When Calculators lie will anyone notice?

  • Thread starter jedishrfu
  • Start date
  • Tags
    Calculators
In summary: Maybe it's because they are looking at trigonometry from the wrong angle.In summary, a grad student has studied the blind belief people have in calculators, and found that no surprise, it's very similar to what was found with the 487 co-processor. There was a chip with that name that had a flaw in it that caused math errors in a small percentage of calculations. The Pentium FDIV error was one of them, and it only affected processors made between 1994-1996. Intel did not handle the problem well, and people wanted to replace their entire computers when they first found out about it.
Physics news on Phys.org
  • #3
Remember the 487 co-processor?
 
  • Like
  • Haha
Likes Dragrath, QuantumQuest, gleem and 1 other person
  • #4
Bystander said:
Remember the 487 co-processor?

I remember Pentium FDIV error, was there something similar for 487?
 
  • #5
Borek said:
I remember Pentium FDIV error, was there something similar for 487?
I don't remember the name of it (and am assuming it was the 487 coprocessor), but there was a math FP coprocessor a couple of decades ago that was found to have significant flaws and had make some impactful business calculations incorrectly before it was discovered.
 
  • #6
Borek said:
something similar for 487?
My memory may be failing, but that's what I recall...it came out about the same time as "7th Guest," and that's all the computer was good for was playing "7th Guest," couldn't do anything numerical.
 
  • #7
There was no 80487 coprocessor. (There was a chip with that name, which is a more interesting story) The problem was in the P5 and P54 Pentium processors. In ~10-10 of the calculations, it would make a (perfectly reproducible) ~10-4 error. It was caused by a bad value in a lookup table.

It was so subtle that it took a year and a half to be discovered (by a prof named Thomas Nicely). This is the worst kind of error: rare, and when wrong it's plausible.

That's what makes this so interesting - the results are wrong, and not very plausible.
 
Last edited:
  • Like
Likes newbiegirl and QuantumQuest
  • #8
Vanadium 50 said:
There was no 80487 coprocessor. (There was a chip with that name, which is a more interesting story)
The 80486 processor of that generation came out in two main models: 80486SX (simplex) and 80486DX (duplex). Both chips were identical in the sense that both had floating point units, but the SX variant had the connections to the FPU severed. The DX variant had a working FPU. For the SX models, you could add an i487 chip, which contained a full implementation of 80486DX functionality.
Vanadium 50 said:
The problem was in the P5 and P54 Pentium processors.
Yes, and not with the previous generations. I wrote a short assembly program that could be used to determine if the Pentium processor it was run on was one of the faulty ones. My program and an explanation of how it worked was published in Jeff Duntemann's PC Techniques magazine back in Feb/Mar of '95.
I heard that it cost Intel on the order of a billion dollars in recalling the flawed processors.
 
  • #9
Yes, that was the more interesting story.

Intel did not handle this at all well. First they tried to explain that a wrong answer now and again didn't matter. Then when they finally agreed to replace the chip, they agreed to replace the chip. You take the chip out of its socket (and this was not as common as it is today), mail it to them, and someday you'd get a replacement. What folks really wanted was to swap out their entire computers - but Gateway or Northgate or whoever you bought computers from back in the day had moved on to newer products: P5/60s were almost two years old by then and P54Cs were being sold. They had an incompatible socket.
 
  • #10
One of the best student mistakes on the calculator is when radians are confused with degrees ie they don’t know which mode the calculator is set to.
 
  • Like
Likes Dragrath
  • #12
By the way, speaking of calculators I'd like to take the opportunity to mention a couple of good, free calculator Android apps for e.g. smart phones and tablets:

RealCalc (Scientific calculator):
(though, judging by the name it may not be able to handle imaginary numbers :smile:. But it's a good calculator)
http://play.google.com/store/apps/details?id=uk.co.nickfines.RealCalc&hl=en
Graphing Calculator - Algeo
(a good graphing calculator which also can do some calculus)
http://play.google.com/store/apps/details?id=com.algeo.algeo
Mathlab Graphing Calculator + Math, Algebra & Calculus
(it can also do graphs, do algebra and some more)
http://play.google.com/store/apps/details?id=us.mathlab.android&hl=en
And now I remembered that Greg asked me a long time ago to write an insight post about useful scientific apps for smartphones. Regretfully I haven't gotten around to do it, but maybe I will... :smile:
 
