Cryptocurrencies, fad or future?

I gather if you want anonymous and untraceable transactions this is the way to go. Which does not appeal to me, I'm quick to say...

Mike.

Also the ultimate vehicle for market speculation: expensive to produce but worthless.

There is a big market for CPUs and GPU chips and memory and stuff for mining, too. When the coin thing crashes, some semi suppliers will be in trouble.

Expect some heavy iron on ebay.

My understanding is these are not anonymous or untraceable by the government. Other citizens maybe, but then deposits to bank accounts are both for citizens.

Rick C.

Tesla referral code ----+

I hope so I wanna build a sweet GPU-accelerated FORTRAN supercomputer!

Your understanding is incorrect. If you have the private key, you control the coins.

If you can't keep the private key secret from the government, or the IP address where your transactions originate is traced to you, that's a different problem. It's not because the coins themselves are traceable. Use TOR or something if you care.

For me, the whole thing is too much trouble. If my government won't protect me, I have bigger problems that them knowing what I buy and how I pay.

Clifford Heath.

AFAICT Bitcoin mostly exists to evade Chinese currency controls. It's also the only way Julian Assange can pay his bills.

Apart from that, and the general coolness factor that is now past its sell-by date, ISTM it's mostly for tax evasion and money laundering.

Cheers

Phil Hobbs

Protect you from what exactly???

Rick C.

Tesla referral code ---+-

I want LT Spice to run on GPUs. I want sliders to change part values and see waveforms change in real time. 1000x my current CPU power would be a start.

And generating CO2. You forgot the CO2. That's what it's *really* good at.

A recent review by an Australian government-funded study found that there is nothing currently being done that can be done better using blockchain. Telling!

Clifford Heath

I have a friend who got in cheap and early and bought a house in Truckee with appreciated coins. I think he sold too soon, could have bought the entire town.

Right, with the Australian government's deep understanding of crypto issues, that settles it. ;)

Cheers

Phil Hobbs

LOL, yes. But they accidentally hire good people sometimes, mainly because they can't tell the difference.

Clifford Heath.

Running SPICE on a GPU is possible it's parallelizable but AFAIK not trivially so in the way shading or mapping an image to a polygon is, where each pixel can often be rendered totally independently of the others, some pre-work has to be done which slows it down.

With current GPU tech the speedup I saw advertised for some GPU-SPICE implementation on a single modern GPU vs. a modern multi-threaded CPU implementation utilizing four cores was 3 - 5x depending

which shows that with cash going away, it becomes real easy for governments to confiscate your money and stop your income if they want to

and the usual pump and dump scams

I think I'm already running four cores, so the partitioning problem is at least partly solved.

We tried running LT Spice on an Amazon compute farm, with lots of cores, but it wasn't much better.

You can buy time on big FPGA arrays from Amazon. Get the core bits of SPICE rewritten in VHDL and you could have your real-time simulations.

Clifford Heath.

Even a 3x speedup is pretty significant/useful, no?

Rick C.

Tesla referral code ---++

I run LTspice on a 24-core AMD machine, and the number of threads has only a minor effect on performance. Amdahl's law again.

Cheers

Phil Hobbs

It may be worth experimenting with limiting the number of cores permitted to run. I have a sneaky suspicion that like with chess problems if you run too many cores it goes IO bound and slows down as you add more cores after a certain point. Essentially the machine does a lot of extra work in parallel that is later discarded.

On my 8 core machine that happens at about 6 cores active. The last two cores merely create extra heat and lower overall performance. It takes a very special type of problem to run well on a parallel implementation.

Experimenting with the maximum number of cores permitted to run on the task is worthwhile on non-trivial computational problems. I have known several where the system worked best when there were a very particular number of CPUs allowed to run - paradoxically the most important of them is the one which allocates work to the others that do the grunt work.