The General thread for Computers And Internet

User avatar
RTH10260
Posts: 8196
Joined: Mon Feb 22, 2021 10:16 am
Location: Switzerland, near the Alps
Verified: eurobot

The General thread for Computers And Internet

#1

Post by RTH10260 »

From the Canadian LMG Linus Media Group of geeks



User avatar
bill_g
Posts: 3318
Joined: Mon Feb 22, 2021 5:52 pm
Location: Portland OR
Occupation: Retired (kind of)
Verified: ✅ Checked Republic

The General thread for Computers And Internet

#2

Post by bill_g »

That was fun. Thx!


somerset
Posts: 444
Joined: Mon Feb 22, 2021 12:06 pm
Occupation: Lab Rat

The General thread for Computers And Internet

#3

Post by somerset »

RTH10260 wrote: Thu Nov 24, 2022 6:48 am From the Canadian LMG Linus Media Group of geeks

What a lot of folks don't realize is that many high performance chips are already made the way Intel is proposing (including some of Intel's chips going back 20+ years). It's common to design just one fully-featured chip, then "de-feature" it with wirebond or fuse options. This sounds wasteful, but the cost of laying out, validating and fabricating two different processor designs is usually far higher than the cost of wasted silicon real estate. Easier to just make one die and sell it as two (or more) products. The idea of making this "de-featuring" reversible is new, and maybe a little bit controversial from a marketing perspective, but not surprising. What you're really paying for is IP, and making it available via subscription is essentially the same as software licensing.


User avatar
RTH10260
Posts: 8196
Joined: Mon Feb 22, 2021 10:16 am
Location: Switzerland, near the Alps
Verified: eurobot

The General thread for Computers And Internet

#4

Post by RTH10260 »

somerset wrote: Thu Nov 24, 2022 1:00 pm
RTH10260 wrote: Thu Nov 24, 2022 6:48 am From the Canadian LMG Linus Media Group of geeks

https://youtu.be/awLbarjUTdg
What a lot of folks don't realize is that many high performance chips are already made the way Intel is proposing (including some of Intel's chips going back 20+ years). It's common to design just one fully-featured chip, then "de-feature" it with wirebond or fuse options. This sounds wasteful, but the cost of laying out, validating and fabricating two different processor designs is usually far higher than the cost of wasted silicon real estate. Easier to just make one die and sell it as two (or more) products. The idea of making this "de-featuring" reversible is new, and maybe a little bit controversial from a marketing perspective, but not surprising. What you're really paying for is IP, and making it available via subscription is essentially the same as software licensing.
In a recent clip they had just discussed a new cpu that Intel will bring to market in the near future (consumer grade, but highend, rather than servers) that contains certain features one has to subscribe to or get a separate license. Iirc higher cpu speads and overclocking will be among them. I understand that in the past the same chip was branded slightly differently if certain quality control levels coulld not be reached but satisfied some lesser grade (i guess reliability with speed was mostly the case).


Reddog
Posts: 169
Joined: Mon Feb 22, 2021 2:29 pm

The General thread for Computers And Internet

#5

Post by Reddog »

Not sure if this is a hijack or not :hijacked: ?
Best I can guess I think it belongs here. Good to see technology progressing.


User avatar
p0rtia
Posts: 2916
Joined: Mon Feb 22, 2021 9:55 am

The General thread for Computers And Internet

#6

Post by p0rtia »

:clap: :clap: :clap:


humblescribe
Posts: 876
Joined: Mon Feb 22, 2021 3:42 pm
Occupation: Dude
Verified:

The General thread for Computers And Internet

#7

Post by humblescribe »

Reddog wrote: Thu Nov 24, 2022 2:55 pm Not sure if this is a hijack or not :hijacked: ?
Best I can guess I think it belongs here. Good to see technology progressing.
Ahh, the Turbo Encabulator canard updated for the 2020s. Love it!

The video showed a brief clip from the original Turbo Encabulator video. The actor in that video from the '70s was Irving "Bud" Taggart. He was a local actor from Detroit who narrated a lot of industrial and educational films. The Turbo Encabulator was a gag devised by some engineering students in Merrie Olde England after WWII, and it sort of became an inside joke as it spread throughout universities. Taggart and his filming crew decided to make their video after they had filmed an industrial video for an auto manufacturer.

