Performance test results

3dcad · November 2016

I thought it might be helpful for people looking for new hardware if they see results with different sets.
Take test here: https://cad.onshape.com/check and post your result.

Image: https://us.v-cdn.net/5022071/uploads/editor/bb/wx87p42mo34f.png

Not bad for couple years old custom pc with about $1500 parts. A lot better result than $2500 brand new Thinkpad..

brian_jordan · November 2016

No surprise that I get a bit frustrated with Onshape... seems it might not all be down to a slow rural broadband connection.

Image: http://www.bjordan.org.uk/onshape/perf.jpg

Maybe a Black Friday present to myself and this i5 Windows 10 Dell is in order?

NeilCooke · November 2016

You may find that repeated tests give slightly different results and some browsers (like Safari on Mac) will report higher numbers. Remember it is only an indication, but it looks like Rami's got a nice machine

bradley_sauln · November 2016

To level set between the two results so far. Here is a screenshot of the results and set up of the MacBook Pro I give demos on through the Chrome browser. @3dcad, 362+ million triangles, wow!

Image: https://us.v-cdn.net/5022071/uploads/editor/gg/qdjrqxnmuqjh.png

emagdalenaC2i · November 2016

Dell Precision M6800 2.5 years old. Not too bad ;-)

Image: https://us.v-cdn.net/5022071/uploads/editor/9z/p16zbu1dnzxa.png

matthew_menard · November 2016

I don't know if it was a fluke, but the compatibility check page crashed Firefox on me when I tried it yesterday. Here's my results from using a Thinkpad P70 with a Quadro 5000m graphics card.

Image: https://us.v-cdn.net/5022071/uploads/editor/xx/jabaeynf2ii3.png

I also got two wildly different results when I used it right now. First run had .2 million measured triangles per second and 100 something million measured lines per second. After refreshing, I got the above results.

Is there a point of diminishing returns for local graphics power using Onshape? 3dcad's results are fairly close to mine with a 4 generation old gaming card. These results could be very important for anybody considering using Onshape for enterprise use. If you'll get equal results out of a mid range vs high end computer, there could be some considerable cost savings without sacrificing performance. Also, any word on how CPU plays into performance vs GPU?

edward_petrillo · November 2016

I'm almost sorry I checked my 6-month old HP Chromebook: It's 2 orders of magnitude below what's been posted above, yet it suffices for the majority of my work. So how do these numbers relate to what a user actually sees? Does a 100-fold difference in these parameters mean a workspace opens 100 times faster? 10 times faster? Ditto for pan, zoom, rotate.

I'd like to see a video of a "high performance" versus "low performance" workstation doing the same task, so I can interpret these numbers.

matthew_menard · November 2016

edward_petrillo said:

I'm almost sorry I checked my 6-month old HP Chromebook: It's 2 orders of magnitude below what's been posted above, yet it suffices for the majority of my work. So how do these numbers relate to what a user actually sees? Does a 100-fold difference in these parameters mean a workspace opens 100 times faster? 10 times faster? Ditto for pan, zoom, rotate.

I'd like to see a video of a "high performance" versus "low performance" workstation doing the same task, so I can interpret these numbers.

Does the Chromebook use the Android Onshape App? I've got a really cheap Android / Windows 10 dual boot tablet. The Android App runs very well on it, but using Onshape in a browser in Windows 10 is pretty laggy.

Is there any way to submit these performance test results to get a better aggregate look at hardware performance for Onshape? It would be interesting to see what the best cost:performance hardware is.

emagdalenaC2i · November 2016

There are several factors to improve the performance. It is better with Firefox than with Google Chrome, but I have notice that the performance is significantly better with Firefox 48.0 than with the last release Firefox 50.0

Also, I don't know why show the message "Your browser does not disclose this information" instead of my NVIDIA information.

