![]() |
|
|
|
Get off my lawn!
|
Help building a monster computer
First off, no Macintosh will work. We run a mapping program that is only Windows based. Nothing wrong with a Mac for a average home computer but I need a monster that I can put together myself.
Right now my computer runs Win 7 Professional and has a single i7-4770K CPU 3.5 GHz 8 Megs cache and 32 gig of RAM with two NVidia GTX 980 graphic cards (4 gig each) and a 1 TB SSD for the boot and a 500 Gig SSD as a data drive and some spinning drives for long term storage. It is taking a week at a time to do just parts of a project. I am looking for a good source for the parts for a killer monster computer. Think Tim Taylor! Arf Arf Arf. It looks like 2 physical CPUs are the limit with Windows 7, 8 or 10 unless I go to Windows Server software. Win 8 & 10 can handle 512 Gig of RAM and I really think our big bottleneck is RAM. 32 Gig is just not enough as it is pegged out on RAM usage almost all the time. Where can I even find a good motherboard with 2 physical CPU slots and maybe 8 or 12 cores each. Since Win 8 & 10 can handle 512 Gig how do we even find a 256 gig motherboard and the memory to get it there? Maybe a good intermediate or at least cheaper step would be a motherboard that will take one more i7-4770K CPU so I only need one more and just more RAM like my current DDR3 memory. 64 Gig or at least 128 as a start. I would love to find a vendor that will sell me a motherboard, 2 CPUs and the RAM to get to 128 Gig. My current case can handle the motherboard and the power supply is up to it. I prefer to build my own computer. Any suggestions would be great.
__________________
Glen 49 Year member of the Porsche Club of America 1985 911 Carrera; 2017 Macan 1986 El Camino with Fuel Injected 350 Crate Engine My Motto: I will never be too old to have a happy childhood! |
||
![]() |
|
Too big to fail
|
We have one of these in the data center; ours is 48-way with 1Tb of RAM Sadly, we don't actually use it for anything these days. It used to be used as a 'ringer' system to produce benchmark numbers.
HP ProLiant DL980 G7 Server
__________________
"You go to the track with the Porsche you have, not the Porsche you wish you had." '03 E46 M3 '57 356A Various VWs |
||
![]() |
|
Registered
|
Check a couple boxes on the order form for a Mac Pro and it will do 7 teraflops. I wouldn't rule it out.
How to install Windows 7, 8, or 10 on a Mac using Bootcamp | Digital Trends
__________________
techweenie | techweenie.com Marketing Consultant (expensive!) 1969 coupe hot rod 2016 Tesla Model S dd/parts fetcher |
||
![]() |
|
Registered
Join Date: Dec 1969
Location: chula vista ca usa
Posts: 5,694
|
Well this site will get you started: AMD Dual CPU Motherboard: Research Calibex - Seller Reviews, Ratings
Many of my "gamer" students use dual CPU mother boards and run dual video cards but their biggest concern is cooling of everything. Good luck and make sure you have a large enough power source. Here is an interesting review: Gigabyte GA-7PESH1 Review: A Dual Processor Motherboard through a Scientist |
||
![]() |
|
Banned
Join Date: Jan 2005
Location: cutler bay
Posts: 15,141
|
you need a server board
but I would just sell your current rig and start over fresh as the server type are a different socket/chip from your i7 haswell 1150 socket 4770 chip you need a xenon type dual socket old 2011 or the newer 2011v3 based mother board +chips with 8 memory slots and 4 video card slots I fact it may be cheaper just to buy a built server then build one as they can buy in bulk and you can't plus you get a warranty and support that don't come with a self build |
||
![]() |
|
The Unsettler
|
You might be better off just building a small cluster / grid / rendering farm.
Probably cheaper and it would be expandable by adding more boxes as needs change. What's the app you're running?
