New Rig Build for AM - Dell Optiplex 9010 - video card questions

[a.quaihoi]

Hi all,

 

since there is not going to be any more Mac versions for the next Catalina Os, I’m building a new PC rig for AM and some other CAD apps, I got a super cheap ( free ) 2012 Dell Optiplex 9010 min tower, the offer was too good to refuse, its was used as print server for a Xerox printer, it has a 3.7 i7 processor, 20gb of ram and currently a Nvidia  GT 640 1gb card, I can max out the ram to 32gb, but was wondering about the video card, the case is a Mini tower and it can have half length graphics cards in the PCI slots - I can Frankenstein it and cut out the HD enclosures to fit a full length card if necessary. I’m thinking about upgrading to a AMD RX 580 8gb or GTX 1070 8gb, I’d also be required to get a higher power PSU for these cards. I am testing AM on its current configuration now with the GT 640 card and creating basic models from scratch its nice and responsive, but if I load a model say, thom model or any other models which are a bit complex, AM really bogs down in speed and frame rate in shaded wireframe view when modelling , I have set the Nvidia driver to performance over quality but it really slows down once models get complex. What are people experiences with video cards and performance in AM, what would be a card that gives fast modelling and frame rates in shaded wireframe mode with textures ?  Which is the best card for this, Nvidia or AMD from people doing complex models and scenes ? I really want a card that can deliver faster performance for high path count when modelling. Is AM just relying on cards for modelling or do I need more RAM as well ? 

[robcat2075]

Quote

...but if I load a model say, thom model or any other models which are a bit complex, AM really bogs down in speed and frame rate in shaded wireframe view when modelling ...

I'm surprised to read that... on Thom?  Plain yellow Thom?

The specs you have are all way better than the computer I'm on and i don't have any trouble with Thom. Something must be wrong.

One thing to check...  go to Tools>Options>Global and check "Limit real-time subdivisions"

That stops you from going past 4x4 subdivisions with the Page Up key in shaded mode.

Let me know if that changes anything.

 

Edit... I'm looking at me video card specs I have 2GB rather than your 1GB. maybe that would be why?

experiment... try setting your monitor resolution to something small like 1024x768, run A:M, load a model and see if it is still slow.

[Fuchur]

Hm, Thom really is not a complex model...

This is my current pc setup: https://www.patchwork3d.de/blog-5-en/amd-ryzen-my-new-pc-for-content-creation-with-hash-animationmaster-and-games-part-23-783
And it has been pretty much great for A:M. A RX 580 beats my RX570. The 1070 is better if your CAD software supports CUDA, if not the 580 has more "raw power".

What is a little bit odd: An i7 as a printserver? That is pretty much overkill... but good for you ;).

Did you install the latest driver for the current video card? The default one of Win10 are not meant for 3d graphics and that might be troublesome or your card is running very hot?
Are the fans spinning as they should?

Best regards
*Fuchur*

[a.quaihoi]

Hi Rob and Fuchur, forgot to mention, Im running Win 7 64bit, AM 18 x, the current card is the older GT 640 1gb

14 hours ago, robcat2075 said:

I'm surprised to read that... on Thom?  Plain yellow Thom?

Yes, loading old fellow Thom and viewing in shaded + wireframe makes it super slow, I remember a B&W Mac G3 tower being more responsive with more patch count models running on the old ati cards

 

14 hours ago, robcat2075 said:

One thing to check...  go to Tools>Options>Global and check "Limit real-time subdivisions"

Yes I will have to check this out, will have to do it tomorrow as its fathers day here and I quickly snuck into check the post 

 

14 hours ago, Fuchur said:

pretty much great for A:M. A RX 580 beats my RX570. The 1070 is better if your CAD software supports CUDA, if not the 580 has more "raw power"

I have been reading and watching about these two cards for the last to days, theres only a minor speed bump from RX570 to RX580, the AMD Card draws a bit more power than the Nvidia 1070, but here's the thing we can get a RX580 / 70 for $250 aud over here and all Nvidia 1070 xx is like $500 - $600 aud,  I suppose now that Fuchur has pointed out that the Nvidia card has CUDA for the other apps I suppose that justifies the price over the RX series, I was using Fusion 360 - but it is really really slow on mac, so I'll see how this goes on the pc, also going to learn Design spark mechanical - from trials this appears to work much faster than Fusion 360, I don't know the specs for these programs if they use CUDA or not, I will check and probably this will decide if I go AMD or Nvidia

14 hours ago, Fuchur said:

What is a little bit odd: An i7 as a printserver? That is pretty much overkill... but good for you ;).

The print server was driving " Memjet " large format printers by Xerox, I think this hardware was requirements for Caldera Rip software, the machine originally had Linux on it - well the main HD still has this and the RIP on it with a Dongle,  I think the HD is a propriety Xerox unit as it cannot be erased with the usual methods, so I  put in a SSD with Win 7 and it works fine . . . one thing I have not done is boot into bios and check settings - could AM slowing down have something to do with any Bios setting ? 

