Maxwell Render | Official public forum

Hi,

I know that this maybe should be posted on the OSX forum, but I thought it might be interesting enough for everyone to know.

I just received my new 12 Core Mac Pro - 2 x 2.66Ghz Xeon X5650 (not quite the top model) with 12 Gb RAM. Obviously I was keen to run a benchwell test.

I ran it in both 2.1 and 1.7.1 as follows:-

2.1 - 3m34s
1.7.1 - 3m49s

This seems much faster than I was expecting. Is there anything odd about these results? Am I doing something wrong do you think? Believe me, I am not complaining it is way faster than my last Mac!

I have a SSD for the boot drive and a RAID 0 for storage of files. I was interested to see if the increased write speeds would make a difference to the render time. I tried writing to both and the difference was nothing or very very small.

You will see even bigger differences with V2/1.7 when comparing time vs noise. V2 will be cleaner at the same SL than 1.7, so you have to compare render time/noise instead. You can see in your images also if you look at the dark parts of the object.

glassbathroom wrote: This seems much faster than I was expecting. Is there anything odd about these results?

No, sounds about right.

glassbathroom wrote: I have a SSD for the boot drive and a RAID 0 for storage of files. I was interested to see if the increased write speeds would make a difference to the render time.

Not really...

BTW Benchwell site is gone...

MS wrote:BTW Benchwell site is gone...

Yes, Benchwell is currently out of action. I think Next Limit are planning on introducing something similar or taking over the running of it.

Thanks everyone for their comments and input. I am loving the speed.

I'd be interested in the benchmarks for the single 6-Core « Westmere » 3.33 GHz, in comparison to my machine as both have about the same processing power (8*2.66 = 21,28 GHz vs 6*3.33 = 19,98 GHz). It could make a nice slave rendering machine, or even replace my actual mac which would become the render slave.

there is a thread for the 980x results. They were right around 6 minutes on benchwell.

http://maxwellrender.com/forum/viewtopi ... 97&t=34224

glassbathroom wrote:

MS wrote:BTW Benchwell site is gone...

Yes, Benchwell is currently out of action. I think Next Limit are planning on introducing something similar or taking over the running of it.

I was thinking who is an owner of benchwell.com domain now...

MS wrote:

glassbathroom wrote:

MS wrote:BTW Benchwell site is gone...

Yes, Benchwell is currently out of action. I think Next Limit are planning on introducing something similar or taking over the running of it.

I was thinking who is an owner of benchwell.com domain now...

I don't think Next Level is as it seems to have been dropped and picked up by a huge domain collector.

Max.

maggie wrote:Hi!

We are working on getting the Benchwell back. I hope to come with more news on this soon.


Cheers,
Maggie

Good news!!

yeah

Yeah, Benchwell is a very useful tool. I'm looking for a render machine and Benchwell would have come in very handy.

Just to mention that our free RANCHecker for Maxwell utility has an integrated benchmark, you may want to take a look at it even if you are not a RANCH customer

I also believe that one or two new benchmark scenes should be created by the Maxwell community. They should be more representative of actual production work done with Maxwell (architecture, design...) and should better show some of Maxwell 2 new features. The benchwell, although it has historical value, seems to me a bit too 'simple' nowadays...

Cheers,

Fred

Thanks for the input!

RenderFred wrote:Just to mention that our free RANCHecker for Maxwell utility has an integrated benchmark, you may want to take a look at it even if you are not a RANCH customer

I also believe that one or two new benchmark scenes should be created by the Maxwell community. They should be more representative of actual production work done with Maxwell (architecture, design...) and should better show some of Maxwell 2 new features. The benchwell, although it has historical value, seems to me a bit too 'simple' nowadays...

Cheers,

Fred

Totally agree with you Fred.
Benchmarks should focus not only on actual production areas, but even more on the time sensitive parts. See also my remarks in the preview renderer topic. It would be great if these were integrated in one benchmark suite where a few scenarios (archviz, design, studio usage) would be available with one model per scenario.
Timing should be done on interface issues (time to replace HDRI map f.i.) and render time up to a certain quality standard defined in three classes: passable raw images for internal usage, end user ready low standard, web based and end user high quality print / film ready.

There could be more but it should be kept simple with numbered scenarios in a dropdown list at a maximim of 9 like I propose. You can then get output per sector so f.i. weighted quality arch viz which you can easily compare. Full benchmark for evaluation of new machines and 9 scores for the underlying results.

Integrating all of this with a few preview steps like i suggested in the other topic: f.i 30 sec, 2m30s, 15 minutes would be easy.

This could easily be maintained in a simple database where you just look at the figures which interest you. The benchmark itself could just provide individual tests or the ´run everything´ take a few hours type of run.

Max.