Jump to content

About CPU size

Go to solution Solved by badreg,

What you are describing is essentially a GPU. With parallel workloads, you can copy and paste as many cores into however big a die you want.

 

However, with general purpose CPUs, the workloads are not parallel, so when you add cores, you also add latency, which reduces the performance of the system.

Hi! Before i ask my question, I want to precise that i did research answers on the forum and a little on the web aswell.

 

We now have central processing units with 14nanometer process nodes and even less, this make a lot more place to add more "power" to the CPU, and i guess that this is why we can have 8 cores CPUs under a heatspreader that once was toping only one or two cores

In addition to that, we see bigger CPUs like ThreadRipper or the new Xeon W-3175X going for 32, 28 cores, etc.

 

Is there a way to make a CPU as big as a laptop with like 1500 cores? I know that it sounds stupid and that improving a CPU is more than multiplying the nodes, but hey, i can't get it out of my mind today so i came here (naturally ?)

By the way, i also typed "Biggest CPU" / "Biggest processor" online, and i did not found what I was looking for, I even found things I definitely did not want to see.

Link to comment
Share on other sites

Link to post
Share on other sites

What you are describing is essentially a GPU. With parallel workloads, you can copy and paste as many cores into however big a die you want.

 

However, with general purpose CPUs, the workloads are not parallel, so when you add cores, you also add latency, which reduces the performance of the system.

Link to comment
Share on other sites

Link to post
Share on other sites

Theoretically it is possible. However taking advantage of this would be pretty much impossible. 

 

plus the amount of heat it would put off at higher clock speeds would be immeasurable. 

 

Not to mention the cost to manufacture it, plus you have other smaller things that would have to manage it. 

 

 

Current Build

Spoiler
  • CPU
  • Motherboard
  • RAM
  • GPU
  • Case
  • Storage
  • PSU
  • Display(s)
  • Cooling
  • Keyboard
  • Mouse
  • Sound
  • Operating System

 

Link to comment
Share on other sites

Link to post
Share on other sites

3 minutes ago, badreg said:

What you are describing is essentially a GPU. With parallel workloads, you can copy and paste as many cores into however big a die you want.

 

However, with general purpose CPUs, the workloads are not parallel, so when you add cores, you also add latency, which reduces the performance of the system.

Oh yeah, I totally did not think about that, so the multiple cores are not the real improvement for workloads, but more the performance of each core?

 

2 minutes ago, Emanbaird said:

Theoretically it is possible. However taking advantage of this would be pretty much impossible. 

 

plus the amount of heat it would put off at higher clock speeds would be immeasurable. 

 

Not to mention the cost to manufacture it, plus you have other smaller things that would have to manage it. 

 

 

Yeah, i certainly did not think of that as a consumer product and more like a science experiment ^^. As for the heat, even a motherboard made for this purpose would be stupidly expensive and ineffective. 

Well, you made my day, I can think of another not-so-great-imaginative-idea now ! 

Link to comment
Share on other sites

Link to post
Share on other sites

Imagine how many pins there would have to be lol, imagine bending one

actually you'd probably have to make this an SOC or an all in one motherboard and cpu 

Current Build

Spoiler
  • CPU
  • Motherboard
  • RAM
  • GPU
  • Case
  • Storage
  • PSU
  • Display(s)
  • Cooling
  • Keyboard
  • Mouse
  • Sound
  • Operating System

 

Link to comment
Share on other sites

Link to post
Share on other sites

1 minute ago, Emanbaird said:

actually you'd probably have to make this an SOC or an all in one motherboard and cpu 

Yes, i actually pictured it more as a futuristic all-knowing computer in an episode of Stargate SG-1 or Doctor Who. xD

Link to comment
Share on other sites

Link to post
Share on other sites

16 minutes ago, RobinHtk said:

Oh yeah, I totally did not think about that, so the multiple cores are not the real improvement for workloads, but more the performance of each core?

It's not so much the performance per core, but the architecture required to manage a large amount of cores. A CPU core doesn't do anything until it is given instructions. Intel uses a ring bus for consumer chips and a mesh interconnect for HEDT and server chips, while AMD calls their interconnect "Infinity Fabric". Look up how these buses work, and it will be clear how efficiency reduces when core counts go up.

 

A crude analogy would be needing more fuel to achieve a certain velocity with a rocket, but then also needing more fuel to account for the additional fuel, and so on.

Link to comment
Share on other sites

Link to post
Share on other sites

Thanks for the information, I know what I am gonna do tonight ! I only "recently" fell into electronics, I have always been a "nerd" but i never did take the time to actually learn all I could about it. Since that last 3 years I have learn so much, thanks to help me build up on that knowledge :)

 

EDIT: By the way, I am french and pardon me for my english, i hate to verify my sentences on a lowsy online translator so some of my words might not be the best choice xD

Link to comment
Share on other sites

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×