EEVblog Electronics Community Forum
General => General Technical Chat => Topic started by: Beamin on September 21, 2018, 01:54:10 pm
-
So Pc's are x86 CPU's while apple android and most other things are ARM? Wen did this happen in the 1980's? I remember seeing that in the early 80's you bought the software that matched your processor. Commodore programs for commodore, apple for apple, PC for PC(sort of wasn't there different types of DOS because there wasn't just a few chip manufactures like intel and AMD?). When did it become standardized or rather was it dictated by the market and not a standard? Who makes the chips for apple? intel but not using their x86 architecture?
What is used inside an arduno or a pic? I know these aren't x86 or ARM but are they simpler versions of this?
Does software dictate processor or is it the other way around? Will we ever see another architecture come out hypothetically or have we found that "round is the best shape for the wheel: With the ARM being a circle wheel and x86 being a 900 sided polygon that we add sides to as new technology comes out?"
Or is a turing machine a turing machine and to come up with something else just doesn't make sense based on binary transistor based gates making the abstract 1/0 come into physical form.
I think if quantum computers become common you won't actually own one, but rather you will buy a monthly plan where you get a % of it's resources or it's time and your device will just be a terminal. The wireless mesh network could make this a great idea with nothing centrally controlled or if we keep the monopoly with ISP's like they do now it would be a dark and expensive monopoly. I heard ISP's want to change their business model to the Cable TV business model where you buy a package like you do with TV; movie channel package, sports package etc. They would offer a search engine package, a social media package, or start selling things alacart. This would make the internet very expensive (high profitable) and lose our internet freedom. Having worked at comcast they would love this idea. When Netflix stopped mailing DVDs and started streaming the CEO said "Netflix is making an extra 1 bazillion dollars a year and WE ARE ENTITLED TO THAT MONEY". Funny he used "entitled" since that party is all about ending "entitlements" like the social security that we are in fact entitled to because unlike comcast we paid into it. Damn right we should get our entitlements. But that's the end of my political rant couldn't help myself.
When making chips do they all follow a basic design that we found to work the best over the years, unless you want something very specialized you use an ASIC. But it seems even ASIC can be replaced by FPGA unless you need maximum efficiency like a bitcoin miner where each watt of heat = lost profit or makes it unprofitable.
Seems like the future of things will all contain and AT style arduno chip at its core and you just build your device around it. Alarm clock throw an $1ATmicro in it. Dishwasher At micro. Then repairing them would be possible as it could just be interchangeable models. A whole industry could come from this: A company needs a product based on this chip and companies could offer the firmware to make it work leaving the company to better use it RD on making the product instead of writing code. If I was an EE and I wanted a star trek house everything would work like this.
-
Nice question.
Actually no architecture, instruction set or hardware implementation is completely superior to another.
x86 systems became a "de facto standard" for personal computers in the mid '80s mainly because they were offered in IBM based computers. The strong IBM brand made that product largely accepted and appreciated, regardless of the actual performance or real need. The first personal computers have been bought out of curiosity; I know very few people who actually got advantage of having an 8MHz computer at home! Everyone else was just playing - it's no mistery the first massively successful electronic products were videogames, starting with Pong, through C64, Amiga, Nintendo, etc. just to name a few.
But again, regarding personal computers we are talking about a very specific market, which boomed in the '90s and only by the mid 2000s met its plateau. With the mature advent of Pentium 4 processors, personal computers basically made everything we expected them to do (internet surfing, sound, picture and video streaming, videogames, what else?), so the "Megahertz war" suddenly stopped. The Desktop PC market is basically dead, and even the "Enthusiast" customer base is starving.
Oh, I remember the days when we waited for the new GPU to be released on the market, being it a GeForce3, 4 or whatever it was; then rushing to buy some more RAM to allow our game to run better... how much money has been wasted, in retrospective!
From the mid 2000s, the IT world keyword switched from "performance" to "battery life". We saw therefore the rise of low power CPUs, the return of the ARM platform, the development of incredibly powerful embedded processors and memories, but again, now this market also stalled.
Does anyone need a phone with 16GB of RAM? No, and in fact no one is producing it.
So, to summarize it all, development is stringly driven by the market need. People wanted 200W TDP 3.0GHz processors? We made them. People now want octa-core 2.2GHz processors that run on 1W? We make them.
Does anyone want or need a 20GHz processor? No. We don't make it. People wanted performance on real daily tasks, therefore we optimized our CPUs with extended instructions set, so we could avoid pumping up only the clock.
Many other architectures have been used during the years, even if they passed almost unnoticed:
- The touchscreen controller on your smartphone is very likely running on an 8051 core, which is 40 years old. It if works, it works; no need to change a perfectly running, stable and free System (the license expired!)
- The Arduino boards run on an Atmel microcontroller, which is based on an AVR core architecture. AVR was developed by two students long time ago, and gradually saw success in the industry and nowadays became famous in the "maker" movement.
- PIC microcontrollers started off as simple-to-program RISC devices, and gradually evolved up to the 32-bit Monsters we have today.
Each of them has a specific market, which justifies their existence.
-
I think if quantum computers become common you won't actually own one, but rather you will buy a monthly plan where you get a % of it's resources or it's time and your device will just be a terminal.
Sounds a lot like timeshare. We haven't done that in 30+ years. But there was a time when 8080 and Z80 code wasn't written on the host, the programmers signed in to timesharing systems. By 1980 this was on its way out because the host systems had enough memory for higher level language compilers like C, Fortran and PL/I.
So, my tower does something over 300 GFLOPs, why on earth would I need access to a quantum computer. I have just one application that could use more horsepower and, since it is the free version, it won't use more than 4 cores. If it were capable of using them, I would simply buy a 32 core processor and call it good.
When making chips do they all follow a basic design that we found to work the best over the years, unless you want something very specialized you use an ASIC. But it seems even ASIC can be replaced by FPGA unless you need maximum efficiency like a bitcoin miner where each watt of heat = lost profit or makes it unprofitable.
Well, sure, the basic transistor cell is pretty well known but the overall architectures of the x86 and ARM are vastly different. ARM is what x86 would like to be but it simply wasn't possible back in '73 when the 8080 came out and when the 8086 was introduced later, there had to be some commonality in hardware. Code needed to be more or less source level compatible.
You really can't talk about the various hardware without understanding the architecture and, perhaps more important, the instruction set. The ARM instruction set is designed specifically for pipelined processors and one clock per instruction. It is a RISC architecture so you won't find 'move string' instructions in an ARM processor.