__________________
"I want my two dollars" "Goodbye and thanks for the fish" "Proud Member and Supporter of the YWL" "Brandon Won" |
||
![]() |
|
![]() |
Get off my lawn!
|
Pix4D Mapper
__________________
Glen 49 Year member of the Porsche Club of America 1985 911 Carrera; 2017 Macan 1986 El Camino with Fuel Injected 350 Crate Engine My Motto: I will never be too old to have a happy childhood! |
||
![]() |
|
Registered
|
You may want to ask the manufacturer if the software will take advantage of multiple cores and if so ask if there is a limit on the number of cores supported. Prior to version 10.1 ESRI ArcGIS did not support multiple cores.
|
||
![]() |
|
abides.
|
I was just about to suggest this. It really pisses me off when my computer has one core maxed out, and the other eleven are idle.
__________________
Graham 1984 Carrera Targa |
||
![]() |
|
Registered
|
Windows 7 has a lot of overhead. One of the more CPU intensive processes that is really not necessary is the RAC task(https://technet.microsoft.com/en-us/magazine/gg452701.aspx). RAC runs on startup and then every hour and will take a whole core by itself. To disable the RAC task go to task scheduler->Microsoft->windows->RAC and disable the RacTask task. This is one of several things we do to tune our virtual desktop pools so we can run with single cores with out impacting system performance.
|
||
![]() |
|
Get off my lawn!
|
All 4 cores and all 8 threads are often running at 100% and the RAM is at 100% for most of the process.
From my reading on the software the best bang for the money is just more RAM. I would really love to find a motherboard that would take my current CPU and let me run 64 or 128 gig of ram. I will check out disabling the RAC task.
__________________
Glen 49 Year member of the Porsche Club of America 1985 911 Carrera; 2017 Macan 1986 El Camino with Fuel Injected 350 Crate Engine My Motto: I will never be too old to have a happy childhood! |
||
![]() |
|
Slackerous Maximus
Join Date: Apr 2005
Location: Columbus, OH
Posts: 18,150
|
Have you costed out renting time in the cloud? AWS will let you spin up a Windows 2008 Server box with 32 processors and 244 gb of memory (Perhaps more, just my cursory look). Dump the data to an S3 bucket, fire up the server and do the work, then shut the server down. You only pay for the time you use.
__________________
2022 Royal Enfield Interceptor. 2012 Harley Davidson Road King 2014 Triumph Bonneville T100. 2014 Cayman S, PDK. Mercedes E350 family truckster. |
||
![]() |
|
Get off my lawn!
|
We are processing many gigabytes of data. We would have to get a really fast internet connection to make that viable. Local processing is the only logical way to do it right now.
__________________
Glen 49 Year member of the Porsche Club of America 1985 911 Carrera; 2017 Macan 1986 El Camino with Fuel Injected 350 Crate Engine My Motto: I will never be too old to have a happy childhood! |
||
![]() |
|
Slackerous Maximus
Join Date: Apr 2005
Location: Columbus, OH
Posts: 18,150
|
Send them the disk. They do it onsite.
AWS Import/Export - Cloud Data Transfer & Migration Services I don't work for Amazon, just offering up a suggestion. I know a guy that owns an streaming service, and every week he has to post 2000+ hour long podcasts, so he just overnights them a disk.
__________________
2022 Royal Enfield Interceptor. 2012 Harley Davidson Road King 2014 Triumph Bonneville T100. 2014 Cayman S, PDK. Mercedes E350 family truckster. Last edited by HardDrive; 08-19-2015 at 09:12 AM.. |
||
![]() |
|
Super Moderator
|
Agree with the above. AWS or Google Cloud will scale out on demand. You might find that it ends up vastly cheaper if the horsepower is only required occasionally.
We did an experiment where an process that ran on 62 cores for about 8 hours submitted to Google cloud finished in 20 minutes at the cost of $112 for the run. An enterprise class machine with 62 RISC cores and Terabytes of data could cost upwards of $2-3M. That's a lot of submissions!
__________________
Chris ---------------------------------------------- 1996 993 RS Replica 2023 KTM 890 Adventure R 1971 Norton 750 Commando Alcon Brake Kits |
||
![]() |
|
Registered
Join Date: Oct 2013
Posts: 1,621
|
I wish you could run adobe premiere pro on a cloud machine. I wait hours and hours for exports and transcodes damnit
|
||
![]() |
|