Stacking cpu cores?

[zFox]

In theory is it possible to create a cpu architecture to stack 2 cpu cores for example,  at 4.0ghz each to create a single 8.0ghz core?

 

Is current technology anywhere near that?

[WereCatf]

17 minutes ago, zFox said:

In theory is it possible to create a cpu architecture to stack 2 cpu cores for example,  at 4.0ghz each to create a single 8.0ghz core?

That's not how CPUs work in the first place, so no. No matter how you place the cores in relation to one another, or how you connect them, they'll still be individual cores. To get a single 8GHz core, you have to make a single 8GHz core.

[Enderman]

No.

You can put them in parallel (aka multithreading), not in series.

[zFox]

1 minute ago, WereCatf said:

That's not how CPUs work in the first place, so no. No matter how you place the cores in relation to one another, or how you connect them, they'll still be individual cores. To get a single 8GHz core, you have to make a single 8GHz core.

Will making extremely efficient cpus with the lowest nm possible give us the ability to create a stable 8.0ghz single core? What do we have to do to reach 8.0 ghz?

[WereCatf]

7 minutes ago, zFox said:

Will making extremely efficient cpus with the lowest nm possible give us the ability to create a stable 8.0ghz single core? What do we have to do to reach 8.0 ghz?

If I knew that, I'd be a billionaire.

[19_blackie_73]

6 minutes ago, zFox said:

Will making extremely efficient cpus with the lowest nm possible give us the ability to create a stable 8.0ghz single core? What do we have to do to reach 8.0 ghz?

we will for sure get there, as you may know we started in the MHz area and are now in the multiple GHz area. So getting to 8 is in my opinion only a matter of time.
When and what will be necessary dunno, I know too little about designing circuitry at that level, but eventually we'll get it

[Enderman]

7 minutes ago, zFox said:

Will making extremely efficient cpus with the lowest nm possible give us the ability to create a stable 8.0ghz single core? What do we have to do to reach 8.0 ghz?

There already are people that overclock CPUs past 8GHz stable.

[zFox]

1 minute ago, Enderman said:

There already are people that overclock CPUs past 8GHz stable.

Isn't that with insane cooling solutions? Like liquid nitrogen and other below freezing solutions.

 

I am talking about being able to run under a normal high quality air cooler or a normal  all in one liquid cooler.

 

One thing I was curious about,  is there all in one liquid coolers that have a built in chiller to bring down the water temps?

 

[Enderman]

4 minutes ago, zFox said:

Isn't that with insane cooling solutions? Like liquid nitrogen and other below freezing solutions.

 

I am talking about being able to run under a normal high quality air cooler or a normal  all in one liquid cooler.

 

One thing I was curious about,  is there all in one liquid coolers that have a built in chiller to bring down the water temps?

1) Yes, and as efficiency increases it will be easier to get higher frequencies, as can be seen from the history of literally every CPU ever in the past decade.

 

2) Obviously we're not there yet and nobody can tell you what exactly needs to be done to make it happen if the hundreds of phD engineers working at intel haven't found the answer yet.

 

3) You might want to go take a look at how a chiller works and how big it is, then maybe you will see why it cannot go inside an AIO.

[Master Disaster]

3 minutes ago, Enderman said:

There already are people that overclock CPUs past 8GHz stable.

There really aren't. Anybody overclocking to that extreme is using exotic cooling and is only after stability for one benchmark run. Running 8GHz stable is no where near a reality.

 

37 minutes ago, zFox said:

In theory is it possible to create a cpu architecture to stack 2 cpu cores for example,  at 4.0ghz each to create a single 8.0ghz core?

 

Is current technology anywhere near that?

It might never happen, current CPU tech is dangerously close to its limits. To go faster we need new technology and that might mean an entirely different way of measuring. Honestly nobody knows.

 

4 minutes ago, zFox said:

Isn't that with insane cooling solutions? Like liquid nitrogen and other below freezing solutions.

 

I am talking about being able to run under a normal high quality air cooler or a normal  all in one liquid cooler.

 

One thing I was curious about,  is there all in one liquid coolers that have a built in chiller to bring down the water temps?

 

Sub ambient coolers are not suitable for mainstream use as running below ambient means condensation which means fried boards (without protection).

[Enderman]

Just now, Master Disaster said:

There really aren't. Anybody overclocking to that extreme is using exotic cooling and is only after stability for one benchmark run. Running 8GHz stable is no where near a reality.

Yeah I never said it didn't require extreme cooling.

It's still stable, that's what required for it to be an actual record.

