PC setups, configurations, benchmarks and general recommendations

Because the interest is high among new users of SQX (i saw many topics with the same goal), i thought it would be a good idea to concentrate everything hardware related in one topic where new users but also more experienced ones, can easely find in one place what they are looking for. What to buy, what configuration, air or liquid cooled…etc.

This topic is designed to help users in choosing the most apropriate hardware configuration according to their budget, a new computer or an upgrade. Also, posting benchmarks for processors, general raw performance benchmarks as a guide but also examples of settings in SQX and the resulting performance for others to have an idea what expectations they can have from a certain processor.

I am sure the contribution of the more experienced users will be greatly appreciated by everyone.

For starters, in the following days and weeks, i will post different results (pairs and timeframes) on my configuration, an AMD 3700X (3,6Ghz 8 cores) and 32GB RAM.

To be relevant, we can agree on a minimum criteria, for example, each setting must be run at least 24 hours, and we can agree to count the accepted strategies/hour (in databank).

The architecture on which the processor is built is very important because besides the number of cores, there is the power consumption and heat output.

The number of cores is important but on the same level as architecture and a CPU benchmark, helps you compare two processors with same number of cores but with very different architecture especially at more than 5 years apart. Only the number of cores, taken out from the whole context, doesn’t tell you much.

When choosing the configuration of the system, you start with the processor based also on the budget and once the processor has been chosen, you choose the power supply and the motherboard (depending also on the form factor) and then the computer case if its standard ATX or micro ATX. The mini ATX or mini ITX are too small for SQX generating system.

Your have:

extended-ATX 30.5×26.9 cm
standard ATX 24,4×30,5 cm
micro-ATX 24,4×24,4 cm
mini-ITX 17×17 cm
mini ATX 15×15 cm

its simple…start with some rebubrished workstation from ebay for hundreds of USD…spending money for some newest PC specs without be profitable is wasting of everything 🙂

if anyone want to start with some newer benchmark, i have prepared some basic build config to download

its random generation, build process will run for 1 hour and after that the crucial value is “STRATEGIES PER HOUR” not the accepted ones

the build task is on EURUSD data cloned to EST07

my results with 2xXEON E5-2680, 2.7 GhZ, 32 cores, 128 GB RAM, full performance:

B122 – 808000

B124 – 800000

B126beta – 735000

B126 is slower by something like 10%, because of always turned on MAE/MFE, without possibility to turn it OFF

i dont understand, why we need to have turned something automatically ON, if we dont used it in building process – and 10% difference is not small

hankeys wrote:

its simple…start with some rebubrished workstation from ebay for hundreds of USD…spending money for some newest PC specs without be profitable is wasting of everything 🙂

i would might add: a refurbished workstation…with the minimum configuration of dual CPU and 16 cores (in total, 8 cores each CPU) and minimum Ivy Bridge. To some it might be difficult in the first second to tell which architecture it is, but they are easily recognizable:  v2 = Ivy Bridge, v3 = Haswell, v4 = Broadwell.

An ideal or recommended configuration would be a dual CPU E5-2630 v4 2.2 Ghz 10 cores (20 cores in total).

Great idea with the newer bench, i will run it as soon as i would find time.

the bench is EURUSD…on which timeframe?

timeframe doesnt matter in the benchamrk if everyone use the same settings – we are watching only for raw strategies per hour…only that value is comparable and only with random generation

but its prepared for H1 timeframe with selected TF only precision

everything is set in the CFX file – you need to use only your EURUSD data, because my naming for EST07 EURUSD is “EURUSD_M1_UTC2”

hankeys wrote:

everything is set in the CFX file – you need to use only your EURUSD data,

ok

speaking of hardware, i saw you wrote you have a 2 x XEON E5-2680, 2.7 Ghz 16 core (32 cores in total). I wasn’t paying very close attention, can you post an Ark Intel site link ? i am asking because out of curiosity i looked for that processor but i cant find it. There is a Broadwell v4 but at 2.4 Ghz and 14 core and a v2 Ivy Bridge but with other specs.

My run of hankeys benchmark.cfx from post #250667

First my system

·         HP Z620 Workstation, 16-Core 2x 8-Core 2.60GHz Intel Xeon E5-2670, 96GB RAM -Refurbished ($699 on Amazon when purchased in April 2019)

Next Benchmark Strategies per hour

·         B122           –  786,299

·         B124           –  764,393

·         B126 beta –  682,190

So I ran the hankeys benchmark.cfx Build126 beta again but this time setting maximum memory to 64GB,

(originally I took the default of “Let program determine maximum memory”),

This change increased Strategies per hour to 754,020 from 682,190. Or did it? Was it the memory setting change or just rerunning the benchmark?

I also ran the benchmark test with the given .cfx file

B124 – 756.254

it reached aprox 600.000 in the first 8 – 10 min and then increased very slowly until the end. Memory usage: aprox 2 GB.

Conclusions:

-memory usage on this particular test is very small, so Michael, in my view its more the rerun than increasing the memory.

-CPU-Z database with benchmarks seems almost mirrors exactly the performance in using SQ so its a very handy, easy tool to asses computer configuration (CPU performance).

i am attaching a few photos of my system

the M.2 SSD can be partially seen on the motherboard between the video card and CPU cooler

ivan, looks like a very clean build and based on the benchmark.cfx test comparison your newer 8 core rig vs my 16 core Intel has me definitely taking another look at an AMD solution. But I am not yet ready to make a move. My little puppy has been running pretty much 24/7 for weeks now with an occasional break to install software/updates. I keep toying with the idea to do a build-out, something I haven’t done in 20 years and leave my workstation untouched.

in my case, the situation was a bit different because i live in eastern Europe, Romania where you cant find refurbished workstations at decent prices, only at middlemen or with expensive shipment

another question i asked myself: ok, this is 7 year old technology and the workstation was used 24h a day for at least 6 years (maybe not in full load all the time). How long (how many years) can i expect that particular workstation to work (given the 3 or 6 month guarantee from the reseller)

and the last question: when it comes to upgrade, what would be the total cost ? because although technically possible, no one upgrades the processors on these dual CPU computers. I have to sell the workstation all together and buy another one. For example, in the custom desktop case, the answers are more easy to answer: i brought the 3700X with 294 EUR and i can sell it after a year with around 190 EUR. I can buy then a 3900X with some 500 EUR (i put the price of next year, not presently). So with aprox 310 EUR i can increase by aprox 80% the processing power and replace only the CPU, not any other component. A very easy and cheap procedure it self.

So when buying a present day platform, you use the CPU 2 years and then upgrade and use it another 2 years.

In the case of the refurbished workstation, lets say i brought it with 600 – 700 EUR ( i am using EUR because living in Europe its easier to me). After using it 2 years, or 3 at max, who can i sell it to and at what price? can i get 200 EUR for it? those workstations were over evaluated from the start in the first place. So i am almost certain that it would cost me double to upgrade a workstation and a custom build desktop.

but there is another issue to which i hope i will find the answer in the next year, here on this forum, with someone who has for example a Threadripper and is willing to perform a test

i had so far a laptop with a mobile 4 core Core i7 4700HQ at 2,4Ghz. When i transferred the workload on my currently 3700X, i discovered to my surprise and horror that increasing 3 times the processing power (this is the difference between the two), did not shorten 3 fold, the time needed to generate a strategy. In other words, although in such short tests and benchmarks you get 3 times the number, when i generated for example on USDCAD or USDJPY on H1, ordinary strategies with ordinary settings, the difference was not proportional with the processing power difference.

On the old mobile 4700HQ i used to get 14 days to get a decent functional strategy, with the 3700X maybe in 10 days. Although we are talking about a processor 3 times more powerful.

So in my view, the complete benchmark or comparison will be with the same data and .cfx file reaching 3.000 or 5.000 in databank on a certain market and timeframe in how many days (or hours).

I hear you loud and clear. I personally have approached my pc purchases as a sunk cost to be given away at end of life, then again I’v only personally bought a couple.  When I was working they were always provided as part of the job and the life cycle was out of my control. My i7 laptop I’ve had for 6+ years before the Xeon workstation was brought on board and I still use the laptop, just not for development. I swapped the Samsung ssd out of my laptop to the workstation for a boot drive (it was faster than the ssd provided). I could very easily get caught up in the latest and greatest but the Xeon is driving home the point of expected vs. actual throughput as your previous posts have pointed out. I hope that made sense without going into minute detail.

As far as doing a build out of a new system, maybe next year, I get the sense next year is going to be a change year.