Planetside Software Forums

Hello everyone. I posted a week or two ago about creating a new TG2 benchmark and I've now completed it. The inspiration for it was three-fold:

First, with the newer quad core and Core i7 CPUs and TG2 now being multithreaded, most older benchmarks simply render too fast to see real differentiation.

Second, none of the other benchmarks I've seen incorporate all the aspects of a TG2 scene, including water and reflectivity, populations, and complex terrain and surfacing.

Third and finally, for a long time now I have had a dream of creating a really good benchmarking website with system tracking, multiple benchmarks, easy benchmark entry, search functions, etc. and recently I was lucky enough to stumble across an already running, in-progress site that I am now contributing to as a partner. The site already includes not only my TG2 benchmark, but benchmarks for many other 3D applications like Maya and 3DS Max. You can even contribute your own benchmark scenes for others to try!

So if you're interested in TG2 benchmarking, or 3D application benchmarking in general, please drop by and check it out. We've got a lot of good basic functionality in there, and are working on a lot more exciting stuff. If you have any feedback or ideas, please let me know. We're aiming to make this the best source for 3D application benchmarks on the web, all in one place.

http://www.3dspeedmachine.com/

Thanks for checking it out!

- Oshyan

Hi Oshyan,

there's no object for the population included in the archive, as far as I can see, but the demo render (which is actually a very pleasant scene) apparently has object populations.
Nice idea with the new benchmark site.

Cheers,
Frank

Yes, they're populations of tweaked grasses (built-in). Sneaky, I know. ;)

By the way this scene is fairly demanding since it is meant to deal with newer Core i7's and beyond. I hope the scene will last a year or more as a relevant benchmark. So on a Core 2 Duo, as you can see from the results, it takes nearly an hour. I realize that's a long time, and for some of you with even slower systems it will be over an hour, but for those with fast Core i7's it is less than 10 minutes, and this just highlights the difference in speeds, which is exactly what is needed.

- Oshyan

Looks good! I'll give it a shot when I get a chance.

ok, I've added my result there.

Me too, once I get my activation link in the email. Who's JavaJones? 53 minutes seems a little long for a core 2 duo 2.4GHz. Mine's a 2.93 and it did it in 21:34.

I think he works for Starbucks?  And maybe Planetside?   ;)

a little bit of feedback from my side:

the website is not exactly easy to use as far as the registration is concerned. also the graphical design doesn't seem optimal on the header and menu. On the homepage, below the header there's a banner ad and the menu item link that says "register" actually line breaks into the ad space (because it doesn't seem to fit in the space of the header menu). Took me some time to realize to find the registration link because of it.

Hope you'll be able to have that improved.
Cheers,
Frank

Quote from: Oshyan on February 28, 2009, 03:04:54 PM
Hello everyone. I posted a week or two ago about creating a new TG2 benchmark and I've now completed it. The inspiration for it was three-fold:

First, with the newer quad core and Core i7 CPUs and TG2 now being multithreaded, most older benchmarks simply render too fast to see real differentiation.

Second, none of the other benchmarks I've seen incorporate all the aspects of a TG2 scene, including water and reflectivity, populations, and complex terrain and surfacing.

Third and finally, for a long time now I have had a dream of creating a really good benchmarking website with system tracking, multiple benchmarks, easy benchmark entry, search functions, etc. and recently I was lucky enough to stumble across an already running, in-progress site that I am now contributing to as a partner. The site already includes not only my TG2 benchmark, but benchmarks for many other 3D applications like Maya and 3DS Max. You can even contribute your own benchmark scenes for others to try!

So if you're interested in TG2 benchmarking, or 3D application benchmarking in general, please drop by and check it out. We've got a lot of good basic functionality in there, and are working on a lot more exciting stuff. If you have any feedback or ideas, please let me know. We're aiming to make this the best source for 3D application benchmarks on the web, all in one place.

http://www.3dspeedmachine.com/

Thanks for checking it out!

- Oshyan

Hi Oshyan,

Ok, I went there and resistered. Now, how to I perform the test. I clicked on the image and the picture uploaded. How do you test the render/

Bob

@choronr - you need to click the text "Download Scene" under the test picture to download the test scene in a zip file.