There has been considerable debate over the years on RISC vs CISC and I'm not convinced it is settled yet.
Hardware is meaningless! People could care less what processor they have. RISC, CISC, ARM, x86? Who cares? It is the software the user sees. Email, Chrome, Word, Excel, and so on. These are what the user wants. At the present time, most large software programs are targeted at x86. ARM is playing catch-up on speed and throughput but that's because the x86 has had a 30 year headstart. It is also because of the cost of creating a second version of the code to run on an ARM. Microsoft has had miserable success getting Windows onto an ARM device. They target resource limited devices like cell phones and there just isn't enough speed and memory to run Windows.
Microsoft tried Windows RT on an ARM device (Surface RT). Nobody cared! It didn't have enough horsepower to be interesting.
As ugly as the x86 architecture is, relative to the clean ARM architecture, it's going to be around for a very long time. It is fast, capable and usable.
https://www.infoworld.com/article/2610369/processors/intel-vs--arm--two-titans--tangled-fate.html (https://www.infoworld.com/article/2610369/processors/intel-vs--arm--two-titans--tangled-fate.html)
-
This new lap top is 1.6/1.8 Ghz like the last one, but I do notice a difference in i3 and i5 especially turn on time. Why 1.6/1.8? One for power saving and one for when its plugged in? The worst time to buy a laptop was the early 2000's when they cost thousands of dollars and in two years they were too slow to run software so you literally had to replace it. Now it seems like everything is less then 1000$ and performance/disk speed/memory doesn't matter all that much for normal use. That's a good thing although the novelty of "look how much faster" is gone. Last time I filled a hard drive or had to delete something was on a netbook.
So what architectures exist now? ARM and x86 for cpus and AVR for MCU's? Aren't Arms just upgrades RISC with the R standing for RISC?
Do ASIC have a general design that follows some other architecture or are these completely built from the ground up? Seems like you would need a computer program to make them as things are too complicated to design by hand anymore. I remember in the 90's (early?) intel said the layout of their chips was comparable to a street map of NYC. Now in guessing its 1000 NYC's or bigger? I saw an interview where someone at AMD said the product they sell isn't really their chips but rather the software that can be told do XYZ and then makes a chip that does that doing complex algorithms. The first part of chips not to be layed out by hand was the memory since it's just repeating cells. On early chips you can see under a microscope the hand layed out regions are the small messy looking ones and computer "autorouter" ones are the neat repeating pattern parts. Shouldn't this mean that autorouter software should be better then hand drawn since they can/have to do this for chips? Or are the companies that make chips just not willing to sell software that powerful?
-
Although x86 and ARM currently dominate the 32 bit and 64 bit markets, RISC-V is an emerging architecture that is looking to displace ARM from certain applications.
-
Although x86 and ARM currently dominate the 32 bit and 64 bit markets, RISC-V is an emerging architecture that is looking to displace ARM from certain applications.
Seems more likely to displace MIPS, purely from a licensing-cost perspective; RV basically is a slightly modified MIPS with the same (lack of) performance characteristics.
-
So Pc's are x86 CPU's while apple android and most other things are ARM? When did this happen in the 1980's? I remember seeing that in the early 80's you bought the software that matched your processor.
I dunno. Seems to me like in about 1980 you had essentially two choices of CPU: a Z80/8080, or a 6502. Z80s came from a bunch of different vendors, and mostly ran some form of CP/M. 6502s showed up in the more "personal" computers (Apple, Commodore, Atari) and ran a wide variety of proprietary operating systems with proprietary peripherals (especially video and sound.)In 1981 IBM introduced their PC with an x86 and boring but easily reproducible hardware. In 1984 Apple introduced the Mac with a 68k process, and somewhere in there a bunch of unix-like 68000 "workstations" (SUN was founded in 1982.) These eventually replaced the 8bit systems.Both x86 and 68k (and IBMPC+Clones and Apple) led to a long string of successor "derivative" chips. RISC looked favorable and Motorola was a bit un-focused, so 68k gave way to PPC and sometimes MIPS while Intel/AMD/Via/etc continued with the x86. x86 seemed to win the performance war on desktops, and ARM won the RISC battles by focusing on phones.
So I'd say that there have always been "approximately two choices" of CPU architecture (heh. And one of them has always been Intel.)
If you have a broader perspective, you find a lot more options. But there's also a much bigger difference between a low end x86 and a high-end x86 (or an ARM CM0 vs an Apple ARM phone-processor) than there used to be between a 6800 and a 6502 (for example.) If you set your performance goals a bit lower, there are MANY MANY choices (nearly all of which are much faster than the 1980s computers.) 6502 and Z80 can still be purchased (from the original semiconductor companies, even!), there's 8bit and 16bit PICs, AVR8, 16bit MSP430, 32bit AVR32, PIC32, 68k, PPC, and MIPS chips are still around and serve niche markets. ST8, 68xx, and a wide variety of 8051 clones and derivatives. Many DSPs. Renesas with at least two non-ARM architectures (previous Hitachi and NEC, IIRC.) FTDI, Maxim, and probably a bunch more with their own architectures. RISC-V. FPGA cores. Graphics, Neural Network, and AI processors.
Not all of those would be "good choices" for a new design, and many don't do well on a "cost effectiveness" scale (a modern 65c02 (CPU-only) is the same $6 as an ARM Cortex-M4 with 1M Flash, 256k RAM, and peripherals.) But you have quite a selection.
-
I've seen nobody mention that x86 (i.e 32 bit) is becoming nearly obsolete, if not already so? If you buy a new Windows machine today will be running a 64 bit version of Windows on 64 bit hardware (x64). The x86 architecture is limited to 2 GB of base user memory and is unreasonably limited in today's world.
Also for a long time Apple used to run on PowerPC hardware and only switched over to Intel when the price/performance benefit of Intel became an undeniable reality.
What's more there are now rumors that Apple may switch away from Intel in the near future. Each time the driver is that Apple is trying to deliver the best possible integrated system and is not just focused on one bit of the platform.
With Microsoft it is a bit different. Microsoft doesn't sell complete products, it sells the OS and applications. Therefore Microsoft needs the widest possible compatibility across a variety of hardware and thus needs to embrace (and promote) some kind of hardware standardization and consistency.
-
The Desktop PC market is basically dead, and even the "Enthusiast" customer base is starving.
