How fast is your Inventor PC really?

[Raider_71]

Hi guys,

We have had to do some testing on a bunch of Inventor PC's recently to determine which of the PC's needs to be replaced. Obviously we needed to find out which of the PC's are the worst eprformers as there was only budget to replace 50% of the design PC's. So we thought the Darwin theory will come in handy right... 🙂

Anyways I started searching on the net for toppics on how to benchmark an Inventor PC. Then I thought whats the point of using gaming benchmark tools because Inventor is not a game and there are more aspects than just graphics performance when it comes to percieved performance on an Inventor PC right.

So we decided to create our own Inventor benchmark tool which tests various aspects of an Inventor PC to give us an overview of our PC's performance. This then helped us make a decision as to whcih pc's to replace.

We have made the tool available free of charge to anyone interested in checking how their PC stacks up to their peers or friends. 🙂

Please download it here and post your results here as well if you want. Would be interesting to see what beast workstations are out there.

I would like to say thanks to Kirk #karthur1, for helping in testing the app.

Please feel free to send any suggestions our way. There is an email link in the app.

Download and Install

The application will work with Inventor 2014 to 2016 only.

IMPORTANT: After installation there will be an Inventor Bench icon on your desktop that looks like this: 

My resluts:

HP Elitebook 8560w with an SSD upgrade.

[Neil_Cross]

You'll have to be more specific with the Xeon models, these companies offer woefully unsuitable Xeons in their workstations which you need to avoid for Inventor so you'll need to list the exact Xeon model number for receiving feedback.

Regarding Lenovo... I've no idea how they managed to cap out on 32GB RAM, that doesn't make any sense.  Most CPUs now (i9's and Xeons) support well over 128GB RAM.  I've got a Lenovo P520 in the office which I could put 512GB RAM in if I wanted to.

[Anonymous]

Neil, 

You are the person I had hoped would respond since you are highly knowledgeable about configurations.  

HP: Intel® Xeon® E-2226G Processor (3.4GHz, up to 4.7 GHz w/Boost, 12MB cache, 6 core) 
Dell Precision 7920: Intel Xeon Silver 4110 2.1GHz, 3.0GHz Turbo, 8C, 9.6GT/s 2UPI, 11MB Cache, HT (85W) DDR4-2400

Dell Precision 7740 Mobile Workstation: Intel® Xeon E-2276M, 6 Core Xeon, 12M Cache, 2.80GHz up to 4.70GHz Turbo, 45W, vPro
Lenovo: Intel Xeon W-2145 Processor with vPro (8 cores, 3.70GHz, 11MB Cache)

As I stated, the rep from Lenovo told me that the cap for the ThinkStation P520c was 32 GB of RAM (I don't want to go lower than 64 but will push for 128 GB If I have to).

[Neil_Cross]

Well, that's all a bit of a mixed bag.

Avoid the Xeon E2226G, it doesn't have hyperthreading which seems like a crazy thing to go for when you don't have to.  Dodge that.  HP offer a great variety of workstations so I wouldn't rule them out entirely based on one poor configuration.

Avoid any Xeon with a metal in its name, Silver, Gold, Platinum etc none of those are suited for Inventor.  Even Intels own website specifically states that they're for data centre servers, I've no idea why these workstation vendors deem it appropriate to offer them in PCs unless they're contractually obliged to shift a certain number of them.  Avoid them like the black plague of death.

The third one is a laptop? As far as laptops go the Xeon 2276M is probably as good as it gets, there is a 2286M which is marginally better but the 2276M is certainly a good choice for a mobile workstation.  I've got a Lenovo P53 laptop here which has a i9-9980H CPU in, it's scoring around 12.5 on the bench test.

The Xeon W-2145 is a decent CPU for Inventor, I have that in the office on test at the moment and was shocked at how good it was.  Again there are better CPUs out there but it was a great all-rounder, it scored around 11.5 on the bench test and was pretty capable.  Lenovo are sending me a newer Xeon W-2255 workstation next week, I'd be inclined to look towards that instead of the W-2145 but we're probably talking marginally different again.  