BTW, I've got a problem with the test render scene. As soon as I open it I get a number of warning messages (see attached). I'm using Build 1.10.23.1. Anyone else had the same happen or know what the problem is?

Regards,
Paul

You can ignore that, because the error messages relate to a feature that doesn't exist in your version of TG2, yet. However, this feature is not used in the test scene anyway. So just ignore the errors and warning for now.

Regards;
Frank

Hi Frank,

Thanks. I'll do a render and post my time.

Cheers,
Paul

Thanks for the feedback (and benchmarks) everyone. Frank, I don't see the line break Register link issue you described. What browser are you using, and can you provide a screen shot? Btw, congrats on being top of the heap so far. ;D

- Oshyan

Quote from: stunnedmullet on March 01, 2009, 05:04:59 AM
@choronr - you need to click the text "Download Scene" under the test picture to download the test scene in a zip file.

BTW, I've got a problem with the test render scene. As soon as I open it I get a number of warning messages (see attached). I'm using Build 1.10.23.1. Anyone else had the same happen or know what the problem is?

Regards,
Paul

OOh some of these look very intriguing :) Looks like theses a few new features coming up, could "lat_long_at_apex in class: planet" be something about using a different co-ordinate system? And could "ray_detail_region_padding in class: render" be about controlling the shadow clipping issue?
Looking forward to the next installment :)

Richard

Where do I submit my results?  I registered, but can't seem to login.

Once you register, login at the top, and entering your username/password in the fields that appear there. If your login doesn't work properly, you will see the "register/login" buttons again at the top - it unfortunately does not give an error messaage yet. Once it logs in correctly, you will see a message at the top indicating you are logged in. Then the first thing to do is create a new system, go to My Systems, add your system data and click Submit at the bottom. Finally, go to Benchmarks and Scenes and find the TG2 benchmark at the bottom, use Download Scene link below the benchmark example image, run the benchmark, and then while still logged-in use the Edit My Benchmarks area on the TG2 scene file page under the "Select from the lists below to filter benchmarks." section on the right, enter your benchmark information and press Save Benchmark.

- Oshyan

Thanks Oshyan.  I was blind in one eye and couldn't see out of the other.  I have submitted my benchmark - 20m46s.

interesting to see that, at the moment, the 4 top results are using vista 64bits. tomorrow i'll test under 8 core macOS, at work.

I guess.  That really doesn't show anything about Vista...yet.  Let's see a Windows XP 64-bit with these specs run it, but with a faster processor than Q6600 - Intel Core 2 Quad and 4G of RAM.  Or with 8G of RAM.  Right now the top 4 have faster than a Q6600 processor.

Quote from: mt_sabao on March 01, 2009, 05:46:04 PM
interesting to see that, at the moment, the 4 top results are using vista 64bits. tomorrow i'll test under 8 core macOS, at work.

Maybe tomorrow I'll overclock mine just for the hell of it and see what I can get. I've seen the E8400 clocked to 3.8 to 4.o Ghz.  NTune makes it rather easy.  If you don't see me for awhile you can guess what happened? LOL

Quote from: calico on March 01, 2009, 06:38:25 PM
I guess.  That really doesn't show anything about Vista...yet.  Let's see a Windows XP 64-bit with these specs run it, but with a faster processor than Q6600 - Intel Core 2 Quad and 4G of RAM.  Or with 8G of RAM.  Right now the top 4 have faster than a Q6600 processor.

Quote from: mt_sabao on March 01, 2009, 05:46:04 PM
interesting to see that, at the moment, the 4 top results are using vista 64bits. tomorrow i'll test under 8 core macOS, at work.

I have a Q6600 @ 2.4 GHz and 8 GB RAM and it took 20 minutes.

Added results for my Core 2 Quad with 8GB of RAM on Vista x64. Looks like Vista doesn't provide any benefit. ;D

- Oshyan

Just ran another render after bumping my cpu up to 3.5 Ghz.  18min 6sec.... Saved a little time, core temps got up to 46c, of course no problem with memory. See the screen shot with the NVIDIA Monitors and finished scene.  Not sure yet how my machine will run with a larger more demanding scene though. Certainly wouldn't render overnight with this OC without closely watching the temps.

