Performance Xeon workstation and SQX
12 replies
Johny55
4 years ago #240863
Hello,
can anyone write me please a performance PC in a standardized test (SQX Main menu / Settings -top right / benchmark)?
Especially for a single-processor workstation. Best for Xeon E5-26xx series (CPU, version CPU or GHz, RAM,DDR3 or DDR4, number of strategies).
Thanks
Johny55
4 years ago #240936
The price of 2x Xeon X5650 is very interesting. If you have someone, please write about the performance in SQX.
Thanks
mabi
4 years ago #240938
Looking at budget Single Xeons or older dual systems like X5650 I would instead go for a Threadripper with higher GHZ. The 1950X is cheap now and outperforms any older models of XEONs when making strategies in SQX. I find it is easy to compare performance in SQx between CPu’s. Just multiply total core count with Ghz and you have it.
hankeys
4 years ago #240947
2x XEON E5-2680 – 80000 strs in benchmark
You want to be a profitable algotrader? We started using StrateQuant software in early 2014. For now we have a very big knowhow for building EAs for every possible types of markets. We share this knowhow, apps, tools and also all final strategies with real traders. If you want to join us, fill in the FORM.
mabi
4 years ago #240952
Dual Systems with more then 64 Threads are splitted in groups by Windows and SQx can only use one Group. So 2 SQX has to be used to utilize both CPU,s or You have to disable Hyper threading. So i run 2 test in my dual system
1x XEON E5-2673 V4 – 82000 strs in Benchmark
1x XEON E5-2673 V4 -77000 strs in Benchmark
1x XEON E5-2698 V4- 73000 ”
1xXEON E5-2698 V4- 66000 ”
Notice that one CPU always performs better then the other in dual socket systems.
Johny55
4 years ago #240955
mabi
4 years ago #241039
Threadripper 2950x – 84000
mabi
4 years ago #241109
Threadripper 2950x – 84000
Threadripper “gaming Mode” – 93000 ( simultaneous multi threading turned of gives higher usage and output on all processes as WF, Montercarlo and Generation)
sbecm
4 years ago #241155
hankeys
4 years ago #241273
these numbers will tell us nothing, we dont know what are you generating…this is not comparable if everyone is not using the same settings
also benchmark from SQX could get results from 60-80 k, so it could differ, because its running only for few seconds
for this purposes we run the same project for 1 hour in one of the cooperation groups and get these results
this is comparable
You want to be a profitable algotrader? We started using StrateQuant software in early 2014. For now we have a very big knowhow for building EAs for every possible types of markets. We share this knowhow, apps, tools and also all final strategies with real traders. If you want to join us, fill in the FORM.
ivan
4 years ago #242468
i think the idea is interesting but it should be posted on general questions, it doesn’t have much in common with presales questions
i agree with those who said that a test must be identical in order to compare 2 different proceesors
in any case, from the tests i performed on SQ3 on mobile processors, i would say to any person interested in such comparisons that any dual CPU configuration with Xeon processors from Haswell up or a single CPU configuration with Intel 9900K or AMD Threadripper would be sufficient. This means at least 16 cores on any configuration, single or dual.
Anything below would take a lot more time. This is something very general that any new “miner” should be aware of.
Besides this, any other tests or detailed comparisons would be more like decorative than functional. I wouldn’t waste too much time into such comparisons.
Timisoara, Romania
3900X 3.8 Ghz 12 cores, 64GB RAM DDR4 3000Mhz, Samsung 970 EVO Plus M.2 NVMe
Gianfranco
4 years ago #242470
you have to see how the SQX program is written if it takes advantage of all the available cores… I seem to have read can distribute in parallel calculation up to 24 cores … So I think for now you can not do more or a little more if they do not better distribute the parallel calculation example on 128 or more cores…. but when you have millions of billions of combinations to process 3.5 ghz 0 4 ghz
little changes. maybe instead of 40 hours it takes 35 hours then the difference and a few years ago I was racing on a forum with big overclock and liquid cooling and I saw that the super top cpu is never convex in performance price ratio
if there was a possibility to use TESLA nvidia cards in the pc that I use for SQX would be a nice leap forward but the program should support the CUDA libraries of nvidia and here we talk about 2000/3000 cuda core … then you start to think
Thank you
Gianfranco
ivan
4 years ago #242858
the essence here is that mostly newcomers are interested in this subject and they are (or should be) focused on systems with maximum 24 cores. Main reason is that systems either new or used with more than 24 cores are too expensive for most newcomers
and the main question of these people is how a reasonable system looks like and in what time interval can they expect to create a decent, working final strategy on a particular pair.
i know this because last year i was also a newcomer and i put the same questions and i was looking for those answers on SQ3.
An aproximate answer but sufficient for most would look like this: with the system X (…enter configuration) you can expect a decent final stategy on Y settings (enter parity and timeframe…etc) in Z hours, days or weeks
Timisoara, Romania
3900X 3.8 Ghz 12 cores, 64GB RAM DDR4 3000Mhz, Samsung 970 EVO Plus M.2 NVMe
Viewing 12 replies - 1 through 12 (of 12 total)