NFL, NBA, etc. They're still disappointing, but no one thinks to compare them to the 5 year old that's chasing butterflies in the outfield at a t-ball match.

 

[Insert_Nickname]

I don't think it's surprising that communist countries make worse products. Go watch some of the old Top Gear episodes where they drive Soviet cars. Anyone from Russia or China that could design a great CPU got the hell out a long time ago.

While not completely Soviet, I have driven an East German Trabant. So at least in the same league. It was fun enough, but definitely not something I'd want to rely on on a daily basis. It's like being thrown back to the 50s driving wise. If you should ever find yourself in Berlin, I still think you can get the experience first hand. There are still some around.

About it's single redeeming feature is that you can fix it with nothing more then a screwdriver. Which you'll be doing a lot.

So these Chinese CPUs have to be the IT equivalent of such cars. Fast forward to the past.

 

[Ranulf]

Yes, my poorly made point, was that the i5 got its cheeks clapped. That means the i3, HT was fairly weak compared to real cores, was going to have a horrible time.

Your timing on the 3258 could not be better. One of my favorite newer tech tubers just did a revisit. It ain't pretty.

I like this ice tech guy, good video.

 

[burninatortech4]

Atom N455. Found it in the dumpster at work and figured I could save it with XFCE and a small SSD. Nope! Total POS.

 

[Shivansps]

The single cores in-order Atoms have no saving.