CA Stress Testing Results

[Chrisb222]

These times are until the operation and the displayed view were both fully completed. Several seconds elapsed between the operation completion and the view update.

 

Open standard camera, :36

Drag wall surface up, :46

Undo above, :44

Build roof planes, 3D, 1:19

Build roof planes, plan, :27

 

Mac Mini (Late 2014)

MacOS High Sierra 10.13.6

2.6 GHz Intel Core i5 processor, 2 cores

8 GB 1600 MHz DDR3 memory

Intel Iris Graphics 1536 MB

Apple ATA HDD 1 TB

(2) Samsung S24D300 1920 x 1080 Displays

[TheKitchenAbode]

Here are the results for my 360 Spectra.

 

Open standard camera = 36 seconds

Drag wall surface up = 41 seconds

Undo above = 42 seconds

Build roof planes in 3D View = 73 seconds

Build Roof Planes in Plan View = 22 seconds

[HumbleChief]

36 minutes ago, TheKitchenAbode said:

Larry, does the laptop have a discrete graphics card or just the Intel HD chip? 

Discrete GTX 960M

[TheKitchenAbode]

Just now, HumbleChief said:

Discrete GTX 960M

Thanks once again Larry.

[TheKitchenAbode]

So far we have 4 systems reporting. As can be seen the relationship between the results on one system are aligned with those on another system. In other words, everyone's undo time was essentially identical to their initial do time. Building the roof planes took about double the time as it took to open the standard camera view. Dragging a wall up took about 15% longer than it took to open the standard camera view.

 

Concerning comparison between each system the differences are highly aligned to the CPU specs of each system. If one was watching Task Manger it can be easily seen that likely 95% or more of the processing time was CPU based, the video card only jumps into action at the very end. Also, if watching core activity it can be seen that the majority of CPU processes are single or lightly threaded and as such high core count CPU's are not likely to fair much better. What appears to be most important here is how well the CPU performs as a result of it's base frequency and boost(turbo) frequency in conjunction with how many cores and for how long it can sustain it's maximum throughput. For most single and lightly threaded operations this is most applicable to the first 4 cores.

[MarkMc]

Here's mine. For each action I reset HWmonitor then took a screen shot after so the table represents that action ONLY. I had resource monitor open, and later when back and redid them with just the disk showing so I could pay more attention. Interesting that Chief did not always show up at all in that, also interesting was that with nothing but the resource monitor open system disk usage spiked quite a it while Chief,when it did show up, did not equal the spikes in system disk usage.

Also I don't trust Chiefs built in timer which may skew the results you are getting. In some cases it was just too fast to read before it went away. In most others it simply did not come close to matching real world results on a stop watch, I mean silly off.

Open Camera-17 sec first time, repeats were 9

Zoom-instant

Drag up-21 heavy disk

undo 12

build roof-36 Heavy disk

zoom out -instant

All operations that took significant time hit the CPU -all cores at or near 100%, all clocks at max turbo. GPU-never fully used. Had fans on full to make sure there was no throttle, though I almost never get any.

 

[TheKitchenAbode]

Here is another stress test model. This one looks at the effect of very high surface counts. The model contains 5,000 BBQ symbols for a total surface count of just under 19 million. The test here is a simple one.

 

1.) Open Plan.

2.) Open Camera View, Record Time to Display.

 

The results may surprise you.

 

This one should be of greatest interest to those with high core count processors as the processing here is fully threaded.

 

My 4 core/8 thread 6700K displays in 34 seconds.

 

The rule in respect to core/thread performance is in general, to reduce the time by half you need double the cores. Theoretically to drop my 38 seconds to 19 seconds will require 16 cores that can run at my 4.1 Ghz rate. To get this down to 8 seconds would require 32 cores, 4 seconds 64 cores, 2 seconds 128 cores and to 1 second 256 cores.

 

BBQ Parade.plan

 

[TheKitchenAbode]

9 minutes ago, MarkMc said:

Here's mine. For each action I reset HWmonitor then took a screen shot after so the table represents that action ONLY. I had resource monitor open, and later when back and redid them with just the disk showing so I could pay more attention. Interesting that Chief did not always show up at all in that, also interesting was that with nothing but the resource monitor open system disk usage spiked quite a it while Chief,when it did show up, did not equal the spikes in system disk usage.

