Boardsort.com

Your complete E-scrap resource website. LEARN - SELL - PROFIT
It is currently Mon Jan 25, 2021 10:36 pm


Post new topic Reply to topic  [ 10 posts ] 
Author Message
PostPosted: Tue Dec 29, 2020 12:36 pm 
Site Admin
User avatar

Joined: Tue Feb 03, 2015 6:57 pm
Posts: 8588
Location: Low DOS
The end of an era?
IA86, and the greater x86 market started in 1971 with the 8-bit Intel 8008 and it’s semi-16-bit successor in 1973: the 8080.
50 years! That’s a long time.
Let’s take a quick run down memory lane before I get to the question.

So what, really, was the evolution of x86. What big leaps need to be remember?
Obviously Intel, a memory manufacturer, had a big hand in the last 50 years.
1 We start with the intel 80 architecture in 1970/1971.

2 The big leap to a 16 bit address in 1974 created low-cost computing components to the market.

2.5 1977 controller manufacture AMD releases 29000b family. Ignored by the mini and microcomputer market at the time it played a big part in some early kit computers including Altair and Apple. But the 292s would play a big part later, instruction translators.

3 1982 after ignoring the general public IBM launches both Gem/Glass/Unix compatible risc units and DOS/CPM Compatible x86 Personal Computers.

4 1984 after being trashed in a Super Bowl commercial by competitor Apple IBM goes all in with competitors Microsoft and Intel. The x86 platform dominates through 20 years of updates.

5 1989 AMD releases 32-bit x86 chips with K extensions, adding 29k instructions to the processor and creating the beginnings of at-home virtualisation.

6 1999 in a poor attempt at changing the market intel moves to IA64. A new and incompatible architecture unlike then current RISC or CISC chips. AMD releases AMD 64 x86 instructions 100% backwards compatible with x86, taking over the innovation of x86 permanently.
Though not speed or electrically competitive with Intel’s offerings moving forward it would be AMD that would release new instruction innovation through the rest of x86s life.

7 2004 AMD adds 290k instruction set to high-end desktop and server chips. Furthering the VM market divide.

8 2008 intel adds Detect instruction set to Core CPUs. Reducing the large gap with AMD in VM performance.

9 2017 AMD releases Ryzen CPUs and guts intel performance advantage. Through today (2021) intel has yet to match speed and stability since.

10 2020 Apple releases Apple Risc Desktop Processors, the M1. Based on ARM and instruction set compliant, with proprietary extensions.

10b 2020 Apple announces Rosetta 2. The pre-load translation software instruction set allows nearly completely native x86 on ARM.

10c various Wine, Fuse, and Parallels developers discuss working with Apple for cross-compile portability.

11 AMD announces 29run instructions for a 2021 release. Bringing 29k to their Arm SOC.

12 December 2020 timings and benchmarks show M1 a minimum 25% real-world faster than Core7 chips, on par with Core9 and Ryzen7

12b 2020 benchmarks on M1 show 75-100% gains in speed in Photoshop and and Final Cut. 200%+ gains with ultrazip compression and AV1 encoding.


So here’s the discussion point.
Where does intel go from here? They’ve had die shrink issues for years. They won’t cut pricing more than they have and the i9 is just barely competitive with the lower cost AMD Ryzen 7 and M1.

On top of that a whole M1 based Mac with 16GB ram costs less than an i9 CPU!
Along with Apple iBuyPower offers mid-range complete R5 and R7 systems for less than low end i9 chips.
AMD has been chipping away at what is left of the non-RISC x86 server market for years.

I won’t throw Intel away like some tech writers have; but they are really in a spot now. They’re also IBM-level stubborn as a corporate bureaucracy!

Granted AMD and Apple are at the same level of stubbornness but AMD and Apple fans will buy the products regardless.
Dutifully upgrading and swapping generation after generation (myself included).

Like IBM intel doesn’t have that dedicated fan-base.
It Killed IBM. Is Intel next?

