What's the future of CPUs in the next 5 years?

[Lord Szechenyi]

since we are getting close to the maximum (or rather, minimum) size of CPU dies when using silicon, will we see some other material, or will we see popularity for systems with more than 1 cpu, or perhaps make the cpu bigger?

[Radium_Angel]

1 minute ago, Lord Szechenyi said:

since we are getting close to the maximum (or rather, minimum) size of CPU dies when using silicon, will we see some other material, or will we see popularity for systems with more than 1 cpu, or perhaps make the cpu bigger?

The future?

Intel and AMD will go down in flames.

Via will rise again.

Along with Matrox.

 

Hey, if I could predict the future, I'd be a lot richer than I am now.

[Lord Szechenyi]

1 minute ago, Radium_Angel said:

The future?

Intel and AMD will go down in flames.

Via will rise again.

Along with Matrox.

 

Hey, if I could predict the future, I'd be a lot richer than I am now.

That doesn't make any sense.

EVERYONE would be affected by this, since (AFAI) all cpus are manufactured in the same way

Spoiler

But don't worry, i still found your joke funny

 

[WereCatf]

3 minutes ago, Lord Szechenyi said:

will we see some other material

How do you expect anyone to be able to answer this? We have no way of knowing until someone comes up with a new, suitable material!

[Radium_Angel]

1 minute ago, Lord Szechenyi said:

That doesn't make any sense.

EVERYONE would be affected by this, since (AFAI) all cpus are manufactured in the same way

  Reveal hidden contents

But don't worry, i still found your joke funny

 

Personally, I'd love to see the resurrection of custom iron like SGI and Sun systems, it won't happen, but I'd kill to have a modern non-x86 SGI on my desk

[Lord Szechenyi]

1 minute ago, Radium_Angel said:

Personally, I'd love to see the resurrection of custom iron like SGI and Sun systems, it won't happen, but I'd kill to have a modern non-x86 SGI on my desk

i had no idea what you were talking about, so i checked it out and WOW i was impressed.

but the problem is, they stopped researching that kind of CPU decades ago (probably for the best) so even if we did make these now there would be lots of research time needed to get remotely close to the same level as today's cpus are

[mahyar]

imo they will switch to some other srmi ccondactor like germanium (unlikely)

[Radium_Angel]

5 minutes ago, Lord Szechenyi said:

i had no idea what you were talking about, so i checked it out and WOW i was impressed.

but the problem is, they stopped researching that kind of CPU decades ago (probably for the best) so even if we did make these now there would be lots of research time needed to get remotely close to the same level as today's cpus are

Yeah, even today those old SGI and Sun systems go for prime dosh on ebay, and have been hopelessly outclassed by even 2nd-gen Intel chips. Shame really, there was an elegance in SGI systems you won't find today.

[Ian Greenhalgh]

Considering how little mainstream CPUs have changed in the last 5 years, I don't expect much to change radically in the next 5 years.

[Biomecanoid]

Since everybody is talking about his hopes and dreams  I want 3DFX to rise from the dead and instead of direct3d and opengl, glide becomes the norm.

[Lord Szechenyi]

5 minutes ago, Biomecanoid said:

Since everybody is talking about his hopes and dreams  I want 3DFX to rise from the dead and instead of direct3d and opengl, glide becomes the norm.

GOD PLEASE NO

we are already having issues nowadays with old games not running correctly i don't want Today's Games having compatibility issues in the future

[Lord Szechenyi]

21 minutes ago, Ian Greenhalgh said:

Considering how little mainstream CPUs have changed in the last 5 years, I don't expect much to change radically in the next 5 years.

Fair Point, but something quite big is going to (and must) happen to be able to make better CPUs

[Internet Person]

Some say graphene is the future. Graphene chips can in theory run at frequencies of 1000+ GHz, use less power, dissipate heat faster and conduct electricity better than current materials. It's all in experimental stage so there is no telling if it will work out. It could just be hype.

[Lord Szechenyi]

23 minutes ago, Internet Person said:

Some say graphene is the future. Graphene chips can in theory run at frequencies of 1000+ GHz, use less power, dissipate heat faster and conduct electricity better than current materials. It's all in experimental stage so there is no telling if it will work out. It could just be hype.

well there really isn't any other viable option

[IkeaGnome]

Ryzen 103 7900000XX

I52 142450000k

 

 

/s

 

Should be interesting to see. Especially with where we've come in the last 5 years.

[Internet Person]

14 minutes ago, Lord Szechenyi said:

well there really isn't any other viable option

 

Yeah, the other option is stagnation. We hit the GHz barrier. We already have 16+ cores consumer CPUs so I don't see much in this direction. New materials is the only way.

