The first is that they would spend all their time waiting on memory. If you run a CPU at 5 ghz, you will have 500 cycles before you get the bytes from memory you want.
1% of a modern CPU is running the instructions, the other 99% is trying to keep it busy. You would then need to double the transistor count for cache, but even then you will be waiting on memory for everything that you haven't already used and put in cache. To counter act that you need prefetching so you can get from memory ahead of time etc. This goes on and on until you either build a modern CPU or a modern GPU.
The other problem is that this doesn't solve a problem anyone has. People aren't really clamoring for more raw math or instructions any more. Most of it goes to waste by running scripting languages that run at 1/100th the speed of naive C. Most C runs at 1/100th the speed of a real optimized program. It's as if someone built a super powerful engine and everyone just decided to fly jumbo jets to work instead of using the same cars and less gas.
Then finally there are modern GPUs which already can do one million floating point operations in the time it takes to get a single byte from memory (the single byte is really two cache lines which is the minimum and this also includes the memory not being in the TLB).
It's all memory latency and software optimization. The need for raw math is not a general computing problem any more, it has all been relegated to GPUs.
Waiting for those periods only occurs when a modern cpu is fetching memory outside its caches. When its all in cache it runs at basically full speed.
> To counter act that you need prefetching so you can get from memory ahead of time etc. This goes on and on until you either build a modern CPU or a modern GPU.
Thats all true but what im suggesting is if you had MORE a lot MORE humble cpus even if they run at a much less than GHZ, doing a little work the sum of the humble cpus would actually become more than the modern fast cpu.
Think a lot of slow donkeys would still get move more than a ferrari, for less money and less fuel.
> Then finally there are modern GPUs which already can do one million floating point operations in the time it takes to get a single byte from memory (the single byte is really two cache lines which is the minimum and this also includes the memory not being in the TLB).
Thats all true but the cost of that performance is a lot of heat and silicon.
What if that power was given to a million 6510s ?
I said that exactly. Half the cpu will be cache and it only helps once data is in the cache, which it won't be if you are running through big arrays. Big arrays would be necessary to supply a bunch of tiny cores, unless of course you had prefetching and other mechanics.
When its all in cache it runs at basically full speed.
L3 cache will not run at full speed. The latency is much better but you still have 40 cycles. That means that every lda instruction will wait for 40 cycles when in old computers it would take one cycle. That's just for anything already in cache, not the 500 cycles you would stall for something outside of cache. Don't forget that there are no cache lines and no out of order execution so you pull down a single byte after 500 cycles and if you read the next byte you wait 500 more cycles.
Thats all true but what im suggesting is if you had MORE a lot MORE humble cpus even if they run at a much less than GHZ
Your idea is clear, it's not a misunderstanding, it's just something that has been considered a long time ago and dismissed.
Think a lot of slow donkeys would still get move more than a ferrari, for less money and less fuel.
People don't want slow donkeys, they want to move themselves with a ferrari or jet, or huge amounts of stuff on a giant ship. Donkeys are obsolete. They don't cross water, they don't move fast, they don't go straight.
Thats all true but the cost of that performance is a lot of heat and silicon.
So what?
What if that power was given to a million 6510s ?
You would have the same number of transistors and you would be able to do 0 floating point calculations because those CPUs had no floating point units. You can't even multiply integers. You can add and subtract integers. Now you have to sit and wait on every memory load, you have lower Ghz and every math operation is going to use lots and lots of cycles because you can't even multiply integers. Good luck with that.
The 6510 did not have any cache. Im suggesting the cost/benefit of cpus that do not have any cache, no branch prediction, no out of order, just the minimum to execute its ISA.
Again cutting things down to as simple as possible and hope for the best.
> Your idea is clear, it's not a misunderstanding, it's just something that has been considered a long time ago and dismissed
It was dismissed because x86 has a very large impossibly expensive range of software that would be lost on any new different arch.
> You would have the same number of transistors and you would be able to do 0 floating point calculations because those CPUs had no floating point units.
The 6510 could do floating point in software. Im not saying that the solution must be a million 6510, one could add a simple FPU and advance some of the features. Im just suggesting an arch without all the extras that require lots of transistors like Out of order execution, caches and all that.
Right, again this is clear. Although you did talk about cache earlier, it's not a mystery what you are trying to say. It's simple, easy and doesn't work.