One more thing of note. There is a huge movement in open-source and cross-compatibility sectors of software right now to cross the idealogical barriers and move to software portability!
Linux and BSD developers, often in brutal competition with each other, are working on updates posix specs to make binary compatibility possible. Microsoft is working to bring a WINE style layer directly to Windows and move software out of the “Subsystem for:”’
Apple is documenting internals of Rosetta 2 for developers. Fuse is getting official recognition.
Parallels and Code Weavers are working directly with Apple l-attached developers to bring windows to Arm Mac.
Microsoft has not officially announced anything but they already have Arm Windows and an x86 translation kit.

All this portability considerably easier on the ARM platform.

So: is Intel dead? What do you think?

_________________
-- my grades are my own and do not represent an offer from boardsort, nor are they guaranteed. Please keep that in mind.


Top
 Profile  
 
PostPosted: Thu Dec 31, 2020 11:59 am 

Joined: Fri Feb 14, 2014 3:42 pm
Posts: 64
Location: Troy, NY
On a side note, the story of Intel and AMD being the sub-contractor to produce many of Intel's chips because they couldn't meet demand is an interesting one, and Intel's lawsuit against AMD when they caught wind that AMD was making Intel's chips better than Intel is even more fascinating. More to it than that, but an over generalization.
That said, after the 486, Intel quite literally had to change names to differentiate itself in the marketplace (among other reasons) knowing AMD was hot on their asses. AMD started off poorly with their own designs, think Duron and Athlon series, which were OK, but not on Intel's level. Look at them now, people cream when they hear of the new Ryzen 9000 and line up to buy, like you said, they now have not just a cult following, but a sizable user base, the way Intel used to.
I hate to throw out "too big to fail", but I don't think Intel CAN fade into obscurity, the demand for processors is just too large for one company to produce them all, as much as either would like that scenario.
I envision two possible scenarios:
1) Intel has a complete role reversal and becomes sub-contractor for AMD to remain relevant and to bring in some much needed income.Maybe they make AMD's chips better than AMD. But, that leads the way for:
2) Intel lurks in the shadows for a while, developing the next big thing. Then we see another momentum shift in Intel's favor. Another new architecture? I think one will be needed very soon.
Will the cycle ever end? Probably not. Pure speculation, of course, but we've see the trends since the 70's. How far can we go with what we have now? Will there be an i19?
Do 5G towers give us Coronavirus? No. But it's interesting speculation.


Top
 Profile  
 
PostPosted: Thu Dec 31, 2020 3:10 pm 
Site Admin
User avatar

Joined: Tue Feb 03, 2015 6:57 pm
Posts: 8588
Location: Low DOS
I’m in agreement that “too big to fail” is used too much. And I think Intel really does fit the correct use.

The problem I see is Intel can’t get their manufacturing size down.
They’re stuck. At 10-14nm.
The can’t push more performance per watt at this level which has created a bottleneck for them.
AMD took back the crown, temporarily with the FX 9590. While intel was making sub 100watt chips AMD did the unthinkable and put out a 295watt super chip! And it destroyed everything else.
It didn’t last long but it was a sign of things to come for AMD.

Part of what makes the current line SO powerful is the fact they run at 42watt full tilt when the rest of the industry, including pre-M1 Apple, was staying low power sub 30wt.
Most current Core chips idle at 12-15wt and push into the 25-30 range.
AMD has never been “green” despite being the “Green brand”.

I see three logical roads for intel.
1) Like you said they could sublicense AMDs chips and do what AMD pulled in the 90s.
AMD’s contract ended (and the legal issues began) before the pentium but after they already had the “pentium papers”. The design and structure documentation. The k5 faltered but the K6 and K6II were equal and better to the P2 design.
If Intel licenses the R system they get the tech specs for the smaller DM process with it!

2) Intel could pull an FX move and pump out a high 14nm 200wt TDP super chip that destroys the R series. To hell with power savings.

3) they could sell or auction the CPU branch of the company. Including the x86 specs. Intel still is a major memory manufacturer and also has a large software development branch. Spinning off, then dumping, the CPU branch could provide them with much needed cash to ramp up the memory production program. They are currently the largest supplier of GDDR5 already. With more cash and refocused production volume they could actually complete with Samsung.

Alternatives? A few.
They could license ARM and make chips for BSD systems. An Intel based Darwin server could rival Apple in it’s own home theatre!

