Alright I posted something yesterday and no one took a bite at it... Let me put it another way.
We have a monster assembly we need to put together (100,000+ parts).
What we have now = 16GB RAM, 512 SSD HD, 1 GB Nvidia Quadro 2000M, i7 2.3GHz Precision Laptop
What do you suggest for building a beast, it has to be Dell (I know save that one for another day) and it will be a desktop workstation not a laptop. I was thinking Dual Xeon 8-core monsters, 32 GB (Maybe 64GB) RAM, 2.5GB Quadro Card, SSD HD. Should I go for the Tesla Co Processor? Anything else to help with Large assemblies?
Hi
What we can suggest is to visit these page:
http://usa.autodesk.com/adsk/servlet/syscert?id=18844534&siteID=123112
There is also an FAQ that you can also review:
http://usa.autodesk.com/adsk/servlet/index?siteID=123112&id=16318759
And..:
http://usa.autodesk.com/adsk/servlet/ps/dl/item?siteID=123112&id=15402497&linkID=9242018
Hope it helps in choosing the right machine
Inventor is still mostly single core for modeling/assembly environments so get the fastest processor you can find (3.5+GHZ), 16+G of RAM (fast ram 1600+), an NVIDIA 2G graphics card and SSD.
That is pretty much what we have now...
We are going to have to go to 32 GB RAM (Hitting the 16 level now). We are going to upgrade the Laptop we have now to the 1600Mz 32 GB RAM as a stop gap I will let you know if he sees a difference their.
My questions:
Will the 2 GB Video Card make a difference?
On the processor I have been told in the past a i7 actually will run better than two Xeons as you said Inventor really is not good at hyperthreading yet. Is this still true or is their a point where I can see a boost?
Money is no issue for the quote part of this, if they bulk I can always go down. This is delaying a 1/2 Billion $ project I really don't think they care at this point, I can spent what ever it takes to get the best machine for the job.
Sergio, I really need someone who has experience with Assemblies that have over 100,000+ parts to help me here. We are way past Minimum requirements. I have worked with other 100,000+ customers in the past and this even surpass them and they are using the Autodesk Substation Design Tool. I am almost to a mainframe at this point...
Anyone else deal with super large assemblies?
@warrenandy wrote:That is pretty much what we have now...
No its not..
BIG (HUGE) difference from a 2.3Ghz laptop to 3.5+ghz desktop...
Attempting very large assemblies on a laptop IMO is STUPID. A laptop will NEVER perform like a desktop computer.
There is only so much you can do with hardware then you need to start looking into stuff like LOD,etc..
http://danmiles.blogs.com/files/ml111-1p_dan-miles-1.pdf
Oh and "Dream Machine" and "Dell" is an oxymoron
You better fight that battle now...
I didn't say it was the same I just was saying it was close, they are workstation class machines w sandy bridge processors.
I know, I have been on them about converting all the subs to Shrinkwrap parts. They keep telling me they tried it a couple years ago and it messed with something they were doing. I think it was a procedure issue but I was not here for that. Being the new guy I have to take baby steps. Part of specking (and buying one) the new machine is to show them how much money they can save if they will let me teach them how to use Shrinwrap properly...
Are you kidding, corporate IT? I have been working with these guys from the outside (training / support /consulting) for years. We just talked them into Windows 7 64-bit late last year, hired me to come in and help full time in April. I am just happy they have Precisions and not Optiplexes... I have to play nice and not stir up trouble by going after real workstations, playing the hand I am dealt...
@warrenandy wrote:they are workstation class machines
= marketing jargon..
I guarantee most "gamers" out there have a computer that would whip a "workstation class" machine anyday.
Sounds to me like proper large modeling training is really whats required here.
And just so its clear.. Number of parts really means squat.. Its part complexity thats the real issue.
In my experience anything over 100,000+ unique parts puts you into a new class of design. True I have seen more complex parts take more processing than larger assemblies, but not that case here. We design the machinery here it is fairly complex (lots of robotics and automation).
Also I am talking 100,000+ unique parts (not occurrences), and they have to include all of the bolts to ensure clearances.
