64 core

Hi, excuse my ignorance and thanks in advance for any advice.

I need an upgrade! For context, I trade stocks for a living and I use a VPS to run my backtesting data programs.

The VPS is 32vCPU / 64GB RAM / 500GB SSD but is too slow.

I believe VPS cores are not the same as actual cores??

Anyway, I only need a build to run one program that processes lots of data, no visual or graphics requirements.

I’ve been looking into 64core threadrippers, not too sure if they are needed, or the tradeoff between threads and cores etc.

I’m new to all of this.

Can anyone recommend a DFY build or requirements for processing speed of data.

I’d like the speed of the system to be at least 5 x faster than the VPS i am using.

Thanks in advance

I am a full time stock trader as well but why so much processing power. Why do you need a trading script or anything virtual running?

Sometimes the analogy of K.I.S.S. is more then enough

hate to hi-jack this thread, but what sort of performance/services are you running as stock trader?

I am playing with the concept of starting a very tiny datacenter. and while I do not plan on targeting stock traders as a target market, but still curious as to the requirements.

I personally just use a provider and do the work on my computer with nothing fancy(just stop losses and other various trading settings) if my internet goes down(which is rare) or traveling/vacation I just use my phone since it’s all on there too.

Trying to complicate things only create problems

there is a ton of unknowns here but i will take a stab at it while calling out my assumptions.

you have an on site computer doing some data crunching task, FOREX maybe?, that you call your ‘VPS’.

this computer has a SATA based SSD all configured with the OS and everything on it.

realistically this is a database application, feeding the output to ‘something’ locally, hopefully over ethernet, probably wireless.

you need faster storage, that PC might have the ability to use NVME. that would be a MASSIVE performance increase here.

you might also need faster connectivity internally at your work area.

1 Like

Could you give us some more information about the programs that you are running? Are you sure they are CPU limited? It is possible that your performance is held back by storage, or not using hardware acceleration. It is not really possible to help you without some more information. More cores could be better if this is a CPU limited scenario, but you might not get 5x performance.

Let’s start with something very basic. What is your data flow or more precisely what are your inputs and outputs? After that, describe how this data is moved back and forth, as something of the nature you are describing requires networking. After that, we can get into the hardware side of a computer or computers.

Spec a VPS with better storage. What’s the IOPS and peak storage bandwidth on that host? The reply saying NVMe is right in that it would be better than SSD, but it’s rarely clear what’s actually connected to a cloud-hosted server.

You also don’t list the size of your backtesting dataset, or if you’ve profiled the code doing the calculations to find the bottleneck. “Throw more money/hardware at it” is one solution, but more threads/cache/RAM bandwidth (feeding data to the calculation) won’t help if your data is on a slow disk and has to come from disk because it’s too large to fit into RAM. Faster storage won’t help if you’ve got an inefficient implementation of the algorithm used for your backtesting calcs.

Note carefully: performance optimisation needs to consider the system as a whole.

K3n.