3dcad · November 2016

@emagdalenaC2C
I got better result with Chrome than FF. And also overall performance actually using Onshape is better with Chrome for me.
I suppose it's the complete setup that counts rather than individual part of it..

emagdalenaC2i · November 2016

@3dcad as @NeilCooke mention at the end of the article "BROWSER WARS: IS YOUR BROWSER TOUGH ENOUGH?"
"At the end of the day, it’s your choice and Onshape gives you plenty"
I use Chrome and Firefox. Now we have some metrics to improve the performance a little more

michał_1 · November 2016

That test might be misleading. Here are 6 resaults, done one after another (F5):

michael3424 · November 2016

Here's mine, from an HP Z400 workstation with a Xeon 3.06 GHZ processor and 18 GB RAM. Looks like my expensive Quadro 600 graphics card isn't doing much for me. Or maybe its my Comcast ISP.

billy2 · November 2016

my .9kg (2 lbs) mac:

Image: https://us.v-cdn.net/5022071/uploads/editor/si/qbes0ivt3wrj.png

I guess it shows what I've known all along, buy nvidea. Why macs now use intel graphics chip I have no idea, kinda of ruins the pleasure of owning a mac.

matthew_menard · November 2016

Would it be possible to add connection speed and ping to the results page? Will a poor connection affect the performance test results?

Evan · November 2016

The performance check works by giving the browser and GPU work to do, then measuring the framerate at which it can do that work. Browsers limit the maximum framerate for a page to 60 frames per second. So the test searches for your system's limits by increasing the workload until the framerate starts to drop.

A performance measurement that ends up in the “grey” zone does not necessarily mean that you will experience slow performance. A system that has comparatively low performance may run just fine with models of moderate size. As a design grows in complexity, the amount of geometry tends to grow with it. It is for these larger models where you will see benefits of a more powerful GPU, like faster view manipulation and selection.

For this particular performance test, internet connection speed does not come into play. The data are loaded and tested locally on your machine to gauge system hardware performance. However, in general connection speed is a factor for things like model load times.

The accuracy of the performance measurements are subject to the general load on your system, and to some particulars of how the browser runtime interacts with our software. For best results when running the check:

Cease activity in other applications that may be putting load on the CPU/GPU
Make sure the browser is maximized, and the browser and the browser tab are in view for the duration of the test

brian_jordan · November 2016

I was unimpressed by the performance of my graphics card compared with other results being posted here (see the second post in this thread) so I took a Black Friday punt, bought and installed a new card...
http://www.bjordan.org.uk/onshape/perf2.png
Not only do the numbers look much better but, subjectively, Onshape feels crisper particularly when I look at larger models.

PeteYodis · November 2016

If people are looking to spend some money on an upgrade this holiday season and running a desktop, I've got my eye on an nVidia GTX 1050 card that should have a nice bang for the buck for updating one of my desktop rigs. Thought I'd pass along a good deal on sub $200 card that should really perform well. One of the nice things about Onshape, is that you don't need overpriced/underwhelming workstation class graphics cards.

https://www.amazon.com/EVGA-GeForce-GAMING-Support-04G-P4-6253-KR/dp/B01MF7EQJZ/ref=sr_1_1?s=pc&rps=1&ie=UTF8&qid=1480292666&sr=1-1&keywords=gtx+1050&refinements=p_85:2470955011

3dcad · November 2016

@PeteYodis
Can you post your results with that..?

PeteYodis · November 2016

@3dcad I will sometime around Christmas

adam_mercier · November 2016

Z600 with 2 X5570 24gb ram & GTX 970 4GB
Setup price with second hand Z600 about 800€ 1 year ago
My T430 with HD4000 is pretty bad

Image: https://us.v-cdn.net/5022071/uploads/editor/xu/av8dqu3rodgl.png

BLJ_ENGINEERING · December 2016

ThinkPad W541
i7-4910MQ @2.9GHz
32 GB
Win 7...

Nvidia settings? only ever played NES 8 bit so...