You obviously can't keep it stable forever since LN2 costs money and you constantly need to monitor the temps (unless you buy kingpin's roboclocker)

[Master Disaster]

2 minutes ago, Enderman said:

1) Yes, and as efficiency increases it will be easier to get higher frequencies, as can be seen from the history of literally every CPU ever in the past decade.

 

2) Obviously we're not there yet and nobody can tell you what exactly needs to be done to make it happen if the hundreds of phD engineers working at intel haven't found the answer yet.

 

3) You might want to go take a look at how a chiller works and how big it is, then maybe you will see why it cannot go inside an AIO.

1) Going below 5nm is impossible with current techniques because of quantum tunneling.

 

2) The solution is either new material or an entirely new CPU but yes, neither is anywhere near close.

 

3) I assure you, size and complexity is not the reason.

[WereCatf]

9 minutes ago, 19_blackie_73 said:

we will for sure get there, as you may know we started in the MHz area and are now in the multiple GHz area. So getting to 8 is in my opinion only a matter of time.
When and what will be necessary dunno, I know too little about designing circuitry at that level, but eventually we'll get it

One thing people keep not understanding is that reaching extremely high frequencies isn't the end-all-be-all of CPU-speed: instead of pushing for ever higher frequencies, Intel, AMD and others may instead opt to focus on their instruction-sets to require fewer clocks to execute -- the end result will still be faster CPUs, even if the clockspeeds remained the same. The addition of the various kinds of SIMD-instructions (AVX, for example) in the newer CPUs follows this thinking; they don't increase clockspeed, instead they do more calculations in the same clocks.

[Master Disaster]

5 minutes ago, Enderman said:

Yeah I never said it didn't require extreme cooling.

It's still stable, that's what required for it to be an actual record.

You obviously can't keep it stable forever since LN2 costs money and you constantly need to monitor the temps (unless you buy kingpin's roboclocker)

A phrase used in the extreme overclocking world is "if your car explodes as it crosses the line it still wins the race". Again I assure you, stability beyond one run is neither required nor a concern.

 

Ok slight correction, one run plus a screenshot.

[zFox]

Just now, Enderman said:

Yeah I never said it didn't require extreme cooling.

It's still stable, that's what required for it to be an actual record.

You obviously can't keep it stable forever since LN2 costs money and you constantly need to monitor the temps (unless you buy kingpin's roboclocker)

What about someone who lives in a colder area near a stream of water,  to create a funneling cooling loop system that uses the running water to come into the loop to cool cpu and GPU,  and then funnels back into the river. 

 

I know this idea is out there,  but it's very interesting to me,  would condensation be a issue this way?

[WereCatf]

Just now, zFox said:

What about someone who lives in a colder area near a stream of water,  to create a funneling cooling loop system that uses the running water to come into the loop to cool cpu and GPU,  and then funnels back into the river. 

 

I know this idea is out there,  but it's very interesting to me,  would condensation be a issue this way?

Corrosion and blockages from the impurities in the water would be far bigger issues.

[zFox]

3 minutes ago, WereCatf said:

One thing people keep not understanding is that reaching extremely high frequencies isn't the end-all-be-all of CPU-speed: instead of pushing for ever higher frequencies, Intel, AMD and others may instead opt to focus on their instruction-sets to require fewer clocks to execute -- the end result will still be faster CPUs, even if the clockspeeds remained the same. The addition of the various kinds of SIMD-instructions (AVX, for example) in the newer CPUs follows this thinking; they don't increase clockspeed, instead they do more calculations in the same clocks.

How much headroom is their to improve instruction sets?

[Master Disaster]

2 minutes ago, zFox said:

What about someone who lives in a colder area near a stream of water,  to create a funneling cooling loop system that uses the running water to come into the loop to cool cpu and GPU,  and then funnels back into the river. 

 

I know this idea is out there,  but it's very interesting to me,  would condensation be a issue this way?

If it's below ambient it will cause condensation.

[zFox]

Just now, WereCatf said:

Corrosion and blockages from the impurities in the water would be far bigger issues.

Perhaps to have the water runs through a filter system to catch most of unwanted things.

Of course this is not your everyday thing, I just wonder if someone has ever tried it,  I like the idea of using nature to cool a pc

[Enderman]

3 minutes ago, Master Disaster said:

1) Going below 5nm is impossible with current techniques because of quantum tunneling.

 

2) The solution is either new material or an entirely new CPU but yes, neither is anywhere near close.

 

3) I assure you, size and complexity is not the reason.

1) I never said anything about process size. The only thing I mentioned is efficiency.

2) Why are you still talking about process size

3) Go take a look at what a water chiller is on google images.

 

1 minute ago, Master Disaster said:

A phrase used in the extreme overclocking world is "if your car explodes as it crosses the line it still wins the race". Again I assure you, stability beyond one run is neither required nor a concern.

