Ram Speed on render node

Become a member of the CGSociety

Connect, Share, and Learn with our Large Growing CG Art Community. It's Free!

Thread Tools Display Modes
  06 June 2013
Ram Speed on render node


not many clues about this one... I'm trying to build a render farm, but as usual, not much resources around.

I've heard it makes no much difference, but not about numbers and such.

16gb 1333mhz vs 2333mhz. You think it's possible to measure what's the difference between them at render time?

  06 June 2013
1-4% difference depending on the scene
  06 June 2013
Thank you, man! I really appreciate it

Do you have more info on what you just stated?
  06 June 2013
there's been a lot of websites that have benchmarked different memory speeds with different programs. Typically memory speed or latency are not a large bottleneck for the majority of things. It helps, but not much.

It's typically the last thing someone would prioritize since the return on investment is minimal.
  06 June 2013
Thank you! that makes a lot of sense
  06 June 2013
The cas timing -might- make some difference, although usually more so in artificial memory tests than it does in rendering, where 1% would already be above average.

The clock of the ram, as long as it's CPU aligned, makes no difference whatsoever unless you're worried about forward compatibility (moot since there are switches from ddr3 to 4 and 5 happening in the next few quarters), or overclocking (something not entirely recommended on a farm).
Come, Join the Cult http://www.cultofrig.com - Rigging from First Principles
  06 June 2013
Wow, that's quite something. So 1% it is.

And I had no idea I wasn't supposed to overclock a farm. I guess that's a complete different topic

  06 June 2013
The processor will also be a factor in how much the memory speed affects performance. For example slower memory on a render node with 16 cores at 3.1 GHz (like the Xeon E5-2687W) would affect performance more than slow memory on a render node with 8 cores at 2.0 GHz. If going with low end processors I'd go with low end memory, if going with high end processors I'd go with high end memory too.
  06 June 2013
most memory bench and stats that i've read are comparing differences in nano seconds... hardly something to be overly concerned with to me.

stability and cost are more prevailing factors when I am looking at it, as most other components can reap far great performance benefits in real-world operations.
-- LinkedIn Profile --
-- Blog --
-- Portfolio --
  06 June 2013
Hey man,

I bumped the RAM speeds on my computer from 1333 to 1600, and my cinebench score went from like 7.08 to 7.11.

Overclocking is fine on your own system at home, but its probably not a good idea to do it at work. If your working at a small company and an overclocked computer breaks, you will be blamed for it. Even if what broke had nothing to do with the overclocking.

  06 June 2013
Originally Posted by jtvergarav: And I had no idea I wasn't supposed to overclock a farm. I guess that's a complete different topic

It's not that you're not supposed to, it's just that it warrants some deep thinking before you do, and it depends farm to farm and what your costs are like.

With farm density you have concerns like climate control, power consumption, power cycling, scalability vertical and horizontal and so on. When you bring it down to the money, which a farm should be entirely about, it's rarely good ROI.

If your farm is four small boxes in the laundry room and you live with your parents and they foot the electric bill, it might be good ROI. If you have a 24x rack packed with microblades and you run a variety of jobs on it, then it porbbaly isn't
Come, Join the Cult http://www.cultofrig.com - Rigging from First Principles
  06 June 2013
ThE_JacO, I'd be curious which microblades would even allow overclocking . Most blade anything I've ever seen are running server motherboards with server CPU's that don't support it - let alone the warranty issues at stake with the vendor.

dual xeons are still at best 20-30% more power efficient than an overclocked single CPU system when considering the rendering times. 20 dual xeon systems floored for a solid month would translate into around $100 savings a month vs a similar-performing overclocked render farm. Considering blades and rackmount servers cost usually 3-5x as much as an overclocked consumer-grade PC, the power savings won't offset the hardware costs unless you're going to use those machines for 15+ years with no future upgrade plans.

The physical space, warranty, and tech support staff issues are for sure a real concern. Those become more important as a farm gets larger.

overclocking a small render farm isn't recommended unless you know what you're doing. It takes some knowledge, testing, and refining. You're giving up thermal headroom. Any machine can lock up or restart if it runs too hot for too long. If you overclock high, you'll have at most 10-13 degrees C of headroom before the CPU starts throttling. Xeons at stock speeds will give you 30-40 degrees C of headroom so they can go a lot longer if your A/C unit breaks. It's wise to invest in a USB temperature email/phone alert monitor so team members can be alerted if the room temp goes above a threshold.

I've been overclocking our small render farm for almost 10 years. Had a few fans, motherboards, power supplies, hard drives, and ram die over the years, but no thermal nuclear meltdowns or melted components (contrary to a lot of the misconceptions). Meanwhile, I've had professional servers have the exact same issues, just less often. Everything still hinges on the A/C and those little system fans no matter what hardware it is.

Last edited by sentry66 : 06 June 2013 at 12:28 AM.
  06 June 2013
I was talking in hyperbolic examples, I somehow doubt he's got 4 cases stashed up over the washing machine in the laundry room either

All I was saying was that my comment about overclocking was in relation to the ram discussion a couple posts before, and shouldn't be taken to heart with no context when discussing the farm environment itself.

There are times when a little bit of it is simply free money, and times when it's not, just consider all the factors when you choose
Come, Join the Cult http://www.cultofrig.com - Rigging from First Principles
  06 June 2013
Thread automatically closed

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.
CGTalk Policy/Legalities
Note that as CGTalk Members, you agree to the terms and conditions of using this website.
Thread Closed share thread

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

vB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Society of Digital Artists

Powered by vBulletin
Copyright 2000 - 2006,
Jelsoft Enterprises Ltd.
Minimize Ads
Forum Jump

All times are GMT. The time now is 09:41 AM.

Powered by vBulletin
Copyright ©2000 - 2018, Jelsoft Enterprises Ltd.