PDA

View Full Version : Which is better for LW, less cores fast speed or more cores less speed?



yoshiii
02-03-2015, 04:47 PM
Hello

What is better for Lightwave, have a four core processor that hyperthreads with fast per core speed or a six core processor that has little bit lesser speed than the four core?

I am choosing between the i7 4790k and the i7 5820K.

Snosrap
02-03-2015, 06:55 PM
Rendering benefits from more cores. Modeling benefits from faster CPU speeds.

jasonwestmas
02-03-2015, 07:26 PM
More Memory that is faster and Faster Hard Drives are better for Lightwave too.

magiclight
02-04-2015, 02:17 AM
I would go for the 4790 even though the 5820 is a little faster, but memory for the 5820 cost both arms.
a 4790 with 32 gig ram or even better 64 gigs give you most for your money at the moment.

For modeling video card speed is king, CPU is not very important for that.

yoshiii
02-04-2015, 07:18 AM
More Memory that is faster and Faster Hard Drives are better for Lightwave too.


Rendering benefits from more cores. Modeling benefits from faster CPU speeds.


I would go for the 4790 even though the 5820 is a little faster, but memory for the 5820 cost both arms.
a 4790 with 32 gig ram or even better 64 gigs give you most for your money at the moment.

For modeling video card speed is king, CPU is not very important for that.

If the more cores the better, then the FX8350 should do fine with LW right? Does hyper threading from the Intels' actually work in LW?

I have a 120gb Samsung Evo ssd. I am going to add another one, maybe 256 and a WD Black hdd.

I cant start with 32 gig of ram but I can start with 16 gigs.

I have a GTX660, was going to get a GTX970 but wonder if maybe better to get a GTX980?

jasonwestmas
02-04-2015, 07:40 AM
A high end video card is not really useful unless you plan on doing GPU rendering with Octane or doing video game previews in LW or something like that. Lightwave does have a DirectX11 shader now that you can use.

erikals
02-04-2015, 07:47 AM
for Modeler only,
less cores fast speed is faster

for Layout render,
more cores less speed is faster


afaik... but it depends on many factors, not sure just how fast a cpu with fewer cores is these days

yoshiii
02-04-2015, 07:48 AM
A high end video card is not really useful unless you plan on doing GPU rendering with Octane or doing video game previews in LW or something like that. Lightwave does have a DirectX11 shader now that you can use.

Ok so Light Wave is mostly cpu bound? Is it worth it to get Octane?

erikals
02-04-2015, 07:51 AM
for many, yes...


https://www.youtube.com/watch?v=hAw3Qvk85ag

if you do a lot of compositing, or cartoons, exteriors, LightWave is still a very good choice, and often better

in "Snow White and the Huntsman" for example, 95% of the shots are exterior shots...

jasonwestmas
02-04-2015, 07:55 AM
Ok so Light Wave is mostly cpu bound? Is it worth it to get Octane?

Yes, CPU for the most part. Nothing in native lightwave does GPU caching and hardware Rendering is limited to OGL and DX11. I think Octane is worth it if you get a card with 6 GB or more VRam. But I love to use a lot of textures and displacements so I tend to stick to CPU for cost effectiveness and using Huge amounts of memory. If you don't use a lot of textures and crazy polygon counts GPU rendering is a faster way to render imo. Octane in particular has awesome physical accuracy too if you like that sort of thing. If you do more stylized rendering then CPU rendering is more flexible for that.

ianr
02-04-2015, 07:55 AM
At the moment no Sims inside Lightwave use GPU But if you may use Pyro
in the shape of the Turbulance plugin, then it gets heavy lifting from a bigger GPU
If you don't use pyro ,then as the ofthers say :
Fast SSD & FAST MEMORY, for the time being.
Always a problem when to jump on that Hardware Train!

erikals
02-04-2015, 07:59 AM
more Octane LightWave videos... >
http://vimeo.com/user2182154/videos

yoshiii
02-04-2015, 08:07 AM
Yes, CPU for the most part. Nothing in native lightwave does GPU caching and hardware Rendering is limited to OGL and DX11. I think Octane is worth it if you get a card with 6 GB or more VRam. But I love to use a lot of textures and displacements so I tend to stick to CPU for cost effectiveness and using Huge amounts of memory. If you don't use a lot of textures and crazy polygon counts GPU rendering is a faster way to render imo. Octane in particular has awesome physical accuracy too if you like that sort of thing. If you do more stylized rendering then CPU rendering is more flexible for that.


At the moment no Sims inside Lightwave use GPU But if you may use Pyro
in the shape of the Turbulance plugin, then it gets heavy lifting from a bigger GPU
If you don't use pyro ,then as the ofthers say :
Fast SSD & FAST MEMORY, for the time being.
Always a problem when to jump on that Hardware Train!



more Octane LightWave videos... >
http://vimeo.com/user2182154/videos

Hmm, I can afford a GTX970 and maybe a 980, only 3.5 and 4 gigs of ram on those.

Thanks for the video links

If Lightwave cares more about how many cores, it should ok to use a FX8350 with good results right?

magiclight
02-04-2015, 08:28 AM
Yes, Lightwave will use all the cores, the more you have the faster it will render

jasonwestmas
02-04-2015, 08:37 AM
If you are willing to pay for a GTX 980 I think getting a CPU that is twice as good as a FX8350 might last you longer if you are planning on rendering alot in Lightwave.

yoshiii
02-04-2015, 08:40 AM
If you are willing to pay for a GTX 980 I think getting a CPU that is twice as good as a FX8350 might last you longer if you are planning on rendering alot in Lightwave.

