Best 8, 10 OR 12 Core Xeons for Video Editing?

L

lsudvm

n00b
Joined
Oct 15, 2010
Messages
60
Looking for advice on 8 10 or 12 core xeons. There are so many choices and am wondering what would be the best bang for the buck - v1, v2, v3 - on used Xeons for a new Video Editing system. Right now on 2 5680s Westmeres that were used to upgrade my 5550s.

anyways whatever advice is appreciated.
 
is there a budget? That obviously influences things.

Also, what software? what resolution? what is your storage like?
 
mnewxcv said:
is there a budget? That obviously influences things.

Also, what software? what resolution? what is your storage like?
Click to expand...

Budget - no exact dollar amount but i'd like to be as thrifty as i can by going used with the CPUs and anything else i can. Already have the case a Cosmos II. Software - Premiere, Handbrake. Storage not an issue as i have dozens of terabytes.
 
lsudvm said:
Budget - no exact dollar amount but i'd like to be as thrifty as i can by going used with the CPUs and anything else i can. Already have the case a Cosmos II. Software - Premiere, Handbrake. Storage not an issue as i have dozens of terabytes.
Click to expand...

as for storage, I meant more in line with will it be sata drives, sas, ssds, RAID? Storage can become a major bottleneck for renders/exports if it cannot keep up with the fast cpu/ram. In any event, you can expect the latest 12 core xeons to be faster than older 8 core xeons, naturally. The 5680s are not bad chips, but the newer xeons will greatly improve on single core performance and therefore overall performance. A single 8 core v1/2/3 may be quicker than your dual hexacores.
 
mnewxcv said:
as for storage, I meant more in line with will it be sata drives, sas, ssds, RAID? Storage can become a major bottleneck for renders/exports if it cannot keep up with the fast cpu/ram. In any event, you can expect the latest 12 core xeons to be faster than older 8 core xeons, naturally. The 5680s are not bad chips, but the newer xeons will greatly improve on single core performance and therefore overall performance. A single 8 core v1/2/3 may be quicker than your dual hexacores.
Click to expand...

OK Got it on storage - I'd like to do NVME or M.2 for the boot drive/os drive. Media will bve exporte to and save to My HGST 4 and 6TBs. I work with 720 video mainly as i edit and upload news clips - this all really a hobby that i make a little money from. I see the 8 core sandys are pretty cheap - so a pain of those would be faster than my westmeres - any idea by how much? Any idea if the C602 boards can do NVME?
 
you could look into a pair of 2690 sandy chips. A pair of those should provide nice gains over the westmeres. I am not sure how the ivy bridge or newer xeons stack up, though it may be nice to find a board that supports the ivys as well. Hopefully someone else can chime in for your other questions.
 
dual used e5-2670s are by far the best value at 60 each however if your program can use all the treads of many cpus concider haveing 4 e7-4850's giving you 40 cores and 80 threads
 
I just watched this the other night, he compares the GPU''s and CPU for exporting from Premiere,
 
I don't have premiere on my new Intel setup but I did do a handbrake test and it used all 32 threads but it looked like only 50% usage across all of the threads.
My old 3770K and my 4790K max out handbrake and premiere on all 8 threads.

3770k-encoding-files.jpg
Premiere CS6 4790K rendering test.jpg
 
cdabc123 said:
what what about quad xeon setups? would it drop down to 25% then?
Click to expand...
That likely depends on the number of threads not the number of cpus

Edit: another reason may be it cannot write to the disk fast enough to keep up with the cpu therefore it isnt running at 100%
 
mnewxcv said:
That likely depends on the number of threads not the number of cpus

Edit: another reason may be it cannot write to the disk fast enough to keep up with the cpu therefore it isnt running at 100%
Click to expand...

If you look at the Task Manager, Disk 4 "J" is the source and Disk 1 "K" is the destination. It shows 5% on the Source and 14% on the Destination, if it were an I/O bottleneck, wouldn't the drives show closer to 100%?
 
Zepher said:
If you look at the Task Manager, Disk 4 "J" is the source and Disk 1 "K" is the destination. It shows 5% on the Source and 14% on the Destination, if it were an I/O bottleneck, wouldn't the drives show closer to 100%?
Click to expand...
I missed that! Perhaps handbrake has a 16 thread limit and spread it across 32 hence 50 percent.
 
For a 4P setup first of all you need to run a MS Server os or Linux. Now Handbrake is available for Linux, fortunately. But if you use 720p you could run the final transcoding job on the Intel IGP. Quick Sync is a hardware encoder and is terrific in terms of speed. I have seen one test on anandtech doing transcoding at 220fps for 720p on a 3770k. I also have seen Final Cut X at work and it seemed one order of magnitude above premiere in terms of speed and efficiency. So maybe you should consider some other alternatives not just upping absolute cpu firepower. And I think Handbrake does not scale linearly after 16 threads. Why do I say that? Well i have seen 100% cpu time on a 2p Opteron rig with a pair of 2419s but not the same thing on a SR-2 with 2xL5639s.
 
I was encoding some old DVD files last night and CPU1 got pretty hot, the list shows 2 complete items but I was stopping after each one to let the cpu cool down.
I ended up making handbrake use only CPU2 and that cpu only got to the 70's during the 2 hours it was processing.
Also, using only CPU2, cpu load was up to 90%, so it was doing the task in the same amount of time vs both CPU's enabled.

intel-dual-xeon-handbrake-hot-core-temp.jpg
intel-dual-xeon-handbrake-hot.jpg
 
How much faster would it go to encode it using the Quick Sync on that 4790?
 
You must log in or register to reply here.
Back
Top