They could dump the entire IC aspect. Intel is one of the largest firmware and backend distributors.

They could dust off Unix 7, port it to OC and sell it as a binary compatible cross platform C+/OC based platform running Weyland and wX.

They could pull an Apple and open source the core base of Unix 7. They’re the largest rights and patents holder! Talk about sticking a fork in the software competition! Then licensing the display server and RS/SWS system alone!

Sadly I think they will pull an IBM and say nothing’s wrong. As they flounder on like a fish on the beach they will eventually drift into obscurity.
Cyrix chips still live on under VIA. You just don’t know it. WE lives on as a rebranded product line under VIA as well. (VIA appears to still be the dumping ground abandoned IC tech).

One last thought. They could just buy VIA. And go all in on IOT. VIA’s 9nm platform for the licensed i860 and and 80386/7 chips in so many devices today could be a gold mine. Bringing that back in house could be a great asset

No doubt though, Intel is in trouble and needs to do SOMETHING.

_________________
-- my grades are my own and do not represent an offer from boardsort, nor are they guaranteed. Please keep that in mind.


Top
 Profile  
 
PostPosted: Tue Jan 05, 2021 6:20 pm 

Joined: Fri Feb 14, 2014 3:42 pm
Posts: 64
Location: Troy, NY
Ramp it up! To hell with power savings!
Load up on D cell batteries and get ready to BOO Santa! #'Merica

Putting on my tinfoil hat and thinking the unthinkable... maybe they dump all of their IP and hit the ground running with something crazy; and not even in the same market segment. Look at Coleco.
Leather goods to video gaming/home computing (RIP Adam) to Cabbage Patch Dolls? Made buku bucks in the process.
Maybe they cash in on the nostalgia (a really good place to cash in, BTW) and open up Intel World, where folks can reminisce about the good old days of the 8086? Charge $100 bucks a head, sell some Intel swag, and they are back baby!
Now that's[i]sinister.[/i]


Top
 Profile  
 
PostPosted: Tue Jan 05, 2021 6:34 pm 
Site Admin
User avatar

Joined: Tue Feb 03, 2015 6:57 pm
Posts: 8588
Location: Low DOS
Lol.
Nothing is off the table in tech anymore.
Atari went all in with cryptocurrency and it’s open alt coin. AND is selling a $500 PC with a Ryzen chip with a built in VCS!
atari vcs 2021

Anything is possible now. If intel taped the agreement with ECS and a few game studios and put out a retro 486DX4 box for $399 with 100 games it would definitely sell. Especially if they pick up king’s quest, doom, and duke nukem!

What bothers me, in my c86 view, is the IS is now dying. Microsoft has announced in house ARM development. AMD is pushing for a server class super chip by late 2021. When windows, Mac, and many flavours of BSD and Linux run on ARM just fine where’s the draw for intel x86.
90% of Intel’s processor development is x86. A Power derivative makes up a good chunk of the rest!
If they don’t do something fast, and if AMD looses interest in the next few years in AMD86; it could be the end of the architecture completely.

_________________
-- my grades are my own and do not represent an offer from boardsort, nor are they guaranteed. Please keep that in mind.


Top
 Profile  
 
PostPosted: Tue Jan 05, 2021 7:41 pm 

Joined: Fri Feb 14, 2014 3:42 pm
Posts: 64
Location: Troy, NY
Yeah, like I said, they are essentially boned if nothing changes. They have to have forseen this many moons ago: since the founding of the company they have had to deal with huge pressures and one-upsmanship throughout.
I think, like we both think, either they assimilate or be destroyed, unless they have that ace up their sleeve with something radically new.
I think they should look into new architecture and make designs on a chip, make those designs available and open-source.
I was always partial to Sierra titles, myself. King's Quest, Space Quest, Quest For Glory, etc.
I can't recall who has the rights to those old Seirra titles, but if Intel could secure rights, and also with interplay (Wasteland) Looking Glass (System Shock), like you said, they could crank out an all in one system with preload crap on it and rake in some much needed revenue for their skunkworks project (assuming they have one).