It's still stable for that run, it does not magically explode after 1 benchmark run. There is just no need to keep it going 24/7.

If it's stable for that one run it can remain stable afterwards if you want to keep wasting money on LN2 and your time monitoring temps.

 

1 minute ago, zFox said:

What about someone who lives in a colder area near a stream of water,  to create a funneling cooling loop system that uses the running water to come into the loop to cool cpu and GPU,  and then funnels back into the river. 

 

I know this idea is out there,  but it's very interesting to me,  would condensation be a issue this way?

Yeah that's possible.

Condensation will happen on the tubing so you still need to insulate anything that carries the river water inside your home.

You won't get 8GHz with 2C river water though, you need like -200C for 8GHz.

[WereCatf]

Just now, zFox said:

How much headroom is their to improve instruction sets?

That question doesn't make sense. There is no headroom, period. There being headroom would imply that there's some sort of a magical limit as to what sorts of instruction-sets can or can't be made, but that's not how it works at all. There is no way of answering such a question.

[WereCatf]

Just now, zFox said:

Perhaps to have the water runs through a filter system to catch most of unwanted things.

Of course this is not your everyday thing, I just wonder if someone has ever tried it,  I like the idea of using nature to cool a pc

No, even with a filter, the impurities would still accumulate in the system and eventually block it. The way to do it would be a heat-exchange, instead: you'd have one system circulating water from the river through bigger pipes that won't corrode and in which there are no small heatpipes to go through that could get clogged, and a separate system that runs proper cooling-solution inside a closed loop, and these two system then run in parallel at one point, exchanging heat from one to the other -- the water from the river and the cooling-solution for the PC would never, at any point, get mixed.

[19_blackie_73]

12 minutes ago, WereCatf said:

One thing people keep not understanding is that reaching extremely high frequencies isn't the end-all-be-all of CPU-speed: instead of pushing for ever higher frequencies, Intel, AMD and others may instead opt to focus on their instruction-sets to require fewer clocks to execute -- the end result will still be faster CPUs, even if the clockspeeds remained the same. The addition of the various kinds of SIMD-instructions (AVX, for example) in the newer CPUs follows this thinking; they don't increase clockspeed, instead they do more calculations in the same clocks.

yeah that's true, but intel still already passes the 5GHz number despite new additional stuff. And with each generation they tend to increase the clockspeed a bit so 8Ghz will come some day, maybe not tomorrow or the day after tomorrow, but eventually. Remember 4th gen i7 ended by 4.4 turbo, so they already pushed it another .6GHz. I won't say that increasing the cpu speed alone makes everything a gazillion times faster, you need other stuff to improve too.

[Master Disaster]

9 minutes ago, Enderman said:

1) I never said anything about process size. The only thing I mentioned is efficiency.

2) Why are you still talking about process size

3) Go take a look at what a water chiller is on google images.

 

It's still stable for that run, it does not magically explode after 1 benchmark run. There is just no need to keep it going 24/7.

If it's stable for that one run it can remain stable afterwards if you want to keep wasting money on LN2 and your time monitoring temps.

 

 

1) How else do you increase efficiency without a node shrink?

2) I wasn't talking about node shrinks

3) Again, miniaturisation is not the problem. It has been done (see phase changing), it's just doing it is pointless without protecting the other equipment.

 

You obviously have no idea how extreme overclocking works. I'm not even bothering arguing on this one because what you said is word salad. Just know one thing, in a very large amount of case it in fact does "explode after one run", figuratively speaking of course. In the VAST MAJORITY of cases it explodes before one run.

 

[Enderman]

8 minutes ago, Master Disaster said:

1) How else do you increase efficiency without a node shrink?

2) I wasn't talking about node shrinks

3) Again, miniaturisation is not the problem. It has been done (see phase changing), it's just doing it is pointless without protecting the other equipment.

 

You obviously have no idea how extreme overclocking works. I'm not even bothering arguing on this one because what you said is word salad. Just know one thing, in a very large amount of case it in fact does "explode after one run", figuratively speaking of course. In the VAST MAJORITY of cases it explodes before one run.

 

1) lol there are tons of things that affect the efficiency of a CPU that have nothing to do with node size (for example gate shape, path layering, or even different materials as you mentioned yourself). Shrinking nodes is not done just for efficiency, there are many other reasons why they do that.

 

2) yeah you were, changing materials is solution to not being able to go below 5nm, hence why you mentioned it and were still talking about process size.

 

3) go ahead and try to stick a phase compressor in an AIO, see how that goes, lol.

 

4) seems like I know more about water chillers and LN2 overclocking than you do....