Login | Register 
FEATURES
APPLICATIONS
DOWNLOADS
EDUCATION
BLOG
WIKI
FORUM
STORE

GPU memory to system memory TOP2Chop benchmark.

Discuss graphics hardware, computer hardware, and input/output devices to use with TouchDesigner

GPU memory to system memory TOP2Chop benchmark.

Postby barakooda » Tue May 14, 2019 6:27 am

Hi there dear community,

I`m sharing file that takes 1280 X 1280 texture and convert it to chops.
please share results of cook time , cpu and gpu models.
hopefully it will give some estimation and benchmark for GPU memory to system memory on different hardware ( intel vs AMD vs NVIDIA \ quadro vs rtc\grforce ) .
give it 10 second and be on the same window for that duration.
--------------------------
I use this software for hardware performance test and details.
https://www.passmark.com/download/pt_download.htm
--------------------------

GPU : NVIDIA GeForce GTX 1070 (laptop)
CPU : Intel Core i7-7700HQ @ 2.80GHz (laptop)
RAM: 32GB 2XSamsung M471A2K43CB1-CRC
Drive:Samsung SSD 960 EVO 500GB
OS: Win 10 1709 build 16229.785
TD: 2019.14650

top_chop_benchmark.toe result:

1280X1280
Latency: ~58ms


top_chop_benchmark.MalcolmFix.toe result:

1280X1280
Latency: ~58ms


------------------------

GPU : NVIDIA GeForce GTX 1070 (laptop)
CPU : Intel Core i7-8750H
RAM: 32GB 2X SK Hynix HMA82GS6CJR8N-VK, wk/yr: 13/2018
Drive:Samsung SSD sm961 1tb
OS: Win 10 1709 build 16229.492
TD: 2019.14650

top_chop_benchmark.toe result:

1280X1280
Latency: ~54ms

--------------

GPU : NVIDIA GeForce rtx 2080 TI
CPU : Intel Core i7-8700k
RAM: 64gb 4XCrucial Technology CT16G4DFD8266.C16FD1
Drive:Samsung SSD 960 pro 1GB
OS: Win 10 1803 build 17134
TD: 2019.14650

top_chop_benchmark.toe result:

1280X1280
Latency: ~46ms




thanks
Barak.
Attachments
top_chop_benchmark.MalcolmFix.toe
Added Malcolm version to the same test.
avoiding generating noise every frame on the GPU.
(3.93 KiB) Downloaded 54 times
top_chop_benchmark.toe
(3.93 KiB) Downloaded 58 times
Last edited by barakooda on Wed May 15, 2019 12:21 am, edited 10 times in total.
User avatar
barakooda
 
Posts: 651
Joined: Fri Nov 27, 2009 5:08 pm

Re: GPU memory to system memory TOP2Chop benchmark.

Postby lucesintetica » Tue May 14, 2019 6:40 am

GPU: NVIDIA GeForce RTX 2070 (desktop)
CPU: Intel I7 - 2600k (desktop)

Latency: ~240
lucesintetica
 
Posts: 40
Joined: Mon Nov 19, 2012 5:21 am

Re: GPU memory to system memory TOP2Chop benchmark.

Postby Moohr » Tue May 14, 2019 7:04 am

GPU: nVidia RTX 2080
CPU: Intel i5 9600K
RAM: 16GB
Drive: nvme SSD
OS: Win 10 1809 build 17763.475
TD: 2019.14650

2048X2048 test:
Latency: ~135ms

1280x1280 test:
Latency: ~54ms

MalcomFix test:
Latency: ~54ms
Last edited by Moohr on Wed May 15, 2019 4:40 am, edited 2 times in total.
User avatar
Moohr
 
Posts: 3
Joined: Tue Feb 17, 2015 2:11 pm

Re: GPU memory to system memory TOP2Chop benchmark.

Postby ben » Tue May 14, 2019 8:24 am

This is a great idea, but you might get a larger set of results if you drop the test to 1280x1280 so anyone can test on any machine easily using Non-Commercial. It's still large enough to get a good idea of relative performance.
User avatar
ben
Staff
 
Posts: 2557
Joined: Thu Nov 08, 2007 6:55 pm
Location: Toronto

Re: GPU memory to system memory TOP2Chop benchmark.