Top
 Profile  
 
PostPosted: Tue Jan 05, 2021 9:28 pm 
Site Admin
User avatar

Joined: Tue Feb 03, 2015 6:57 pm
Posts: 8588
Location: Low DOS
Epic actually owns a lot of SIO titles but mostly junk.
The franchises are all over the place. It’s kind of sad. One of the many reasons I support drastically cutting copyright lengths. 20-30 years later and what massive percentage of software is locked up forever?

In topic though; intel is stuck at 10nm. AMD and Apple already have production ASICs and controllers on 5/6nm. Via had ICs, though not CPUs or ASIC level complicated, at 5nm since 2019.
There’s no doubt the M1 crushed the i7 line. It’s competitive with R5/R7 and i9.

I think the fastest way Intel could keep x86 and themselves alive would be to licence the vII spec from AMD. The in-hardware translation machine in the Ryzen CPUs and APUs. It’s what allows Sony and Microsoft to play RISC based games without rewriting them. PS3 with CELL and 360 with Power3.

Or mimic AMD’s FX move and put out a 14nm or 24nm core supreme with 32 AMD64 cores clocked at 5Ghz power concerns be damned. That /could/ compete with the threadrippers.

Apple and AMD are both set to make new products announcements in May. Intel, generally, has till then to figure something out.
The games industry has gone all in behind Rosetta 2 on M1. Including Microsoft.

This could be the year we stop thinking about PC ON Apple and start thinking about PC IN Apple. Fuse and Wine forks are already workable. Though not stable.

As a side track; if Apple were to licence Rosetta2 to Windows developers for porting windows apps to ARM regardless of OS they could single handedly kill x86.

_________________
-- my grades are my own and do not represent an offer from boardsort, nor are they guaranteed. Please keep that in mind.


Top
 Profile  
 
PostPosted: Wed Jan 13, 2021 3:32 pm 
Site Admin
User avatar

Joined: Tue Feb 03, 2015 6:57 pm
Posts: 8588
Location: Low DOS
Bumping this
New news...

Using someone else’s manufacturing?
Quote:
“Bob Swan’s response, that it is a possibility, would have seemed inconceivable at the beginning of last year. However, Intel’s recent revelation that it will outsource some leading-edge chip production in the wake of its delayed 7nm node, and the fact that the company still hasn’t decided just how much, what, or where it will outsource, has led to serious questions that have resulted in bruising stock valuation declines for the world’s top chipmaker.”

Looks like they found a short term solution.

_________________
-- my grades are my own and do not represent an offer from boardsort, nor are they guaranteed. Please keep that in mind.


Top
 Profile  
 
PostPosted: Fri Jan 22, 2021 10:24 am 

Joined: Thu Jan 21, 2021 10:19 pm
Posts: 1
I think we'll see Intel return to the top spot, or at least, to being competitive.

This all feels to me much like it felt during the early P4 era where AMD cpus were just dramatically better, and then intel came back strong with the core series chips a few years later.


And to be honest, I thought AMD was going to be the one failing right now, until the ryzen came out it looked like they were going to be playing the catch up game forever with their much smaller r&d budget compared to Intel.

I'm generally an AMD fan - but it's definitely better for everyone to have more players in the chip game.

Long term, idk. "X86 is dead" has been shouted for at least 30 years as well, but here we still are. The future is hard to predict.


Top
 Profile  
 
PostPosted: Fri Jan 22, 2021 3:15 pm 
Site Admin
User avatar

Joined: Tue Feb 03, 2015 6:57 pm
Posts: 8588
Location: Low DOS
Terrh wrote:
... "X86 is dead" has been shouted for at least 30 years as well, but here we still are. The future is hard to predict.


True that. I just wonder how long intel can remain in the game. It’s not just performance alone anymore.
If these outsourcing agreements work out intel has bought themselves a few more years.
The Startup and Ars have both commented recently that “PC users are in denial” about the M1 specs.

In recently had the pleasure of using one of the development prototype 128-but HPs for 48 hours; which worked out to about 6 hours of actual use sadly.

Knowing it was coming I took a few days ahead of time to make some suggested modifications to 7zip (which are now in the Keka repository wiki). Expanding the pipelines for a massive parallelisation and to handle 2tb ram.