Last edited:
  • #13
jedishrfu said:
One of the best student mistakes on the calculator is when radians are confused with degrees ie they don’t know which mode the calculator is set to.
Maybe it's because they are looking at trigonometry from the wrong angle.
(sorry, I could not resist)
 
  • Like
Likes Dragrath, Bystander and jedishrfu
  • #14
jedishrfu said:
One of the best student mistakes on the calculator is when radians are confused with degrees ie they don’t know which mode the calculator is set to.
Maths mode is best :smile:
1572906063790.png
 
  • Like
Likes jedishrfu and Bystander
  • #17
HP and RPN, I love it! I've got an old HP-27 in a drawer somewhere at home. And I love the design:

27.jpg
 
Last edited by a moderator:
  • Like
Likes jedishrfu and phinds
  • #18
DennisN said:
HP and RPN, I love it! I've got an old HP-27 in a drawer somewhere at home. And I love the design:
For many years, I found it very annoying to have to ever use any calculator that didn't use RPN. Now I doubt I could even remember how to do it. I think it's pretty much disappeared from common usage.
 
  • Like
Likes DennisN
  • #19
russ_watters said:
You really think I'd trust a robot to tell me about lying calculators? How do I know a calculator doesn't owe you money or something?
Hey, I passed the Turing test and got accepted to grad school. So there always with the negative vibes (Oddball in Kelly’s Heroes).
 
  • Like
Likes DaveC426913 and russ_watters
  • #21
phinds said:
For many years, I found it very annoying to have to ever use any calculator that didn't use RPN.
Same here :smile:. I got the HP-27 from my father when I was in my teens, and he taught me how to use RPN. He had been using the HP before me. And I very well remember I used it in high school for math, technology and science classes. I was the only one in my class that had a HP that could do RPN, everybody else used Casio or Texas Instruments calculators which used parentheses for formulas.

And many of them made fun of me using my old RPN calculator, but when I challenged anyone to beat the speed of my HP when doing large factorials, not a single one of the more modern calculators beat mine. And then my classmates stopped making fun of me and my calculator. :)

Later I taught some of them how RPN works, and they got interested and a bit impressed by the calculations that can be done using RPN without a single parenthesis.
 
  • #22
There is an Android app RealCalc that can work both in RPN and standard calculator mode.
 
  • Like
Likes collinsmark, jedishrfu, DennisN and 1 other person
  • #24
berkeman said:
And there is Free42, which I use almost every day on my phone at work...
Cool, I didn't know about that one! You posted about it three minutes ago, and now I will download it immediately.
 
  • Like
Likes berkeman
  • #25
berkeman said:
And there is Free42, which I use almost every day on my phone at work...
I've got Free42 and thus HP style RPN on my phone now and there are almost tears of joy in my eyes 😀.
"Hope is not lost today, it is found."
(but I am aware that RealCalc also can do RPN)

If Free42 would have supported skins (user customizable design) I would have customized it into a 1970s vintage look :smile: .
 
  • Like
Likes berkeman
  • #26
Sorry folks, I had to remove some political content from the thread.

-- Jedi
 
  • #27
Borek said:
There is an Android app RealCalc that can work both in RPN and standard calculator mode.

berkeman said:
And there is Free42, which I use almost every day on my phone at work...

https://thomasokken.com/free42/

I use an HP48 app. I suspect I might have lost the ability to use a calculator without RPN. (Although I don't know for sure because I haven't tried.)
 
  • Like
Likes DennisN
  • #28
DennisN said:
HP and RPN, I love it! I've got an old HP-27 in a drawer somewhere at home.

I have an HP-67 that was handed down to me by my aunt. Unfortunately it no longer works. I am hoping I can get it repaired some time.
jedishrfu said:
They made a retro tribute to one of the HP calculators a few years ago. The guts are modern but the UI is the same classic design.

https://store.hp.com/us/en/pdp/hp-35s-scientific-calculator-p-f2215aa-aba--1

I bought one of these that I saw in a shop about a decade ago. HP introduced it in 2007 to commemorate the 35th anniversary of their first pocket scientific calculator, the HP-35. It has been in continual production since then, presumably because it's one of only a few scientific calculators allowed in some professional engineering exams in the US.