Again cutting things down to as simple as possible and hope for the best.
These old CPUs are extremely efficient with transistor budgets but they do almost nothing.
It was dismissed because x86 has a very large impossibly expensive range of software that would be lost on any new different arch.
This is completely wrong. If this architecture worked, there would be a 100,000 original x86 cores on a chip. You can look at the evolution of x86, super scalar computers like cray machines, the successors to the 6502, chips from sun, dec, SGI etc. and see what they added and why.
There are new CPUs running riscv that you can buy right now and they don't do anything like this even though it is running on a completely new instruction set.
The 6510 could do floating point in software.
32 or 64 bit floating point numbers using just 8 bit integer add and subtract with bit shifts. Kind of seems like the thing that would be much better done in hardware. Maybe do 8 or 16 floating point operations in a single cycle.
one could add a simple FPU
The first FPU was 45,000 transistors. Now each chip isn't 3,500 transistors any more. Don't forget you are still moving single bytes around with every instruction, many instructions take more than one cycle to complete, you have almost no registers and the ones you have are all 8 bit and because you have no cache, every memory load takes hundreds of cycles while the cpu sits and does nothing. Also even with this FPU some instructions take 100 cycles to complete and some take 1000 cycles to complete.
https://en.wikipedia.org/wiki/Intel_8087
Im just suggesting an arch without all the extras that require lots of transistors like Out of order execution, caches and all that.
Again, it's obvious that you are suggesting that and with a little research it's obvious why it doesn't work out.
WHY ?
> These old CPUs are extremely efficient with transistor budgets but they do almost nothing.
Yes one does very little very slowly but if you add enough of them, it could be power efficient in a round about way.
> The first FPU was 45,000 transistors. Now each chip isn't 3,500 transistors any more. Don't forget you are still moving single bytes around with every instruction, many instructions take more than one cycle to complete, you have almost no registers and the ones you have are all 8 bit and because you have no cache, every memory load takes hundreds of cycles while the cpu sits and does nothing. Also even with this FPU some instructions take 100 cycles to complete and some take 1000 cycles to complete.
Sure the 8087 was complex, but 45k and 3k for a total of roughly 50k is minute compared to the billions of a modern x86 or m5. What would the performance be if those billions were divided into cpus with a budget 50k or even 100k ? That would buy tens of millions of equivalent 6510.
> many instructions take more than one cycle to complete, you have almost no registers and the ones you have are all 8 bit and because you have no cache, every memory load takes hundreds of cycles while the cpu sits and does nothing.
You are too focused in mixing the limitations of a 6510 and the weaknesses of modern cpu the cache requirement.
It doesnt have to be a 6510, it could quite easily be a 68000 or equivalent but enhanced with a FPU and given more registers. My suggestion was again was these million cpus dont have to share a single memory, they could all have local memories.
Old cpus that did mHZ did not take 500 cycles for a memory access. For being SMART you shoudl know this basic fact, the 6510 for example fetched a memory byte every other cycle. If most instructions require several memory reads for the opcode and possibly operands like memory addresses and if each memory read took 500 cycles, the c64 which ran at 1mhz would take seconds to clear its text mode screen. Did it take seconds for the c64 to clear screen memory and not im not talking about a basic loop ?
> Again, it's obvious that you are suggesting that and with a little research it's obvious why it doesn't work out.
Where are your proofs that your statements are true ?
Demanding that you are right because you say so is very boring and nobody learns when you dont share your knowledge.
What do you mean why? Did you read anything I wrote?
That would buy tens of millions of equivalent 6510.
Not tens of millions. How do you use those cores? They load one byte at a time and it will take 500 cycles to get there. Want a 64 bit address? 4000 cycles for something that happens in less than one cycle on a modern cpu because the cpu can do multiple things at once. Then you can only add and subtract 8 bit numbers.
Want floating point numbers? Now you need 15 times the transistors, loading each float takes 2000 cycles and doing the float math takes 1000 cycles. Don't forget to write it back one byte at a time, that's another 2000 cycles. A single mid range CPU with SIMD is 320,000 faster. Feeding all those cores doesn't come for free and you have no virtual memory which means you have no OS and no multi-tasking.
You are too focused in mixing the limitations of a 6510 and the weaknesses of modern cpu the cache requirement.