People say that, however if you look at the numbers at least as recently as a year ago, far more desktop PCs are sold today than were sold during the peak of the PC golden age in the late 80s-early 90s. The market has shrank greatly but it is still bigger than it was in the pre-internet era. The PC is a mature commodity now, and during the biggest boom a lot of people had a PC just to get on the internet sine that was the only way to do it. Now a lot of those people get by with a mobile device, the people who used a PC for productivity are still using them.
They are mature though and upgrading frequently is no longer necessary. Used to be a 2-3 year old PC was almost hopelessly obsolete, now it's reasonable to use the same one for 10 years or more.
-
Also for a long time Apple used to run on PowerPC hardware and only switched over to Intel when the price/performance benefit of Intel became an undeniable reality.
No. Apple publicly stated that the principal reason for the switch was that they couldn't get a low enough power consumption in PPC chips for the targetted performance of their next generation of notebooks. That's supported by the fact that at the time of the switch their notebooks were running G4 processors, a whole generation behind the G5 processors they were using on desktop and deskside systems. The G5 processors had enough performance, but you'd have cooked your thighs if you used one in a laptop. I can tell you from experience that even the G4 notebooks were uncomfortably warm to keep on your lap.
-
So I'd say that there have always been "approximately two choices" of CPU architecture (heh. And one of them has always been Intel.)
You omit many of the processors in the workstation market 1980-2010 and all the processors in the minicomputer space of the 70s through mid 90s.
SPARC, MIPs, HP PA-RISC, DEC alpha, Motorola as the principal players in the workstation processor space, with a nod to Symbolics, Xerox, and Three Rivers.
More processors than you can shake a stick at in the minicomputer space: Data General, PR1ME, DEC, Honeywell, Interdata, Tandem, Varian, Wang, IBM system/38, HP, Microdata, Harris, General Electric, Packard Bell, Raytheon, Westinghouse and many more, all with unique processor architectures.
-
x86 (i.e 32 bit) is becoming nearly obsolete,... new Windows machine 64 bit hardware (x64).
Isn't x64 considered just another x86? It didn't stop being x86 when they went from the actual 8086 to 80286, or when they completely ditched the horrible segmented memory scheme to get the 386, or started building in co-processors with the 486. (Well, I suppose Intel might want us to believe that each step was "new", but you're talking the same base instruction set and stuff. Sort of(?))
The x86 architecture is limited to 2 GB
4GB? At least, 32bit windows supports up to 4G of physical memory (minus a bit that they need for IO Space and stuff.) I guess that there's an annoying sign bit problem that might show up for individual applications trying to use more the 2G.
You omit many of the processors in the workstation market 1980-2010 and all the processors in the minicomputer space of the 70s through mid 90s.
Yes; I specifically wanted to focus on the more "personal" computers - things practical for an individual to own. I should have include Sparc, though. And we left out mainframes and supercomputers, too. IBM still makes mainframes...
Apple ... to Intel when the price/performance benefit of Intel became an undeniable reality.
No. Apple publicly stated that the principal reason for the switch was that they couldn't get a low enough power consumption
Power consumption is part of "performance." The G5 Macs needed to be marvels of mechanical engineering just for the cooling; the 970MP in the 2005 G5 was a 95W CPU, at a time when the Intel equivalent (Xeon) was about 65W... (although that probably put Apple ahead of the game for their 8-core Xeon Mac Pros...)
-
So Pc's are x86 CPU's while apple android and most other things are ARM? Wen did this happen in the 1980's? I remember seeing that in the early 80's you bought the software that matched your processor. Commodore programs for commodore, apple for apple, PC for PC(sort of wasn't there different types of DOS because there wasn't just a few chip manufactures like intel and AMD?). When did it become standardized or rather was it dictated by the market and not a standard? Who makes the chips for apple? intel but not using their x86 architecture?
I haven't seen anyone clarify some misconceptions here:
Let me see if I can provide some historical perspective to help clarify what I hope to clarify:
Back in the day when there were Commodores, and Apples, and TI99/4A's and Atari 800's and so on... when you bought software you were buying software for the Platform, not for the CPU. By platform, I mean the CPU and the hardware around it. To help clarify some more: The Commodore VIC20, the Apple II, the Atari 800, and several others all used the 6502 processor. However you couldn't buy software for the "6502", you bought software for the platform, in this case, the specific computer you were using.
Fairly quickly, most computers started having a real operating system such as CP/M or DOS or later Windows or the Classic Mac OS. These operating systems were released for a specific hardware platform. DOS/Windows for the combination of an intel 8086 CPU with a rather specific set of peripherals. MacOS originally for a Motorola 68K processor and again a specific set of peripherals. With the introduction of these operating systems you cared less about the underlying hardware since the operating system hid most of the complexities. So you ended up buying an operating system for the platform, and software for the operating system, with the caveat that you often had to match the processor as well.
Nowadays both a desktop mac and a PC are both using the same underlying hardware architecture. So (with a bit of coercion) you can run MacOS on PC hardware and Windows on Mac hardware. For all intents and purposes these are the same hardware, just different operating systems. You still have to buy software for your specific operating system, CPU, and of course having enough storage. It should be noted that there is a lot more variation of the peripherals supported in modern operating systems than the early ones, but it still is very much hidden from the person writing the end software.
Now, let's talk about Linux and to a lesser extent Unix. Much of what I say about Linux applies equally well to some of the Unixes, so I'm not going to go down that rathole, just realize that Unix is similar to linux in some ways.
Linux is an operating system which is fairly easy to port to other platforms. Linux runs well on numerous CPU's and supports numerous peripheral sets. Linux is what is underneath Android. It's the OS for the raspberry Pi. And probably a lot of other things you don't even realize
In the Linux world, PC hardware (8086) and ARM processors seem to rule, however there is lots of Linux running and being sold which runs on other processors such as MIPS and RISC-V and PowerPC and on and on behind the scenes. Generally for modern Linux to run, you generally need a fairly beefy processor which usually takes quite a bit of power when running. With ARM cores being relatively low power while still running Linux with a fairly decent performance spec, it seems like a logical outcome for Arm to have much of this market share for Linux based devices. The fact that ARM seems to have won over MIPS for the cell phone market adds to this as well.
BUT... If you look at units shipped every year, you're going to find that ARM and x86 - which seem to be the most recognizable processors - aren't even close to the most popular processors on the planet. The reason is that there are billions and billions of microcontrollers embedded in everything - so many that it's hard to count. I will admit that ARM has a large (growing) chunk of this space, but there are still billions and billions of tiny non-ARM 8, 16 and 32 bit processors shipped for use in everything. And yes, the arduino uses an 8 bit AVR, not an ARM.