I used the same OC Thunderbolt 3/USB3.2 m2 dock on the Hap and the M1 Mac mini. In both cases I used 5 512GB m2 discs as a ram drive using rdsk.
The Mac with OS11.1 and the HP with the custom BSD4/nBSD9.8

I found in 7zip that they were surprisingly close.
I then tried throwing Big Buck into handbrake with the nightly source for Handbrake with AV1 enabled. I set the 1080/60 input at 1.6MBps in to AV1 1080/60 at 799Mbps out. In both cases I stripped the audio.
After 1 hour (not a complete conversion set) the Mac was the clear leader.
It took 12minutes per pass. Vs the HP which took 21minutes per pass.
This isn’t a surprise though. It’s well known risc is the better choice for video conversion.

I came to some personal conclusions though from this.
First is that Apple has just moved the entire tech industry forward two generations. The M1 is by far the most advanced ARM/RISC-V variant to reach the consumer level. A super-computer level chip is now in desktops.

Second 128-bit x86 computing is competitive here. Unfortunately unlike 64-bit x86 there is zero existing infrastructure for it. If the PC industry moves at its standard pace it will be a decade before anyone comes out with 86-128 and again it will likely be AMD to move the industry to a new ISA.

Third, that’s unlikely to happen. AMD has recently moved to testing a new chiplet design for their Graphics processors and ASICs. What this tells us is that they are moving away from focusing on x86. Not only would this be a potential GCU explosion moment; this design is easily ported to Risc. As such we may see a desktop level risc chip from AMD in the next year or two.
They are already a large supplier of risc formats.

AMD, not intel, has always been the innovator. Tithe rift with intel that started with the 386 was because AMD made their license chips better, and faster. And for less cost. The 5x86 matched the pentium the same year both were released. The K5 and K6 ran circles around the pentium counterparts.

When intel hit an innovation bottleneck AMD moved forward with AMD64. An expansion to x86.
Intel licensed it from AMD.

Intel currently is a low end budget risc supplier.
The largest buyer of intel risc chips is, intel. Lol. They make controllers for their own solid state drives.
AMD on the other hand remains in the top 5 risc suppliers making chips for iot devices. Optical drives. A/D converters. They’re a large supplier for the tape industry. One of three for LTO system controllers.

Qualcomm has a scale issue. They’re targeted on a single set of platforms and do massive volumes. It’s unlikely to see them move into a desktop processor market.

Raven, the risc branch of VIA, is locked into decade contracts. It’s unlikely they could even try to move in another market.

Intel is still a massive memory supplier. I honestly think the best thing they could possibly do right now is sell x86 outright to Apple or AMD.
Apple has long worked on JIT IA translation. It would be the end of x86 but it would make it likely we have comparability of x86 software on RISC for the near future. For all the Apple bashing hatred online, they are a “for the good” company. They have a good track record of open sourcing tech for the greater good. From Darwin OS to Bonjurno. Much of the original Rosetta is open source.
Or
They could sell x86 outright to AMD! AMD is one of the companies that has toyed with 128-bit instructions. They’re probably in the best position to make the innovative leap. That would put 86 back on a competitive level with RISC.

One thing that probably keeps some intel heads awake right now is the risk of Apple opensourcing M-Risc. Ouch.
Apple is no stranger to cross compatibility. They os’d low level code and opened the kernel specs for Power rather early. On Intel we had Rosetta and they worked directly with groups for translators. Fuse, parallels, etc. They are doing that again already with M1. Parallels has an alpha, fuse has partially made the move, LAMP is already up and running.

We’ve now seen many intel apps work better translated on M1 than on native intel Macs. Not all, but many. (MS Access sucks on M1)!


Everything else aside this is an exciting time for tech. This is now the 4th major computing change of my lifetime and I’m happy to see it.
[list=We had the 8-bit explosion and the home computer.
The GUI revolution
AMD 64
Now M1 RISC[/list]
(And the failure of list on PHPBB. Lmao.

Despite COVID if you like sci-fi and tech this is a great time to be alive!

_________________
-- my grades are my own and do not represent an offer from boardsort, nor are they guaranteed. Please keep that in mind.


Top
 Profile  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 10 posts ] 

All times are UTC - 5 hours [ DST ]


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Jump to: