SQX : Performance per core / Number of cores? Threadrippers / I9 ?

I want to ask dev team and adv users what is better in terms of performance for new SQX.

Actually, new threadrippers are out with many many cores 16/24/32  but with lower clock speed per core.

At the end of the month will be out new i9  where clock per core is much better but they have also fewer cores. What do you recommend?

Is SQX optimized for using many many cores or all work will do JVM?

how much ram did you install?

ive got a 1950x, originally 32gb ram.

Having problems with my ram at the moment so only running 16gb.

I’ve noticed slight performance decrease but StrategyQuant works fine with 16gb just adjust settings.

I think 32gb is more than enough with SQ.

thanx;)

The i am sure the ram usage buildup issue will be resolved by dev in the mean time i have solved the issue with this little free program which fixed the problem. Been using it for about 1 week.  https://www.koshyjohn.com/software/memclean/

I set it to clean at 80 % and evry 5 min since thats the only settings there is.

This is a great thread, unfortunate nobody has posted in 3 years. I just did a build just to run SQX. Using an i9-12900K with 64GB of Ram.

It is cranking out strategies in short order. I am actually running 4 custom projects simultaneously. I know the training video’s all reference RAM as being the limiting factor while I have found it’s actually processing power.

Fxgold54 wrote:

This is a great thread, unfortunate nobody has posted in 3 years. I just did a build just to run SQX. Using an i9-12900K with 64GB of Ram. It is cranking out strategies in short order. I am actually running 4 custom projects simultaneously. I know the training video’s all reference RAM as being the limiting factor while I have found it’s actually processing power.

I did build the same pc and found the same, processor is the limiter today. Ram could be 32 gigs and it will be the same.

Hey GermanMazza;

I changed the SQX memory configuration from “Let program determine max memory” to manually setting the max memory to 60 GB ( out of 64GB) and that did speed up SQX and now it is using 47% of memory.

I love the i9, although took me a few steps to get it properly cooled. I am using the Fractal design S36 AIO as a front intake and throttled down the CPU voltage to 1.25. Curious how you did/do it?

ありがとう

I tried SQX on a Ryzen 4600G with 32GB and it was too slow. I’m now looking at building either a Ryzen 5950X or an i9-12900K with 64GB RAM.

I also considered the M1 Ultra Mac Studio, but more familiar with Windows and can probably get two Windows for that price.

On the i9-12900K, how long does the PC crunch before finding the strategies? Curious to know the performance difference between the 5950X and the i9-12900K on SQX. The 5950X has more cores, so in theory it should outperform the i9-12900K as most software cannot take advantage of the extra cores and threads.

The benchmark tests show the i9-12900K outperforming the 5950X as they can’t take advantage of the extra cores.

Hi, which processor is better for running SQX, an AMD Ryzen 9 7940HS with 8 cores and 16 threads, or an i9 13900H with 14 cores and 20 threads? Considering the AMD is more powerful in multicore than the Intel? So, the real question is, what matters more, the number of cores and threads or the overall multicore performance of the processor?

ChatGPT’s  answer. “To determine which processor is better for running StrategyQuant X (SQX), we need to focus on how the software utilizes computing resources. SQX is highly parallelizable, meaning it can take advantage of multiple cores and threads effectively, especially during tasks like strategy generation, backtesting, and Monte Carlo simulations.

Here’s how to evaluate your options:

—

1. Key Comparison: Ryzen 9 7940HS vs. i9-13900H

AMD Ryzen 9 7940HS

Cores/Threads: 8 cores, 16 threads.

Base/Boost Clock: High clock speeds, especially on fewer threads, with excellent multicore performance.

Multicore Performance: Superior to the i9-13900H in some benchmarks.

Efficiency: More power-efficient, generating less heat during sustained workloads.

Architecture: Built on the Zen 4 architecture, known for performance efficiency.

Intel i9-13900H

Cores/Threads: 14 cores (6 performance + 8 efficiency), 20 threads.

Base/Boost Clock: High single-thread performance with a mix of P-cores and E-cores.

Multicore Performance: Slightly lower overall multicore performance compared to the Ryzen 9 7940HS.

Power: Slightly less efficient, can run hotter during prolonged multicore tasks.

Architecture: Intel’s hybrid architecture (P-cores for heavy tasks, E-cores for light tasks).

—

2. What Matters for SQX?

(i) Multicore Performance:

Multicore performance is more critical than the number of cores and threads, as SQX utilizes all available cores to perform computationally heavy tasks like backtesting and strategy generation.

If a processor delivers higher multicore performance, tasks will complete faster, even with fewer cores/threads.

(ii) Thread Utilization in SQX:

SQX scales well with threads, but there are diminishing returns beyond a certain number (often around 16-20 threads for most users).

Having many cores is beneficial, but efficient and powerful cores (like Ryzen’s) can often outperform processors with more but less efficient cores.

(iii) Thermal Management and Sustained Loads:

SQX runs long simulations that stress the CPU. Processors that throttle due to heat will underperform in such tasks.

AMD Ryzen 9 7940HS is typically better at sustained loads due to its power efficiency.

—

3. Conclusion:

Go for the AMD Ryzen 9 7940HS.

Its superior multicore performance aligns with SQX’s demands.

Efficient cores ensure consistent performance over long periods without thermal throttling.

Although it has fewer cores/threads than the i9-13900H, the Zen 4 cores are more powerful per thread in multicore scenarios, making it better suited for heavy, sustained SQX workloads.

Additional Considerations:

Ensure sufficient RAM (16GB minimum, 32GB recommended for large-scale SQX

tasks).

Use fast storage (NVMe SSD) to handle large datasets efficiently. ”

@njoyim hope Thank you very much for your clear, well-developed, and well-founded response!!! But above all, for the speed with which you have done it.