-
BUT... If you look at units shipped every year, you're going to find that ARM and x86 - which seem to be the most recognizable processors - aren't even close to the most popular processors on the planet. The reason is that there are billions and billions of microcontrollers embedded in everything - so many that it's hard to count. I will admit that ARM has a large (growing) chunk of this space, but there are still billions and billions of tiny non-ARM 8, 16 and 32 bit processors shipped for use in everything. And yes, the arduino uses an 8 bit AVR, not an ARM.
The 8 bit AVR Arduinos are the "classic" ones - there are also Arduinos based on ARM. In the embedded world, 32 bit seems to be dominated by ARM with MIPS a distant second, 16 bit seems to be mostly MSP430 and dsPIC/PIC24, and 8 bit mostly AVR, PIC, and STM8 plus a bunch of other architectures like 8051.
-
With Microsoft it is a bit different. Microsoft doesn't sell complete products, it sells the OS and applications. Therefore Microsoft needs the widest possible compatibility across a variety of hardware and thus needs to embrace (and promote) some kind of hardware standardization and consistency.
Are you overlooking the Microsoft Surface products? I doubt that they actually 'build' them but they certainly do 'sell' them.
And maybe we should include Xbox even though it is not a PC.
-
On the high end POWER is still competitive, these days with machines from third parties (not just IBM) such as the TALOS-II series (https://www.raptorcs.com/content/TL2WK2/intro.html).
SPARC still exists, but given the death of Solaris it's becoming even more of a niche platform.
-
I've seen nobody mention that x86 (i.e 32 bit) is becoming nearly obsolete, if not already so? If you buy a new Windows machine today will be running a 64 bit version of Windows on 64 bit hardware (x64). The x86 architecture is limited to 2 GB of base user memory and is unreasonably limited in today's world.
(...)
I'm not really sure I get what you are saying. x86 is an architecture and an instruction set, while x64 identifies the address width. An x64 machine runs x86 code no problem whatsoever. Am I not seeing something here?
-
BUT... If you look at units shipped every year, you're going to find that ARM and x86 - which seem to be the most recognizable processors - aren't even close to the most popular processors on the planet. The reason is that there are billions and billions of microcontrollers embedded in everything - so many that it's hard to count. I will admit that ARM has a large (growing) chunk of this space, but there are still billions and billions of tiny non-ARM 8, 16 and 32 bit processors shipped for use in everything. And yes, the arduino uses an 8 bit AVR, not an ARM.
Not to forget the gazillion 4-bit microcontrollers used in things like remotes and calculators. Bit suprisingly HP was still making calculators based on 4-bit controllers in 2002
-
I'm not really sure I get what you are saying. x86 is an architecture and an instruction set, while x64 identifies the address width. An x64 machine runs x86 code no problem whatsoever. Am I not seeing something here?
The x64 instruction set is a superset of x86. It could be that the practice of exclusively using the x86 instruction set with hardware that only supports x86 instructions would become extinct, though it seems that the use of x86 for more basic applications is on the rise too and that this will actually increase its use in scenarios where x64 isn't really needed. One example would be the new iPhone baseband processor, which turns out to be a x86 processor instead of the more traditional ARM processor.
-
I would think that if AI or some as yet unknown computing pardigm is developed, I suppose there could be a new architecture that is more appropriate than the existing x64 or ARM architectures. I understand that existing architectures are Von Neumann or Harvard. Perhaps newer architectures will be more brain like and use neurons and synapses and be more plastic and maybe even analogish. Hmmm, I wonder what chips the T100 terminators used?
I wonder too if the virtual and physical address spaces are large enough in the current architectures to handle some advanced ideas. I don't think Moore's law will actually allow us to build enough memory to need a 64 bit address, but perhaps some super array processor would allow a common address space that would use a big chunk of it. Or perhaps there's some sort of sparse addressing algorithm that does't actually need all the memory to be present, but if it were just a few bits longer...
Security may also dictate a change in the architecture. The recent exposure of vulnerabilities where kernel data could be revealed by processor timing is an example. Burroughs used a stack architecture in their mainframes that had tag bits in parallel with the data. The processor could only perform the correct operation with the data. ie., if the data were floating point, it couldn't perform an integer or bitwise operation on it. (Cripes, maybe the tag will be a blockchain!)
Cheers,
-
So Pc's are x86 CPU's while apple android and most other things are ARM?
Mostly now yeah. But:
- Apple and other manufacturers (with support from Microsoft) are looking into adopting ARM platforms for their laptops. Some of the newer "Smartphone" chips should reportedly compete with Intel Celeron's and Pentium's, which is plenty for a lot of people
- There are also x86 Android machines with Intel Atom's, but they haven't really caught on.
What is used inside an arduno or a pic? I know these aren't x86 or ARM but are they simpler versions of this?
The original Arduino uses the AVR core.
The PIC has some variety in cores, like PIC16 and PIC18s which are slightly different. However their general design philosophy remains the same.
Both chips have 8-bit CPU's and different approaches to the microcomputer problem (for example Harvard vs Von Neumann architectures).
ARM is yet another flavour..
And x86, in theory, also.
However, after many years of CPU generations, people noticed two groups started to appear: RISC and CISC processors. RISC use very simple instructions, so to do anything meaningful you need many of them. CISC tries to combine a complex operation into 1 instruction. It turns out that building CISC processors is really hard, so you often see modern x86 implementations translating CISC instructions to RISC using microcode in the CPU.
That is not to say RISC processors can't have microcode as well. It's a common approach to implement the controller for a datapath (https://en.wikipedia.org/wiki/Datapath).
Does software dictate processor or is it the other way around?
Software does mostly dictate what we use. The only reason why we like to keep x86 around is because all our software runs on it. The industry has more computing demands, which runs on x86, so we want faster x86 chips.
We ofcourse also want faster microcontrollers and ARM SoCs in smartphones, but those applications are heavily tied to power budgets. You can't fit a 65W CPU in a smartphone without it catching on fire or burning your hand. Why would a vendor then build a 65W ARM CPU if nobody wants to buy it; as it can't fit smartphones, and PC software doesn't run on it (yet).
If a player like Microsoft is able to make the Windows platform agnostic (i.e. the whole eco-system runs on ARM and x86), then you will probably see this situation change.
