If you were Bill Gates and had the funds, could you build your PC from scratch SCRATCH?

[Aleksbgbg]

If you had as much money as mr. Gates does, then would you be able to build your own PC from start to finish? I'm talking manufacturing your own one-off CPU, GPU, mobo etc.

 

Could you make, I don't know ... a 50 core 25GHz CPU or something?

Or is it that for some reason you wouldn't be able to ...?

 

Just wondering ... speak your thoughts

[theninja35]

No, you'd still need the know-how on forming a company, hiring the right people, and general production. Money does not mean brains...or anything, really.

[vanished]

So you're saying could you hire a team of engineers to design - from scratch - something to match or beat current top level stuff from Intel/AMD/NVIDIA, construct the manufacturing for it, build it, and then tear it all down... I don't think I've ever heard of a worse price/performance pitch

[Aleksbgbg]

1 minute ago, Ryan_Vickers said:

So you're saying could you hire a team of engineers to design - from scratch - something to match or beat current top level stuff from Intel/AMD/NVIDIA, construct the manufacturing for it, build it, and then tear it all down... I don't think I've ever heard of a worse price/performance pitch

That's the point  you have much funds  

[DXMember]

there is no single living person that can make a modern computer mouse

[Megahurt]

I don't know about 25GHz ... but if you outsourced everything, I guess you could.

[legonate416]

5 minutes ago, Aleksbgbg said:

If you had as much money as mr. Gates does, then would you be able to build your own PC from start to finish? I'm talking manufacturing your own one-off CPU, GPU, mobo etc.

 

Could you make, I don't know ... a 50 core 25GHz CPU or something?

Or is it that for some reason you wouldn't be able to ...?

 

Just wondering ... speak your thoughts

Well considering our technology isn't really even there yet, I don't think it's possible as of right now. Most CPU architecture upgrades are mainly IPC and not clock-speed related. And I'm not positive but I have a feeling that 25ghz isn't even possible with the materials we have for making chips. And to make those 50 cores to be actually worth something instead of AMD's Bulldozer line, that's one huge wafer size. 

[Aleksbgbg]

1 minute ago, legonate416 said:

Well considering our technology isn't really even there yet, I don't think it's possible as of right now. Most CPU architecture upgrades are mainly IPC and not clock-speed related. And I'm not positive but I have a feeling that 25ghz isn't even possible with the materials we have for making chips. And to make those 50 cores to be actually worth something instead of AMD's Bulldozer line, that's one huge wafer size. 

Yeah I was kidding, the idea is ... could you make it suit you?

[legonate416]

1 minute ago, Aleksbgbg said:

Yeah I was kidding, the idea is ... could you make it suit you?

No. I'd just buy the best of the best and be done with it. 

 

6950X and 4 1080s in SLI

 

That's just pocket change for a guy like Gates. 

[Godlygamer23]

4 minutes ago, legonate416 said:

No. I'd just buy the best of the best and be done with it. 

 

6950X and 4 1080s in SLI

 

That's just pocket change for a guy like Gates. 

I would not buy 4 1080s in SLI. Money-growing-on-trees or not, what a waste of funds that could be used in a much better fashion.

[minibois]

Well you would first have to throw a couple billion in the R&D team if you want a 50 core 25GHz CPU and wait a whole bunch of years.

 

But if you wanted them to make you a special CPU, based of something that exists now that for example says "for my absolute favorite Bill, Mr. Gates" on it, you could. If you paid Intel/AMD enough

[samcool55]

No, i wouldn't, i would get whatever is available and probably end up with a sick server-ish system like linus has.

And being rich, it probably means lonely so i would get multiple systems and pay people to play with me...

 

[Bananasplit_00]

well if you had all the monies then just throw it at AMD or Intel and they would probably make you a one off CPU, throw some more at ASUS and they will probably make a mobo that fitts the CPU and then throw some more at AMD/NVIDIA and they will make you some GPUs and shits. throwing wads of money at companies solves a lot of problems you know

[Coaxialgamer]

 You mean acquire a team of engineers that can compete with AMD , intel and nvidia's experience ?Design a new architecture ( which takes years ), find the means of manufacturing it , selling , etc  and remain profitable ( if you plan to sell ) ?

 

As for the 50 core cpu at 25ghz  . there is a reason we don't have one on the market. Not because intel/AMD are lazy/greedy , but because it isz NOT FEASIBLE . Transistor heat and power requirements skyrocket past 5ghz ( main reason why we moved to more cores vs faster cores = power wall)  and cannot sustain it under anything else than extreme cooling ( also , even under ln2 , no cpu has ever hit over 8.5ghz . A transistor is a transistor , so even a new architecture wouldn't change this).

Second : cores take up space on the die . Realistically , you cannot manufacture a chip more than roughly 600mm². Now , cores don't take up much space on a cpu ( skylake core is like 20mm² . What does take a very large amount of die space is cache . 50 cores would require a buttload of cache to keep cores fed with data. Now , you could make much simpler cores ( like gpu cores or xeon phi cores) with much less cache , but then you wouldn't necessarily be able to do some workloads . Not to mention the problems in optimizing software for 50+ threads would be a nightmare.

[legonate416]

7 minutes ago, Godlygamer23 said:

I would not buy 4 1080s in SLI. Money-growing-on-trees or not, what a waste of funds that could be used in a much better fashion.

I'm quite aware of 4 1080s being a waste, thank you. 

 

The point I was making is that it's much cheaper to get the best possible experience with today's parts than to design your own and build with those. 

 

And, worth it or not, 4 way SLI is faster than 3 way or 2 way. Not by much, but the difference is there.

[Sauron]

You would ahve the theoretical funds needed to build a pc from start to finish. However, it would be garbage.

