Upgrading my pc focusing on Mudbox Performance


#1

Hello folks!
I have a 2009 built PC which i use as a freelance for 3d sculpting, modelling and illustration.

CPU: Intel Core i7 920 D0 BOX, Quad Core, 2.66 GHz,
CU: Noctua CPU-Cooler NH-U12P SE2 - Sockel 1366/1156
MOBO: AsRock X58 Extreme, Intel X58, LGA1366, SLI, CFX
RAM: GeIL EVO ONE 6x2GB Kit, DDR3-1600, CL7-7-7-24@1
SSD: OCZ SSD Vertex 2 Extended Cap. 120GB, SATA-II,
HDD: WD VelociRaptor WD3000HLFS, 10’000rpm, 16MB, 30
HDD: Samsung HD103SJ, F3, 7200rpm, 32MB, 1TB, SATA-II
DVD: LG GH22NS, 22x DVD±R Brenner, SATA, schwarz
PSU: Corsair CMPSU-850HX, 850W, SLI/CF, ATX2.2/EPS
Case: Cooler Master RC-690-KKN1-GP, inkl. Mesh-Gitter
GPU: MSI GTX-275 Lightning (GT200b) 1792MB GDDR3, REMOVED
GPU: EVGA GTX780 SuperClocked

I changed my graphic unit 3 days ago, and i think i picked a good one… the 780 sits on the upper corner of the “sweetspot” on the “Bang per buck” curve, i believe.

Having always been a very unsecure person, I can not settle and i think i might have done a mistake. Many ramble about how the Titan Black is much better because of floating points. I dont even know what floating points: do i need them or not for Mudbox sculpting?

I have noticed that since i have plugged in the 780, the viewport is extremely smooth, which is great and just as i expected, but i see that the sculpting tools are not as fast as i tought they would be. What im wondering is, do the calculations of the brushes go to the GPU or to the CPU? In that case, it would seem my CPU might now be the bottleneck.

Question 1: did i pick a good GPU?
Question 2: Any toughts on how to proceed to improve performance? I have already overclocked my i7 920 to 3.7ghz…

thanks a lot for your answers and time :slight_smile:
R.


#2

If maximum performance and interactivity is important to you, you might look at a GPU that is optimized specifically for OpenGL and Mudbox…

http://www.cgchannel.com/2013/11/group-test-amd-and-nvidia-professional-gpus-2013/


#3

Who are those many and what the hell are they on about? :slight_smile:
Maybe they refer to the fact that the Titan is the only GTX card that doesn’t have a completely crippled Dual Precision Floating Point handling (any post 5xx card is crippled in those regards), but if that’s what they mean then they have no idea what they’re on about, given that outside of the scientific field practically nothing relies on crunching doubles.

I have noticed that since i have plugged in the 780, the viewport is extremely smooth, which is great and just as i expected, but i see that the sculpting tools are not as fast as i tought they would be. What im wondering is, do the calculations of the brushes go to the GPU or to the CPU? In that case, it would seem my CPU might now be the bottleneck.

ZB doesn’t use the GPU at all, MB does, brutally so, but you can’t get much better than what you have GPU wise. The 920 is getting a bit long in the tooth, especially bus wise, so it is possible it’s throttling transfers, but in all honestly I couldn’t tell you for sure. The difference between the PCI-E 2.0 you run and 3.0 in more modern set ups hardly has any impact (tested and counter tested), but other factors do come into play.
You’re also a bit vague “not as fast as you thought” isn’t really terribly helpful :slight_smile:
They should be between slightly smoother and considerably smoother than interacting with the viewport alone, if they’re chugging harder than that you definitely have a problem. As for how fast even just the size of the brush and the mesh can make a difference.

Question 1: did i pick a good GPU?

Most definitely yes.

Question 2: Any toughts on how to proceed to improve performance? I have already overclocked my i7 920 to 3.7ghz…

You could go to an Ivy Bridge i5 without breaking the bank, but it might not make a ton of difference. Again, it’s pretty hard to answer such specific question with such vague data :slight_smile:


#4

Your 920 at 3.7GHz is still a decent performer, though a new cpu will now give a worthwhile speed boost if you go down that route. For single threaded tasks like sculpting you can expect a new chip at stock speeds (3.5GHz) to be about 30% faster, or if you overclock the new chip to about 4.5GHz then it will be around 60-70% quicker than your 920.


#5

Sculpting isn’t a single thread bound problem, it’s perfectly possible to thread it very widely, and Mudbox mostly squeezes the GPU for the heavier part of that task anyway.
If he’s suffering for the 920 he’s mostly suffering the bridges and the bus management, not really lack of cycles.