Hey Oshyan, Should I post this time on the site? Or is it setup for overclocked cpu's.

Nice result Buzzzzz. If you want to enter that into the site you'd make a separate system entry for the machine overclocked. Everything would be the same except the CPU frequency would be different. When you select your CPU it will fill in the speed automatically based on the default clock, but you should be able to override that. In the future we'll have better support for overclocking of a single machine, to make the results more easily and directly comparable.

- Oshyan

Oops! Wrong nic. Ignore that man behind the curtain.  ;D

Quote from: Oshyan on March 01, 2009, 08:50:51 PM
Nice result Buzzzzz. If you want to enter that into the site you'd make a separate system entry for the machine overclocked. Everything would be the same except the CPU frequency would be different. When you select your CPU it will fill in the speed automatically based on the default clock, but you should be able to override that. In the future we'll have better support for overclocking of a single machine, to make the results more easily and directly comparable.

- Oshyan

OK, thanks Oshyan, I'll see if I can enter it with the OC.  Cool site BTW!

So 4 more GB of RAM shaves 46 seconds...it seems.  Right?

Quote from: Tangled-Universe on March 01, 2009, 07:14:04 PM
I have a Q6600 @ 2.4 GHz and 8 GB RAM and it took 20 minutes.

It's not likely that more RAM (above 4GB especially) would have any significant effect. There are many other factors which may be involved, which is why it's always good to have as many benchmark results as possible.

- Oshyan

Maybe the fact I was doing other things at the same time...

nice site, added my results, would be nice if as many as possible added theirs since im considering upgrading cpu  ;)

Interesting results so far. My upgrade philosophy is to always double double my power (see Moore's law (http://en.wikipedia.org/wiki/Moore%27s_law)) and its good to see that the entry level i7 is almost exactly twice as fast as the Q9450. The down side here however is the i7 upgrade path is pricey as it has to include a new board and new ram. Looking at the very least at £500 just for the processor, board and ram.
Time to start saving  ;D
richard

i found the i7 a good upgrade path if you still own a mobo that cannot be upgraded to anything decent at all. With that, probably your RAM is not to today's standard either. Essentially, if your whole machine is dusty crap, why not go for the i7? :D

That being said, my old box was a AMD X2 4200, on a 939 CPU interface (I think), and I don't even remember what the RAM was :-)
even if I would have upgraded to a newer Intel Quad, I would have had to buy almost everything new - anyway.

Frank

Quote from: FrankB on March 02, 2009, 11:42:31 AM
i found the i7 a good upgrade path if you still own a mobo that cannot be upgraded to anything decent at all. With that, probably your RAM is not to today's standard either. Essentially, if your whole machine is dusty crap, why not go for the i7? :D

Frank

Hmm I don't think my machine is dusty old crap (dusty for sure) but it is at the end of its shelf life. The trouble is that at a certain point you have to make the leap, hence my upgrade philosophy. Its just that other than the hard drive, graphics card, case and power supply there's sod all I can transfer to anew build. So £500 it is then. lol
richard

btw, everytime I wrote "you" I actually didn't mean you personally. Used this more in generic sense, speaking to noone in particular :-)
That being said, I don't assume your personal box dusty old crap. But *my* old one certainly is :-)

I agree about the leap thing. A PC usually lasts 2 years with me. After that time, I cannot play any recent games anymore and others with newer machines render me away easy. That goes on my nerves of course, hence after 2 years the pain is big enough to justify a decent leap :D

Cheers,
Frank

Anyone out there got a top end AMD?
I've always wanted to see how good they are.
richard

ps its cool, didn't take it personally :)

I've got a Phenom II at work. Can I use my license in two places? Not for commercial use, just playing around in my lunch break.

I have  Intel Core2 Duo 3 Ghz and 4 Gb DDR2.
Rendered 800X600 .5 resolution  34:35
       "       640X480        "          14:14

I see how the Distribution Shader is used with a population now...thanks.
The scene is complicated for me, but interesting to study.
Norm.

You don't need the licensed version to run the benchmark, the free version works just fine :)
Richard

Oh baby, this is great, wait till i get home I'll set a new record for longest time.


