How much CPU is too much CPU?

[#1]

Welcome to the Community.

The speed and the number of processors are completely intertwined the amount of data you are processing and how complex of an analysis you are performing to be able to give a blanket answer to the question. 

I personally always buy as fast a processor that my budget will allow.

If you supply some examples of the size of data tables and the types of analyses you will be doing, someone in the Community may be able to give you some good estimates. 

[#2]

To pile onto @txnelson 's reply, when I worked in industry pretty much not a week went by, as a statistician, where somebody didn't ask me the "How many samples do I need?" question. So we'd walk through the standard sample size determination questions first...and almost without exception the answer was "Too many for our budget." So we defaulted too "How big is your budget?" answer/question.

[#3]

I am ashamed to admit, I am not a statistician, nor am I using JMP for complex analysis. What little analysis I do is simple, and the files are small, and my laptop can handle it. The reason for my question is for the "mindless" plotting of large amount of sensor data.

My equipment collects data from 443 sensors every 100ms, 24/7. When I attempt to look at more than couple of days of data Excel taps out, therefore I am forced to use JMP. Currently, I am trying to plot and visualize 45 days of data. My files is just under 3GB with about 72million rows. Every time I want to do any manipulation of the data or even save the file, it takes over a 60s. 

To speed up this process, I want to build a workstation, and I was wondering if I spent $17,000 for an Intel® Xeon® Platinum 8490H Processor, would I be utilizing the full capabilities of the processor, or is that processor just overkill? Could I spend $6,000 on an Intel® Xeon® Platinum 8568Y+ Processor and be just fine?

I hope my questions make sense, and I look forward to everyone's comments and suggestions.

[#4]

Hi @RedMonster796 ,

  Do you have to review every single data point? Is each 0.1s really that important? Could you get by with cutting it back to every 10s or something? That certainly is a lot of data to load into memory, but I honestly am not sure that spending $17k on a processor is going to be worth it.

  Definitely go for a fast processor and one that has multiple cores, but the processor alone isn't the whole story. You should also consider how much RAM your machine will/should need. I have 16GB in my work notebook and haven't had any problems dealing with files that are a million rows tall. However, considering that your data sets are 72mil and 3GB in size, you might consider upping your RAM and getting at least 32GB of RAM -- in which case you'll want to make sure that both your motherboard and processor can access all 32GB of RAM, otherwise you're losing out on resources for your computer.

  Another thing to consider and have in mind is your GPU. Considering that you're wanting to visualize up to maybe 72mil rows, you should probably get a better GPU than a standard default that the system would come with. This way, you should be able to optimize the processing (math for statistics) from the graphical display of the data.

  Again, you'll want to make sure that the motherboard can work with and handle a higher end GPU, as you'd need to verify that about the CPU and RAM as well.

  I'd still come back to the task of evaluating whether every 0.1s, 24/7 is really necessary for your data analysis. Or, if you can identify any kind of trigger points in your data that you can then run a JSL script that creates a smaller data table that saves -- I don't know -- 1000 data points prior and 1000 data points after said trigger event. It seems to me that the majority of the time, you probably aren't needing to evaluate things every 0.1s and can probably get away with a larger time step. But, that's just a guess as I don't know what kind of system you're dealing with and how crucial each data point might be.

Hope this helps!,

DS

[#5]

I agree with @SDF1 and I will add that I would also upgrade to an M.2 Solid State Drive.  If the millions of rows of data are being read in to JMP from your local system, having very fast local data storage is a good thing.

I will also add that having multiple monitors will also make your analytical experiences far more satisfying.

[#6]

Let's just say I have to look at every data point every 100ms...

System I'd like to build has the following HW:

2 x Intel® Xeon® Platinum 8490H Processor (1.90 GHz up to 3.50 GHz) $34,600

2 x 128 GB DDR5-4800MHz (RDIMM, ECC) $5,298

3 x 4 TB SSD M.2 2280 PCIe Gen4 Performance TLC Opal $1,587

Even if I have to look at every single data point every 100ms, is the system described above similar to an 18 wheeler hauling a toy Tonka truck? I don't know if you've ever seen an 18 wheeler with an empty open trailer with a toy Tonka truck strapped down in the middle of the trailer. It's a cute sight, but one would hardly buy an 18 wheeler just to haul a toy Tonka truck, and that is the scenario I am trying to avoid.

I hope that makes sense. 

[#7]

Like most things in life, the answer is it depends.

Some platforms in JMP only use a single thread so having many threads wouldn't add much value, but having that single thread running at a faster frequency would add value.  Some platforms in JMP do take advantage of multiple threads so having more threads/cores does help.

With the Torch add-in in JMP Pro 18, now for the first time a GPU will be utilized for image classification and deep learning which does those tasks orders of magnitude faster than a CPU. 

[#8]

From everyone's responses so far, I get the feeling, spending $35K for two 8490H (60 cores/120 threads per CPU), is an overkill, yes? My laptop was upgraded to 48GB of RAM, but maybe my laptop processor is getting too old. It's 3 years old. This is why I want to build a workstation dedicated just for JMP. I just don't want to submit a request for a $40K workstation if is too much of an overkill. It would be nice to have, but I don't want to waste money if the expensive CPU will not even be utilized.