Hacker News Reader

I don't quite buy that you could really have said the same about the 32 bit to 64 bit transition. 4 Gigabytes is a far cry from 18 exabytes! It is going to be a long while till we scale up to that level.

Yeah, but I remember when computers had 1KB RAM, and 64KB was more than the largest mainframe and seemed insanely massive.

And then there was the time, a handful of years later, when "No one will ever need more than 640KB".

And then 4GB was stupidly massive, and no-one could want, let alone afford, that much RAM.

The same with CPU clockspeed (my first computer ran at 2MHz), storage capacity (I paid several thousand dollars for my first 5MB hard drive).

Which is not to say that the curve won't flatten out as we approach the number of atoms in the universe, but ... there's never quite enough of anything.

Wouldn't a computer with 18 exabytes of RAM be totally impractically massive, even with maximal possible miniaturization? I agree with warehouse-scale "thinking machines" going beyond 64 bits, but standard hardware is hard to believe.

Current chips are basically 2D. Not quite, in that they do have 3D features, but those 3D features are essentially still implementing individual cells in a 2D circuit.

And those 2D circuits fit, eg. 4Gbits of DRAM into a single layer chip. The actual chip inside the plastic packaging is quite small -- let's say 4x4mm.

Now, you can can pack multiple layers of chips into a single package. Samsung is doing at least 12 layers for HBM, for instance, and they're only 60um thick each.

So you've got a chip that is roughly 65536 cells wide, and 65536 cells long, but you can only stack it 12 cells high.

Let's get that to 65536 layers: roughly a 4mm cube. That's now 256Tbits per package. Put 36 of those on a DIMM, and you've got an 8PB DIMM, and 512 PB in a 64 DIMM slot server.

Perhaps we can double the density? That gives us 4EB in a server.

This totally glosses over myriad difficult issues, but it's a very, very long way from being inconceivable when you consider the difference between a 512 byte 2114 RAM chip that was popular when I got my first computer, and what we have today.

If security starts to depend more and more on a random address NOT being real addressable memory the move to 128 bits may come sooner rather than later.

I suspect it’ll happen because a 128 bit wide bus will be useful for calculations.

The width of the bus, the width of the register, and the size of the pointer are all different values.

A 64-but cpu has registers that are larger than 64-bits (eg 512bits for AVX512) and the memory bus itself is already either 128 or 256bit already.

128bit pointers seems questionable for scalar processing and vector processing already has wider registers/GPU/TPU for acceleration.

From a security perspective 64bits of random space (actually 48ish) is pretty darn good and increasing that space for security is not something I’ve seen compelling papers on.

Bus width is orthogonal to the bitness of a CPU. QPI, the bus Intel CPUs use, is a 20 bit bus. HT, the equivalent AMD technology, uses a 32 bit bus, even on 64 bit CPUs. HBM2, used on certain AMD video cards, uses a 1024 bit bus. The Commodore 64 had 8 bit registers, an 8 bit data bus, and a 16 bit address bus.

We do not need 128 bit data types in hardware. We barely even need data types with 64 bits of precision. It crops up every now and then- for instance, 32 bit floats cannot store a latitude/longitude value precisely enough, however a 32 bit integer can store a longitude value to about 1 foot of precision if -2^32 is 180 degrees west and 2^32 is 180 degrees east. (this has the nifty benefit of 2s complement integer overflow doing the right thing, although only with longitude, not latitude. but I digress.)

There are a handful of situations where you do need more precision than that- RSA and DH come to mind. But in those situations, you'll need arbitrary precision arithmetic anyway. A 128 bit register won't help a whole lot- certainly not enough to offset the cost of doubling the size of all your ALUs, even if the machine is a special purpose one that does nothing but sign SSL certs all day.

The article mentions this (but it is easily missed): while current chips are 64 bits, in practice the memory addressing is only 48 bits, with plans to extend it to 57 bits for server products [1]. Taking this into account, there's a clear slowdown in the growth of the number of bits: 7 years from 16 to 32, then 18 years for anther 16 bits (32 to 48), and now ~18 more years for just 9 more bits. Seems like exponential slowdown, which makes sense given that the size of memory is an exponent of the number of bits.

[1] https://en.wikipedia.org/wiki/Intel_5-level_paging

You're counting the years per bit incorrectly, the comment you reply to gets it right - the new generation is needed when the current space runs out, not when the next limit is needed in its entirety. So the ~18 years you mention are the time from needing 33 bit addresses to needing 49 bit addresses. (And 7 years is the time from needing the 9th bit of address space to needing the 17th).

Moore's law for RAM capacity has begun slowing over the past few years. The 64 TB system you mention is from six years ago, right about when Moore's law began tapering. If Moore's law had continued doubling RAM capacity every eighteen months, we should have seen four doublings, up to 1PB. But that's not a machine we're seeing.

Notably, the bandwidth of RAM does not and has not been following Moore's law, improving linearly, and the latency is more or less constant. So as workloads scale up in size, performance will get worse and worse and worse, until "just buy more RAM" no longer solves the problem. Similar to how swap partitions used to be a thing, but performance is now so bad that swap can't be used as cheap RAM anymore.

https://www.semanticscholar.org/paper/Understanding-and-Impr...

While that’s 64TB in a single address space, it isn’t a desktop computer. An apples to apples comparison would consider the 1975 Cray-1 which had a 64 bit CPU, and similarly there have been 128 bit CPU’s.

64bit didn't come in 2003, it came in 1991 in MIPS, in 1992, in Alpha, and in 1995, in SPARC.

Nintendo 64 and Playstation 2 were using 64-bit CPUs before PCs.

x86 was really late in the game.

Yes, you can argue a lot about the years, and about what constitutes a x-bit machine. The first 32-bit CPU was from 1948, for example (https://en.wikipedia.org/wiki/Manchester_Baby), the IBM/360 had 128-bit registers in 1969 (https://en.m.wikipedia.org/wiki/IBM_System/360_Model_85) and the Pentium Pro is considered 32-bit, but has a 36-bit address bus.

I think the analogy with Moore’s law still holds if you pick machines used in comparable roles. Reason is that you can’t make much use of more memory unless your computer gets faster (an exception would be extremely sparse hash tables, but those would need amounts of memory we don’t have yet)

Assuming an extremely generous 2 cycles/byte, a 1MHz 6502 would take 2 seconds to clear 1MB of RAM, or 2,000 to clear 1GB. That’s over half an hour.

That first Alpha also only had a 34-bit address bus, limiting it to 16GB of physical RAM (https://en.wikipedia.org/wiki/Alpha_21064#Address_unit)

I picked x86 because it was used in more or less the same role for decades and because it made the 16-32-64 steps.

He explicitly said x86.

They did not; they just quoted a Wikipedia article on x86. But then, if you ignore everything but x86 the argument is much weaker, because technology did not stop and wait for Intel.

It would not be that surprising if x86 ceases to be dominant in the future, at which point when the x86 became 64-bit is about as relevant as, say, the release of the PowerPC 620 (first 64-bit PPC chip, released in 1997).

The post said "I think a better argument is to look at history" and then left out 12 years of history. x86_64 was on the trailing edge of that history.

I literally already have too much RAM on my computer for a 32 bit processor to address it all.

More than 64GB? Because that's the limit of PAE introduced with the Pentium Pro (36-bit address bus), which was otherwise considered a 32-bit processor.

Nope, only 32 GB. But maybe next computer?

My laptop has 128GB

I'll be quadrupling my threadripper to 256GB. Can't have the the riff raff getting to where I am, you know.

This is not "No one will ever need more than 640k".