Welcome to our community

Be a part of something great, join today!

Register Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

8Core Nehalem / Red Rushes madness - absolute madness!!! HELP

Same problem here

Same problem here

jacknusa said:
So, after reading the glowing reviews of the 8core 2.26 I went ahead and bought one of those to replace the now long in the tooth 2xdualcore 2.66 Mac Pro that's been our main Red Rushes render machine.

With the double number of cores, and with the faster Nehalem architecture I was expecting a rough 3x (give or take) speed increase running Red Rushes.

Result: the new 8core is almost 3x SLOWER than the old 4core!

I did not measure things but I'm experiencing the same problem, I certainly did not get faster with exactly the same computers.
Click to expand...
 
A special thanks goes to Apple for releasing the new Mac Pro. The dual i7 core Intel processor model just cut transcode times in half (roughly).

So if you open an r3d in RED Alert! and make a "look" using REDspace or REDlog and curves/saturation, etc. it saves the look to the .rsx file. Then open a bunch of these in RED Rushes and output to whatever format you want (say ProRes or Avid)... your dailies/edit files crunch out MUCH quicker.

Processing speed is RED's friend.

Jim
Click to expand...

If your result differs from this statement there should be some problems with your system or ... but this is far less believable , with J. statement.

Running multiple clips in parallel helps maximize throughput. Different options (debayer resolution, denoise, etc) all affect the speeds.
Click to expand...

of-course but he said that is using the same setting and files for both his system! so?
 
I've never touched a Mac (okay a couple of times).

This sounds like memory usage. How is the memory allocated, and how much memory per core? Your system settings might lead you to the holy grail. Check the internet about optimizing performance. You may need more memory (and/or faster memory than you purchased).
 
I just got an 8-core Nehalem a few weeks ago (2.26 GHz w/ 16 Gb RAM) and I'm also experiencing a lot slower speeds than I anticipate. From my experience there's a lot of speed hidden by selecting the right combination of clips and cores per clip. I'd be happy to download an R3D and do a test. Could be rather interresting to see what speeds people are getting on their gear.

On another note - the most disturbing thing about Redrushes is the fact that it's so damn buggy and unreliable. I was going to deliver some files to a client the other day and set it up to render files over-night only to find out it had crashed in the middle of the whole thing not even giving me an error message I could report to RED. I started it again and it kept crashing at the same file. Sigh. I couldn't really blame the client for being pissed about not getting their files in due time.
 
I am on a 2.66 - 12gb. Redrushes is extremely unreliable for me too.... random crashes.

I think it is memory related. - if you check activity monitor you can see RR eating up all the ram - when "free ram" becomes low_ I suspect it crashes... but i am not sure:(
 
I agree with you Seth.

I definitely and always prefer to use full debayer.
Yes for less visible noise, but also very natural and silky precise edges.

Antoine
 
more ram

more ram

Ok, I got some more RAM today, I will install and run a new test tomorrow.

Some other people in this thread have also reported "slower than expected performance" with their new Nehalem macs - I'm wondering:

a) are your dissapointing results also tied to full debayer processing like mine (at half debayer the Nehalem is performing more or less as expected in my case)

b) does processing with or without RSX have such a huge impact as in my case (see my original post).

c) Are your "slower than expected" results such as min that a 2-year old 4core beats out the new Nehalem 8core 3 to 1 when processing full debayer with a RSX?

j.

P.S. I'm also finding RedRushes very "chrashy" on the Nehalem - on the old 4core it was rock solid - rendered for 25 days straight once;-)
 
Wow, this is an enlightening (and disappointing) thread. I was just ogling a friend's new 8-core, thinking that his tower would be a veritable beast in comparison to my 2.66 quad. Now I'm happy that I'm rocking older technology.

To some of the other responses in here, I've found RedRushes to be super solid, no problems to report so far. Still on QT 7.5.5...

Also, while I would consider a half-res debayer acceptable for a 4k file, would definitely push for a full debayer for 3k and lower -> 1080p.

Hope you can figure this out J...excited to hear about your findings with the updated ram. Good luck, one and all.

Cheers,
Clint
 
More RAM did not help - 4K does the same thing - Activity monitor dumps.

More RAM did not help - 4K does the same thing - Activity monitor dumps.

Ok, so I put in another 4GB of RAM in the 8core Nehalem for a total of 10GB - No change, still over 5minutes/100frames for a 3K Full res Debayer with an RSX.

