do I need a new computer to get faster optimizations?

• 03/30/2020 at 11:12 AM #123912
  nonetheless

  Since upgrading to Premium, trying to optimize thousands of permutations @ 200k has become painfully slow.

  My desktop is about 8 years old (i7-3770, 3.4ghz, 8gb RAM). Would a faster processor help? or is there a bottleneck at the PRT server?

  

  nonetheless

  Participant

  Topics: 83

  Replies: 1237

  Been thanked: 978 times
  03/30/2020 at 11:19 AM #123913
  fifi743

  Hi nonetheless
  Reinstalling Windows is the best solution.
  I do it every 3 to 4 years.

  

  fifi743

  Participant

  Topics: 10

  Replies: 916

  Been thanked: 427 times
  03/30/2020 at 11:26 AM #123915
  nonetheless

  Yes, that’s good advice. I used to do it regularly but haven’t done so on this machine. So you think the problem is more with the computer than with the PRT server?

  

  nonetheless

  Participant

  Topics: 83

  Replies: 1237

  Been thanked: 978 times
  03/30/2020 at 11:32 AM #123917
  GraHal

  Maybe we should do a side by side trial (You and A. N Other etc) as to how long it takes to optimise x combinations over x bars??

  I suspect it is what it is and the PRT are throttling response so we all fit in / get some bites of the cherry?

  

  GraHal

  Participant

  Topics: 214

  Replies: 8833

  Been thanked: 2202 times
  03/30/2020 at 11:43 AM #123924
  nonetheless

  That would be interesting to try.

  I used to just run the variables separately which is obviously much faster. Then I started noticing that you get vastly different results by running them all together, even if it takes forever.

  Twice in a row I left it to run overnight and both times PRT closed down on me before I ever saw the results. 🤬🤬🤬

  

  nonetheless

  Participant

  Topics: 83

  Replies: 1237

  Been thanked: 978 times
  03/30/2020 at 11:43 AM #123925
  Francesco

  I also have a 7 years old PC (AMD FX-8320, 8GB RAM) but i’m not so slowed in 200k and optimizations. Maybe implementing an SSD in your configuration could help.

  

  Francesco

  Participant

  Topics: 32

  Replies: 439

  Been thanked: 200 times
  03/30/2020 at 11:49 AM #123927
  nonetheless

  Francesco wrote:

  i’m not so slowed in 200k and optimizations

  So, if you were running 5000 permutations how long would you expect it to take? For me about 5 hours.

  

  nonetheless

  Participant

  Topics: 83

  Replies: 1237

  Been thanked: 978 times
  03/30/2020 at 11:49 AM #123928
  GraHal

  nonetheless wrote:

  both times PRT closed down on me

  Next time turn off and on before you set it going overnight?

  PRT is supposed to only switch off after 18 hours.

  Francesco wrote:

  but i’m not so slowed in 200k and optimizations

  There we are … ideal partner to do a trial with on the same optimisation … goes without saying, but though I’d say it anyway! 🙂

  

  GraHal

  Participant

  Topics: 214

  Replies: 8833

  Been thanked: 2202 times
  03/30/2020 at 11:51 AM #123929
  Francesco

  nonetheless wrote:

  Francesco wrote:

  i’m not so slowed in 200k and optimizations

  So, if you were running 5000 permutations how long would you expect it to take? For me about 5 hours.

  Well it depends by the timeframe for sure, and i think probably also by the complexity of the code.

  

  Francesco

  Participant

  Topics: 32

  Replies: 439

  Been thanked: 200 times
  03/30/2020 at 11:53 AM #123931
  GraHal

  nonetheless wrote:

  how long would you expect it to take? For me about 5 hours.

  I know you are only getting a feel for how long, but it would be so useful if we could do a side by side trial … same System optimising on two PC’s etc.

  Would be easy to set up … load up and then both say on here … START Now!? 🙂

  I’m more than willing to participate, but I’ve only got 100k bars.

   

  

  GraHal

  Participant

  Topics: 214

  Replies: 8833

  Been thanked: 2202 times
  03/30/2020 at 11:54 AM #123932
  nonetheless

  GraHal wrote:

  Next time turn off and on before you set it going overnight

  Thanks GraHal – I should have thought of that. 18 hours should be enough to get me through, didn’t realise it was a fixed length of time.

  

  nonetheless

  Participant

  Topics: 83

  Replies: 1237

  Been thanked: 978 times
  03/30/2020 at 12:00 PM #123936
  nonetheless

  Francesco wrote:

  depends by the timeframe for sure

  And I think MTF takes longer, no? as it has to run through each TF separately.

  GraHal wrote:

  same System optimising on two PC’s

  It is a cool idea, but at the mo I’ve got a long one running that I can’t possibly interrupt.

  What I’m also trying that I didn’t even realise was poss, is run 2 simultaneously. They don’t seem to conflict with each other, or not much.

  

  nonetheless

  Participant

  Topics: 83

  Replies: 1237

  Been thanked: 978 times
  03/30/2020 at 12:03 PM #123939
  Francesco

  Yes of course! For example, try to backtest a vectorial code with over 1k lines and with multiple TF. It takes long long long time…

  

  Francesco

  Participant

  Topics: 32

  Replies: 439

  Been thanked: 200 times
  03/30/2020 at 12:59 PM #123948
  Nicolas

  Since backtests/optimizations are made server side, your computer performance is not involved into any slowdown. In fact, the time to get the results of an optimization depends on many factors:

  1. code complexity
  2. iterations quantity
  3. Walk Forward active or not
  4. free available slot on servers

   

  4 users thanked author for this post.

  ArnoldB, nonetheless, Francesco, GraHal

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20863

  Been thanked: 4774 times
  03/30/2020 at 1:22 PM #123951
  Francesco

  So can also Internet speed be a crucial factor?

  

  Francesco

  Participant

  Topics: 32

  Replies: 439

  Been thanked: 200 times
• Author
  Posts