[Coaxialgamer]

You already can .If you wanted wanted to build your own cpu , i would get an fpga or make something with TTL . Would be a lot of work , but doesn't require a lot of money . It would also be ( virtually ) useless.

[Mr.Meerkat]

18 minutes ago, Godlygamer23 said:

I would not buy 4 1080s in SLI. Money-growing-on-trees or not, what a waste of funds that could be used in a much better fashion.

like food?  

[Godlygamer23]

1 minute ago, Mr.Meerkat said:

like food?  

That or maybe charity. 

[BrightCandle]

If we limit how far we talk about "from scratch" and assume we use TSMC/Samsung/Intel to actually produce the silicon we design then its certainly plausible within years to design and build your own system and platform, multiple companies have done so relatively recently (The raspberry Pi, Parallela etc). You would be limited to the state of the art in processes assuming you wanted the sort of volumes these guys deal in (100,000 to millions of chips) so you need a mass market because not only is the design process going to be millions to billions but the product costs at volume are astronomical as well.

 

However if we talk about building a modern silicon plant from scratch and all the tools and everything else then I don't think Bill Gates actually has enough funds to do it. It takes a country worth of people to run a silicon industry and its spread through the world and rebuilding all of that from scratch is unsustainable for any human being today regardless of funds.

 

One likely area of interest that might produce favourable performance gains with a very different type of CPU would be clock less. We have known for a while a way to make CPUs that don't require a clock, they waste a lot of transistors on sycnhronising and capturing the results but there are very real benefits to running as fast as the individual transistor will actually go and it should be possible to develop a CPU that could be 10+Ghz equivalent. I am sure Intel and others have research labs working on these so you really would need a substantial development team but its got potential as an approach.

 

The other possible avenue is that of using Germanium instead of silicon. Its got almost no infrastructure so you likely want to focus mostly on the process technology itself over decades of research and production but we know it can sustain better clockspeeds and lower power consumption, but the last time anyone made an actual chip out of it was the 1980s. This is technically a possible approach although I would argue for making the process work as a business first, selling it to ARM partners and such first to recoup some costs before pursuing a CPU/Motherboard etc business

[straight_stewie]

Just now, Aleksbgbg said:

Could you make, I don't know ... a 50 core 25GHz CPU or something?

Considering the best and brightest in this industry can't do that and have decided that without moving away from transistor based technologies we are nearing the end of Moore's law, I doubt it.

You can, however, already build your own CPU out of FPD's or discrete logic gates (like 74xx series chips). You can even use premanufactured cpu's like z80 or early Intel ripoffs and build your computer around those out of FPD's or discrete logic gates. It's called a "homebrew" computer. If you're willing to read alot and teach yourself a few (read: Alot of) things, you can even buy an x86 processor, and manufacture your own motherboard, ram, and other i/o to make a complete system.

[TetraSky]

Well, it is possible, but you wouldn't be able to do it yourself.
You'd need to hire/contract Intel/AMD (if you want x86) to make a custom chip along with a custom motherboard for said chip. But you wouldn't have any support afterward.

If you wanted like 50+ cores, they would probably need to make the cpu a lot bigger, which means a custom heatshink will be needed...  And really, the "custom" build will be based off existing CPU architecture and not some brand new one that they just came up with, as that would just take way too much time and effort (and money, surely even Bill's fortune will go down the drain very fast at that point) 

 

Overall, this will cost a sh*t ton of money, but somewhat possible. This is essentially what consoles are, custom build based on existing tech.

 

But you also have to be realistic. A 50 cores 25GHz CPU would consume a lot of power, output a lot of heat and will most likely be massive. That would just be bad.

 

[Coaxialgamer]

4 minutes ago, BrightCandle said:

If we limit how far we talk about "from scratch" and assume we use TSMC/Samsung/Intel to actually produce the silicon we design then its certainly plausible within years to design and build your own system and platform, multiple companies have done so relatively recently (The raspberry Pi, Parallela etc). You would be limited to the state of the art in processes assuming you wanted the sort of volumes these guys deal in (100,000 to millions of chips) so you need a mass market because not only is the design process going to be millions to billions but the product costs at volume are astronomical as well.

 

However if we talk about building a modern silicon plant from scratch and all the tools and everything else then I don't think Bill Gates actually has enough funds to do it. It takes a country worth of people to run a silicon industry and its spread through the world and rebuilding all of that from scratch is unsustainable for any human being today regardless of funds.

 

One likely area of interest that might produce favourable performance gains with a very different type of CPU would be clock less. We have known for a while a way to make CPUs that don't require a clock, they waste a lot of transistors on sycnhronising and capturing the results but there are very real benefits to running as fast as the individual transistor will actually go and it should be possible to develop a CPU that could be 10+Ghz equivalent. I am sure Intel and others have research labs working on these so you really would need a substantial development team but its got potential as an approach.

 

The other possible avenue is that of using Germanium instead of silicon. Its got almost no infrastructure so you likely want to focus mostly on the process technology itself over decades of research and production but we know it can sustain better clockspeeds and lower power consumption, but the last time anyone made an actual chip out of it was the 1980s. This is technically a possible approach although I would argue for making the process work as a business first, selling it to ARM partners and such first to recoup some costs before pursuing a CPU/Motherboard etc business

GaN transistors would be an alternative .  Potentially a good alternative to germanium as it can be used with existing manufacturing tech, and has many advantages over Si

[Tech_Dreamer]

Skip all the the slow marketing bullsh*t & get the unreleased bleeding edge chip  with as much performance i can squeeze out of a single dye& mobo & rock the worlds fastest pC!

[BuckGup]

Or just donate tons of money to companies so there is more competition and everyone innovates faster helping everyone.