When would you say that x86 computers became "black boxes"?

RetroComputingCollector

RetroComputingCollector

When do you think x86 computers became "black boxes" that are no longer yours to work with and own? I would say it was when the Intel ME and AMD PSP were introduced. I think the state of modern consumer computing is terrifying. The ironic thing is that the IBM PC architecture succeeded because it was open. Now it has just been completely closed up again. My ideal future computing architecture would be RISC-V based, with standardized support chips, memory layout, expansion slots, ports, graphics, and firmware. Considering how the IBM PC was thrown together in under a year, someone should be able to throw together a next-gen open standard using already available components in just a few years.
 
Year of the RISC-V Desktop...

Nothing like stray architecture to keep nerds amused. They need to add quantum to the name.
RISC-Q or maybe RISC-AI
 
I agree AMT/ME sucks but look at other end of the spectrum. APU2 with coreboot/SeaBIOS.
You want features nobody is forcing you to buy them. There are alternatives.
 
There will always be closed-up, proprietary components in bleeding-edge hardware. Even in the original IBM-PC, the BIOS was closed source and properiety as what-not.

But why would you need all of your hardware components to be “open” anyway? If the interfaces are well-documented, you are ready to go. For example, the main interface to the CPU is machine language, for which you can download the relevant PDFs freely on the internet.

But I agree: other components like the graphics card or GPU, are more difficult. If you depend on these beyond user-consumer-level, you are in trouble.
 
Holger said:
There will always be closed-up, proprietary components in bleeding-edge hardware. Even in the original IBM-PC, the BIOS was closed source and properiety as what-not.
Click to expand...
I Know, but it was documented. Also, as this idea was for a "next-generation" system, why not have one improvement?
 
RetroComputingCollector said:
I Know, but it was documented. Also, as this idea was for a "next-generation" system, why not have one improvement?
Click to expand...
Indeed, this improvement would be a meaningful, documented standard. It does not need to be “open” in the sense that every manufacturer would need do disclose his or her own source code.

Maybe the problem is that the industry is no longer able (or willing?) to produce such standards.
 
RetroComputingCollector said:
Considering how the IBM PC was thrown together in under a year,
Click to expand...
IBM was huge and had research labs and was already successful with 8080 among other things. They did not invent much. Just standardized a loose knit concept. From computer fairs to retail sales.
Remember what the original cost? They got their R&D costs back...
 
I wouldn't say the entire architecture is a black box. Documentation can be had for the basic chips and chipsets. Though the hardware is much more complex. Take CPUs as an example. And each CPU is different from each other. Not from basic design but peripheral aspects. You'll see little workarounds in the kernel.

Peripheral devices are certainly black boxes. I've been working on bge(4) WOL for a while. The sketchy chipset documentation hasn't helped me resolve the issue. Other developers have had the same complaints. I get the sense everyone is trying to protect their intellectual property to the point that documentation is either non-existent or useless.

ACPI and UEFI interfaces aren't documented to the detail I would like. Though I hunt for it, then give up only to go looking for it again.

But for major components like CPUs and busses it's all there in almost too much volume.

Short answer: it depends.
 
cy@ said:
[…]
Peripheral devices are certainly black boxes. I've been working on bge(4) WOL for a while. The sketchy chipset documentation hasn't helped me resolve the issue. Other developers have had the same complaints. I get the sense everyone is trying to protect their intellectual property to the point that documentation is either non-existent or useless.
[…]
Click to expand...
If you tell Broadcom you will buy a couple of million pieces of their network hardware, maybe they will let you sign an NDA and help you.

Sorry for the sarcasm.
 
Phishfry said:
IBM was huge and had research labs and was already successful with 8080 among other things. They did not invent much. Just standardized a loose knit concept. From computer fairs to retail sales.
Remember what the original cost? They got their R&D costs back...
Click to expand...
What about something like the Commander X16? It was thrown-together in the same way by a much smaller amount of people. You don't need to be Big Blue to throw together a computer. Sure it will be diffucult, but not impossible. And a much more complex modern computer design could still be done, it would just take a lot more time. You could also try to make it as similar to our PCs as is allowable. Given that Amigas can use Radeon cards, you could use pre-existing hardware for that. I might sound idealistic, but I am speaking in the words of possibility, not feasability.
 
Phishfry said:
Well 8086 was the first CPU to contain microcode. Was that it???
Click to expand...
Microcode existed much earlier. The IBM 360 had it in the early 60s. The Motorola 68xx and 680xx had it.

In the late 70s, I built my own Z80-based computer, together with a friend. We wire-wrapped the CPU board, the IO boards, and bought the floppy controller from kits. We knew every chip and every wire. But in reality, we still dealt with problems of incomplete documentation. The manual for the 1797 floppy controller chip didn't completely describe it, and in other's people code, we found them doing things that shouldn't have worked (but did). The Z80 CPU had undocumented instructions; we've never found them being used, but people talked about them. So even back then, computers were "light gray" boxes.
 
ralphbsz said:
The Z80 CPU had undocumented instructions; we've never found them being used, but people talked about them. So even back then, computers were "light gray" boxes.
Click to expand...
I wouldn't say those were undocumented to mistreat users. In that case they probably have a good reason to leave them undocumented, (Unpriviledged Pentium HCF instruction) or have some other reason like deprecated in design, or that the behavior varies between chip revisions and second-sources. I would say that black box behavior is specifically for abusing users.
 
RetroComputingCollector said:
I wouldn't say those were undocumented to mistreat users. In that case they probably have a good reason to leave them undocumented, (Unpriviledged Pentium HCF instruction) or have some other reason like deprecated in design, or that the behavior varies between chip revisions and second-sources. I would say that black box behavior is specifically for abusing users.
Click to expand...
Many of them were simply not intended. You had stages chained into a pipeline (one clock mostly), and the instruction would activate or deactivate groups in there. So when you activate the adder and the binary inverter together, you get a result that would not be a substration but one off. Not something you would expect in an instruction set, but sometimes you would have use for x=x-(y+1).

Also, did the pentium really have the HCF? I only heard about that being in a MIPS core which was custom made for the Puzzle Palace.

And these NDAs are funny. If norse legend is true, we will all meet in Valhalla and swap stories about this.
 
Back then, I'm pretty sure "undocumented" just meant you could write microcode the processor would accept and do something with but it's not something the manufacturer intentionally put in. In other words, it's random input producing random results.

Microcode was something lower level than assembly and was used when processors were built with TTL chips. You had various length instruction sets that fed directly into the TTL chips that produced various results. Microcode writers were once called masochists at one place I worked but I thought it was super cool and you could still write microcode on Intel and 68xxx chips back in my day. I'd bet you still can today.

I wire wrapped computers back in the day for work and at home. Used to get my kid sister-in-law to come over and unwrap some of the boards for parts.

"The Soul of a New Machine" is a great book some of you may like.
 
RetroComputingCollector said:
When do you think x86 computers became "black boxes" that are no longer yours to work with and own? I would say it was when the Intel ME and AMD PSP were introduced. I think the state of modern consumer computing is terrifying. The ironic thing is that the IBM PC architecture succeeded because it was open. Now it has just been completely closed up again. My ideal future computing architecture would be RISC-V based, with standardized support chips, memory layout, expansion slots, ports, graphics, and firmware. Considering how the IBM PC was thrown together in under a year, someone should be able to throw together a next-gen open standard using already available components in just a few years.
Click to expand...

Even the quite open and good documented Raspberry boards contain black boxes.
See the sad state of Raspberry wifi with FreeBSD ..

 
You must log in or register to reply here.
Back
Top