Semiwiki on LinkedIn Semiwiki on Facebook Semiwiki on Twitter Register / Log in 
Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/threads/the-worst-cpus-ever-made.24292/page-2
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2030770
            [XFI] => 1060170
        )

    [wordpress] => /var/www/html
)

The Worst CPUs Ever Made

hist78 said:
The bug causing unexpected and uncontrollable crashes in the Core i9-14900K and 13900K was not the only major issue. Crash incidents were reported as early as late 2022 and continued throughout 2023 and 2024. Intel finally admitted that there were issues in April 2024 and released a microcode patch in August 2024.

By then, the damage to Intel’s reputation and credibility had already been done.

I believe all processors have issues, limitations, and bugs. The best way to deal with them is to address them honestly and quickly. The longer the situation drags on, the worse the company will emerge from the incident.
Click to expand...
+1

Intel lost a lot of long term mid-tier customers because of these Raptor Lake issues. The issues spilled over into workstation "Xeon" branded chips using the Raptor Lake architecture, too.

(Though I still worship at the altar of clock speed personally - it was exciting to see the 6.2 GHz 14900KS launch, before these issues became obvious :)).
 
B_Vaughn said:
Do not remember the chip #, but there was an intel chip that got its math results wrong. So then it was into enough they tried to float a are you doing something requires highly accurate math results...
Click to expand...
I believe this is the original "Pentium FDIV" error: https://en.wikipedia.org/wiki/Pentium_FDIV_bug

It was a defining moment for the company as they handled it very poorly, but after they learned their lesson - transformed from serving only businesses to serving businesses and consumers directly.
 
B_Vaughn said:
Do not remember the chip #, but there was an intel chip that got its math results wrong. So then it was into enough they tried to float a are you doing something requires highly accurate math results...
Click to expand...

It was the Intel Pentium processor, and the year was 1994.


"Finally, Intel’s CEO Andy Grove gave in. On December 20, 1994, Intel announced a full recall program, offering free replacements to anyone who asked — no proof required. The total cost? Around $475 million."

"To their credit, Intel learned from the debacle. They revamped their quality assurance processes, changed how they handled product defects, and — most importantly — never again underestimated the power of the internet to turn a niche technical problem into front-page news."

But, from the 2022–2024 Intel 13th and 14th Gen Raptor Lake CPU crash bugs, it seems Intel has forgotten the lessons it learned from the tiny math bug in the Pentium processor back in 1994.
 
Great list. I also concur about Crusoe. Not only was the performance bad, due to a gamble on the TSMC 90nm process and killing all production in the IBM fab. The company couldn't ship product for about a year due to this and many OEMs were waiting for their processors to put into the boxes they had built. The company never really recovered from this in spite of the much better performing Efficion later on.

One other processor that is almost as disastrous as Itanium is Millenium. Funny how the last 4 characters are identical. It should get at least a dishonorable mention. It was about half a decade behind schedule and never made it into production for being obsolete.
 
Coolchips said:
Great list. I also concur about Crusoe. Not only was the performance bad, due to a gamble on the TSMC 90nm process and killing all production in the IBM fab. The company couldn't ship product for about a year due to this and many OEMs were waiting for their processors to put into the boxes they had built. The company never really recovered from this in spite of the much better performing Efficion later on.

One other processor that is almost as disastrous as Itanium is Millenium. Funny how the last 4 characters are identical. It should get at least a dishonorable mention. It was about half a decade behind schedule and never made it into production for being obsolete.
Click to expand...
Who made this "Millenium" processor?
 
I would say worst were Intel Atom processors. Bay trail, Cherry trail, Braswels... They failed in mobile market and Intel decided to dump them into laptops. They had low IPC, low clock frequency, terrible memory bandwidth (10GB/s). Result was that it took almost hour to boot Windows and fully load Outlook/Skype/Chrome...

Reason why I think they are the worst is that Intel knew they were worse than previous generation (low-end 'big' cores) but decided to dump them on unsuspecting victims anyway. Which was not the case with more exotic architectures such as Itanium.

Pentium 4 were not that bad. Biggest complain was IPC, but they had higher frequency so result was that they were faster anyway.

And Buldozers were great, actually. You could get 4 GHz 8-core Bulldozer cheaper than 2-core Intel.

Then there is something like trend in all modern CPUs. They claim to have 10W TDP, but they will draw 30-40W sustained... This is problem because lot of laptop manufacturers are following TDP in their thermal design and result is that they are too loud and they hit thermal throttling too often.
 
Jozo035 said:
I would say worst were Intel Atom processors. Bay trail, Cherry trail, Braswels... They failed in mobile market and Intel decided to dump them into laptops. They had low IPC, low clock frequency, terrible memory bandwidth (10GB/s). Result was that it took almost hour to boot Windows and fully load Outlook/Skype/Chrome...

Reason why I think they are the worst is that Intel knew they were worse than previous generation (low-end 'big' cores) but decided to dump them on unsuspecting victims anyway. Which was not the case with more exotic architectures such as Itanium.
Click to expand...

This is true - early Atom chips deserve a place on the list of worst CPUs. Intel actually sold them at a loss for years (a decade?) despite their small die size.

Jozo035 said:
And Buldozers were great, actually. You could get 4 GHz 8-core Bulldozer cheaper than 2-core Intel.
Click to expand...

While Bulldozer was sometimes an "OK" or even "good" deal for consumers, it was absolutely terrible deal for AMD itself. The chip had extremely slim margins and ate into AMD's future R&D because of this. This is why Bulldozer deserves a "Faildozer" rating :).

Bulldozer had a die size of 315mm2, compared to 216mm2 for the quad core Sandy Bridge. Roughly speaking, that would give Intel about 60% more yield per wafer than AMD, meaning Sandy Bridge was a LOT cheaper to manufacture. Even worse, the quad core Sandy Bridge (2600K) significantly outperforms the FX 8150 in most applications while using a lot less power. This is even before overclocking - where Sandy Bridge could reliably overclock by 35-40% while Bulldozer was lucky to add 15% over stock.

www.techpowerup.com

AMD FX-8150 3.60 GHz with Windows Patches Review

After settling on the market, with all the quirks and bugs supposedly fixed, all the hype and disappointment blown away, we put AMD's FX-8150 under the scope. Benchmarks are done with and without the Windows 7 hotfix and in depth overclocking should resolve any doubts you have about AMD's...
www.techpowerup.com

(P.S. I agree that Intel dual cores stayed on the desktop market for way too long...)
 
Xebec said:
While Bulldozer was sometimes an "OK" or even "good" deal for consumers, it was absolutely terrible deal for AMD itself. The chip had extremely slim margins and ate into AMD's future R&D because of this. This is why Bulldozer deserves a "Faildozer" rating :).

Bulldozer had a die size of 315mm2, compared to 216mm2 for the quad core Sandy Bridge. Roughly speaking, that would give Intel about 60% more yield per wafer than AMD, meaning Sandy Bridge was a LOT cheaper to manufacture. Even worse, the quad core Sandy Bridge (2600K) significantly outperforms the FX 8150 in most applications while using a lot less power. This is even before overclocking - where Sandy Bridge could reliably overclock by 35-40% while Bulldozer was lucky to add 15% over stock.
Click to expand...
Well, sucks for them. :D

Plus it was on SOI process so that large silicon was even more expensive.
 
You must log in or register to reply here.
Back
Top