Will we ever see another architecture come out hypothetically or have we found that "round is the best shape for the wheel: With the ARM being a circle wheel and x86 being a 900 sided polygon that we add sides to as new technology comes out?"
Actually, Intel has gone 1 step further and placed a bet on VLIW processors with the Itanium series. They abandoned them a few years ago. Again it's not that it was a inherently bad architecture, but there were some difficulties, above all we had x86-64 that was far more easy to adopt.
In addition; I see CPU architectures are something where the "ultimate" design is an unicorn. Let's face it: you cannot make memory buses infinitely big, memory sized infinitely large, have an infinite amount of instructions or registers. In the end you decide to make an instruction 32-bits because that's what the current technology can cost efficiently support. Then you have a trade-off how you're going to devote those bits to what. Some trade-off may prove better in 1 scenario, but there a dozen more scenario's...
Or is a turing machine a turing machine and to come up with something else just doesn't make sense based on binary transistor based gates making the abstract 1/0 come into physical form.
If a system is Turing complete, it means it can simulate a Turing machine. A Turing machine in turn can model any algorithm you can think of. It will do it very slowly, and will never be used in practice, but at it's fundamental it's possible.
You can build hardware accelerators that will calculate only 1 task, and do it well (and power efficient), but you can't use that as a general purpose computer..
The Turing machine itself has nothing to do with logic gates. The Turing machine is described by computer scientists using sets and tuples. There is no notion of 1s and 0s strictly necessary here.
I think if quantum computers become common you won't actually own one, but rather you will buy a monthly plan where you get a % of it's resources or it's time and your device will just be a terminal. The wireless mesh network could make this a great idea with nothing centrally controlled or if we keep the monopoly with ISP's like they do now it would be a dark and expensive monopoly. I heard ISP's want to change their business model to the Cable TV business model where you buy a package like you do with TV; movie channel package, sports package etc. They would offer a search engine package, a social media package, or start selling things alacart. This would make the internet very expensive (high profitable) and lose our internet freedom. Having worked at comcast they would love this idea. When Netflix stopped mailing DVDs and started streaming the CEO said "Netflix is making an extra 1 bazillion dollars a year and WE ARE ENTITLED TO THAT MONEY". Funny he used "entitled" since that party is all about ending "entitlements" like the social security that we are in fact entitled to because unlike comcast we paid into it. Damn right we should get our entitlements. But that's the end of my political rant couldn't help myself.
I think you need to be less paranoid.
Secondly, I don't think we will have quantum computers in the hands of consumers for decades to come, if there will ever be a practical implementation for them.
When making chips do they all follow a basic design that we found to work the best over the years, unless you want something very specialized you use an ASIC. But it seems even ASIC can be replaced by FPGA unless you need maximum efficiency like a bitcoin miner where each watt of heat = lost profit or makes it unprofitable.
"Found to work the best" => proven to improve designs in academia and commercial research centers at the big manufacturers.
The only difference between an ASIC and a "normal" chip is the amount of applications. Both are designed using the same techniques.
FPGAs are similar to design with and more accessible, but both industries have their ins/outs, tricks and gotcha's.
ASICs are more power efficient and you can really work at a gate level. But they only work in sufficient volume.
Seems like the future of things will all contain and AT style arduno chip at its core and you just build your device around it. Alarm clock throw an $1ATmicro in it. Dishwasher At micro. Then repairing them would be possible as it could just be interchangeable models. A whole industry could come from this: A company needs a product based on this chip and companies could offer the firmware to make it work leaving the company to better use it RD on making the product instead of writing code. If I was an EE and I wanted a star trek house everything would work like this.
Companies want to make money to please their share holders. They have no interest in pleasing customers.
Without cynicism, manufacturers have really no incentive to work together. I think the only industry where you see that happen is the automobile industry, in particular the more price pressured market segments. Again this is also purely financially driven: developing a platform for a car that most not cost more than 10k$ is almost impossible to do, unless you can team up to double or triple your effective sale figures.
-
However, after many years of CPU generations, people noticed two groups started to appear: RISC and CISC processors.
Erm, no, nobody 'noticed two groups'. RISC computers were deliberately designed and designated that way. It wasn't some naming of an emergent phenomenon but a specific attempt to design computers around the idea that a Reduced (or simplified) Instruction Set would prove more efficient. The two pioneers were the RISC (https://en.wikipedia.org/wiki/Berkeley_RISC) project at Berkeley run by David Patterson (SPARC and SOAR were direct descendents of this) and the MIPS (https://en.wikipedia.org/wiki/Stanford_MIPS) project at Stanford run by John L. Hennessy, which is obviously the progenitor of the commercial MIPS architecture.
The respective heads of the two competing projects have since written books on computer architecture together which goes to show that competition and cooperation are not necessarily mutually exclusive.
Design and Implementation of RISC I. Séquin and Patterson (paper) (http://www.eecs.berkeley.edu/Pubs/TechRpts/1982/CSD-82-106.pdf)
-
Actually, Intel has gone 1 step further and placed a bet on VLIW processors with the Itanium series. They abandoned them a few years ago.
Hummm. So what explains that we just bought a bunch of systems with IA64 CPUs?
-
Hummm. So what explains that we just bought a bunch of systems with IA64 CPUs?
At this point being locked into VMS but unable to wait for the AMD64 (AKA x86-64) port is the only somewhat sensible reason left.
Itanium is dead, everyone that can get away with dropping support has, ex from Oracle (http://www.oracle.com/us/corporate/features/itanium-announcement-1867201.html) (who got sued by HP for this (https://www.theregister.co.uk/2016/07/01/jury_awards_us3_beeelion_to_hpe_in_oracleitanium_lawsuit/)):
On March 22, 2011, Oracle announced it has decided to discontinue all software development on the Intel Itanium microprocessor. The announcement is based on changing market conditions and follows both RedHat and Microsoft dropping Itanium support over the past year. Oracle continually evaluates customer demand for its software on server platforms.
-
Mostly now yeah. But:
- Apple and other manufacturers (with support from Microsoft) are looking into adopting ARM platforms for their laptops. Some of the newer "Smartphone" chips should reportedly compete with Intel Celeron's and Pentium's, which is plenty for a lot of people
Windows that doesn't run existing Windows apps pretty much eliminates the only reason to use Windows as opposed to Linux. In a world where Linux has more apps than Windows does, almost nobody would use Windows. And that's the situation with ARM.