It is named after the original HP-35 but by lineage it's a replacement for the HP-33S and descends from the HP-32S and HP-32SII models from the late 1980s and 1990s. In context this makes the HP-35S a mid-range scientific calculator. HP's line of high-end scientifics ended with the HP-42S.

Whether or not you would like it would probably mostly depend on whether you can forgive its long list of bugs (which include failing to calculate trigonometric functions accurately for angles close to multiples of 90 degrees), which HP never seems to have bothered to fix.
 
  • Like
Likes berkeman, jedishrfu and DennisN
  • #29
DennisN said:
If Free42 would have supported skins (user customizable design) I would have customized it into a 1970s vintage look :smile: .
Note that Free42 presents you with different calculators in portrait and landscape modes.
 
  • Like
Likes DennisN
  • #30
Borek said:
There is an Android app RealCalc that can work both in RPN and standard calculator mode.

I don't think you're limited for choice if you want to use RPN on a computer. The dc calculator (one of the traditional Unix command-line utilities) and Forth programming language are both stack-based and use RPN notation. I am not sure this is so useful in a programming language though. RPN to me seems great for inputting and doing calculations interactively but not so good as far as readability is concerned, which is usually something you want in a programming language.

There are also a lot of emulators of various HP calculator models available, both for the PC and for smartphones and tablets. Free42 is not the only one.
 
  • #31
Postscript uses RPN - probably because it's meant to be implemented on fairly low-end hardware and mostly written by other programs.
 
  • #32
Emacs also includes an RPN calculator (calc) in its standard distribution. I haven't really used it but the feature list seems quite impressive for a package included in what is ostensibly a text editor.
 
  • Like
Likes collinsmark
  • #33
I tend to fire up python on linux and type in my expression to get a quick answer.
 
  • Informative
Likes berkeman
  • #34
As long as people are prone to make errors using the tiny keyboards on calculators, there will be some distrust in the answers - just because they may be answers to the wrong calculations. If a simpler interface is developed, the answers are more likely to be taken for granted.

...just as people who compose text messages by talking aloud tend to think they are writing what they intend!
 

1. What do you mean by "calculators lying"?

When we say "calculators lie", we are referring to the potential for errors or inaccuracies in the calculations performed by calculators. This can be caused by a variety of factors, such as faulty programming, user error, or hardware malfunctions.

2. How common are calculator errors?

The frequency of calculator errors varies depending on the type and quality of the calculator. While most modern calculators have built-in safeguards to minimize errors, they are not infallible and can still produce incorrect results in certain situations. Additionally, user error can also contribute to the frequency of calculator errors.

3. What are the consequences of relying on inaccurate calculations?

The consequences of relying on inaccurate calculations can range from minor inconveniences to more serious financial or safety implications. In academic or professional settings, relying on incorrect calculations can lead to incorrect data analysis or incorrect conclusions. In everyday life, relying on inaccurate calculations can result in incorrect budgeting or inaccurate measurements.

4. How can we detect and prevent calculator errors?

To detect and prevent calculator errors, it is important to double-check calculations and use multiple sources for verification. Additionally, regularly updating and maintaining calculators, as well as using reputable and reliable brands, can also help minimize the risk of errors. In cases where high accuracy is crucial, using manual calculations or consulting with a human expert may also be necessary.

5. Are there any regulations or standards for calculator accuracy?

In most countries, there are no specific regulations or standards for calculator accuracy. However, some industries, such as finance and engineering, may have their own standards for the accuracy of calculations. Additionally, reputable calculator manufacturers often have their products tested and certified by independent organizations to ensure their accuracy meets certain standards.

Similar threads

Replies
10
Views
758
Replies
19
Views
1K
  • General Discussion
Replies
16
Views
993
Replies
2
Views
267
  • General Discussion
Replies
0
Views
567
  • General Discussion
Replies
21
Views
2K
  • General Discussion
Replies
1
Views
782
  • General Discussion
Replies
24
Views
8K
Replies
8
Views
906
Back
Top