Are x86 CISC based processors doomed !?

[moderategamer]

It would seem that x86 chipsets are stagnating getting to the limits of what they can achieve through die shrinks and with the evermore popular parallel computing tasks it's looking like ARM RISC based chips could be the future with it's low power low transistor design. x86 instruction sets have become bloated and often complex instructions take more than 1 cycle to complete whilst ARM instructions often run in a single cycle. ARM chips are catching up with their x86 counterparts and with large companies like Microsoft and apple making a push for ARM does this spell doom for Intel and AMD x86 chips or will they prevail . It will take a complete over hall of current software to make such a change possible but it's looking likely that this task will be undertaken. 

 

Here is an interesting Video that gives some food for thought about this, what are your thoughts is it the end of an era of power hungry x86 CISC based computing or do we need to put away the tinfoil hats for when the aliens come?. 

 

 

 

[GoldenLag]

Well. For them to actually be in trouble. ARM would need to be able to emulate x86 and AMD64 and beat traditional x86 CPUs. 

 

 

Otherwise there wont be any adoption. 

[moderategamer]

1 minute ago, GoldenLag said:

Well. For them to actually be in trouble. ARM would need to be able to emulate x86 and AMD64 and beat traditional x86 CPUs. 

 

 

Otherwise there wont be any adoption. 

Well Microsoft and apple are actively supporting RISC so it's not unimaginable that they could completely re write their operating systems and if that happens it would force the hands of other software companies. ARM can emulate x86 already it's just not exceptionally fast but at the same time it's not exceptionally slow.

[moderategamer]

2 minutes ago, PrinceNorris said:

Are there any advantages of using x86 nowdays? We have x64 for pretty much anything not mobile, ARM for anything mobile.

I think the main advantages are just the fact that it's better developed in terms of software and hardware. Although ARM has the potential to be faster the fact is that we have come so far under the x86 architecture that it will be a challenge to make the switch 

[Bananasplit_00]

Risk V is my hope for the future, but I don't think we will get there for quite a while

[GoldenLag]

1 minute ago, moderategamer said:

ARM can emulate x86 already it's just not exceptionally fast but at the same time it's not exceptionally slow.

But not AMD64. Hence its not very useful.

 

Also that is the top of the line ARM CPU holding their ground against the lowest end x86 CPUs.

 

2 minutes ago, moderategamer said:

Well Microsoft and apple are actively supporting RISC so it's not unimaginable that they could completely re write their operating systems and if that happens it would force the hands of other software companies

You know how long it takes to resrite a operating system? And for ut to be any succesful it will need to be able to emulate older instructionsets. 

 

Windows works because its compatibility range is huge.

[moderategamer]

6 minutes ago, GoldenLag said:

But not AMD64. Hence its not very useful.

 

Also that is the top of the line ARM CPU holding their ground against the lowest end x86 CPUs.

 

You know how long it takes to resrite a operating system? And for ut to be any succesful it will need to be able to emulate older instructionsets. 

 

Windows works because its compatibility range is huge.

I'm not saying it would be easy but if ARM can provide a better path to compute times and the hardware can catch up which it's looking like they will the industry will be forced to do that. I mean Adobe have already created photoshop for ARM something they don't stand to make any real money from but are pretty much preparing themselves for the future. And it's not like they have to literally re write the entire operating system they have to re write the compilation of their OS sure it's not as easy as that but essentially it's very possible.

[GoldenLag]

1 minute ago, moderategamer said:

I mean Adobe have already created photoshop for ARM something they don't stand to make any real money from but are pretty much preparing themselves for the future.

they stand to make a lot of money from that....... you have seen the push from Apple right? they wouldnt do it unles there was a lot of money involved. same with Windows running ARM. dont you see the huge profit margins at their disposal?

2 minutes ago, moderategamer said:

I'm not saying it would be easy but if ARM can provide a better path to compute times and the hardware can catch up which it's looking like they will the industry will be forced to do that.

i mean both ARM and RISC is used for serverstuff because in the given appliccation  it is the best way to go. 

 

if ARM beats x86 outright in a majority of software scenarios, the industry will move to that. hence Windows being emulated in ARM. 

[moderategamer]

2 minutes ago, PacketMan said:

Well, as long as we can emulate most programs, there could be a mass adoption. The promise of getting laptops with actually huge battery life without having to buy some XPS, Macbook, Gigabyte Aero, etc for more than $1500 and so, is very appealing

And smartphones could benefit too, with the new foldable phones and an external keyboard we could have the new generation of mobile devices for business, students, etc

Yeah the power saving is really the big calling card not to mention better parallelisation. Mobiles already run on the ARM Architecture so no need for anything to change their.

[straight_stewie]

Current ARM chips are no where near the performance of a modern x86 chip in scientific computing workloads.

