BallBench - Physics Based Benchmark.

V: 0.2

Hello my friends.
We have a new version ready.


Change Log:
- 2 modes:
1. Physics Test - No graphics intensive shaders, textures, lights. No particles. Only physics.




2. Combined Test - Added FX like water and particles, dynamic multiple lights. heavier advanced shaders and new geometries.




- PhysX version upgraded from 2.8.3 to 3.3. Way more efficient. 80%-90% improvement.

- High scores save for each test.

- SLI\CF support.

- 64 BIT.

Share your scores.

Download link:
https://drive.google.com/file/d/0B-iGj1OnRxA4X01jc2dLWm1UUlk/view?usp=sharing


---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
V: 0.1



Hi everybody!
I have developed a new benchmark today in order to check my computer physics capabilities.
I would like to share it with you and see what your machines are capable of.

Its a very simple app that generates balls in a glass container.
each ball checks collisions, velocity and gravity every frame.
more balls in the scene = more physics calculations every frame = frame-rate drop.
when the FPS counter < 10 ball generation is stopped and you get the score.
balls number * total frames = score.

This software is free for use and share.

Please share your scores with us so we can compare our machines.

Have Fun!

Download Link:



My results:

Auto-Select SLI on = 77,640,255



One of my 880m's dedicated to PhysX = 78,637,584



CPU calculate PhysX = 80,127,206

 

Brilliant! Thanks for sharing and Rep +1

Would you mind also posting this in the Alienware 18 benchmark thread? I'm sure a few members will enjoy testing this out.

 

GTX 860m Maxwell: 13,988,130 "Good"

edt: Got about the same setting it to the Intel IGP...

 

750 Ti SC PhysX


4930K @ 4.5GHz PhysX

 

I love this benchmark. My screen capture didn't work with my first run. I did a second run with a mild OC @ +110/330 and my machine abruptly shut off. I'm a little scared to use it more since the program was only using a single GPU with CPU and their should be plenty of power available. I'm wondering if it's somehow drawing too much power?

 

Could be a nice non-boring stress test if done right IMO

 

I really like it. It's quick and mesmerizing. I like how the FPS slowly tank with every ball. Just for now I will wait more input from sa7ina before I use it more.

 

This is great! Thank you for sharing with us. I like it a lot. +1 Rep. :thumbsup:

Below are my results. Notice the improvement with AFR2 versus "NVIDIA Recommended" for SLI on the last two sets. I ran 3 set on "Good" (default) and the last one with "Fantastic" settings, just for good measure. The first was at 4.3GHz and stock GPU clocks with GPU PhysX and the rest at 4.5GHz using CPU for PhysX and a GPU overclock of +157/+500 and maxed out power target. Max temps on CPU and GPU(s) were 79°C and 64°C, respectively, using only HWiNFO64 to run the fans full blast.

Spoiler

4.3GHz; GPU PhysX; Stock GPU clocks; NVIDIA Recommended SLI rendering mode; Graphics Quality Good


4.5GHz; CPU PhysX; GPU +157/+500; NVIDIA Recommended SLI rendering mode; Graphics Quality Good


4.5GHz; CPU PhysX; GPU +157/+500; Alternate Frame Rendering 2 SLI rendering mode; Graphics Quality Good


4.5GHz; CPU PhysX; GPU +157/+500; Alternate Frame Rendering 2 SLI rendering mode; Graphics Quality Fantastic

 

Yeah lots of boob physics would do the trick for me.

 

Lol, if only someone could hack DeadOrAlive assets and turn it into a benchmark.



Judging from the...loft and bounce...I'm pretty sure the guy that wrote those physics hadn't encounter too many breasts in his life :laugh:

 

Mmmm Dead or Alive. Not sure if it's a decent fighting game, but all I've ever known it for is the babes.

 

Yeah it was pretty good. At least back in the days of Dreamcast.

 

OK guys, let's keep it PG-rated... http://forum.notebookreview.com/site-suggestions-help-announcements/109941-forum-rules.html

Spoiler

The NotebookReview.com forums have grown tremendously over the last few years and we thank all of you for being a part of this growing community.

In order to keep the forums running well and serve as a great place to get help with notebooks, please follow these simple rules for the forums here. If you have any questions or problems, please feel free to contact one of the Admins or Moderators.