-
In a world where Linux has more apps than Windows does, almost nobody would use Windows.
Sorry, but I don't think so. Over 95% of the population has no clue about computers and would be lost in the middle of the ocean if confronted with Linux. As much as I dislike Windows 10 and where it has gone, I understand that it caters to the majority of computer users. I can cope OK with Windows 7 as a "techie". Most non-technical people I know find Windows 7 to baffling and frustrating.
-
Android and ChromeOS have achieved mainstream status years ago. And Windows on ARM has failed over and over because there are next to no apps available.
-
In a world where Linux has more apps than Windows does, almost nobody would use Windows.
Sorry, but I don't think so. Over 95% of the population has no clue about computers and would be lost in the middle of the ocean if confronted with Linux. As much as I dislike Windows 10 and where it has gone, I understand that it caters to the majority of computer users. I can cope OK with Windows 7 as a "techie". Most non-technical people I know find Windows 7 to baffling and frustrating.
But he's not talking about the Windows 10 you're talking about, he's talking about the Windows 10 that is effectively Windows RT, ie it's called "Windows", it resembles Windows, but it doesn't run any of the existing x86 software, which is the whole reason anyone uses Windows. Naturally Microsoft has confused the hell out of everybody by offering multiple products called "Windows" that are not at all the same thing.
-
But he's not talking about the Windows 10 you're talking about, he's talking about the Windows 10 that is effectively Windows RT, ie it's called "Windows", it resembles Windows, but it doesn't run any of the existing x86 software, which is the whole reason anyone uses Windows. Naturally Microsoft has confused the hell out of everybody by offering multiple products called "Windows" that are not at all the same thing.
But this comment has nothing to do with what NiHaoMike said, or what I responded to.
NiHaoMike said, "If Linux had more apps than Windows, then few people would use Windows."
So we are talking about Linux. And it doesn't matter how many apps Linux has, people still won't use it. Because running x86 software isn't the reason people use Windows. Having lots of apps isn't the reason people use Windows either. If you think that's why people use Windows you also have missed the point entirely.
-
People do use Linux. A lot. Why else do they sell so many Android devices?
-
Comparing Android to even ubuntu is a bit twisted...
-
Both are Linux...
-
However, after many years of CPU generations, people noticed two groups started to appear: RISC and CISC processors.
Erm, no, nobody 'noticed two groups'. RISC computers were deliberately designed and designated that way. It wasn't some naming of an emergent phenomenon but a specific attempt to design computers around the idea that a Reduced (or simplified) Instruction Set would prove more efficient. The two pioneers were the RISC (https://en.wikipedia.org/wiki/Berkeley_RISC) project at Berkeley run by David Patterson (SPARC and SOAR were direct descendents of this) and the MIPS (https://en.wikipedia.org/wiki/Stanford_MIPS) project at Stanford run by John L. Hennessy, which is obviously the progenitor of the commercial MIPS architecture.
The respective heads of the two competing projects have since written books on computer architecture together which goes to show that competition and cooperation are not necessarily mutually exclusive.
Design and Implementation of RISC I. Séquin and Patterson (paper) (http://www.eecs.berkeley.edu/Pubs/TechRpts/1982/CSD-82-106.pdf)
Unfortunate wording. There are ofcourse more ways to build a computer, like a single instruction machine, but in general RISC and CISC were "competing" architecture foundations.
There were genuine benefits to x86 and perhaps CISC at first. For example the variable sized instruction size saved memory, which is not as scarce as it was back in the day.
Actually, Intel has gone 1 step further and placed a bet on VLIW processors with the Itanium series. They abandoned them a few years ago.
Hummm. So what explains that we just bought a bunch of systems with IA64 CPUs?
They introduced the "last series" in march 2017, with no upcoming generation planned. They may sell these chips for the upcoming years to support these business and enterprise solutions. But without any upgrade path, you could very well speak of "abandoned".
Mostly now yeah. But:
- Apple and other manufacturers (with support from Microsoft) are looking into adopting ARM platforms for their laptops. Some of the newer "Smartphone" chips should reportedly compete with Intel Celeron's and Pentium's, which is plenty for a lot of people
Windows that doesn't run existing Windows apps pretty much eliminates the only reason to use Windows as opposed to Linux. In a world where Linux has more apps than Windows does, almost nobody would use Windows. And that's the situation with ARM.
Quantity is only a number that doesn't say much. It's always about the specific applications.
For example, if you do PCB design for a living then odds are high you're still on Windows. If you need more advanced tools than Eagle or KiCad can offer, and you can't afford Cadence stuff, then you're basically out of luck on the Linux platform (unless you want to hassle with VM's, Wine, or whatever).
Similarly a lot of people only need a webbrowser, email, calendar, office suite and perhaps be able to play the odd game. Some of these even can run in the browser; look at ChromeBooks.
MS tried Windows RT a few years back, and I think it flopped because who wants to deal with double versions and licensing for e.g. Office or other programs. If an ecosystem really becomes seamless across x86 and ARM I think you'll see more people stepping over.
-
Both are Linux...
Yes but to the average person who uses android they would say "what is Linux?". All people see is what desk top/home screen looks like and how the top file/edit/view tab is set up. The innerworkings might as well be magic to them. I think Linux would be more popular with average people if there was one common flavor that 90% of people would use. As it is to get a program working it depends on which one with things working on somethings better then others I have found.
-
NiHaoMike said, "If Linux had more apps than Windows, then few people would use Windows."
So we are talking about Linux. And it doesn't matter how many apps Linux has, people still won't use it. Because running x86 software isn't the reason people use Windows. Having lots of apps isn't the reason people use Windows either. If you think that's why people use Windows you also have missed the point entirely.
Well the software is the reason everyone I know runs Windows. The people who don't need any legacy Windows software are the ones getting by these days with mobile devices, which are virtually all either a variation of Linux (Android) or BSD (iOS). There is still some momentum in that virtually all new computers come with Windows installed, but the attempt to phase out "legacy" Win32 software has played a substantial part in the plummeting sales of Windows PCs, and it's exactly why Windows RT flopped like a lead balloon. It was called Windows, it looked like Windows, but it couldn't run the enormous library of Windows software, so people soundly rejected it. Clearly people run Windows because they need it to run the software they use.