That one instruction cycle thing only applies to a subclass of ARM processors, meant to go into microcontrollers. Microcrontrollers are very different than general purpose computing: We like to trade high throughput for low latency. What that means is that, over time, the x86 processor will get "more work" done than one of the single cycle pipeiline ARM chips. However, the single cycle chip will have a deterministic execution time for a given program, something necessary for reasoning about human safety critical computing.

To put the above paragraph into shorter words: Realtime computing is very different than scientific computing. Comparing processors aimed at different solutions is comparing apples to oranges.

But hopefully, with a shred of luck, and a dash of intelligent design, somewhere in about 10-15 years, a processor that can emulate x86/64 and all current extensions better than the best x86 processors of that day can, will be designed. That's what will be required to make the switch.

As Bill Gates said in the Tidal Wave Email:

Quote

The PC wasn't perfect. Aspects of the PC were arbitrary or even poor. However a phenomena grew up around the IBM PC that made it a key element of everything that would happen for the next 15 years. Companies that tried to fight the PC standard often had good reasons for doing so but they failed because the phenomena overcame any weaknesses that resisters identified.

Simply put, too much work has gone into writing code for x86 based PC systems. In order to switch to a fundamentally new design either everything must be rewritten for it, or the new design needs to also run old software. And I've got to tell you, the likelihood of all software being rewritten for a new design is less than the likelihood that you read this whole reply.

[porina]

ARM could take over, but only in an indirect way as they can't directly attack x86 as a drop in replacement.

 

ARM is already all over mobile devices.

ARM can find niche but significant uses in server as the applications are more specific.

Apple could use custom ARM cores as the basis for their future desktop line to get away from dependence on Intel.

Windows PCs will arguably be the hardest nut to crack.

 

Putting aside software/OS compatibility, my biggest reason for not looking at ARM, at least in the short term, is that it is still orientated at lower power, lower performance compared to what we're used to. I don't know if it is possible scale up ARM cores to give higher performance, possibly at the cost of power consumption.

[manikyath]

x86 (or more specificly, x86_64) isnt doomed yet. the big achilles heel about ARM is that it has some specific workloads it REALLY sucks at. and while x86 is indeed rather bloated, CISC just has its benefits, as well as 30 years of x86 market domination being a thing that'll be hard to win against.

[moderategamer]

24 minutes ago, GoldenLag said:

But not AMD64. Hence its not very useful.

 

Also that is the top of the line ARM CPU holding their ground against the lowest end x86 CPUs.

 

You know how long it takes to resrite a operating system? And for ut to be any succesful it will need to be able to emulate older instructionsets. 

 

Windows works because its compatibility range is huge.

That is the current top of the line ARM processor holding it's own against a low to mid range CPU but in the video it explains that the next generation ARM chip will be 3.5x faster than the current snapdragon so it's not unimaginable that ARM chips could become quite fast enough to rival current gen Intel chips in certain work loads not to mention the fact that ARM has the potential to have greater # of cores.

[Labeled]

only if they can get a good adoption rate, but people are reluctant to upgrade their OS, I find it hard to believe that they'll switch to ARM because of future possibilities, X86-64 won't be around forever probably, but I don't think it is going anywhere for the mainstream anytime soon.

 

windows 10 is just closing in on 40% market share, nearly 4 years after release. Other than smaller usecases, there is no mainstream adoption of ARM based CPUs yet. there isn't any data to even suggest it is doomed yet, and intel and AMD can choose to make certain instructions obsolete and trim down the bloat if they want to.

[manikyath]

1 minute ago, moderategamer said:

in certain work loads

this is the issue with the entire statement.

 

as long as that condition remains, only very few people will feel the need to move ship.

[GoldenLag]

11 minutes ago, moderategamer said:

That is the current top of the line ARM processor holding it's own against a low to mid range CPU but in the video it explains that the next generation ARM chip will be 3.5x faster than the current snapdragon

Based on what? What is "next gen"? Is next gen next year? If so, its highly unlikely we will see something. If next gen is 10 years from now. Is it that special?

15 minutes ago, moderategamer said:

ARM has the potential to have greater # of cores.

How wide are said cores? Because there isnt a limit to how many x86 cores we want. 

16 minutes ago, moderategamer said:

chips could become quite fast enough to rival current gen Intel chips in certain work loads

I mean they allready excist. Hence serveruse. But its a niche

[moderategamer]

2 minutes ago, GoldenLag said:

Based on what? What is "next gen"? Is next gen next year? If so, its highly unlikely we will see something. If next gen is 10 years from now. Is it that special?

How wide are said cores? Because there isnt a limit to how many x86 cores we want. 

I mean they allready excist. Hence serveruse. But its a niche

It would help if you bothered to watch the video before asking me questions that are covered in said video, not a dig would just be nice XD 

There is absolutely a limit to how many cores we can have on x86 due to the high transistor count and trouble reducing die sizes. there comes a point where things are just too far apart. That's not to say ARM chips don't have to follow the same rules it's just ARM chips use far less transistors and are already on small dies they can just physically afford more cores plus the instruction set is better suited for parallelised tasks.

