Hardware for backtesting

[Paracelsus]

Hello!

I need advice about the optimal hardware setup (CPU, RAM, etc.) for backtesting on "tick-by-tick" e-mini data, and info about the approximate size of a 3-month "tick-by-tick" data file. Thanks!

[NinjaTrader_Bertrand]

Paracelsus, unfortunately I can't advise you on the file size, but tick data is quite resource intensive to work with so with the continuous drop in hardware prices I would go for a state of the art multicore machine offering 4-8 GB of quick RAM as solid backtesting machine.

[strategy1]

When you say quick ram you mean Frequency correct?

So Frequency speed is more important than Low Latency when in comes to backtesting?

For example 12GB DDR3 PC12800 9-9-9-24 will be better than 12GB DDR3 PC10666 7-7-7-21 ?

[NinjaTrader_Austin]

strategy, we can't make specific recommendations, but in general I would go with a very high-end PC for backtesting a bunch of tick data.

[strategy1]

Is frequency speed more important than latency.

I understand you cannot provide specific recommendations.

I also understand that a high end pc is necessary for backtesting.

My question was in reference to Bertrand stating in his response to another poster to get 4 - 8 Gig of quick ram.

My question was what does he mean by quick ram?
Does he mean high frequency or low latency?

There are some applications for which Low Latency Ram works better than High Frequency (but also higher latency) ram?

My question was which is better for backtesting?

All things being equal is it better to get 12GB DDR3 PC12800 9-9-9-24 or 12GB DDR3 PC10666 (7-7-7-21)?

Both of these kits are exactly same price at this time.
Here is the first one: http://www.newegg.com/Product/Produc...82E16820231358
Here is the second one: http://www.newegg.com/Product/Produc...82E16820231405

As you can see one's frequency is about 20% faster than the other. However it's latency is about 25% higher.

For someone building a new pc it is important to know which one is better for backtesting especially when both of them cost the same.

Thanks

[NinjaTrader_Bertrand]

strategy1, this question can't unfortunately be generalized easily - talk to the PC builders what RAM they would recommend for your specific motherboard / chip used and which offers also some OC potential then if you desire to go down that path.

[naxo79]

Slightly OT, but:
I just bought a new Core i5 2500K with HyperX RAM, overclocked it etc. and now when I run strategy optimizer in 64bit it freezes the whole NT application, also control panel. So eg. I cannot see how much time is left for the task to complete. Other applications are responsive. Any advice on this?

[NinjaTrader_Bertrand]

Does it only happen with your custom strategy or would it generally freeze for you even with for example the SampleMACrossOver used? Do you notice any errors in your log tab after you regain access to the application?

[naxo79]

No errors or anything and it completes the testing successfully, its just that NT is very unresponsive during the process. On my other PC it isnt. Can try also sampleMAcrossover.

[NinjaTrader_Bertrand]

Thanks, please give the Sample a try on your end as well - it would help if we could review your trace / log files from the machine of the issue. Could you please contact us via Help > Mail to Support?

[naxo79]

Had forgotten a print statement in the code which jammed the NT UI. My bad sry.

[NinjaTrader_Bertrand]

Glad to hear you got it resolved.

[naxo79]

Would NT run on this http://www.nvidia.com/object/persona...computing.html if I got it?

[NinjaTrader_Bertrand]

I would believe it should run on it as I see supported OS would be Windows based as well from NVIDIA's website. However we would not make specific use of any custom GPU code needed to fully utilize the power of those workstations.