General Rules:

• While debating and discussion is fine, we will not tolerate rudeness, insulting posts, personal attacks or purposeless inflammatory posts. Our decision is final in these matters.
• Please refrain from posting meaningless threads, one word (or short) nonsense posts, or the such.
• Multiple or repeated posting in order to increase your post count is not allowed.
• Cross posting, the act of posting the same thing across multiple forums, is not allowed. Please just post once in the most appropriate forum.
• Advertising, spamming and trolling is not allowed. This includes using the forum email and Private message system to spam other members.
• We do not allow discussion of anything related to medical help
• Please wait a reasonable amount of time before bumping posts, at least 24 hours.
• We also do not allow posts that are sexual or violent in nature or violate NotebookReview.com community standards.
• Additionally adult/sexual topics and those concerning sexual orientation as used in a derogatory, inflammatory, or insensitive manner are not allowed.
• The NotebookReview.com moderation team shall be the sole arbitrator of what does and what does not violate community standards.
• Discussion of illegal activities such as fraud, software and music piracy and other intellectual property violations are not allowed. There is a zero tolerance policy for illegal activity being discussed.
• Do not post content from private correspondence, such as email or chat, within the forums. Posting any other persons private information such as email, phone number or address could result in banning.
• Discussion of cracking passwords or bypassing security measures is prohibited.
• Each member is allowed one login account. Registering with multiple accounts is not allowed.
• We reserve the right to remove offensive posts without notice.
• If you are going to post non-English on these forums, please also post an English Translation of your post.
• You cannot post any affiliate or referral links, or post anything asking for a referral.
• You cannot post advertisements or notices for contests.
• Foul language isn't necessary to discuss notebooks and help people, so it won't be tolerated. Any attempt to bypass the profanity filters will result in an infraction against the user and that post being edited or removed. Using asterisks or any other symbol to disguise bad language is considered bypassing the filter.
• Hidden links in posts or signatures are assumed to be malicious and members posting such content will be permanently banned, no exceptions.
• Please, no rep begging.
• Do not post EPP codes or anything similar not for public use.
• If you are going to post material that is not your own, properly credit the source(s); failure to do so is considered plagiarism.
• Please do not post huge images. Anything over 500x500 should be thumbnailed or linked only. See our guide for where to host images here and how to resize images here.
• Image replies are not allowed (posting just an image in response to a question)

Any threads/posts of these above types will be deleted.

Also while these rules cover most common situations, they cannot anticipate everything. Consequently we reserve the right to take any actions we deem appropriate to ensure these forums are not disrupted or abused in any way.

We reserve the right to ban or infract anyone who willfully violates the forum rules. Bans will not necessarily be permanent, but enacted based on the seriousness of an offense to these rules.

*snip*

This forum is rated PG - lets keep it that way.

*snip*

 

Disregard.

 

It seems we got a bit carried away. Probably best we not hijack the thread anymore that we have either.

Back to the bouncing balls

 

First thing first.

Glad you liked it and thanks for the Rep point!
you can post it anywhere you like.
i don't think the power draw is a problem for such short period, but indeed it squeezes every possible physics calculation capabilities from your machine.


HTWingNut & n=1
Your low scores are because you ran it in windowed mode what limited your FPS to 60.
do a second run on full screen.

It is not that interesting test but it is doing it's job.
nice work for a 10 hours development...i am happy
and be happy to hear if you have any suggestions.

Thank you my friend glad you liked it.
very interesting scores you have there i need to analyse it...thank you for that too.



This is a very early stage development build (took me 10 hours to develop) and i would like to hear from you my friends any suggestions(even if it's a boobs stress test
10X alot!

 

Is it supposed to limit the FPS to 60 at the start? Seems like that would bias the result quite a bit. Also would be nice to have a screen grab built into the benchmark.

Re-ran 1080p "Good":

GTX 860m Maxwell - GPU PhysX: 14730442
GTX 860m Maxwell - CPU PhysX: 13945226 (i7-4810mq @ 3.2GHz)

edit: I just saw your post sa7ina. I was running in full screen and it was still limited to 60fps.

 

If you run it on windowed mode i could not get it to go beyond 60FPS.
real results only on full screen.
make sure that windowed is unchecked.

 

60 FPS cap when windowed is because DWM is enforcing V-Sync?

 

Window mode is unchecked:

<iframe width='640' height="360" src="//www.youtube.com/embed/TaomE0WTQfE?rel=0" frameborder='0' allowfullscreen></iframe>

 

The Desktop Windows Manager is forcing all windowed apps to the same refresh rate...i guess it will be 120FPS in an 120Hz screen.

Maybe the v-sync is forced by the driver?