I can see now why you where limited to 32GB RAM, that P520c must have a motherboard or spacial limitation because the platform they're using supports far higher than 32GB.  Not sure why they offered you the C version but if you test the regular P520 it supports much more RAM.

It's super confusing I know.  There are now hundreds of different configurations you can get, and in a lot of cases it's super marginal as to the difference between them all, but there certainly are ones to avoid.

As a general rule, go for a (desktop) CPU with -

- 6 or more cores

- Hyperthreading

- Turbo frequency 4.5GHz or higher

- Base frequency 3.5GHz or higher

- Launch date '19 onwards

Xeon List 

If you look at that Xeon list, arrange them by Max Turbo Frequency, that's a good place to start.

Xeon W List 

Same with this list, although these are the Workstation branded Xeons which are more expensive.

[leowarren34]

@Neil_Cross I'm glad they include metals in the xeon names now, significantly clearer than the E number.

The only thing I have to add is avoid scalable chips in general for Inventor (Metals and E5/E7)

The only reason they must be offering these high core low speed chips is for scalable processes but for most intents and purposes is irritating when it comes to buy as most non savvy people will order by price high to low and spend as much as their budget allows.

[nutral]

I don't agree on needing to have hyperthreading, hyperthreading is for when an application scales well with cores, as part of the cpu is doubled (a bit like a highway that is 2 lane but starts with part of it being 4 lanes).

Inventor is not well threaded except drawings and even that is not well threaded. Hyperthreading would only help a bit if you have more than 8 drawing views on a 8 core cpu.

The pc's we have running at work have 9700k, which as a turbo clock of 4,9ghz and no hyperthreading, but they run inventor really well.

Also, for some cpu bound operations, memory speed does matter, especially for AMD CPU's.

I've upgraded my gaming pc with a 3700x, altough the scores are not that great, somehow the framerates are really good, but the rest of the results aren't.

I still have the stock cpu cooler and slower memory, tomorrow i'll get a new cpu cooler (which helps with turbo boost) and faster memory. Altough i don't think it will help that much.

I still don't think that the high end workstations with Xeon's are the best, altough they are with reliability. I would actually suggest doing this:

Have all the workstations be Built like a gaming pc but with higher end components and get 1 or 2 extra pc's. When someone has a broken component, just swap the pc and move the SSD over.

Recommended Build:

High quality power supply, like the Seasonic prime or focus plus or EVGA

Reliable motherboard tends to differ from chipset, but generally Gigabyte, msi or Asus

EVGA Geforce graphics card, with the better cooling like ICX.

I think a RTX 2060 should be enough, except when running higher resultions like 4K. Then i would suggest a higher end Graphics card.

Faster memory, For amd the max usable is 3600 CL16. and i would suggest the same for intel. I'll check tomorrow with faster memory, what the difference is.

I would recommend getting an intel cpu, because it is still faster for single threaded performance, and most of inventor while modelling is single threaded. The fastest right now is the 9900KS but the 9900K has the same single core turbo. for a cheaper system the 9700K only has a 0,1 ghz lower speed.

If you want to get support in the US, Puget systems i a company that delivers workstations with these specs, and also give good support for larger engineering companies.  They have recognized that inventor pc's just need single thread performance and do deliver systems, they also select their components and check the reliability of them. And they test them before the system goes out. Their reliability article from august 2018 https://www.pugetsystems.com/labs/articles/What-is-the-most-reliable-hardware-in-our-Puget-Systems-w...

[Neil_Cross]

I was answering someone who was in the market for an OEM build.

Also, hyperthreading isn't detrimental to the performance of Inventor.  I've performed several tests with it on and off, anyone can, and it makes absolutely no difference either way, so opting to buy a CPU without it when the majority have it is a ridiculous decision when it objectively helps other applications.   FWIW many vendors still ship their own software suites with application profiles which disable hyperthreading for CAD optimisation, that's an archaic process based off dated methodology born out of a time when CPUs only had 2 cores.  If you've only got 2 cores and you want one of them to handle Inventor, then yes you don't want that core shared with other processes, but now we're talking about 6, 8, 10+ cores the reasons for doing this is entirely obsolete now.