Also I don't trust Chiefs built in timer which may skew the results you are getting. In some cases it was just too fast to read before it went away. In most others it simply did not come close to matching real world results on a stop watch, I mean silly off.

Open Camera-17 sec first time, repeats were 9

Zoom-instant

Drag up-21 heavy disk

undo 12

build roof-36 Heavy disk

zoom out -instant

All operations that took significant time hit the CPU -all cores at or near 100%, all clocks at max turbo. GPU-never fully used. Had fans on full to make sure there was no throttle, though I almost never get any.

 

 

Thanks Mark. As you have a 6700K like mine I see that the numbers are very similar.

 

Not sure why you are seeing a lot of disk activity. There will always be some as other background services can be accessing the disk. I just redid the camera open view with the disk monitor opened and did not see any activity that was differed from idle. You have 32GB of ram there should be no reason for it to be using the disk. Have you by any chance manually changed the memory allocation versus just letting Windows manage it.

 

As far as the clock goes the one in Windows Task Manger is very accurate, certainly for our intended purpose. 

[MarkMc]

2 minutes ago, TheKitchenAbode said:

Not sure why you are seeing a lot of disk activity. There will always be some as other background services can be accessing the disk. I just redid the camera open view with the disk monitor opened and did not see any activity that was differed from idle. You have 32GB of ram there should be no reason for it to be using the disk. Have you by any chance manually changed the memory allocation versus just letting Windows manage it.

No I have not changed settings. Thing is that Chief doesn't show a spike but system goes up first at 125,000 per second, as much as 800,000 per second. I have next to nothing running in the background ever. In this case was Resource monitor and HWmontior. Nothing using net, nothing else could write to disc.

For the BBQ I only tried once, think I got 36 seconds and disk kicked up. Similar results for CPU/GPU as before except temp went up  this time on the CPU- rarely see it that high.

[TheKitchenAbode]

1 minute ago, MarkMc said:

No I have not changed settings. Thing is that Chief doesn't show a spike but system goes up first at 125,000 per second, as much as 800,000 per second. I have next to nothing running in the background ever. In this case was Resource monitor and HWmontior. Nothing using net, nothing else could write to disc.

For the BBQ I only tried once, think I got 36 seconds and disk kicked up. Similar results for CPU/GPU as before except temp went up  this time on the CPU- rarely see it that high.

 I don't think that is anything to be concerned about. If I look at the top graph the scale is only 100KB, that's nothing. If you use Windows task manager it will show two graphs the top one at 100% scale and the underlying one scales according to transfer rates(zooms in). The top one is more important. Also the monitoring program you are using is likely causing some activity if you have asked it to record the results.

[TheKitchenAbode]

8 minutes ago, MarkMc said:

For the BBQ I only tried once, think I got 36 seconds

 That's essentially the same as mine which is what I would expect as we have the same CPU. Not bad for 19 million surfaces.

[Alaskan_Son]

38 minutes ago, TheKitchenAbode said:

Here is another stress test model. This one looks at the effect of very high surface counts. The model contains 5,000 BBQ symbols for a total surface count of just under 19 million. The test here is a simple one.

 

1.) Open Plan.

2.) Open Camera View, Record Time to Display.

 

The results may surprise you.

 

This one should be of greatest interest to those with high core count processors as the processing here is fully threaded.

 

My 4 core/8 thread 6700K displays in 34 seconds.

 

The rule in respect to core/thread performance is in general, to reduce the time by half you need double the cores. Theoretically to drop my 38 seconds to 19 seconds will require 16 cores that can run at my 4.1 Ghz rate. To get this down to 8 seconds would require 32 cores, 4 seconds 64 cores, 2 seconds 128 cores and to 1 second 256 cores.

 

BBQ Parade.plan

 

 

20 seconds at 4.6 GHz

[TheKitchenAbode]

Just now, Alaskan_Son said:

 

20 seconds at 4.6 GHz

 That's right in line with my prediction, especially if you account for the extra GHz over mine.

[TheKitchenAbode]

6 minutes ago, TheKitchenAbode said:

 That's right in line with my prediction, especially if you account for the extra GHz over mine.

 

Just to put a perspective on this, especially concerning the I7 7900K, though it is not hyperthreaded those 8 cores are 100% real cores. Throughput is higher on real cores than the hyper threaded ones. My 6700K only has 4 real cores, the other 4 are what they call hyper threaded.