[Lord Szechenyi]

2 minutes ago, comander said:

5 year time frame - 

I'm speculating:

Hybrid architectures (handful of big cores, large number of small cores)
more specialized hardware/accelerators (think video accelerators, AI inferencing accelerators, etc.)
An extra level in the memory hierarchy (eDRAM? HMC/HBM? maybe the last level of RAM is going to be optane?)
 

 

I also wouldn't be surprised to see more stuff out of ARM. We could end up in a market where AMD/Intel serve high performance needs and ARM is adequate for everything else. Why does my mother need anything more than a video conferencing box + a handful of 20 year old games + office + taxes?

I'm doubtful about any of the ones you mentioned except ONE.

AI may very well be the answer, after all what we are seeing with Nvidia it's very likely that AI might be the best solution

[Guest]

Only 5 years ahead? People might still be using 3rd Gen Ryzen lol.

 

What of 20 years from now? 5 years we can already see from here....

[Lord Szechenyi]

Just now, ShrimpBrime said:

Only 5 years ahead? People might still be using 3rd Gen Ryzen lol.

 

What of 20 years from now? 5 years we can already see from here....

No i mean that we are about 2 cpu generations before we hit a peak (at least with silicon).

i'm sure most people will use current cpus but will applications stop being better? will there be nothing else after ray tracing?

[Guest]

1 minute ago, Lord Szechenyi said:

No i mean that we are about 2 cpu generations before we hit a peak (at least with silicon).

i'm sure most people will use current cpus but will applications stop being better? will there be nothing else after ray tracing?

Depends.

 

Humans are starting to let AI engineer stuff. Perhaps the way we compute may totally change. X128 instead of X64 could even be a possibility.

 

For the hardware, because of something like the above, the physical processor might greatly change. 

 

One thing is certain, the software is behind a bit. 

 

Either way, I'm in for the ride!

[Lord Szechenyi]

9 minutes ago, comander said:

Intel already launched such a part fairly recently and this set up was common place in the ARM world YEARS ago. 
Beyond that Intel has more accelerators on their roadmap and is making a big deal out of it. Also, anecdotally I want to a holiday party and ended up chatting with a CPU designer from a big company that designs processors... they're making accelerators. That's the future. Figure out 10 use cases and make dedicated hardware that does those 10 things VERY VERY well at very low die area/energy costs. This overlaps with neural network inferencing

Same idea with more cache levels - Intel already has it (Broadwell, also Xeon with optane)

 

For most of these it's more of a question of where, how and to what extent - not whether it'll happen. Some of these things will make sense in SOME use cases but not others (e.g. Server vs Laptop vs Workstation vs Desktop will all have different performance needs and cost targets)

 

The main benefit of going from 32 bit to 64 bit was the ability to handle more RAM. There aren't really any performance benefits and there are probably performance drawbacks due to increased overhead. 

2^32 = 4294967296 bytes aka 4GB RAM. 
2^64 = 1.8446744e+19 bytes AKA 16 billion GB of RAM. 

I have yet to find a use case for normal computers that needs more than 16 Billion GB. 
 

honestly it was also for storage

so max file size doesn't have to be 4GB (or whatever number it was)

[Guest]

28 minutes ago, comander said:

 

I have yet to find a use case for normal computers that needs more than 16 Billion GB. 
 

Interesting way to put it.... lol 

 

Who said a processor's main focus was consumer related? 

 

 

 

[Guest]

2 minutes ago, comander said:

Find a use case that needs more than 16 Billion GB RAM in the enterprise. 

For what it's worth, most CPUs are currently limited to 48bit memory addressing in practice... and I think Linux currently has an upper limit of 256TB (aka 256,000 GB, which is nowhere NEAR 1,000,000,000 GB).

Even if you had a 128 bit mapping capacity, getting to 16 billion GB would require 128 million 128GB DIMMs. I don't think that could physically fit in a data center. 

Billion gigabytes...

 

I cant even take you seriously rn lol XD

[Guest]

Just now, comander said:

That's why I'm raising my eyebrows to the guy saying "maybe 128bit CPUs"  - let's get to the point where 64bit MIGHT become an issue... which will be a while. 

 

I mean if each GB of RAM cost $0.10 (not a bad buy) you'd be looking at 1.6 Billion dollar systems (just on RAM) before you'd need to worry about 128bit CPUs. 

Because I was told we'd never be using hard drives with more than say 500mb.

We have flash drives in the TB. 

 

Whomever said we'd even be using system ram in the future? 

 

Perhaps just for speculation, I wasn't being directly serious to a measly 5 years.... or your just being over serious without realization that things change. 

 

And so my data center in 2050 has 500 billion Petabytes of memory over a 200 acre property.

 

So what of it?

 

 

[whm1974]

Carbon Buckyballs and Nanotubes. If Humans can figure out how to make vast numbers of them, that will change everything...