CPU Specs vs Keyshot Performance Estimation

Started by jiyang1018, April 05, 2013, 11:21:31 PM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

jiyang1018

1. Keyshot does not utilize GPU power to render.
2. Keyshot benefits from multiple CPU cores.
3. Keyshot benefits from hyperthreading.
4. Keyshot benefits from high clock speed.
If all 4 above are true, there should be a logical explanation of what type of performance to expect on curtain CPU specs.
With all these in mind, I did a rough table with some data collected from this forum.

If we assume CPU computing power = # of cores * clock speed of each core, and if Keyshot performance (fps of camera.bip) is linear to CPU computing power, fps/cpu computing power should be a fixed value.

As shown in the table below, in Keyshot 4, this value is roughly 1.7, 3 sets of data show 1.6666, 1 set of date shows 2.0. In Keyshot 3.3, this value varies from 1.575 to 2.3. In Keyshot 3.2, this value is 1.45 to 1.6.
I would guess this value should be around 1.7 for Keyshot 4 on most machines. So if you are choosing computers or building computers with Keyshot rendering performance in mind, just simply (number hyperthreads) * (clock speed) *1.7, and result a rough fps you can get in Keyshot.

There are sill variables influencing this value. For example, in my own test, if I close all hardware monitoring software including CPU-Z, my fps increased from 93 to 94.4. That is rough 1.5% increase/variation. Not all people run this benchmark with no other program running at foreground, not mentioning programs running at background. Workstaion with ECC memory may have positive impact on Keyshot performance(uncertain).

Let me know what you think.

NDenekamp

Interesting observation, I was wondering myself what the correlation would be!

I recently measured my machines performance, and it came out around the 47 fps mark.

I'm running 1 quad core i7 at 2.3 Ghz (boost to 3.3), so a total of 8 threads at 3.3 = 26.4 computing powers  ;)

Times 1.7 give an approximate  44.9 fps.

Pretty close, but currently I am test driving KeyShot 4.1 Beta, so this might mean we will see a nice 4.5% increase (or performance factor of ~ 1.78)


N

jiyang1018

I updated some new results to my analysis.

Some results seem to be off a lot from my estimation.

For dejot's result, 3930K by default has 3.8GHz boost clock. Put 3.8 in the table the performance ratio becomes 1.73 which makes more sense comparing with other results.

For peri's result, 4770K, 4 cores, 8 threads @ 4.5GHz with 86.4fps seems insanely good. In my Keyshot 5 result, 3930K 6 cores, 12 threads @ 4.6GHz, when I first unpaused it, it went up to 95+ but became stable around 89 - 90 fps.

Architecture or generation definitely affects result. From previous data, IVE bridge renders slightly faster than Sandy bridge. I would not be surprised if Haswell runs slight faster than IVE bridge. That is generation difference/gain. Xeon processors generally renders faster than i7 processors at the same core number and clock speed. That is architecture difference/gain.

With more data and with more controlled variables, this estimation may become more accurate, and may become a guide of building/buying a computer for Keyshot rendering.

jiyang1018

I have also uploaded the excel file I put together to OneDrive. Feel free to add your result, but keep other results as well.

https://onedrive.live.com/redir?resid=D647C2F1F807AA6A!1729&authkey=!AJ_fbRYMLNSmKIo&ithint=file%2c.xlsx

Rex

Here's my info to add:

- 2x Xeon X5650
- 12 cores/24 threads
- 2.67 GHz
- 100 FPS KS5
- 90 FPS KS4

Arn

Quote from: jiyang1018 on July 13, 2014, 02:36:20 PM
Xeon processors generally renders faster than i7 processors at the same core number and clock speed.
As far as I know there is no technical reason or difference between the two that would suggest this, so if that happens in a large sample group it is an interesting result :)

jiyang1018

Quote from: Arn on July 18, 2014, 08:30:31 PM
Quote from: jiyang1018 on July 13, 2014, 02:36:20 PM
Xeon processors generally renders faster than i7 processors at the same core number and clock speed.
As far as I know there is no technical reason or difference between the two that would suggest this, so if that happens in a large sample group it is an interesting result :)

All I did was collecting and analyzing data samples. Analysis based on smaller quantity of samples usually has greater fluctuation. From the data I collected from other users, at least so far, the is the conclusion, Xeon is faster than i7. I do not have an explanation. My guess is L3 cache.

joseph

#7
Hi,

I made a mistake for my AMD processor, it is FX6300 running at 3.77 Ghz. I have updated my PC with a FX6350 at standard multiplier of 19.5x - 21.5x for turbo mode, Kingston Hyper Fury DDR3 1866 8GB x 2 clocked at DDR3 1600 with bus set to 214 running 1711.8. The net cpu speed range now is 4.173 GHZ for all cores and turbo a half of the cores to 4.60 GHZ. Also replaced the GPU to a Sapphire R7 260X 2 GB video card. I haven't bench mark it yet but the peak fps is 96-114 fps and drops down to 35 fps. Don't have an idea what matters most or the real fps figure. If I change the view point opposite current horizontal position the FPS is higher at a constant rate. Sorry for being late in correcting my error. Haven't checked for quite a time. Regards

satish gobind

#8
Hi All,

Currently I have AMD Fx 6300 processor,
                        ASRock 970 Extreme 4 Motherboard  with 16GB Ram

My rendering is damn slow I couldn't get to see the changes instantly every time I do some changes or zoom it takes time to render up. So I have planned to upgrade my CPU to the below changes Specs on considering the fact that Keyshot performance depends on the CPU cores and threads i.e. CPU power.
                          Intel i7 4790K processor
                          Asus SABERTOOTH Z97 MARK 2
                          16GB RAM
If I upgraded to the above specs will my rendering speed increase ? As only this specs comes within my budget.

Please help me here folks !!

Rex

Quote from: satish gobind on February 04, 2015, 04:46:29 AM
Hi All,

Currently I have AMD Fx 6300 processor,
                        ASRock 970 Extreme 4 Motherboard  with 16GB Ram

My rendering is damn slow I couldn't get to see the changes instantly every time I do some changes or zoom it takes time to render up. So I have planned to upgrade my CPU to the below changes Specs on considering the fact that Keyshot performance depends on the CPU cores and threads i.e. CPU power.
                          Intel i7 4790K processor
                          Asus SABERTOOTH Z97 MARK 2
                          16GB RAM
If I upgraded to the above specs will my rendering speed increase ? As only this specs comes within my budget.

Please help me here folks !!

Since the i7 chip supports hyperthreading you will have a total of 8 threads, compared to 6 for the AMD. So yes, I would expect your rendering speed to increase with the i7 4790k

joseph

#10
Update the performance sheet with wattage column. - Regards

Vasily

Why wouldn't you add flops-based performance measurement? That would be pretty unified for raw performance comparison.

mertmg

#12
Hi all,

I have just started to use KeyShot 5 with Creo. getting there i like it.

but i don`t like my benchmark results.  I`ve same Cpu like Rex but mine is really really so bad.

- 2x Xeon X5650
- 12 cores/24 threads
- 2.67 GHz
-36gb RAM
-QUADRO 4000
-Win7
- 14.4 FPS KS5.3
i`ve re-installed keyshot and it is around 100fps now
what is this?? can anyone help me how to improve it! :'( :'( :'(
-

nutsy

Is it normal that my CPU usage is at 100% Does that mean I should get a better CPU?

guest84672

Yes it is always 100% once you imported a model.