Bud Taggart died in 2021 at age 96.

You can find the original Turbo Encabulator video on YouTube.


"Some cause happiness wherever they go; others whenever they go." O. Wilde
Reddog
Posts: 169
Joined: Mon Feb 22, 2021 2:29 pm

The General thread for Computers And Internet

#8

Post by Reddog »

Yes, I’ve followed several iterations of it. This was the first I’ve seen computer/network related.
It must have been around awhile, but I hadn’t seen it.

To be honest sometimes that’s how the legal jargon here sounds to me. I always wondered why Mason and Burger didn’t punch each other out, after calling the other incompetent. I learned here it has a different connotation than non-legal usage.


somerset
Posts: 444
Joined: Mon Feb 22, 2021 12:06 pm
Occupation: Lab Rat

The General thread for Computers And Internet

#9

Post by somerset »

RTH10260 wrote: Thu Nov 24, 2022 1:12 pm
somerset wrote: Thu Nov 24, 2022 1:00 pm
RTH10260 wrote: Thu Nov 24, 2022 6:48 am From the Canadian LMG Linus Media Group of geeks

https://youtu.be/awLbarjUTdg
What a lot of folks don't realize is that many high performance chips are already made the way Intel is proposing (including some of Intel's chips going back 20+ years). It's common to design just one fully-featured chip, then "de-feature" it with wirebond or fuse options. This sounds wasteful, but the cost of laying out, validating and fabricating two different processor designs is usually far higher than the cost of wasted silicon real estate. Easier to just make one die and sell it as two (or more) products. The idea of making this "de-featuring" reversible is new, and maybe a little bit controversial from a marketing perspective, but not surprising. What you're really paying for is IP, and making it available via subscription is essentially the same as software licensing.
In a recent clip they had just discussed a new cpu that Intel will bring to market in the near future (consumer grade, but highend, rather than servers) that contains certain features one has to subscribe to or get a separate license. Iirc higher cpu speads and overclocking will be among them. I understand that in the past the same chip was branded slightly differently if certain quality control levels coulld not be reached but satisfied some lesser grade (i guess reliability with speed was mostly the case).
Sorta, kinda.

Processor chips used to be "speed binned" to sort them by performance. During wafer sort, chips were first tested at the highest clock speed (i.e., 1.0 GHz). If they failed at this highest level of performance, they would be tested at slower and slower speeds (i.e., 750Mhz, 500Mhz, etc.) until they either passed or didn't meet any spec. This is called "binning." "Bin 1" chips are usually the best, with "Bin 2," Bin 3," etc, etc, being lower performing or failing for various parameters. Bin 99 is typically a total reject. This was recorded in the wafer map, with the "best," "second best," etc., chips assembled, labeled and priced accordingly.

This difference in performance was (is) due to normal variability in process parameters during wafer fabrication. It's very difficult to maintain sub-nanometer consistency across a 300mm wafer, not to mention maintaining that consistency from wafer to wafer and from lot to lot over the course of months. But the whole semiconductor industry has gotten a lot better over the last ~20 years, and what was considered normal process variation in 2005 is now considered unacceptable. It's this improvement in process control that lets Intel now make more chips "Bin 1," and start to "downgrade" otherwise high performing chips to sell at a lower price point. Since the majority of chips are now capable of higher performance, it makes business sense to sell the performance level the customer wants. If you want lower performance, you can get this at a lower cost. And if you want higher performance, that's available too.

I want to point out that none of this performance sorting means providing lower quality or lower reliability chips (quality, reliability and failure analysis is my world ;) ). If anything, there used to be something of an inverse relationship between performance and reliability - Chips that could operate at higher clock speeds also tended degrade faster. Today, the same improvements in process consistency that helped improve yield also help improve quality and reliability.


User avatar
Sam the Centipede
Posts: 951
Joined: Thu Feb 25, 2021 12:19 pm

The General thread for Computers And Internet

#10

Post by Sam the Centipede »

Back in the days of leave blue boxes and whirring tapes in air-conditioned computer rooms, the same was done: it was known as slugging. The customer would but a System/370 Model <something> and if they needed to upgrade it to a System/370 Model <somethingbetter> they'd give IBM a stack of dollars and the IBM engineer would come along, open a cabinet and flip some switches on a circuit board (or remove a jumper or something equally five minutes-ish) and the customer would have a more powerful computer. Possibly some new labels too. No need to install a more powerful electron pump, extra outlets on the channel manifold or a data train to replace the data bus service.