So does LW take advantage of Hyper Threading or only real cores? If only real cores, my choices are going to be the FX8350 or the i7 5820K.

erikals
02-04-2015, 08:42 AM
well, isn't it easier and more cost efficient to go for GPU renders in the long run?

after all, it's easier to add GPU cards, than to replace >
- mother board
- cpu
- ram (possibly)

http://erikalstad.com/backup/misc.php_files/question.gif not quite sure...

yoshiii
02-04-2015, 08:44 AM
Rendering benefits from more cores. Modeling benefits from faster CPU speeds.

Ok that settles it, seems like hyperthreading doesnt appy here so I will choose between the FX8350 and the i7 5820K. If I have it wrong please tell me.

jasonwestmas
02-04-2015, 08:49 AM
well, isn't it easier and more cost efficient to go for GPU renders in the long run?

after all, it's easier to add GPU cards, than to replace >
- mother board
- cpu
- ram (possibly)

http://erikalstad.com/backup/misc.php_files/question.gif not quite sure...

If I had more money to spend then yes, I would make a GPU box with 3 8GB cards in it. That would run me around 6-8 grand. But again it depends on how much texture and displacement space you want. :)

- - - Updated - - -


Ok that settles it, seems like hyperthreading doesnt appy here so I will choose between the FX8350 and the i7 5820K. If I have it wrong please tell me.

you can't really go wrong if you are using 8Cores or more. Just do a search online for 3D rendering bench marks in general to compare.

magiclight
02-04-2015, 10:34 AM
So does LW take advantage of Hyper Threading or only real cores? If only real cores,

Lightwave don't care, the only difference is that hyperthreading is a little slower in the floting point department, it will use all there is, a 4790 will have 8 render threads running even though there are only 4 cores.

spherical
02-04-2015, 05:28 PM
Ok that settles it, seems like hyperthreading doesnt appy here so I will choose between the FX8350 and the i7 5820K. If I have it wrong please tell me.

Yes it does apply. I have an i7-3930K running at 4.5GHz: 6 physical cores, renders with 12 threads. When the person mentioned "more cores", it was in comparison to how fast each individual core is, as Modeler isn't multi-thread aware yet. That means that a faster CPU is better in Modeler, because it will use only one of the cores.

Snosrap
02-04-2015, 10:27 PM
AFAIK the FX8350 is 5ghz with 8 cores and the i7 5820K is 3.3 ghz (3.6 with Turbo Boost) and 6 cores. On paper it seems the FX8350 is the better proc all the way around. In practice I wouldn't have a clue. LW rendering loves cores. The FX8350 with 8 cores would give you 16 rendering threads - that's not bad. I've got a an older i7 920 with 4 cores (8 rendering threads) and at work I've got an older Xeon W3680 with 12 rendering threads and it kicks butt over my home PC in rendering. Granted it's running at 3.3 ghz and my home box is only running at 2.67ghz. Thing is, I absolutely cannot tell the difference in Modelers performance over the two boxes. Cores over GHz is my thought when it comes to LW. :)

vonpietro
02-04-2015, 10:53 PM
why not get one of these monsters
http://www.amazon.com/Nvidia-GeForce-900-12080-0000-000-Ultimate-Graphics/dp/B00KVKAOSM/ref=sr_1_4?ie=UTF8&qid=1423115392&sr=8-4&keywords=titan+video+card

anyone have one, is 2000 bucks worth this beast. Is it like 2 or 3 times faster than anything else out there?

by the way, just how many cores can a person get now adays.
standard laptops are usually 4, and i've heard of people with 12 core machines.
anyone know how much the optimal one would cost? With a titan of course.

get the 32 gigs of ram, you will not regret it - if you spend money on anything - the ram will trump all - i say that because even at the expense of speed, 32 gigs gives the maximum ability to create huge scenes. When you have the extra ram, you'll find ways of utilizing it.
=)

Danner
02-05-2015, 01:43 AM
You can get up to 64 cores on one machine now, I've been using a dual 8core xeon system for 3 years now (32 threads) and it's still better than any single i7.

vonpietro
02-05-2015, 01:46 AM
got a link to one i could buy - whoa thats 64 real 64 virtual -... 128 holy moly, can lightwave handle that many cores? How much =)

http://www.cpu-world.com/news_2015/2015020301_Intel_introduces_embedded_Xeon_E5-2658A_v3_CPU.html
12 core xeon - capable of 765gigs of ram
uses ddr4 - someone said that was expensive though - maybe it can also use ddr3? anyone?

so basically - 24 cores with 760gigs of ram - thats one heck of a machine for lightwave.
anyone got one?
=)

http://www.engadget.com/2007/08/20/tilera-debuts-its-sixty-four-core-processor-melts-faces/
30x speed increase over a 4 core.
anyone recall those overpriced computers that used to run alias... what was its name.. its been so long i kinda forgot =) but seems were back to the RISC processors - not quite windows compatible.
so unless they address that - no one will want to start up that route again.

did some surfing
http://www.pugetsystems.com/featured/64-core-256GB-Workstation-100
4 16 core amd

magiclight
02-05-2015, 09:00 AM
The more cores you get the bigger the ram bandwidth problem gets, dual channel memory is expensive and it still only gives dual channel access, all the cores will need access to memory and they have to share 1 or 2 data channels, cache memory for each core and so on can help, but when you get up to some number of cores the memory bandwith will start to slow everything down a bit, so it's not just to put in a lot of cores, everything around the CPU need to keep up with that also.