Jump to content
Hash, Inc. Forums

NetRender productivity


Recommended Posts

  • Hash Fellow

My new computer (thanks, Steve Shelton!) has an Intel Core i7-12700K with 12 cores, with a base frequency of 3.6 GHz and "boost" up to 5 GHz.

Eight of the 12 cores are high-power "performance" cores and four are low-power "efficiency"cores. The performance cores can be Hyperthreaded to behave as two cores for up to 20 total cores.


I've been giving it a workout with NetRender and testing to see how much extra work gets done with each additional core.

This chart shows the relative thoughput of completed frames as the number of cores allotted to the render increases.

The productivity of one core alone is treated as a baseline of 1.00, so we can look at this chart to see that with six cores working we will get 5.82 times as many frames completed per unit of time as with one core.


I'll discuss this chart more in further posts



  • exciting 1
Link to comment
Share on other sites

  • Hash Fellow
2 hours ago, Fuchur said:

How did you do the graph?

The chart is generated in an OpenOffice spreadsheet.


I made a PRJ and Preset to render the Three Teapots benchmark scene for 100 frames.

I loaded it in NetRender and did it with one core, two cores, three cores, etc .... and saved all the logs. Nothing else was running on the computer during these renders.

After each run I looked to see what the typical render time in seconds per frame was, when all the nodes for that run were rendering. (The time for  the first and last frames for each node are ignored). This number becomes "seconds per frame per node"

"Frames per hour" is (seconds in one hour/seconds per frame per node) * number of nodes

"Throughput" ratio is frames per hour for x nodes /frames per hour for a single node





Seems  like a lot of work to put that together?


Link to comment
Share on other sites

  • Hash Fellow

Notice that with Hyperthreading OFF, each of cores 9-11 adds more to the productivity than when Hyperthreading is ON.

When Hyperthreading is ON the addition of each of cores 9-19 is similar. It may be that when Hyperthreading is ON, the "efficiency cores" have something to do with managing that. The results for cores 9-19 may be a mix of what efficiency cores and hyper threaded cores do.




Q: If there are 8 Performance cores and 4 Efficiency cores why are there scores for only 11 nodes when HT is OFF?

A: The version of A:M I was testing limits the number of nodes to total number of cores minus one. With HT OFF, NetRender sees 12 cores and makes 11 nodes possible.

Likewise, when HT is ON, creating a total 20 logical cores (8 performance + 8 hyperthreaded+ 4 efficiency cores), Netrender allows only up to one less than that total.

Link to comment
Share on other sites

  • Hash Fellow

Considering just the eight standard cores, it is apparent that n cores doesn't quite create n times the first core's throughput.

This is largely because the CPU can't run as fast with more cores working.

With just one core tasked, the CPU can throttle up to its full "boost" speed but as each additional core is spun up the GHz have to come down a bit to manage the heat. It is also possible that there is a slight performance penalty as multiple cores compete for CPU cache and access to RAM.

None-the-less, NetRender scales very well as cores are added... 7.63 is pretty close to 8.0!

I'll take it


Link to comment
Share on other sites

  • Hash Fellow

A boot-up menu of the motherboard lets me enable/disable hyperthreading and various cores.

It isn't possible to turn off Performance cores entirely but it is possible to turn off all but one and leave the four Efficiency cores on.

This chart compares throughput of Efficiency cores to a Performance core. Only three Efficiency cores get tasked by NetRender.



Link to comment
Share on other sites

  • 3 months later...
  • Hash Fellow

Here is some weirdness. I put a job on NetRender but it was rendering painfully slow. The some of the frames were taking way longer than in A:M.

When I got up the resource monitor I found that somehow the slower efficiency cores were getting first priority; if only four frames were rendering they were all that was getting used. And only running the CPU at about half its top speed. 😮

I don't know if this is a Windows problem or a NetRender problem.

Efficiency Cores.jpg



Link to comment
Share on other sites

In the task manager, go to the "advanced" tab (where you can see the processes with "master64.exe" for instance, and RMBon master64.exe. Click on "associate with" (or something like that) and select the processors the software should work with.

At least like that you should be able to put the high performance once in if that is possible. I do not own an Intel CPU, so I can't test it myself.

I think Windows (and maybe A:M) does not recognize that rather new architecture of the Intel CPUs, where there are performance and efficiency cores.
Maybe you can set even more up like energy saving options, etc.?

Best regards

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

  • Create New...