Also FWIW this landscape is changing fast, a lot of the arguments and objective factual information/advice at the beginning of this thread is now no longer true.  For example some of the newer Xeons now are duplicates of the consumer i9's and now yield similar performance as gaming based systems minus any overclocking, that wasn't the case when this thread began.  Now you can get a Xeon and see very close if not identical Inventor performance to that of the consumer core line.

[leowarren34]

Hyperthreading generally helps with background applications so more is free for Inventor.

It doesn't hinder anything so it's a feature that can be used as when needed.

I suspect the 3700x may be suffering due to optimisation favouring Intel due to Intel being the only real competitor in the WS market until now.

Xeons/Quadros etc are binned versions of the consumer chip.

When a chip comes of the assembly line if it has less/few defects then it goes to the WS line and if it's not then it goes to consumer and tweaked. It has been shown that there are sibling CPUs where the chip is identical and the performance is near enough. Main difference being the WS line has more bells and whistles. 

As Neil has said previously when your name is on the line and the money is available go for what is reliable rather than what is cheap. WS products also have drivers and features such as ECC and Lower TDP.

[nutral]

Yes that was what i meant, hyperthreading is not mandatory, it just doesn't really matter, for the 8 or 6 core cpu's that are mostly selected.

If xeon's have the same turbo clocks, it indeed doesn't matter much. the list you sent even has 9900k turbo clocks in them.

I've had al ook at the HP website, and they even have workstations with gaming cpu's and graphics cards in their workstation line ?

https://store.hp.com/NetherlandsStore/Merch/Product.aspx?id=6TT38EA&opt=ABB&sel=WKS

It's still about 1000 euros more expensive than a built to order system. but it does have the 9700k and a RTX2080, i didn't expect to find a workstation with a geforce gpu...

To answer Leowarren,

I don't think things are more optimized for intel. it is probably just that some parts work better than others, the amd for example has more L3 cache, wich means that for instructions that fit into the cache, the amd will be faster than the intel, as the intel has to wait for information to transfer over the DDR4 bus. The AMD ryzen for example does have a higher framerate.

I'm looking at changing the inventor bench, as the framerates currently are at 200fps, at 200fps the graphics card is rarely the bottleneck, as the cpu has to deliver information for every frame with directx. If you're working with cad, it doesn't really matter if you get 200fps or 150fps, but it does when the pc drops to 10fps with a really large model. I would like the test to have a larger model to really drop the framerates.

Running the test, i get about the same score at 1080p and at 4k. At 1080p it uses about 15% of my gpu, and at 4k it uses about 30%.

[Neil_Cross]

@nutral it wasn't clear from the previous post but are you aware that Inventor isn't GPU accelerated? In the same PC you could swap from a Quadro RTX 5000 down to a GeForce 960 and you'd get identical frame rates in Inventor.

The Autodesk OGS used by Inventor is CPU bound... when you orbit a large assembly for example the CPU is performing all the calculations required to present the geometry into the viewport, the GPU obviously holds the data in the frame buffer and delivers it to the display but the GPU itself isn't used to compute anything within Inventor, it's all CPU bound.  The increase in GPU activity when going from 1080p to 4K will be down to the passive increase in pixels being handled by the GPU.  So long story short, using a larger model for benching won't stress the GPU, it just puts more load on the CPU... and the other avenues which handle data transmission.

If you want to see this in action, go to:

HKEY_CURRENT_USER\Software\Autodesk\Inventor\RegistryVersion24.0\System\Preferences\GraphicsUtil

There's a key in there called "UseMultiThreading" if you change that value to 0, reload Inventor, re-run the bench test, you'll get a significantly higher frame rates.

[mluterman]

Thank you, Neil...for helping me get in the 15's with an 8th-gen.  I'm wondering why Inventor doesn't come this way out-of-the-box or they at least tell us about it.  Are there any downsides here/I wonder?  Anyway, thanks again, Mitch

[Neil_Cross]

Wow, I've got a 8700K and can't get anywhere near 15 even with an overclock, what have you got that running at?