[Teddy07]

There isn´t a clear line between cisc and risc anymore because they both want to reduce the power consumed. 

[GoldenLag]

1 minute ago, moderategamer said:

There is absolutely a limit to how many cores we can have on x86 due to the high transistor count and trouble reducing die sizes. there comes a point where things are just too far apart. That's not to say ARM chips don't have to follow the same rules it's just ARM chips use far less transistors and are already on small dies they can just physically afford more cores plus the instruction set is better suited for parallelised tasks.

You know what Zen does? It makes multiple chiplets.

 

You kniw ARM follow the exact same rules at x86 with cores and everything. X86 cores are larger because they contain a lot more instructionbits. Like AVX execution units and much more. 

 

ARM runs into the same issues. 

 

3 minutes ago, moderategamer said:

It would help if you bothered to watch the video before asking me questions that are covered in said video, not a dig would just be nice XD 

I mean i believe i watched it a while back or another similar one. Also no can do because classes. Also the arguments are allways the same. 

[Salv8 (sam)]

1 hour ago, PrinceNorris said:

Are there any advantages of using x86 nowdays?

apart from being able to run x86 and 16-bit applications (yes you can do that in windows.) it's bacillary a dead architecture at this current point.

[moderategamer]

2 minutes ago, GoldenLag said:

You know what Zen does? It makes multiple chiplets.

 

Only problem with that is memory bandwidth if you have two separate chips sure you can handle separate loads individually faster but the two chips can't communicate with each other at useful speeds to help each other may as well just have two separate processors at that point.

[GoldenLag]

4 minutes ago, moderategamer said:

Only problem with that is memory bandwidth if you have two separate chips sure you can handle separate loads individually faster but the two chips can't communicate with each other at useful speeds to help each other may as well just have two separate processors at that point.

Interconnect bandwidth is more than satisfactory. Shown with the 2990wx. 

 

Its latency that is the issue. And Zen is currently handling that rather well. 

 

ARM will have to do the exact same if yields  start tl detiriate due to larger dies.

[moderategamer]

1 minute ago, GoldenLag said:

Interconnect bandwidth is more than satisfactory. Shown with the 2990wx. 

 

Its latency that is the issue. And Zen is currently handling that rather well. 

 

ARM will have to do the exact same if yields  start tl detiriate due to larger dies.

My mistake I didn't mean bandwidth but you know what I meant. I don't follow AMD but I find it hard to believe they can achieve the kind of speeds comparable with L1 cache speeds which are required to provide any sort of useful communication. So unless they share L1 cache which for all I know they could then I stand by my comment.  

[GoldenLag]

1 minute ago, moderategamer said:

L1 cache speeds

do you even know how fast L1 operates?

1 minute ago, moderategamer said:

So unless they share L1 cache which for all I know they could then I stand by my comment.  

you seem to missunderstand how fast L1 operates hard. like it operates real fast. on the 9900k running at 5ghz, the L1 cache essentially also runs at 5ghz. do you understand? 

 

AMD has reached a really low latency between dies. not comparable to what is between traditional 2 socket systems. its not quite L3 either i believe, but its darn fast. 

 

core to core on Zen 1 was 40ns on L3. 

CCX to CCX is 120ns to 110ns based on memmory frequency. 

Die to Die is 240ns. 

 

(sorry i forgot to save sources but here is my last one: https://www.tomshardware.co.uk/amd-ryzen-threadripper-1950x-cpu,review-33976-2.html)

[moderategamer]

2 minutes ago, GoldenLag said:

do you even know how fast L1 operates?

you seem to missunderstand how fast L1 operates hard. like it operates real fast. on the 9900k running at 5ghz, the L1 cache essentially also runs at 5ghz. do you understand? 

 

AMD has reached a really low latency between dies. not comparable to what is between traditional 2 socket systems. its not quite L3 either i believe, but its darn fast. 

 

core to core on Zen 1 was 40ns on L3. 

CCX to CCX is 120ns to 110ns based on memmory frequency. 

Die to Die is 240ns. 

 

(sorry i forgot to save sources but here is my last one: https://www.tomshardware.co.uk/amd-ryzen-threadripper-1950x-cpu,review-33976-2.html)

I'm having trouble here you seem to be proving my point while telling me I don't understand yes I'm aware L1 cache is blisteringly fast that is my point. L2 cache is in comparison very slow obviously not compared to system memory but as far as memory giving instructions to the CPU L1 speeds are pretty much required for useful operation ie two separate chiplets as he called them couldn't hope to communicate at similar speeds as the distance between cache and the cpu is directly responsible for speeds. So yes although the chiplet method provides better speeds than two separate cpu's it's ultimately not that useful. I have/am studying computer architecture/software in University.