Hi everyone,
Could you help me choose a build and make some minor suggestions? —Please?
I am running a large set of excel/tableau spreadsheets that require extensively repetitive calculations. This is not entry level used excel work so I need powerful desktop to save time on calculations.
This would not be a gaming machine, but CPU L3, GT/s, and RAM are a priority. A simple graphics card that can support multiples displays would be nice, but I need a card with functionality, not over the top rendering.. So... I am considering three trek syndicate builds in order of preference:
1) A duel socket Intel Xeon E5-2630 v2 build, such as:
https://teksyndicate.com/videos/build-workstation-pc-our-dual-xeon-build-and-guidelines-help-yours
2) A single socket Intel Xeon E5-2630 v3 build.
3) A i7-5820k build, such as:
https://teksyndicate.com/videos/pistols-x99-build-2015-msi-corsair-fractal-be-quiet-samsung-hgst-thermalright-intel (The MSI mother board recommend has a startling number of bad views!)
Any suggestions? (ty in advance!)
Best,
FBW
Just wondering if Excel multi-threads properly and hence can use lots of cores of if it would be better off with fewer but faster cores...
https://www.youtube.com/watch?v=Q37kluCzYSE (Watch on 2x speed)
> Conclusions > Excel’s multi-threaded calculation can be very successful at reducing calculation times. > The effect of hyper-threading is not as large as that from multiple physical cores, but its still worth having. > The effectiveness of multi-threading is very dependent on the workbook. There will be workbooks where the overhead of analyzing the calculation chains outweighs the gain in calculation speed. > Excel’s default setting to assign a calculation thread to all available logical cores seems sensible.
Types of formulas also affect threading:
https://msdn.microsoft.com/en-us/library/office/bb687899.aspx
(admins, sorry for oversized don't above—feel free to edit!!!)
Could not read, font too small. ;)
Question: Before considering a dual socket workstation: Is there a possibility for you to benchmark your most common calculations using multi-threading, hyperthreading as well as multiple physical cores? It could be a game of diminishing returns, do we know that performance scales well with increased core-count and/or CPU count?
Second question: Is Excel still the optimal tool for the job you are doing? Have you tried outsourcing some of your calculations to a suited script language?
We have actually developed an in-depth solution in Python, but due to legal reasons we are not able to use this program.
Excel is a crutch for us to limp to the finish line.
I'm trying to read up on multi-threading vs hyper threading vs physical cores... It sounds like core number and (and L3 size) outweighs the value of threading.
My condolences for the legal trouble. Losing custody of software is always hard.
You could throw together a test-spreadsheet with dummy data (best with integrated timer) and upload it, so forum members could test it out on their rigs. There might be some benchmark hungry people with high-end rigs around. (I only got a FX-8320E, probably not much help since the core architecture is ancient).
I'd be happy to test on dual core i7 laptop and quadcore i7 desktop (both with 16gb ram and SSD) if you can put something together that's obviously not an issue if someone outside the office sees it.
Is the spreadsheet large in memory terms? Just thinking about how much of it would be paged out to disk. I seem to see Windows paging a lot of stuff out to disk even though there's tons of memory unused. It's helped by the SSD over HDD but better it doesn't page in the first place
Thinking about it I don't actually have Excel/Windows on my desktop :D
Have you thought about renting a server?
Based on your needs, you probably need this only occasionally. When i had to crunch a 60gb sql database to get some patterns out so that tableau would even display the graph i rented a Linode server for 2 days, used a ramdisk and crunched away.
They have an hourly billing option so you pay what you use, and its way cheaper than building your own machine.
20 cores and 96gb of RAM can do wonders for speed, you just VNC into the machine, (or RDP if on windows).
Or rent a dedicated server for a month designed for big data: https://www.ovh.co.uk/dedicated_servers/big-data/mini-HG.xml#options
--EDIT--
Hm, Linode does not support windows, but OVH has a dedicated machine and they have a free windows option as part of the setup.
1 Like
Does it have to be Intel?
Gosh, thank you both for such a kind offer, but I must decline as I need to place an order today. =/
I'm going to go with this CPU:
1) Intel Xeon E5-2630 v3 Haswell-EP 2.4GHz 8 x 256KB L2 Cache 20MB L3 Cache LGA 2011-3 85W BX80644E52630V3
Any recommendations of mother boards?
http://www.cpu-upgrade.com/CPUs/Intel/Xeon/E5-2630_v3_motherboards.html
What board did they use for the (dual?) Xeon build they did for the guy on the video a couple of months ago.
Be interested to hear what difference it makes to processing times once it's up and running.
all of them are way to overkill for what you need, you better check pistol´s new rig!