User avatar
Sam the Centipede
Posts: 951
Joined: Thu Feb 25, 2021 12:19 pm

The General thread for Computers And Internet

#11

Post by Sam the Centipede »

On sorting product by measured quality: this story is probably apocryphal, unless one has a low opinio of statisticians; I don't. Here's the story: college teachers were looking for practical exercises in statistical measurement, calculation of averages, etc. and thought electrical resistors might be something their students could measure with meters and analyze the results: histograms! means! (other measures of central tendency are available), standard deviations! wow, perhaps even kurtosis!

It was then (allegedly) a surprise to realise that a resistor labeled 100O Ω ±10% might measure 992 Ω or 1008 Ω but it would never measure the nominal 100O Ω or even in the 999 Ω to 1001 Ω range because those resistors had been measured during manufacture and packaged as the more expensive 100O Ω ±1% components.

A lesson too in biased samples.


User avatar
tek
Posts: 1869
Joined: Mon Feb 22, 2021 10:15 am

The General thread for Computers And Internet

#12

Post by tek »

Sam the Centipede wrote: Fri Nov 25, 2022 3:42 am Back in the days of leave blue boxes and whirring tapes in air-conditioned computer rooms, the same was done: it was known as slugging. The customer would but a System/370 Model <something> and if they needed to upgrade it to a System/370 Model <somethingbetter> they'd give IBM a stack of dollars and the IBM engineer would come along, open a cabinet and flip some switches on a circuit board (or remove a jumper or something equally five minutes-ish) and the customer would have a more powerful computer. Possibly some new labels too. No need to install a more powerful electron pump, extra outlets on the channel manifold or a data train to replace the data bus service.
Indeed.
In my heyday of doing CPU design for a company that no longer exists as a separate entity, I was asked to partake in such mishegas.
I forcefully argued to my group Vice President that we were not competing with ourselves, we were competing with other companies in our market. He agreed, and put the kibosh on this one instance of slugging - much to marketing's chagrin.


New Turtle
Posts: 429
Joined: Wed Nov 10, 2021 2:43 pm

The General thread for Computers And Internet

#13

Post by New Turtle »

Image
AMD Athlon CPU "The Pencil Trick" Using pencil lead to reconnect the L1 bridges, allowing the processor to be set at any clock frequency.
https://commons.m.wikimedia.org/wiki/Fi ... _Trick.jpg

I've done this before and it worked. These were Athlon "Thunderbird" CPUs, there were 1.0 GHz, 1.2GHz, 1.4 GHz models, but were the same chip at different prices. By putting some graphite over the places where certain leads were broken at the factory, it will unlock the multiplier setting in the BIOS where you can run it at a higher clock frequency.


User avatar
Greatgrey
Posts: 575
Joined: Mon Feb 22, 2021 9:53 am
Location: Unimatrix Zero
Verified: 💲8️⃣

Software Piracy

#14

Post by Greatgrey »

Buy 38 copies of the software, make 550,000 copies.

Oh yeah, it was the US Navy

https://taskandpurpose.com/news/us-navy ... andpurpose




What's the Frequency, Kenneth?
User avatar
noblepa
Posts: 1826
Joined: Mon Feb 22, 2021 2:55 pm
Location: Bay Village, Ohio
Occupation: Retired IT Nerd

The General thread for Computers And Internet

#15

Post by noblepa »

Sam the Centipede wrote: Fri Nov 25, 2022 3:42 am Back in the days of leave blue boxes and whirring tapes in air-conditioned computer rooms, the same was done: it was known as slugging. The customer would but a System/370 Model <something> and if they needed to upgrade it to a System/370 Model <somethingbetter> they'd give IBM a stack of dollars and the IBM engineer would come along, open a cabinet and flip some switches on a circuit board (or remove a jumper or something equally five minutes-ish) and the customer would have a more powerful computer. Possibly some new labels too. No need to install a more powerful electron pump, extra outlets on the channel manifold or a data train to replace the data bus service.
As far as I know, IBM still does that. We had a z/890 (descendent of the System/370). It came with, I believe, eight processors installed. We only paid for the use of two or three. It also had specialized processors that could offload processing for DB/2 or IMS database system, relieving the primary CPU of some of the load.

