Help to speed up rendering times.

[Coolhand78]

Hi Guys,

 

I'm wondering if you can help me, I'd like to speed up my rendering times and general speed in 3dsMax & Vray. I currently have a i7 920@2.67GHz, and 12Gb Ram, I run a single Geforce GTX680 video card as well.

 

what would be the most effective way of increasing my render speeds and general 3ds Max usage..?

 

Any help would be greatly appreciated.

 

I'm a noob with this stuff so be kind

[numerobis]

I would say a i7 4930K (6 cores) or 4770K (4 cores) depending on your budget...

 

http://www.cpubenchmark.net/cpu.php?cpu=Intel+Core+i7+920+%40+2.67GHz

 

http://www.tomshardware.com/charts/cpu-charts-2013/-02-Cinebench-11.5,3143.html

 

...or the most effective way: overclock the 920 to 3,8-4GHz

Edited March 24, 2014 by numerobis

[Coolhand78]

I would say a i7 4930K (6 cores) or 4770K (4 cores) depending on your budget...

 

http://www.cpubenchmark.net/cpu.php?cpu=Intel+Core+i7+920+%40+2.67GHz

 

http://www.tomshardware.com/charts/cpu-charts-2013/-02-Cinebench-11.5,3143.html

 

...or the most effective way: overclock the 920 to 3,8-4GHz

 

Hey mate thanks for the reply, any hints on how to overclock?

[numerobis]

what mainboard do you have?

 

some general guides for i7 920 overclocking:

http://www.overclock.net/t/787341/comprehensive-i7-overclocking-guide-with-explanations-classified-bios-explained

http://www.overclock.net/t/538439/guide-to-overclocking-the-core-i7-920-or-930-to-4-0ghz

http://whirlpool.net.au/wiki/rmp_i7_920_overclocking

http://www.xbitlabs.com/articles/cpu/display/core-i7-920-overclocking.html

 

There are many video tutorials on youtube regarding i7 9xx overclocking

http://www.youtube.com/results?search_query=overclocking%20920%20930%20950&sm=12

 

Some are suggesting to keep Intel Speedstep enabled but many disable it to improve stability. I would keep it enabled for a 24/7 setup.

I think to get faster results it is ok to test stability with Intel Burn Test (IBT) or LinX first. Prime95 can be set up to run more specific tests but it will take more time.

Edited March 25, 2014 by numerobis

[Coolhand78]

thanks man i'll check it out, seems a fairly involved process...?

 

does everyone around here overclock their cpu's? what about graphics cards?

Edited March 25, 2014 by Coolhand78

[Dimitris Tolios]

I do overclock most of my desktops, including GPUs when applicable. The latter will bring no performance benefits for rendering tho (unless it is GPU accelerated rendering ofc, but that's not the case with regular VRay).

 

I would go the mild O/C way + perhaps a slight decrease in render-frame pixel count...I know it is cheating in a way, but you can decrease your rendering time considerably with very slight loss in image quality by lowering your output file size.

 

 

Very few will be able to tell that an image was rendered @ 5000p width vs. 4000p width.

[Coolhand78]

I do overclock most of my desktops, including GPUs when applicable. The latter will bring no performance benefits for rendering tho (unless it is GPU accelerated rendering ofc, but that's not the case with regular VRay).

 

I would go the mild O/C way + perhaps a slight decrease in render-frame pixel count...I know it is cheating in a way, but you can decrease your rendering time considerably with very slight loss in image quality by lowering your output file size.

 

 

Very few will be able to tell that an image was rendered @ 5000p width vs. 4000p width.

 

Thanks for the reply Dimitris, it's mostly to speed up my desktop machine for test rendering, i dont generally render final images on this machine,

we have faster machines on the render farm for that, i'm just finding that currently i'm losing a lot of time testing things, and i was hoping that there might be a way to speed it up a bit... i mean if i'm working on a couple of a jobs a day and testing each of those then i can be spending up to 2 hrs a day test rendering images thats 10 hrs per week. Sure i can limit the amount of times i'm testing my scenes and i do only test them at 600-700px but as i try to increase the quality of the work i do with more complicated textures and geometry the render times blow out particularly on internal images...

[jasonstewart]

I would always render at final size because everything behaves differently if you change the size. The exception to the rule is for general lighting setup other than that full size. Then simply do small region renders for each area you are working on, if you are working on a brick wall why render the whole image just render a few areas of the wall and you can tell if it is good or not.

 

I would recommend overclocking the processor slightly, it is probably your bottleneck at this point. It can be scary (I just did my first one a couple weeks ago) but if you read the guides and stay within the maximum voltages and make sure your temps are good it is fairly safe. I picked up 15% more speed (above turbo boost) on my 4930k and it cost me $0 minus the extra juice it will pull from the wall now under full load.

[numerobis]

thanks man i'll check it out, seems a fairly involved process...?

 

does everyone around here overclock their cpu's?

I have overclocked all of my render systems of the last 15 years. Like Jason said, if you stay within the recommended temps and voltages there should be no problem.

But you have to spend some time to get it right.

 

Concerning the i7 920 overclock i don't know if it is worth the effort in your case. If you have the money i would go with a new 4930K or 4770K system. Even a 4770K @stock will be faster than a 50% overclocked i7 920. And overclocking the 920 is not as easy as for the newer processors. It's not really complicated but it takes some time to get it stable. And the result will be also dependent on your other components (cooler, board, PSU, RAM) especially for the i7 9x0 (and older processors) without unlocked multipliers where you're overclocking the board too.

Edited March 25, 2014 by numerobis

[Dimitris Tolios]

I would always render at final size because everything behaves differently if you change the size.

 

This is true, but after you get a critical point in image size, things get "normalized". The difference from a 1000p to a 2000p render is much bigger than going from a 2000p to a 3000p etc.

 

As far as rendering speed goes, VRay 3.0 also appears to be notably faster than the "old" 2~2.4 versions.

Along with "Render Mask" that allows you to re-render specific geometry or materials maintaining the rest of the "work" in the frame buffer, can also speed up things a lot (much more effective than rendering in regions).

 

As for upgrading vs. overclocking:

numerobis is right, it is not as easy to overclock a 1st gen i7 (like the 920) as it is with a 2nd/3rd/4rth gen i7 (Sandybridge 2xxx / Ivy-Bridge 3xxx / Haswell 4xxx respectively). The jump from the Nehalem (9xx) micro-architecture to Sandybridge was a big one, the rest not so much, but compact ontop of the considerable gains of Sandybridge to add up considerably).

 

Even without overclocking, a new CPU will probably be faster than a mildly clocked 920 in the most part, and you don't even need to get a 6-core like the 4930K to see that, but if you really want to save time on your local rendering tests and/or retouches maybe the hex does worth it.

 

Otherwise - and if overclocking is not an issue - the cheapest quad core haswell with hyperthreading around - actually a Xeon, the E3-1230 v3 for less than $250 - will give you a boost. Add $100~110 or even less for a decent s1150 board (H87 will suffice, Z87 is for overclocking and that's not even an option without a "K" i7) and you are looking at pretty good value.

Edited March 25, 2014 by dtolios