Off memory there was some issues reported with colour overrides on some faces and bodies going a bit weird, colours changing etc but that was a couple of years ago - its not something I would recommend leaving enabled for production work as generally its not really worth it.  Its a completely undocumented setting and for me its not worth leaving on outside of running the benchmark

[leowarren34]

Thats a really good result!

2nd Overall for highest

1st if not overclocked for the Non OC category

[mluterman]

Thanks Neil, that means a lot coming from you.  Stats are: OC'd to 4.7 (prob. won the "chip lottery").  No paging file, 11GB Ramdrive (by Qiling found to be fastest free one).  All temp variables set to Ramdrive and Inventor Undo file set to Ramdrive in Application Options.  No weird colors/artifacts/glitches yet, but I'll keep an eye out.  (PC built by Xi Computers; MSI 370 Mortar game board; that's what they were builidng the single-processor CAD PC's with back in 6/2018).

[mluterman]

Thank you, Leo; keep in mind that's an 8th-gen processor:  I expect the up-and-coming 10th-gens will blow all this away very soon.

[tom_vierling]

I personally wouldnt (not to be pessimistic) but given the architecture and node size has been the same since Broadwell (14nm, "5th gen" CPU's) There's little IPC gain to get other than a boost in single core clocks. 10th gen rumors of the 10 core CPU's have them 25-30% than the 9900k, which has 8 cores....

If intel could get 10nm CPU's out, then we'd start  see some serious gains. 

(A bit off topic I'm sorry)

[mluterman]

Interesting, Tom...thank you for that.  I can put my wallet away now.

[nutral]

I don't think the cpu puts it into the framebuffer, as that would be the software mode, and because cpu's are quite bad at rendering softwaremode is quite slow.

Something weird does happen in inventor, because generally the 3d model is updated by the cpu, and then the model is rendered by the GPU, Because this is done by the GPU, the actual resolution shouldn't matter much, because that is the GPU's task. The CPU just handles the generation and drawcalls to the GPU. This is why high framerates are limited by CPU, because it needs to update information on every frame.

I think something has changed between inventor 2019 and inventor 2020.

Looking at the results with inventor 2020, the score doesn't really change when going to a higher resolution.

inventor 2019: 1080p - 11, 4k - 6

Inventor 2020: 1080p - 11, 4k - 11

The actual usage of the GPU goes up in inventor 2020 because the CPU doesn't seem to be as much of a bottleneck anymore? I am curious, how this would be when running larger models and i'll test this out with GPU utilization.

[leowarren34]

@tom_vierling Architecture has changed since Broadwell quite substantially but not in the same way previously.

Moore's law is way off as yes, as you said, the process has been stuck at 14nm for quite a while. This is also due to Tick-Tick moving to Process(Die Size), Architecture, Optimization.

Cannon Lake will be start of a new cycle at 10nm!

However as you said, the gains are now so tiny that it's no where as fast so upgrades can be slower and better for your wallet. I still run an i7-7800X and it still does fine. Gets between 12-13 in InvBench

[tom_vierling]

Oh for sure, Between Broadwell and Coffeelake, they've made big gains in overall performance due to optimizations and overclocking capabilities (my 4790k could never hit 5ghz as easy as today's chips, it struggles to go past 4.6 without heavy voltages!) but from my memory, 10nm for Desktop applications isnt expected until 2021 which is unfortunate. Because of this I dont think we will see scores break 16 for a while. There's a small chance Ryzen 4000 series (Zen 3) with their 7nm+ architecture passes current single core IPC, which would make up for the lower overall clock speeds Ryzen cant achieve, which might net a high scoring Inventor result (13+). Unless we already have a Ryzen 3000 score that high? Personally haven't been keeping track of all the scores coming in, just been following for updates. 

[Mario-Villada]

---

@RichM1 wrote:
We went with I9-9900k and GTX 1070 Ti a year ago on 8 systems and we never have crashes. Maybe once every 1 to 2 months if that.

---

Would you share your parts? specially which motherboard you used as I am suspecting this might be my issue. We use ASUS PRIME Z720-AR and MSI Z370 PC PRO. the MSI one is giving me the most grieve.