SERVING THE QUANTITATIVE FINANCE COMMUNITY

 
User avatar
outrun
Posts: 4573
Joined: April 29th, 2016, 1:40 pm

Re: Looking for hardware recommendations

June 24th, 2017, 11:15 pm

Cool, that sounds like a very serious machine! The machines on that link also look very good, sounds like you're going to have a lot of fun and speed!
 
User avatar
Cuchulainn
Posts: 61624
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Looking for hardware recommendations

June 25th, 2017, 10:56 am

http://www.datasimfinancial.com
http://www.datasim.nl

Every Time We Teach a Child Something, We Keep Him from Inventing It Himself
Jean Piaget
 
User avatar
Alan
Topic Author
Posts: 10065
Joined: December 19th, 2001, 4:01 am
Location: California
Contact:

Re: Looking for hardware recommendations

June 30th, 2017, 3:31 pm

Thanks, guys for all the good suggestions in this thread. I finally ordered my new system yesterday. Found a local desktop systems builder down the road from me (Xi Computer Corp, San Clemente, Ca)  that was quite helpful. Went with a win 10 pro desktop, Intel i9-7900X cpu, 64GB memory, Geforce GTX 1080 Ti video card.  Should be fun!  

It seems that a "pure compute" GPU (one that supports so-called TCC mode) is, as a practical matter, only designed to be in servers. The main limitation of a video card in MS windows is apparently a windows time-out in 2 secs for GPU kernels (TDR). (There are some work-arounds, registry edits, etc). Anyway, my plan is to live with that until I get more experience with CUDA. Also, now I have so many cpu cores that they will be my first target for parallel stuff.   
Last edited by Alan on July 1st, 2017, 6:23 pm, edited 2 times in total.
 
User avatar
Cuchulainn
Posts: 61624
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Looking for hardware recommendations

June 30th, 2017, 3:49 pm

Looks like 10 cores with hyperthreading?
http://www.datasimfinancial.com
http://www.datasim.nl

Every Time We Teach a Child Something, We Keep Him from Inventing It Himself
Jean Piaget
 
User avatar
Alan
Topic Author
Posts: 10065
Joined: December 19th, 2001, 4:01 am
Location: California
Contact:

Re: Looking for hardware recommendations

June 30th, 2017, 3:51 pm

yes! -- nice, huh? 
 
User avatar
Cuchulainn
Posts: 61624
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Looking for hardware recommendations

July 2nd, 2017, 1:52 pm

Yes :) 

At some stage the C++ connection could be useful, especially if device-independence and OpenCV is important.
#include <vector>
#include <algorithm>

#include <boost/compute/algorithm/transform.hpp>
#include <boost/compute/container/vector.hpp>
#include <boost/compute/functional/math.hpp>

#include <iostream>

// 1. Boost "Compute" Library
// 2. Intel OpenCL install

// djd

namespace compute = boost::compute;

int main()
{
 // get default device and setup context
 compute::device device = compute::system::default_device();
 compute::context context(device);
 compute::command_queue queue(context, device);

 // generate random data on the host
 std::vector<float> host_vector(100);
 std::generate(host_vector.begin(), host_vector.end(), rand);
 for (auto el : host_vector)
 {
 std::cout << el << ",";
 }
 std::cout << '\n';
 // create a vector on the device
 compute::vector<float> device_vector(host_vector.size(), context);

 // transfer data from the host to the device
 compute::copy(
 host_vector.begin(), host_vector.end(), device_vector.begin(), queue
 );

 // calculate the square-root of each element in-place
 compute::transform(
 device_vector.begin(),
 device_vector.end(),
 device_vector.begin(),
 compute::sqrt<float>(),
 queue
 );

 for (auto el : device_vector)
 {
 std::cout << el << ",";
 }
 // copy values back to the host
 compute::copy(
 device_vector.begin(), device_vector.end(), host_vector.begin(), queue
 );

 return 0;
}


Boost has a nice layer ...


BTW Let us now if the electricity bill has any spikes :D :roll: :shock: :oops:
http://www.datasimfinancial.com
http://www.datasim.nl

Every Time We Teach a Child Something, We Keep Him from Inventing It Himself
Jean Piaget
 
User avatar
outrun
Posts: 4573
Joined: April 29th, 2016, 1:40 pm

Re: Looking for hardware recommendations

July 3rd, 2017, 5:37 am

Excellent machine! Congratulations.
 
gvald004
Posts: 1
Joined: July 5th, 2017, 2:41 pm

Re: Looking for hardware recommendations

July 5th, 2017, 2:47 pm

If anyone is in the market for high performing desktops in the cloud (Windows / Linux) definitely check out Paperspace. Super simple to get setup,
 
User avatar
outrun
Posts: 4573
Joined: April 29th, 2016, 1:40 pm

Re: Looking for hardware recommendations

July 6th, 2017, 7:44 am

If anyone is in the market for high performing desktops in the cloud (Windows / Linux) definitely check out Paperspace. Super simple to get setup,
I have a couple of questions:
* What GPUs are in the $0.35 hourly PRO instance?
* If I upload my code for a job and then later delete it, what contractual guarantees to I get w.r.t. NDA of my code?
* are these 3rd part machines (Amazon, IBM, google) or owned by you in your own datacenter?
* can I easily fire up multiple machines and share data between them? E.g. I might have 1Gb of data that all instances needs, but each instance processing it differently. Can I mount or SCP?
Thanks!
 