-
Well the software is the reason everyone I know runs Windows. The people who don't need any legacy Windows software are the ones getting by these days with mobile devices, which are virtually all either a variation of Linux (Android) or BSD (iOS). There is still some momentum in that virtually all new computers come with Windows installed, but the attempt to phase out "legacy" Win32 software has played a substantial part in the plummeting sales of Windows PCs, and it's exactly why Windows RT flopped like a lead balloon. It was called Windows, it looked like Windows, but it couldn't run the enormous library of Windows software, so people soundly rejected it. Clearly people run Windows because they need it to run the software they use.
The new ARM version of Windows runs x86 programs, albeit emulated and a bit slower. Most programs won't mind as they're not that performance hungry. Microsoft has essentially fixed that problem. They've historically aimed for very long term legacy support from an enterprise perspective and they've been rewarded for it.
-
I think they're too late to the party, they shipped RT years before there was any sort of usable x86 emulation and it immediately got a reputation as being crap because it couldn't run any real software.
Even so, once you're going with emulation there is little point to be chained to Windows, Linux or MacOS can run x86 software in emulation too. Hardly anyone has ever run Windows because they like dealing with Microsoft and/or really like it, they run it because it's required by the software they use. For example I'm indirectly involved in two businesses that are entirely Windows based. One is a machine shop that runs Windows because the CAD software that works with the CNC machines is Windows only. The other is a veterinary clinic because the practice management and digital x-ray software (and drivers) is Windows only. It's not because they have some attachment to Windows, it's because they need it to run the software they depend on to run the business.
-
I think they're too late to the party, they shipped RT years before there was any sort of usable x86 emulation and it immediately got a reputation as being crap because it couldn't run any real software.
Even so, once you're going with emulation there is little point to be chained to Windows, Linux or MacOS can run x86 software in emulation too. Hardly anyone has ever run Windows because they like dealing with Microsoft and/or really like it, they run it because it's required by the software they use. For example I'm indirectly involved in two businesses that are entirely Windows based. One is a machine shop that runs Windows because the CAD software that works with the CNC machines is Windows only. The other is a veterinary clinic because the practice management and digital x-ray software (and drivers) is Windows only. It's not because they have some attachment to Windows, it's because they need it to run the software they depend on to run the business.
RT is dead. It's gone. What they're doing now is an entirely new incarnation.
Why people would want to stay with Windows? Because they now get ARM benefits while still having access to their massive software library in what's more or less a seamless experience. That's a world apart from "other stuff runs emulations too".
-
Whilst power users will not agree, we are fast approaching a time in which the web browser is the operating system.
This is why tablets and chromeos have a place in the modern computing world and for alot of people, thats all they will ever need.
This will give completely new operating systems and platforms a chance to succeed, aslong as their first priority is getting a modern web browser running nicely.
-
RT is dead. It's gone. What they're doing now is an entirely new incarnation.
Why people would want to stay with Windows? Because they now get ARM benefits while still having access to their massive software library in what's more or less a seamless experience. That's a world apart from "other stuff runs emulations too".
If it doesn't seamlessly run "legacy" x86 software then it's just RT under another name. Just like Win10 is just Windows 8.3 despite them overcompensating and skipping a version number to show how new and different it is.
And again, the reason people run Windows is because they want to use software that requires Windows. Whether that's run natively on a x86 processor or under some form of emulation, it's still all about the software (and hardware drivers). If their original attempt at completely deprecating so-called legacy x86 software in favor of "modern" apps had succeeded, they would have succeeded in killing Windows entirely. The huge library of x86 software is almost entirely the reason for Windows to exist. People who just want to surf the web and use browser apps don't care what operating system they have, hence the explosion in popularity of non-Windows tablets. "Windows" on phones and ARM tablets failed spectacularly because despite the name, it wasn't Windows.
-
I think Linux would be more popular with average people if there was one common flavor that 90% of people would use.
The variations between the Linux distributions are probably smaller than the differences between Windows XP, 7, 8 and 10. That didn't stop the popularity of Windows. There have never been any "common flavor" of any software that 90% of people use.
-
That's because Windows has long had one very big thing going for it, an absolutely enormous library of existing software, coupled with rather good backward compatibility. A large amount of software written for Windows 2000 will still run today on Windows 10. Windows is the defacto standard, if a vendor is going to produce a piece of traditional software for one operating system, they'll almost always target Windows. If there's one piece of Windows-only software that a person needs, they'll almost always choose Windows as the OS.
Some really innovative operating systems have come and gone over the years, and almost without fail their demise has been due to a lack of software. An operating system is absolutely useless if it won't run *all* of the software you need to use.
-
Some really innovative operating systems have come and gone over the years, and almost without fail their demise has been due to a lack of software. An operating system is absolutely useless if it won't run *all* of the software you need to use.
That's true if the software you need to use is software you own. You don't want to shell out for it, or a functional equivalent, again if you switch operating systems. But does that hold true in a world where all the vendors are trying to move to the rental model of software as a service? If changing to a new (incompatible) operating system means you have to change software, does it matter so much if you have to cancel one software subscription and start a new subscription to something functionally equivalent as long as you can seamlessly move your data across?
-
Some really innovative operating systems have come and gone over the years, and almost without fail their demise has been due to a lack of software. An operating system is absolutely useless if it won't run *all* of the software you need to use.
That's true if the software you need to use is software you own. You don't want to shell out for it, or a functional equivalent, again if you switch operating systems. But does that hold true in a world where all the vendors are trying to move to the rental model of software as a service? If changing to a new (incompatible) operating system means you have to change software, does it matter so much if you have to cancel one software subscription and start a new subscription to something functionally equivalent as long as you can seamlessly move your data across?
That depends. Switching from one software to another is not trivial in many cases, licensing aside it requires one to re-learn how to use it, and in most cases translate their data over to the new product. This also assumes that a functionally equivalent option exists for the OS in question. There are a significant number of products out there that are simply not available for Linux. Think serious CAD/CAM/CNC stuff, drafting and architecture, business management software, software made specifically to operate things like sign cutters, CNC routers, printers, digital xray systems, specialized medical equipment, etc. Then you have hardware drivers for much of the above. In most businesses it is one of these items that dictates running Windows. For home users it is typically games, I've lost count the number of people I've heard say they'd switch to Linux except they're a gamer.
Now in many ways this subscription model and the proliferation of browser based apps *has* made the operating system less relevant than ever. It's one reason the PC market is past its peak and shrinking. Many non-power-users can now do everything they need to do using a tablet, Chromebook, or other non-traditional machine. They are not running software that requires Windows, so they are not running Windows, even though there is "Windows" that runs on mobile devices. Ironically Windows Mobile faces the same challenge that non-Windows desktop operating systems face, it doesn't run certain software that people expect to use. If I have a choice between a phone that has 100% of the apps I want, and one that has 50% of the apps I want, the choice is easy.