Thinking that 3K is perhaps a bit of an oddball format, I ran my tests again with some 4K material - No change, the results for 4K are consistent with what I'm getting for the 3K tests:

4K render on new 8core Nehalem
---
FULL RES, with RSX, scale to 1080 Pro res = 9m20s/100frames
HALF RES, with RSX, scale to 1080 Pro res = 0m11s/100frames


4K render on old 4core
---
FULL RES, with RSX, scale to 1080 Pro res = 3m23s/100frames
HALF RES, with RSX, scale to 1080 Pro res = 0m25s/100frames

Again, just like my 3K results: at Full Debayer with an RSX, the new 8core Nehalem is roughly 3x SLOWER than the old 4core, at Half debayer the new 8core is roughly 2,5x FASTER than the old 4core.


I'm attaching four screen dumps of the Activity monitor window while rendering 3K Full res/Half res respectively. Maybe this will tell someone something about what's going on?


Also, before this thread gets bogged down with more "Why are you using full debayer anyway?" -remarks... remember: this is 3K -> 1080 final output we are talking about, a full debayer does seem prudent in this case, and yes, I can eyeball the difference on a side by side comparison.

So: Any thoughts on what's going on here? Any one else with an 8core Nehalem experiencing the same things?

J.
 
Weel it seems that when using FullDebayer the 8core are not used!


To me is a Redrushes problem or limitation when using FD!
 
jacknusa said:
So - Full Debayer on 8core Nehalem = NO GO?

J.
Click to expand...

No. That's why REDrushes can run multiple clips in parallel to take advantage of all the cores. Load up multiple clips and adjust "Clips at once" and "Core's per Clip". Testing only one clip with RR is misleading because it's meant to be used for rendering multiple clips in parallel.
 
Deanan said:
No. That's why REDrushes can run multiple clips in parallel to take advantage of all the cores. Load up multiple clips and adjust "Clips at once" and "Core's per Clip". Testing only one clip with RR is misleading because it's meant to be used for rendering multiple clips in parallel.
Click to expand...

Ok, I will test and report back - but are you saying it is NORMAL and EXPECTED that a new 8core Nehalem is 3 times slower than a 2 year old 4core when rendering only one clip at full rez debayer?

j.
 
jacknusa said:
Ok, I will test and report back - but are you saying it is NORMAL and EXPECTED that a new 8core Nehalem is 3 times slower than a 2 year old 4core when rendering only one clip at full rez debayer?

j.
Click to expand...

In all our tests the new 8 nahalem is faster than the old 8 core.
 
Deanan said:
In all our tests the new 8 nahalem is faster than the old 8 core.
Click to expand...

so...
a) you are not testing full debayer with RSX, or
b) I've got a lemon of a Nehalem...

If "b" what do I do to fix it? And, where is the lemon hiding - hardware, software?

the other people in this thread that have expressed dissapointment - can you be more precise? And, are you dissapointing results just as in my case tied to full res debayer? (remember - at half res debayer the Nehalem is performing more or less as expexted: a rough 2.5 times faster than the old 4core.)

j.

j.
 
at half res debayer the Nehalem is performing more or less as expexted: a rough 2.5 times faster than the old 4core.
Click to expand...

that's the real strangeness.

If you have already performed an Apple Hardware test you can be sure your Mac is running ok!


Denan is saying
Testing only one clip with RR is misleading because it's meant to be used for rendering multiple clips in parallel.
Click to expand...
but why at half res debayer the Nehalem is performing as expected?
If it's underperforming at full res because why at half res it's not?

Why at half res all the 8 CPU are used almost at their 100% while when dealing with full res only 3 CPU are used?


jacknusa try to post a screenshot of the old QuadCore performing a full res debayer.
 
Deanan said:
Tested that also. I'll retest on Monday.
Click to expand...

please do - I checked the other "Nehalem performance" -threads, and none seem to be testing full debayer...

Is there someone else out there with a 8core Nehalem that can run a test with the following setup:

- full res debayer
- use clip RSX (small adjustment to exposure and brightness)
- scale to 1080 (fit hight)
- output to prores hq

for 4K 2:1 I'm getting over 9 minutes/100 frames, for 3K I'm getting over 5 minutes/100frames (see previous posts for specifics).

j.
 
You must log in or register to reply here.
Back
Top