OmniTrader Archives
OmniTrader 2010 Technical Support
Windows 7 and i7 processor
28 replies, 3180 viewings
|
OmniTrader Archives
OmniTrader 2010 Technical Support
Windows 7 and i7 processor
|
Last Activity 5/17/2022 8:17 AM 28 replies, 3180 viewings |
|
|
| Printer friendly version |
Maciej Veteran   Posts: 297 Joined: 12/19/2003 Location: Paris  User Profile |
I've taken some screenshots of my Apple Macbook Pro running OT2018 in a virtual Win 10 machine. You can see clearly the single thread (process) of OT and confirms that it uses only 1 of the available cores. It will never go above this single core. Other processes (in this case a data downloader among others will use the remaining resources but in this case it looks as if a single additional core will be by and large sufficient so I'd suggest a very fast gHZ (speed) dual core processor will be optimal in terms of price/performance. Furthermore the memory usage is negligible as is the SSD i/o though the latter should be as fast as you can afford. Bottom line is the higher the speed (gHZ) the better. The latest I7 processors can close off some cores and boost the core that needs the speed. I believe its called Turbo mode. This favouring of one core at the expense of the unused cores seems about the only way to speed up a single core process such as OT. I haven't looked but I suspect that the best AMD processors can do the same thing. No point in looking at the ARM processors just yet. Happy to answer any questions.  Attached file : OTPerf20180702.pdf (835KB - 457 downloads) | ||
| Maciej Veteran Posts: 297 Joined: 12/19/2003 Location: Paris User Profile |
Another PDF with this time my main trading PC. It has better SSD (+Optane) and is not a VM but is a Kaby Lake processor. You'll see that its speed is practically twice as fast as my laptop's core speed so unsurprisingly it handles OT much more quickly. I've also added an example of a multi-threading application called FAH (Folding @ Home) which soaks up any spare CPU cycles. This runs simultaneously with OT and does not steal anything from OT. OT cannot run faster than the values shown. If you were to agree with my analysis I'd appreciate you extending this appreciation towards Ed to update the core OT application. In this and age we punters deserve at the very least a 64bit application not withstanding the evident merits of OT. In the absence of OT core improvements I'd recommend that you run in "hot Strategy Finder" mode to get the top 5 systems for your stocks (this will take a long time for sure) and then work on the top 5 or so to improve matters. Luckily I generally manage to work through my 3 core profiles on one machine each day and am ready for the next day's open with a couple of hours to spare. Of course I really do worry that soon I'll have to make serious compromises in order to be ready for the open (luckily I'm an early riser). One obvious way to ensure fast enough processing is to stop adding Add-Ons as they will most likely penalise me. Best Regards Attached file : OTPerformanceI7PC20180703.pdf (2251KB - 386 downloads) | ||
| Jim Dean Sage   Posts: 3022 Joined: 9/21/2006 Location: L'ville, GA User Profile |
Thanks for the info. 99% of the user’s coming to this thread probably don’t need quite that much nuancing on their machines - typically they want to know broad-brush stuff. Ed’s and the N staff are well-aware of the 64bit and multithread and larger-RAM-support desires that many users have expressed for well over ten years now. The internal costs required to do that kind of massive overhaul to the complete bundle of software routines that make up OT (some pieces possibly still using VB6) would be so high that the only way they will be able to afford it is to wait till after they see major financial success via one of their macro ventures such as OmniFunds or iOmniVest, etc. So don’t hold your breath on that. They understand the trade offs. My usual recommendations re speeding up OT - esp when someone is planning extensive modeling runs via SW (ie back to back on a regular basis) are: 1) using multiple VM’s is cheaper overall than using multiple machines. 2) a machine that can run 2-3 VM’s at a moderate speed is morecost effective than one machine that runs one OT instance at blazing speed (since RAM is cheaper than clock speed) 3) with multiple VM’s OT *can* run with just one virtual core but two per OT instance is better (presuming hyperthreading). And 3GB RAM per single OT-only VM instance is adequate, while more than 4 is overkill (since OT maxes out at 2GB for its working live data). 4) when a fast processor is run continuously for hours or days then cooling is important since either the clock speed will auto-reduce as the chip gets hot, or occasional computational errors can creep in (this is especially important for laptops - therefore Xeon chips are better for laptops since they are purpose-designed to run full out, continuously). 5) for long term continuous running, an SSD is not crucial (pun intended) but for most users, an SSD will speed up load times enough to be desirable. OT does not normally do a bunch of multi-cluster/sector hard-drive read writes during its normal execution. 7) i7’s at any speed are definitely better and more cost effective re OT throughput speed than i5’s or lower. I don’t know if i9’s are sold much anymore but is used to be that their performance for the price was not as good as i7’s. 8) although it’s difficult to test, logically, a reasonable “boost” in L3 or L2 cache will help OT speed at a similar level as increasing the chip speed by another notch. 9) keeping the platform “clean” (regular profile rebuilds & occasional data rebuilds) lessens likelihood of OT-error shutdowns. Some users have found that regular OT data-file cleanup/maintenance is advisable as well, though I personally have not seen that need as much - I think it is more likely needed for people using the AI tools with SW). 10) Obviously, keep the OS skinny and free from extraneous stuff running in the background, though if you’ve got two full cores (4 hyperthead cores) and 4-6G RAM per instance, this is less essential - usually. RE VM’s: a couple of expert-VM users in particular have written *extensively* about using VM’s with OT (including how tos and nuances) in the TradeTight forum - click here and study the info in the first several Threads “stuck” to the top of the list. That’s about as many generalizations as I can provide, gleaned from personal research, work published here and in TradeTight by others, general knowledge about OT’s code structure, and general advice provided by N staff and other power users. Hope it helps! [Edited by Jim Dean on 7/3/2018 8:03 AM] | ||
| Maciej Veteran Posts: 297 Joined: 12/19/2003 Location: Paris User Profile |
My response as yours was to Jim Easter's question on buying an old Apple MacPro from 2010. Not for me to judge but he doesn't sound like someone with sufficient capacities or even needs to embark on a series of VMs. In simplistic format; fewer cores rather than many cores, higher clock speeds are better than slower clock speeds, memory not a lot and a SSD because OT uses small volumes. You don't seem to emphasize the CPU speed which is strange. Based on your explanations as well as mine, I suggest that we can surmise for Jin's benefit that the 2010 Apple wil not deliver as much as a similar 700-800 USD on a more modern processor (PowerSupply=150+I7=350+SSD=120+RAM=80+Case=50+Motherboard=100)=850. These prices are off the cuff so need checking and I'm assuming that a monitor/keyboard/mouse cost the same irrespective of processsor. As to Nirvana's knife-balancing act all I can say is that I'm glad that they are fully aware and lucky for all of us that OT does not use the IOS platform where Apple made it obligatory to move from 32 to 64. I sure that Microsoft won't do that will they. |
|
|
|
|
Administrator
Forum Moderator |
Registered User
Unregistered User |
 E-Mail this thread to a friend |
Toggle e-mail notification |