 

Ill try out the subdivision settings and let you know how it goes 

[Fuchur]

Hi!

here are a couple of numbers from my machine (Ryzen 1700, AMD RTX 570).
My machine might be a bit faster, but this should more or less be a work load for the GPU so it should be close to equal if you ask me with a RX580.
In the lower left corner you can see the FPS count. I am using the Thom model and paning the view a little bit to get different values on my FullHD display (with hair and decals enabled, but since thom does not use that, that should not be a difference):

- 800-1000 fps in shaded mode with a subdivision-level of 1.
- 750 - 880 fps in shaded mode with a subdivision-level of 4.
- 290 - 400 fps in shaded mode with a subdivision-level of 16.
- 90 - 140 fps in shaded mode with a subdivision-level of 64.

Any subdivision above that is overkill for any use case if you ask me. (even for realtime displacement maps which thom does not use anyway).

I would say anything over 25 fps is sufficient for pretty smooth modelling work. For gaming (where it really makes a difference if the click is done 10 ms earlier or later 25 fps is not enought and 60 fps are needed, but for 3d model creation, that really is not the case)

The value are of cause pretty high, but Thom isnt very complex anyway and those will drop if you have a much more complex character or something. In that case you will very likely work with something like subdivision levels of 4, but that is pretty smooth already in many cases.

Hope that helps. If you are getting something like 10 fps for subdivision level of 4 or so, something might be wrong at your system.
What do you get?

Best regards
*Fuchur*

[a.quaihoi]

Hi Fuchur, I just bought and installed the RX 580, updated all the drivers, launched AM and  am getting 85 fps in wire frame mode, however when I switch to shaded wireframe it bogs down to 5.7 - 5.8 fps  . . . . . AHA !!!! PROBLEM SOLVED !!!!!  

On 9/1/2019 at 1:56 AM, robcat2075 said:

One thing to check...  go to Tools>Options>Global and check "Limit real-time subdivisions"

That stops you from going past 4x4 subdivisions with the Page Up key in shaded mode.

Let me know if that changes anything.

ok ! so real time subdivisions was checked in the options, however for some reason in the modelling view its set to max subdivisions . . . hence the bogging down, now that I have lowered realtime subdivision down one step  its back to normal speed  49-50fps in modelling and and actions is back up to 80+ fps !!!! Somehow I have forgotten this setting from momentum animations days !!!  Now I have fix up the rigging in the old models ( 2012  - 2013 ) as actions are all gone wonky - anyway I have to go through the models and fix the rigging and redo smart skining etc. THANKS A MIL FOR POINTING THIS OUT TO ME GUYS !!! Its been a while using AM and I have forgotten advanced production routines. Now I feel like I just wasted money on the RX580 . . . but since I got the PC for free I suppose it was ok to spend a little bit of money on it.

I have ordered a ATEN Dual dislay / dual link  switch so I can connect the PC and MAC  simultaneously to 2 x 30"  Apple Cinema Displays for more screen space for better modelling and animation.

Cant wait to get back into AM again and finish off some projects that have been sitting on backup for ages ! 

[robcat2075]

Hurray!

The option to go past 4x4 is new in v19.  I haven't found a use for that so I keep "Limit" checked.

I''m glad you are still with us, even if it means using a PC!

[a.quaihoi]

Yes indeed ! I haven't quite did what Id like to do so far because time and work gets in the way, and obviously the using macs etc. so now Ive get a dedicated system just for AM so I can really get back into, really its such a great program, simple is the best ! 

[Fuchur]

15 hours ago, robcat2075 said:

Hurray!

The option to go past 4x4 is new in v19.  I haven't found a use for that so I keep "Limit" checked.

I''m glad you are still with us, even if it means using a PC!

Try applying a displacement map decal to a model... you will see the result in realtime view and for that you really need to bump that sub count up for it being useful.

Best regards
*Fuchur*

[Fuchur]

11 hours ago, a.quaihoi said:

Yes indeed ! I haven't quite did what Id like to do so far because time and work gets in the way, and obviously the using macs etc. so now Ive get a dedicated system just for AM so I can really get back into, really its such a great program, simple is the best ! 

👍

[robcat2075]

4 minutes ago, Fuchur said:

Try applying a displacement map decal to a model... you will see the result in realtime view and for that you really need to bump that sub count up for it being useful.
 

I have done that but it is so slow in real time that i decided it was faster to do a preview render when I needed to see it.

[Fuchur]

Might have to do with your older GPU/CPU then.
This is how it works for me with the RX 570. It is pretty useable for up to 256 or even 1024 sub (there I get 9 fps in this case... so debateable, but for a progressive render I need 2s or something like that which equals 0.5fps – so I get shadows and stuff like that too... so it depends what you are after).

With higher sub levels it does not make any sense anymore but most situations will not need more than that anyway. The other advantage is for export. You can easily see what the export levels higher than 16 will look like.

It is one of those features which are good to have so you might be able to work without if you need to. The good thing is: You do not need to.

Best regards
*Fuchur*