Training is on the TODO list... Everyone is slammed right now, have to wait until the next cycle.
In general Inventor doesn´t benefit from more then 1 Processors or an incredibly expensive Video Card like Quadro 5000 or Quadro 6000. In a best case scenario you might gain 15% viewport performance while spending more then double compared to Quadro 4000.
So which areas are there for increasing performance and stability without going to a self-built setup?
As much Memory as you can get (use), because of this:
There is only one way in reducing part load in very large assemblies and that´s derived part or simplification (replacement part) in the assembly.
Additionaly a fast processor is key for getting derived parts fast and maybe you should take a look at boxx workstations, because they deliver liquid cooling for their workstations and overclock the processor with warranty up to 4.5 GHz, take a look at this:
http://www.boxxtech.com/products/3DBOXX/4050_Overview2.asp
That´s the mid-range desktop so since money is not a big issue you probably go for the premium workstation:
http://www.boxxtech.com/products/3DBOXX/8550_Overview.asp
I don´t know if multiple GPU´s help Inventor for viewport rendering, maybe some Autodesk employee can help out on this topic, but you can have 4 GPU´s in the 8500/8900 series.
I think the same goes for the Dell T5600 or T7600 workstations at least for multiple GPU´s if it helps anything in Inventor.
All the best
Falk
Faster is better but its not always a big improvement. Here are some rebuild times from when I was testing my new machine with different over clock speeds.
I7 2700k, 16Gb ram (2133Hz), SSD for data, GTX 560 (only because I had one), win7 64
3.6 Ghz - 16.5 sec
3.8Ghz - 15.5 sec
4.0 GHz - 14.3 sec
4.2 Ghz - 13.2 sec
I would have thought that going from 3.6GHz to 4.2Ghz would give better than just 3 seconds. These times are for 1 part only, not some big assembly. You could waste a lot of money for no noticable benefit, especially if you get into water cooling etc for the sake of 0.3Ghz.
Consider also that high end cad would not exist if mid range cad was the best solution for large assemblies. You might also need better CAD to get the best productivity, or if your parts dont have many inter part geometric relationships, a cheaper direct editing CAD that eliminates rebuild times might be the answer Again they exist because there is a strong market for them.
@stevec781 wrote:
I would have thought that going from 3.6GHz to 4.2Ghz would give better than just 3 seconds.
but there 3 sec = 20% improvement = pretty darn good IMO
I would go for the top of the line gaming card. 2 if inventor supports SLI
@Anonymous wrote:I would go for the top of the line gaming card. 2 if inventor supports SLI
Inventor does not support SLI or crossfire.
Back in Inventor 5 and 5.5 I did a 263,000 Part assembly with only 4gigs of ram using 3gig switch. Modeling was not an issue but we never could make a drawing. had to export it in sections and put it back together in autocad.
i may be wrong but...if you use them in crossfire/SLI, windows and the mobo handles the crossfire/SLI bridge and Inventor only sees them as one card, so Inventor doesnt need to support crossfire, the motherboard and graphics driver does.
If your only using inventor then as stated a good gaming card is the go...if you use autocad as well, i would be using the quadro cards (I have a 5000 at work... goes good, my crossfired ata's at home go better for inventor but strugle with autocad)
OP
the mobo is the one that will have the bottlenecks for your other hardware...reasearch chipset and buslines (northside bridge etc)
ps...only use SSD for your operating system and hardware...sata drives give better data retreaval rates...look into stripped on your pc if the data is vaulted on a server or try 4x drives raid 10 for best performance/redundacy in your pc
when you have to open 100000 files at once, small increases per file can = big increase for assembly open time
@mcgyvr wrote:but there 3 sec = 20% improvement = pretty darn good IMO
Sure percentages always look good but if you look at time it's no big deal. Say 8 hours per day, 10 re-rebuilds an hour that's 240 seconds saved which is only 4 minutes per day. Not going to add even $1 to real profits or make any difference to delivery times.
But that's on a really complex part. On simple parts there is no noticable improvement because it's under 1 second at the slowest speed.
Upgrading my machine as made no real difference to my bottom line just as the changes from 2010 to 2012 have made no real difference to my bottom line either.