User avatar
katastrofa
Posts: 8992
Joined: August 16th, 2007, 5:36 am
Location: Alpha Centauri

Re: Looking for hardware recommendations

July 23rd, 2017, 2:10 am

8 or 10 core Intel Core i7 or i9 with 124GB RAM, 8 core AMD Ryzen 1800X with only 64GB RAM (they claim it theoretically works with 2 x 64GB, but my experience with trying to achieve such things with older AMD processors was otherwise) or waiting for Threadrippers? I don't know much about Ryzen architecture (I lost interest in AMD for long years after the experience with an Opteron cluster) - I only know it's been designed from scratch. Apologies for the rambling, but I'm trying to collect my thoughts on getting a new computer and I'm curious what are your feelings about the current market.
 
User avatar
outrun
Posts: 4573
Joined: April 29th, 2016, 1:40 pm

Re: Looking for hardware recommendations

July 23rd, 2017, 11:55 am

Intel has hypertreading which in my experience adds 60% extra virtual cores (when running multithreaded highly parallisable code). AMD doesn't have that.
 
User avatar
outrun
Posts: 4573
Joined: April 29th, 2016, 1:40 pm

Re: Looking for hardware recommendations

July 23rd, 2017, 11:58 am

Investments also have a high decay rate, you wouldn't care less about details on your 2nd day after purchase? Keep your money when you can.
 
User avatar
Billy7
Posts: 282
Joined: March 30th, 2016, 2:12 pm

Re: Looking for hardware recommendations

July 24th, 2017, 1:11 pm

Intel has hypertreading which in my experience adds 60% extra virtual cores (when running multithreaded highly parallisable code). AMD doesn't have that.
I think 60% may be painting too rosy a picture for HT, but I may be wrong of course.
I thought it was less of a benefit last time I checked (many years ago), so I quickly run a couple of simple tests, Monte Carlo pricing, practically 100% parallel efficiency).
Test 1: ST: 162s, MT no HT: 40.45s, MT with HT: 31.45 s. Virtual cores added: 40.45/31.45 = 28.6%
Test 2: ST: 70s, MT no HT: 17.4s, MT with HT: 13.8 s. Virtual cores added: 17.4/13.8 = 26.1%
That's on my now very old CPU (i7-920), it'd be interesting to see if HT implementation has significantly improved in new CPUs, maybe it has.
A brief search online about HT performance seems to confirm my numbers at most, people claiming from no improvement at all in some applications, up to 30% in others. 
 
User avatar
outrun
Posts: 4573
Joined: April 29th, 2016, 1:40 pm

Re: Looking for hardware recommendations

July 24th, 2017, 3:48 pm

Intel has hypertreading which in my experience adds 60% extra virtual cores (when running multithreaded highly parallisable code). AMD doesn't have that.
I think 60% may be painting too rosy a picture for HT, but I may be wrong of course.
I thought it was less of a benefit last time I checked (many years ago), so I quickly run a couple of simple tests, Monte Carlo pricing, practically 100% parallel efficiency).
Test 1: ST: 162s, MT no HT: 40.45s, MT with HT: 31.45 s. Virtual cores added: 40.45/31.45 = 28.6%
Test 2: ST: 70s, MT no HT: 17.4s, MT with HT: 13.8 s. Virtual cores added: 17.4/13.8 = 26.1%
That's on my now very old CPU (i7-920), it'd be interesting to see if HT implementation has significantly improved in new CPUs, maybe it has.
A brief search online about HT performance seems to confirm my numbers at most, people claiming from no improvement at all in some applications, up to 30% in others. 
Good point. I've just checked it and a test a did a long time ago gave a 35% speedup.
 
User avatar
ExSan
Posts: 4552
Joined: April 12th, 2003, 10:40 am

Re: Looking for hardware recommendations

July 26th, 2017, 9:27 pm

I am waiting for Intel 10 cores i9 series. Hopefully less than $1000 when the battle with AMD end
ABOUT WILMOTT

PW by JB

Wilmott.com has been "Serving the Quantitative Finance Community" since 2001. Continued...


Twitter LinkedIn Instagram

JOBS BOARD

JOBS BOARD

Looking for a quant job, risk, algo trading,...? Browse jobs here...


GZIP: On