3dcad said: "..A lot better result than $2500 brand new Thinkpad.."

Oh it was WAY over $2500, but it is certified for SW. Yay.

Image: https://us.v-cdn.net/5022071/uploads/editor/ym/11u90oz4ky1a.png

3dcad · December 2016

Thinkpad X1 ultrabook Yoga

Image: https://us.v-cdn.net/5022071/uploads/editor/9f/vt7fj5fjssjn.png

BLJ_ENGINEERING · December 2016

Did some more testing and my results vary a lot. Maybe the test is a bit short? Feels like the GPU needs to "ramp up". Guess it would still be accurate for normal CAD-operations but maybe not for rendering etc?

It really doesn't matter for me because Onshape works smoothly Even the integrated graphics works ok.
Only tested on my integrated 3K screen. Read that an external screen should/ could force the GPU to work better or at least the entire time.

I'd really appreciate if someone could give me some hints on what settings to use for this NVIDIA K2100M and Onshape (Chrome). Maybe some do's and don'ts?

Image: https://us.v-cdn.net/5022071/uploads/editor/9l/5dfv5nuzcojr.png

john_f_carr · December 2016

Firesale said:

Did some more testing and my results vary a lot. Maybe the test is a bit short? Feels like the GPU needs to "ramp up". Guess it would still be accurate for normal CAD-operations but maybe not for rendering etc?

In normal use -- not sure about the performance test -- Onshape graphics adapt by turning off some rendering features for slow devices. If you load a complex model and start dragging it around you may notice frame rate increase after a few seconds. I don't notice any difference in graphics quality but our graphics developers do. (And they probably don't notice that the chickens' legs in Chicken Run bend the wrong way, but I do.)

bill_daniels · December 2016

3dcad said:

I thought it might be helpful for people looking for new hardware if they see results with different sets.
Take test here: https://cad.onshape.com/check and post your result.

Not bad for couple years old custom pc with about $1500 parts. A lot better result than $2500 brand new Thinkpad..

bill_daniels · December 2016

Huh? Not compatible? It seems to work. Is there something I need to do?

Your browser is not fully compatible with Onshape

Browser

User agent

Mozilla/5.0 (X11; Ubuntu; Linux x86_64; rv:50.0) Gecko/20100101 Firefox/50.0

Performance check

Overall performance*

Measured triangles per second

362.5 million+

Measured lines per second

30.1 million

*For information on graphics performance, visit: Graphics Performance Recommendations

Onshape server region

Region

us-west-2

Report

Download report

Notifications (0 new, 0 total)

delete all mark all as read

BLJ_ENGINEERING · December 2016

@john_f_carr Thank you for that info. Always felt it was good practice to do that. Now, wonder if some elevators actually DO go faster if you hit that button multiple times..

Tony_C_ · December 2016

I guess this would be a good way to see which browser performs the best.

john_f_carr · December 2016

Just to show you what a slow machine looks like. This is an under-$100 Chromebit. It's definitely slower rendering graphics than a machine with a high end graphics card, but you can still run Onshape.

Image: https://us.v-cdn.net/5022071/uploads/editor/tx/jxabnvjd8dv0.png

PeteYodis · December 2016

@3dcad

I recently updated my desktop computer from an nVidia Quadro 600 to an nVidia GTX 1050 Ti. I purchased the GTX card for about $150. Attached are my results of running 7 tests for each card in Chrome with only the test tab open. I did each test after a fresh re-boot just in case... My setup was a Windows 10 (64 bit) machine that I had used for consulting design work with another MCAD program. 8GB Ram.

The GTX 1050 Ti processes on average 7.44 times more triangles per second and 12.38 more lines per second.

And for anyone interested... here is the link to the item I purchased: https://www.amazon.com/gp/product/B01MF7EQJZ/ref=oh_aui_detailpage_o02_s00?ie=UTF8&psc=1

Performance test results

Comments