It's not me, it's physics, sorry. CPU design for the last 35 years has been mostly about minimizing the impact of the latency from memory.
It doesnt have to be a 6510, it could quite easily be a 68000 or equivalent but enhanced with a FPU and given more registers. My suggestion was again was these million cpus dont have to share a single memory, they could all have local memories.
That "local memory" is called cache and it is done with more transistors. What kinds of programs do you think will run on a computer with no main memory? It is literally the only consistent part of computer design over the last 60 years.
Old cpus that did mHZ did not take 500 cycles for a memory access. For being SMART you shoudl know this basic fact, the 6510 for example fetched a memory byte every other cycle.
You think you're going to clock down a cpu to 1 mhz and have that compete with 5 ghz cpus?
You can fetch a single byte at 500 Khz. That's about 1.28 million times slower that a modern cpu. Now your 240,000 fpus (11 billion transistors of a modern / 6800 + fpu transistors) are going to lose on memory alone. Don't forget that you still have every floating point number taking 100 - 1000 cycles and your cycles are 5000x slower than a modern cpu.
Where are your proofs that your statements are true ?
It's not me, I'm telling you what exists and why. You're acting like this is hidden information. The entire history of CPU design is out there to read. I'm not predicting anything, I'm saying what already happened. Every step was brilliant engineers solving problems.
Demanding that you are right because you say so is very boring and nobody learns when you dont share your knowledge.
What have I been writing? If you want deeper information go find it, I'm not going to copy and paste entire books to a person who doesn't want to hear that their wacky ideas don't work.
We haven't even gotten into how someone would program a computer like this, since it would essentially be impossible.
M: CPUs running in the low megahertz such as a 6510 or 68000 or Z80 and so, on do not take 500 cycles to read a byte from memory. Most 6510 instructions are 3 - 7 or 8 cycles and may cost between 1 and 4 reads or writes. Your statement about 500 cycles only applies to CPUS running at GHZ.
Im not talking about about having thousands of millions of simple 6510 like cpus running at GHZ, im suggesting having them run at their usual speed in the low MHZ, where their clock speed is roughly the same as memory bandwidth.
C: That "local memory" is called cache and it is done with more transistors. What kinds of programs do you think will run on a computer with no main memory? It is literally the only consistent part of computer design over the last 60 years.
m: Sorry for the confusion, im not talking about cache im talking about individual or groups of small 6510s like cpus sharing or having their own memory. Look at the old systems such as consoles. Rather than a unified memory, the SNES for example had main memory for the CPU, some memory for the graphics and the sound chip itself had its own memory. In my suggestion you could repeat the same many many times for many many humble cpus.
C: You can fetch a single byte at 500 Khz. That's about 1.28 million times slower that a modern cpu. Now your 240,000 fpus (11 billion transistors of a modern / 6800 + fpu transistors) are going to lose on memory alone. Don't forget that you still have every floating point number taking 100 - 1000 cycles and your cycles are 5000x slower than a modern cpu.
m: The difference between MHZ and GHZ is 1000x not 1000000x. A 6510 is thousnads NOT millions of times slower compared to a single core belonging to a x86.
> speed difference between 6510 and modern x86
> A modern x86 processor is tens of thousands to millions of times faster in real-world computing throughput than a MOS Technology 6510, depending on workload, vector/SIMD use, and core count.
> Key Differences
> Clock speed: The 6510 (used in the Commodore 64) ran at roughly ~1 MHz (0.985–1.023 MHz), whereas modern x86 CPUs run around 3–5+ GHz—roughly 3,000x to 5,000x as many clock cycles per second.
I thought i said it before but theres nothing wrong with my idea supporting simple slow FPU units.
c: What have I been writing? If you want deeper information go find it, I'm not going to copy and paste entire books to a person who doesn't want to hear that their wacky ideas don't work.
m: You are misunderstanding many fundamental statements and not appreciating the spirit of my suggestion. For example the issue of FPUS i never said my suggestion shouldnt have FPUs im simply saying it could have simple FPUs. Modern cpus have a lot of extras that cost transistors, eg multiple levels of cache, Out of order execution which of course requires register renaming and all the extra copies of stuff. If you reduce things as simple as possible with no caches of any kind, no Oout of order, no renameing of registers, none of all the buffers you save a lot of transistors.
Those transistors mean you can have a lot more cores running.