You could also call them up when you approached your busy season, and activate additional processors for a few months, then turn them off when you didn't need the additional power.

More recently, we operated on a z/9. It had the same ability.


User avatar
Sam the Centipede
Posts: 951
Joined: Thu Feb 25, 2021 12:19 pm

The General thread for Computers And Internet

#16

Post by Sam the Centipede »

Interesting to read that noblepa, thanks.

Some people seem to think there's something unethical in providing a machine that can perform strongly but slugging it to a weaker configuration. I don't agree anyway, but your scenario is an analog of temporarily upgrading one's subscription to a service in response to transient requirements.


User avatar
RTH10260
Posts: 8196
Joined: Mon Feb 22, 2021 10:16 am
Location: Switzerland, near the Alps
Verified: eurobot

The General thread for Computers And Internet

#17

Post by RTH10260 »

:doh:

should you own hardware of this brand you may want to burn it down asap



User avatar
John Thomas8
Posts: 2762
Joined: Mon Feb 22, 2021 7:42 pm
Location: Central NC
Occupation: Tech Support

The General thread for Computers And Internet

#18

Post by John Thomas8 »

somerset wrote: Thu Nov 24, 2022 1:00 pm
What a lot of folks don't realize is that many high performance chips are already made the way Intel is proposing (including some of Intel's chips going back 20+ years). It's common to design just one fully-featured chip, then "de-feature" it with wirebond or fuse options. This sounds wasteful, but the cost of laying out, validating and fabricating two different processor designs is usually far higher than the cost of wasted silicon real estate. Easier to just make one die and sell it as two (or more) products. The idea of making this "de-featuring" reversible is new, and maybe a little bit controversial from a marketing perspective, but not surprising. What you're really paying for is IP, and making it available via subscription is essentially the same as software licensing.
A place I worked at in the mid-90s bought some gear from HP (Unix/Motorola 68000s) that had processors turned on by license. It was completely kitted out, but you could only use what you bought the license to use.


User avatar
tek
Posts: 1869
Joined: Mon Feb 22, 2021 10:15 am

The General thread for Computers And Internet

#19

Post by tek »

RTH10260 wrote: Mon Nov 28, 2022 9:30 pm :doh:

should you own hardware of this brand you may want to burn it down asap
This is a pretty amazing screwup.
Well, actually, it isn't amazing at all. When mentoring hackathons, I find security is nowhere near the thoughts of almost all the participants.
These products are just advanced-stage hacks.

Also, Costco:
“Stay off the cloud and protect your privacy with local storage on HomeBase”

This is the Eufy marketing in Costco. Where I bought multiple of these cam devices.

I have been part of the Anker eco system for at least 5 years and never NEVER did I think these guys were anything but on their A game.

Now we find out the videos and data are stored unencrypted right there on the Amazon servers they are using. Not only do Amazon have access to this, but if stored in the USA then so do 5-eyes too. This isn’t tin foil hat stuff, we know what goes of from the Edward Snowden leaks in 2014.
I've actually been an Anker fan.. that's over. Though hard to know the rest of these companies are any better.


User avatar
RTH10260
Posts: 8196
Joined: Mon Feb 22, 2021 10:16 am
Location: Switzerland, near the Alps
Verified: eurobot

The General thread for Computers And Internet

#20

Post by RTH10260 »

TechScape: Enter the multiverse – the chat-room game made of AI art
An exciting multiplayer Discord game asks you to find things in the multiverse through an AI image generator. The hallucinatory results could mark a new frontier for AI art

Alex Hern
Tue 29 Nov 2022 11.40 GMT

The Bureau of Multiversal Arbitration is an unusual workplace. Maude Fletcher’s alright, though she needs to learn how to turn off caps lock in the company chat. But trying to deal with Byron G Snodgrass is like handling an energetic poodle, and Phil is a bit stiff.

Sorry, that was unclear. Byron G Snodgrass is an energetic poodle. Phil is a plant. A peace lily, I think.