 

Using 340.52, installed just recently, did a "clean install". I forced v-sync off in the driver and still same result. It's not even v-sync though, because when the fps starts to drop, the FPS goes down progressively, doesn't cut in half or anything. It's like it's frame limited. I even went into nvidiainspector and forced v-sync off and made sure frame limiter was off. I'm uploading video to show when the fps starts to drop.

Could it be an Optimus thing? Only ones I've seen so far are SLI systems.

EDIT: I just tried it on my desktop with GTX 670 and it also limits the FPS to 60, using same driver. Maybe I'll try another driver and see. Windows 8 on laptop, Windows 7 on desktop.

<iframe width='640' height="360" src="//www.youtube.com/embed/9OBpwwZFt14?rel=0" frameborder='0' allowfullscreen></iframe>

 

This is very odd.
by code the loop starts as soon as possible right after the previous frame rendered to the screen.
i will check it out.
thank you for the QA

 

I had make a profile and force V-sync off in Nvidia control panel . No drama's though :thumbsup:

 

OK
made a few changes and replaced the download link to the new file.
-managed to force v-sync off by default.
-windowed run works on fastest FPS too.

HTWingNut please download and try again.

 

It's Fixed!

GTX 860m Maxwell: 48,281,435
GTX 670 Kepler: 77,140,752

 

No DWM active (windows classic theme), i5 2450M (2.9GHz, EIST and prochot disabled via BIOS), windowed, Win7 x64, GT630M (540M) 2GB DDR3 vsync forced off via nvidia control panel @ driver 340.52.

671MHz core / 900MHz memory
Stock: 1562 balls, 7582 frames
11843084


810MHz core / 1070MHz
OC: 1665 balls (6.18619% increase) 6.19% / 8780 frames (13.64464692% increase) 13.6%
14609920 (18.938064% increase) 18.94%

 

I had the exact same issues as HTWingNut -- full screen and still capped at 60 FPS. Will report back in 9-10 hours after I get off work lol

 

Here is some more interesting data to examine... Screen shots and Excel document are available in the link in case anyone else wants to play with a graph. This is using the version you modified to fix the issue with capped framerate that HTWingNut and n=1 reported. It seems mostly CPU dependent... GPU rendering configuration does have some effect, but it is a small one. It doesn't seem to benefit from brute force on the GPU side of the test. I might run this using the Intel IGFX to find out to what extent the GPU comes into play. If the results are primarily dependent on GPU shader count and computational speed, it would be interesting to see some AMD GCN benchmark results.

Download Data (Link)

 

It's really nice to have a benchmark that begs for a potent CPU for a change... check it out... 26,693,920 on "worthless" Intel HD graphics.

 

IGP HD3000 on i5 2540M / 8GB 1066. Same setting as mentioned above.

1468 balls / 2505 frames
3677340 score


@mr fox, oh wow :s 726% increase when HD3000 -> HD4000


720p "fantastic" setting

 

4930K @ 4.5GHz

750 Ti PhysX:


4930K PhysX:


4930K @ Stock

750 Ti PhysX:


4930K PhysX:


Interestingly 750 Ti performs better when 4930K is underclocked. Using 4930K as PhysX the score is just very slightly lower at stock vs 4.5GHz OC. I ran the bench 3 times each and while the scores vary, the trend is consistent.

So it would seem this bench absolutely ADORES # of cores but not absolute core speed, and GPU doesn't play much of a role. I mean just compare my scores to sa7ina's and keep in mind I'm running a dinky 750 Ti while he's pushing 2x 880M.

 

Seems to run alright in Windowed mode here, i7-4700MQ and iGD HD4600, 800MHz DDR3L CL11.

GPU load drops off quick.

Full screen - 1920x1080 - Good : 2179, 10468, 22.8M

 

Ran the benchmark for reference and comparison.
All tests were performed at 1280x800, full screen.

System
Core 2 Duo 2.4 GHz (P8600)
NVIDIA 320M
4 GB
Windows 7 32-bit.

"Fastest"
7742647

"Fast"
8127405

"Simple"
5233416

"Good"
4955082

"Beautiful"
4297920

"Fantastic"
4057952

Hope this helps. Cheers!

 

Hello my friends!
Didn't had much time to respond but i read it all.

As you all noticed graphically, there is nothing here(2 cheap shaders, and simple ball and cube geometry), it's a physics stress test uses an old version of PhysX running on The CPU.
In the next version i'll use a new version of PhysX with GPU acceleration support for comparison.