-
Power users like me loathe browser apps, the average consumer doesn't seem to mind so much though. There's a whole generation of adults now who have never known life without the internet and many are now growing up using tablets and smartphones from a young age. Desktop and laptop PCs are reverting to more of a niche market, as business/productivity tools. People like me will use a traditional PC into the foreseeable future. Whether or not that PC runs Windows really depends on what Microsoft decides to do with Windows. I don't like where they've gone in recent years at all but I keep Win7 as my primary OS because I have a significant amount of software I like that is Windows-only. For work I have a Mac, I'm actually finding it quite nice but the built in non-replaceable(easily) battery, SSD and RAM and lack of standard USB ports is a non-starter for a machine I'm going to spend my own money on. I hate having to carry around a bunch of adapter dongles. The machine is much thinner than it needs to be, and achieves that by leaving out a lot of convenient features.
-
Power users like me loathe browser apps, the average consumer doesn't seem to mind so much though. There's a whole generation of adults now who have never known life without the internet and many are now growing up using tablets and smartphones from a young age. Desktop and laptop PCs are reverting to more of a niche market, as business/productivity tools. People like me will use a traditional PC into the foreseeable future. Whether or not that PC runs Windows really depends on what Microsoft decides to do with Windows. I don't like where they've gone in recent years at all but I keep Win7 as my primary OS because I have a significant amount of software I like that is Windows-only. For work I have a Mac, I'm actually finding it quite nice but the built in non-replaceable(easily) battery, SSD and RAM and lack of standard USB ports is a non-starter for a machine I'm going to spend my own money on. I hate having to carry around a bunch of adapter dongles. The machine is much thinner than it needs to be, and achieves that by leaving out a lot of convenient features.
Average consumers are very cost driven and therefore bear with browser apps, but that doesn't mean they like them. People don't use the junkiest tablets because they enjoy using them, they use them because they're cheap.
-
But he's not talking about the Windows 10 you're talking about, he's talking about the Windows 10 that is effectively Windows RT, ie it's called "Windows", it resembles Windows, but it doesn't run any of the existing x86 software, which is the whole reason anyone uses Windows. Naturally Microsoft has confused the hell out of everybody by offering multiple products called "Windows" that are not at all the same thing.
You're either talking about Windows 10 S or Windows ARM, neither of which are effectively Windows RT. Windows 10 S runs on normal hardware and is upgradable to Windows 10 and Windows 10 ARM runs on mobile hardware but it's x86 capable. I'm not sure why either is confusing, as they're called differently than the regular Windows 10.
-
Intel x86 architecture has so much legacy, it's been added on to for decades and surely there is an 8088 still stashed inside the i7.
I think ARM will overtake Intel and dethrone that monopoly. Especially since Intel's ATOM processors were a total flop, their one chance to get into phones, tablets, set-top boxes etc. completely bungled and ARM got the turf.
I had a Computing Science course on CPU architectures, and we did a lot of math to calculate the performance of various CPU's. Harvard, von Neumann, cache, pipelining, pre-fetch, bus widths, CISC, RISC etc. It's a lot of fun to calculate that kind of thing but I found it far from reality.
Adding more cores and other CPU modules to improve speed, and increasing clock speeds eventually won out. The compiler's ability to utilize/optimize use of the CISC or RISC op-codes is also a factor. It kind of settled into memory latency as the bottleneck. Good old 68000 died because of 4 clock cycles/access.
CPU monopolies are disliked by all, so I see ARM having the advantage in being multi-sourced. AMD got to second-source Intel from IBM's demand for that.
-
Intel x86 architecture has so much legacy, it's been added on to for decades and surely there is an 8088 still stashed inside the i7.
The 8088 is so tiny by modern standards that I wouldn't be the least surprised if there are a few stashed into modern Intel CPUs for low level management purposes. I'm also imagining that it could be possible to design a modern x86 chip as a modernized 386 to allow it to boot 16 bit code, then the big 32/64 bit cores are activated when the chip is switched into that mode, at which point the 386 is repurposed for power management and whatnot.
-
FPGAs
The world is your Oyster
-
FPGAs
The world is your Oyster
But 10x slower and 10x hotter than custom silicon.
So unless you need to reconfigure in the field, or are a low volume application then a Field Reconfigurable Gate Array is not the technology you are lookimg for....
-
Same as in a real free market economy, the CPU market trends to full consolidation or an oligopoly of 2-3 architectures, most likely two.
Why would anybody need yet another CPU architecture, what is missing from the actual ones ?
The answer is nothing, the current CPU architectures do what is needed to be done and more, if there would have been something worth taking from the cadavers of the dead architectures, it would have been taken already.
EVERYTHING that was possible to be tried, was tried, enormous amounts of cash went into research and the stuff that you see has the optimal features needed, and has them patented.
It could be that there are some corner cases that are covered by the GPU processing, like complex simulations and coin mining, but that's it, no "unknowium" CPU architecture will make your Office and Internet Explorer run faster as it is now and this is what matters.
Of course there will be implementation optimizations, if you're apple or other deep pockets company you ca afford to play around a bit, but on a fundamental level nothing will change.
There also some desperate companies (HDD makers) trying to shave a few cents in costs using Risc V or similar in their controllers, but once the patents expire they will back to ARM.
On the other side, the current programming model and the usage of the existing features it's so unoptimal that is scary, the hardware is waaaaaaaaaaaaaaay ahead of the software, but sadly this will remain like this for a long time, the Moore's law, "who's first on maket wins" and the comodization and bastardization of the programmer profession has lead to mostly garbage programmers that really have no clue of their hardware architecture and treat it as an "appliance", "black box" or whatever abstract model, as far away from the real thing as possible. Because, you know, it's not worth spending time on this shit, the compiler, OS and libraries will take care of all the optimization, just compile it and launch it, we will optimize it later (AKA never).
Of course, deep in the three letters agencies bunkers and in the luxurious offices of Goldman Sachs there are the few remaining real programmers looking at the hardware schematics and detailed specifications of the instruction pipeline and chipset (not available anymore to general public since the times of 486) trying to shave a few nanoseconds here and there, but this is so rare that it can be ignored.
Cheers,
DC1MC