The three work as arbiters, managing a few hundred caseworkers as they carry out the work of the Bureau: scanning through the multiverse for inspiration, information and innovation. Although, if you ask me, the Bureau’s gone a little off-course recently. Is it really a good use of all that technology to set me to work finding the best meal in all of existence?



https://www.theguardian.com/technology/ ... telligence


User avatar
Foggy
Dick Tater
Posts: 6576
Joined: Mon Feb 22, 2021 8:45 am
Location: near 35°55'N 78°33'W
Verified: my air is rarefied 🙄

The General thread for Computers And Internet

#21

Post by Foggy »

As y'all know or are vaguely aware, computers are made with semiconductors.

In a recent post, JohnPCapitalist mentioned something he said were "sub-30 nanometer semiconductors".

My dad's masters thesis was about gain and loss in transistors, the revolutionary new semiconductors that were invented in 1947. At that time, I don't think they could even measure anything as small as 30 nanometers.

A nanometer is a billionth of a meter.

"Let's get small." - Steve Martin


All my ideas are worth twice their weight in pure gold. :towel:

Here, put one on the scale, let's check.
See? Two times ... that was a small one.
Mood: translucent
User avatar
tek
Posts: 1869
Joined: Mon Feb 22, 2021 10:15 am

The General thread for Computers And Internet

#22

Post by tek »

When I was first peripherally involved with semiconductor design, the world was at about 7000nm (7um)
I ceased to be directly involved with semiconductor design back around 750nm (1991-ish).
Even then, I was already at the "OK, I take your word for it" stage..

Down below about 20nm, the traditional 2-dimensional Field-Effect Transistor pretty much stops working. This led to the "FinFET", where the active part of the transistor is a 'fin' that stands up from the surface of the wafer so that you can put a gate (the control electrode) on three sides (instead of one, or in rare cases two, sides)
d-1.jpg
d-1.jpg (98.73 KiB) Viewed 144 times

Tiny stuff.


User avatar
noblepa
Posts: 1826
Joined: Mon Feb 22, 2021 2:55 pm
Location: Bay Village, Ohio
Occupation: Retired IT Nerd

The General thread for Computers And Internet

#23

Post by noblepa »

I read somewhere, a long time ago, that as transistors got smaller and smaller, they became more susceptible to interference from cosmic radiation, especially in high-flying aircraft.

The idea is that when the transistors, or the spot in a chip that contains the semiconducting material gets small enough, one particle (an alpha particle?) can hit the transistor and change the contents of the spot. With larger transistors, the particles don't have enough mass to affect the transistor, but in the very, very tiny ones, the particle's mass is a sufficient fraction of the size of the transistor.

This means that a hit by cosmic radiation, which constantly bombards the earth, could change a bit from one to zero or vice versa. Depending on the bit in question, that could cause the system to malfunction. High-flying aircraft are more vulnerable, because the atmosphere filters out much of the radiation. Even using memory with parity might not prevent the malfunction. It would just allow the system to detect that one bit in a byte or word had changed. It wouldn't indicate which bit.


User avatar
tek
Posts: 1869
Joined: Mon Feb 22, 2021 10:15 am

The General thread for Computers And Internet

#24

Post by tek »

noblepa wrote: Sun Jan 15, 2023 12:59 pm I read somewhere, a long time ago, that as transistors got smaller and smaller, they became more susceptible to interference from cosmic radiation, especially in high-flying aircraft.
Yes. This is why space-rated parts are rarely cutting-edge stuff..
I did a project a couple years ago with these parts, which were a few generations behind similar earthbound parts
https://www.microchip.com/en-us/product ... rant-fpgas

Also, DRAMs are susceptible to cosmic rays and even radiation from the epoxy the chips are packaged in.. DRAMs (memory chips) are probably the highest information-density parts made (so highest density of things to upset)


User avatar
keith
Posts: 2109
Joined: Mon Feb 22, 2021 10:23 pm
Location: The Swamp in Victorian Oz
Occupation: Retired Computer Systems Analyst Project Manager Super Coder
Verified: ✅lunatic

The General thread for Computers And Internet

#25

Post by keith »

Heck, even when memory cores were actual cores, as in donut shaped things that you could see with your eyes and pick up with you fingers and could open the cabinet door and see the wires running through them, they occaisionally got zapped by a cosmic ray.

The first 360 I worked on, the SE would blame a crash on cosmic rays about once a year. Course that was just because he couldnt figure it out... they really aren't that common. And anyway error detection and correction got really good.


According to Woody Allen, Drew Barrymore sings so badly, deaf people refuse to watch her lips move.
Post Reply

Return to “Computers and Internet”