In addition i want to develop a graphic test to squeeze our GPU's: a tower building itself endlessly with complicated geometry, a massive use of multiple expensive shaders and shadows, tessellation + SLI/CF support...later i'll figure a way to combine the scores or develop a combined test.

I want to thank you all for posting your scores and specs, it helps a lot...especially Mr.Fox's organized info
After a small analysis of this data i have noticed that the number of balls is limited by the number of draw calls your machine can handle...that's why all of us got a ball count between 2200-2800 slow or fast machine, but the slower ones did it with a lower FPS.
Arondel my friend, i would like to know the ball count you got in each test if you can...10X...looks like slower machines affected by the GPU power more aggressively.(i get the same results on "fastest" as on "fantastic" presets)

Quality Levels Presets for you to know:
Fastest = AA-Off, AF-Off, Texture Quality-Halved, Pixel Light Count-0.
Fantastic = AA-X2, AF-On, Texture Quality-Full, Pixel Light Count-4.
In the next version i will radicalize the differences between the quality levels.(will be more relevant with the graphic test)

I will update in 10 days because tomorrow i am flying to Ukraine for a week of a crazy party without even taking my phone or any other tech device.

Cheers!
Love you all!

 

sa7ina, did you run BallBench with FRAPS on? Because that's what I used to take screenshots, and seems like it took a huge chunk off of performance. Just ran BallBench without FRAPS and I got over 3000 balls! This was with my 4930K @ 4.5GHz and using 750 Ti to run PhysX:

 

With FRAPS ON I get 2353, with it OFF I get 2255. FRAPS makes no difference, actually a little faster. FRAPS shouldn't make a difference. It's lightweight and does nothing but calculate the FPS.

 

*facepalm*

I just realized all the previous runs were done with the computer in "balanced" power mode, and the 4930K was in all likelihood running at stock speed instead of the 4.5GHz OC. I've since switched over to "high performance mode" and it probably forced the 4930K to run full bore, thus the much higher score.

Before switching modes I was wondering was wrong with my computer and why everything seemed to lag and was so much slower than my laptop. Then it struck me and when I checked, the damn CPU was downclocking at every possible moment and didn't even engage turbo for most tasks and ran at 3.4GHz. Even NBR was slower to load lol

 

haha. I've done that before too. Easy to forget if you periodically change it for use on battery.

 

Honestly I didn't even realize power options existed for desktops. I thought it was just "plug in the PSU, hit the on switch, hit the power switch" and that's it LOL.

 

Oh crud, that was desktop? Yeah you can set it, but usually it defaults to high performance. Either way good thing you got it figured out.

 

Fastest



Fantastic



Sent from my HTC One_M8 using Tapatalk

 

A little teaser.
The next version will have:
- New PhysX version (upgraded from 2.8.3 to PhysX 3.3) now can hold twice as much balls (got almost 5000).

- Developed a new mode, so we have 2 modes:
1. Physics Test - same as it was but with the new PhysX version power. (Replacing the Fastest mode and all others)
2. Combined Test - added FX like water, dynamic lights and particles. Heavier shaders and new geometries. (Replacing the Fantastic mode)



Snapshot of the new mode:



4700+ Balls with the new PhysX version:



I will upload the new version in a few days.

 

You need to turn this into a full fledged 3D benchmark.... "Does your gaming computer have the balls to play?"

 

Ohh Yeah!!!

 

This bench is totally balls.

 

Got to get it to work with SLI and we have a new version.
Next: Tower Graphic Test.
Video:

 

Hello my friends.
We have a new version ready.


Change Log:
- 2 modes:
1. Physics Test - No graphics intensive shaders, textures, lights. No particles. Only physics.
2. Combined Test - Added FX like water and particles, dynamic multiple lights. heavier advanced shaders and new geometries.

- PhysX version upgraded from 2.8.3 to 3.3. Way more efficient. 80%-90% improvement.

- High scores save for each test.

- SLI\CF support.

- 64 BIT.


Share your scores.

Download link:
https://drive.google.com/file/d/0B-iGj1OnRxA4X01jc2dLWm1UUlk/view?usp=sharing

 

Ran v0.1 on my work laptop just for fun. i7-4800MQ+Quadro K610M, 1600x900 windowed, simple rendering:



Interestingly, I got a higher score (over 18m) the first time I ran it at 1920x1080 fullscreen with good rendering, but I wasn't able to get the screenshot.

Edit: v0.2 results

 

Teaser for the new graphics test. W.I.P