[Alaskan_Son]

57 minutes ago, MarkMc said:

GPU-never fully used.

 

Try switching to line drawing rendering technique and spinning the model around a bit.  My GPU reaches as high as 80% usage. 

[Alaskan_Son]

Try taking a front elevation view of the BBQ model and do a CAD Detail From View and you'll see how much memory can come into play too. 

[TheKitchenAbode]

1 minute ago, Alaskan_Son said:

 

Try switching to line drawing rendering technique and spinning the model around a bit.  My GPU reaches as high as 80% usage. 

Mine gets almost maxes out on the line drawing but say vector view only cranks my GPU to slightly over 40%.

[TheKitchenAbode]

7 minutes ago, Alaskan_Son said:

Try taking a front elevation view of the BBQ model and do a CAD Detail From View and you'll see how much memory can come into play too. 

 

Of much greater concern is the CPU processing time, though my 16GB is pushed it's CPU processing that's making this such a long process. Not sure if this is the best test to evaluate CAD view generation, might be more relevant in the house test program.

[Alaskan_Son]

5 hours ago, TheKitchenAbode said:

1.) Open Standard Camera View = 18 seconds.

2.) Drag Wall Surface up = 22 seconds.

3.) Undo Drag Wall Surface up = 22 seconds.

4.) Build Roof Planes = 35 seconds. 

 

1.) Open Standard Camera View = 13 seconds.

2.) Drag Wall Surface up = 18 seconds.

3.) Undo Drag Wall Surface up = 15 seconds.

4.) Build Roof Planes = 28 seconds.

[Alaskan_Son]

11 minutes ago, TheKitchenAbode said:

Mine gets almost maxes out on the line drawing but say vector view only cranks my GPU to slightly over 40%. 

 

Sometimes its even the little things.  Dragging out a marquee for example ramps my GPU usage up to 40%.

[TheKitchenAbode]

1 minute ago, Alaskan_Son said:

 

1.) Open Standard Camera View = 13 seconds.

2.) Drag Wall Surface up = 18 seconds.

3.) Undo Drag Wall Surface up = 15 seconds.

4.) Build Roof Planes = 28 seconds.

 Thanks Michael.

 

I just did a CAD detail from view in this house model, did it in about 2 seconds, additional memory usage was minimal, maybe an extra 1.5 GB

[TheKitchenAbode]

2 minutes ago, Alaskan_Son said:

 

Sometimes its even the little things.  Dragging out a marquee for example ramps my GPU usage up to 40%.

 

Definitely but 40% GPU demand is nothing to worry about, usually things will only start to stutter when the GPU is near maxed out.

 

One thing worth trying if you have the house test up and running. Open up an elevation and try to dimension something, say the width of the house. CA really struggles to find the first dimension location. What's strange is that once it does you can grab a node and it has no problem finding the next dimension point. 

[TheKitchenAbode]

1 minute ago, TheKitchenAbode said:

One thing worth trying if you have the house test up and running. Open up an elevation and try to dimension something, say the width of the house. CA really struggles to find the first dimension location. What's strange is that once it does you can grab a node and it has no problem finding the next dimension point. 

 Just and additional comment, in a CAD detail there is no issue, just in a standard elevation view.

[MarkMc]

54 minutes ago, TheKitchenAbode said:

I don't think that is anything to be concerned about. If I look at the top graph the scale is only 100KB, that's nothing. If you use Windows task manager it will show two graphs the top one at 100% scale and the underlying one scales according to transfer rates(zooms in). The top one is more important. Also the monitoring program you are using is likely causing some activity if you have asked it to record the results.

Not concerned, just a fact to consider. Absolutely impossible that the monitoring is writing 800,000 kbs for 10-20 seconds. Writing to the Undo file OTOH makes sense, since it's system usage. 

[TheKitchenAbode]

Just now, MarkMc said:

Not concerned, just a fact to consider. Absolutely impossible that the monitoring is writing 800,000 kbs for 10-20 seconds. Writing to the Undo file OTOH makes sense, since it's system usage. 

 

It's unusual as I don't see anything like that type of activity on my system nor my Spectra. The only time is every once in a while I will blow the 8GB of memory on the Spectra and it will resort to disk file swapping.