Postby barakooda » Tue May 14, 2019 9:22 am

Your are right,
but maybe there is also bandwidth limitations between cases...
so we will might need different texture sizes cases.

Changing to 1280X1280.
Please guys - on the top ,
update your results.
Last edited by barakooda on Tue May 14, 2019 11:12 am, edited 1 time in total.
User avatar
barakooda
 
Posts: 651
Joined: Fri Nov 27, 2009 5:08 pm

Re: GPU memory to system memory TOP2Chop benchmark.

Postby kedgarvolta » Tue May 14, 2019 11:11 am

GPU: nVidia Quadro P5000
CPU: Intel i7-8700K @ 3.70GHz
RAM: 32GB
Drive: nvme SSD
OS: Win 10 Pro
TD: 2019.14650

Latency: ~52
kedgarvolta
 
Posts: 3
Joined: Thu Aug 11, 2016 9:32 pm

Re: GPU memory to system memory TOP2Chop benchmark.

Postby franzrosati » Tue May 14, 2019 3:02 pm

Alienware r3 13
GPU: nVidia GTX 1060 (optimus)
CPU: Intel i7-6700HQ@ 2.60GHz, 4 core
RAM: 32GB
OS: Win 10 Pro
TD: 2019.14650

2048x2048
Latency: ~150

Custom Build
GPU: nVidia RTX 2080
CPU: Intel i9-9900k@ 3.60GHz, 8 core
RAM: 32GB
OS: Win 10 Pro
TD: 2019.14650

2048x2048
Latency: ~35
franzrosati
 
Posts: 13
Joined: Sun Dec 05, 2010 5:47 pm

Re: GPU memory to system memory TOP2Chop benchmark.

Postby fHainer » Tue May 14, 2019 3:59 pm

Lenovo Yoga X1 2nd

GPU : NVIDIA GeForce GTX 1070 (Desktop in EGPU-Case via TB)
CPU : Intel Core i7-7600 @ 2,8 GHZ
RAM: 8 GB
Drive: Samsung SSD 950 Pro 256 GB
OS: Win 10 1809 build 17763.475
TD: 2019.14650

1280X1280
Latency: ~52ms

2048X2048
Latency: ~123ms
fHainer
 
Posts: 30
Joined: Tue Jul 17, 2012 10:30 am

Re: GPU memory to system memory TOP2Chop benchmark.

Postby malcolm » Tue May 14, 2019 4:21 pm

This test actually doesn't really reflect GPU->CPU time particularly well. The TOP To CHOP node is waiting for multiple things to be able to complete it's cook.
In particular, it will need to wait for the GPU to complete any pending operations it has, including latency for ones that have been queued and not send to the GPU yet, then it needs to wait for the download.
So in this example it's going to need to wait for the GPU to receive any commands that haven't arrived yet, which may have latency, then it needs to calculate the Noise TOP, which can be very expensive, then it can begin the download.
A better test would involve the source TOP not cooking, and instead the TOP To CHOP being forced to cook every frame using a an Execute DAT forcing it to cook(Force=True) every frame.
Overall you're always going to have some latency between when you ask the GPU to do something, and when it actually starts the operation, which will still hide the GPU->CPU bandwidth somewhat.
User avatar
malcolm
Staff
 
Posts: 4400
Joined: Tue Nov 13, 2007 1:11 am

Re: GPU memory to system memory TOP2Chop benchmark.

Postby barakooda » Tue May 14, 2019 5:12 pm

:-)
ok!
I will create another toe file reflect your comment.
thanks Malcolm .

Update:
After creating another version,
I guess its affect will be on weaker
hardware where the time genrating noise on the GPU will
be significant or in higher resolution cases.
User avatar
barakooda
 
Posts: 651
Joined: Fri Nov 27, 2009 5:08 pm

Next

Return to Hardware

Who is online

Users browsing this forum: No registered users and 2 guests