EDIT: 1:34'36"  beat that

Be sure not to enter any benchmarks rendered with changed render settings (e.g. 800x600). The benchmark results depend on the resolution always being the same. :)

- Oshyan

I'd be really interested to see how far one can go with a Phenom Black Edition (reasonably) overclocked... Anyone?

Just read that Vista SP2 RC is leaking around on the internets, and the final will be out by April first-ish.
    Wonder if that will change any times? ::)

PLease sticky this :)
richard

Good idea!

- Oshyan

Cool!  I added my results to the site, and I have a few observations.

I didn't "optimize" my laptop before running the benchmark.  I didn't have Thunderbird running but other than that nothing was disabled, including the wallpaper cycling program I wrote, which changed the wallpaper twice during the render.   ::)

I use Everest Ultimate to monitor chip temperatures, and it also give me a nice per-core utilization feed in real time.  Watching the render, I would have expected 100% on all four cores the entire time.  For the most part all four cores were at 100% but at random times all four would dip down to around 80%, with no other process visibly doing anything.

The most surprising observation to me was at the end of the render, when only the last bit of the grass population was being rendered on top of the already rendered terrain.  NONE of the cpu cores was above 50%, but two of them were up between 45% and 50%.  What I expected to see was the first core around 20%, the 2nd and 3rd around 3%, and the fourth at 100%.  Interesting!!  :)

Also, I'm very happy to see that my 6 month old laptop will hold it's own for a while to come, even against the new procs.  yay!  :D

Rich Allen
(Ricowan)

That's a very nice result for a laptop! I think what you're seeing with the CPU utilization is probably inefficiencies in rendering process. At the end only one core is being utilized as the last tile is rendered, so I think that's to be expected.

- Oshyan

It's more of a mobile desktop than a laptop.  This thing doesn't work very well on my lap.  :D

I guess what was curious was the fact that none of the cores was at 100%, even though the image was still rendering.  Since it only happened during the instance rendering (the terrain was rendered already) hopefully that can help you guys narrow this down.  :)

My PC specs:
Intel Quad Core Q9650 at 3.0 Ghz (not overclocked). 2 GB Corsair Dominator and XFX 260 :D
at 640 x 480 -> 9 mins 40 sec YAY record :P

Here's my desktop: i7 920, Asus P6T motherboard, 6gb 1600mhz corsair DDR3 memory, windows 7 release candidate(build 7100), Nvidia Quadro FX580 video card, and a coolermaster case.
Took me 21:10(800x600 resolution) while cruising facebook, this forum, playing music from pandora, cruising a bunch of other sites, etc... never went over 53% on the processor or 33% on the memory.
(http://lh6.ggpht.com/_kuKNdKTezek/SjCFBl_niKI/AAAAAAAAAGM/lH5-YMDGwYI/s144/rendering%20the%20benchmark.PNG) (http://picasaweb.google.com/lh/photo/nWrVSGZfocWcg2bChBRBpQ?authkey=Gv1sRgCNGC3NC9jsL7bQ&feat=embedwebsite)
Probably would cut time drastically if I just let it sit and render but the life/internet moves too fast for that.

Interesting results, but please don't enter anything that is not done using only the default settings in the benchmark TGD, and with no other major processes running. Otherwise it will skew the test data.

- Oshyan

Maybe this is the right place to post.

I am having trouble registrering at the benchmark site, every time i try, the fields with reg-info is rendered blank and nothing happens.

Try registering through the forums: http://3dspeedmachine.com/forum/

- Oshyan

It seems to work fine, through the forum, perfect

I made another post, http://forums.planetside.co.uk/index.php?topic=6714.0

was not sure where to put it, but you are welcome to delete the above post at the link.

Thank you  ;D ;D ;D

Thank you!
I finally managed to register, and post some benchmarks. :)
((My result for the TG2 scene was 5m 14s...))

My new Core 2 Quad got 10 minutes and 43 seconds :o So much faster than my 21 minute E8400

I just tried to upload my results and nothing happened. The drop down menu for the OS is very hard to see, because only the first part of it shows up and I could not accurately point to my OS. Also, by DEFAULT RENDERER, do you mean the "Quick" Renderer or the "Full" Renderer. I used the "Full" because that would be a more accurate means to judge by. Mine is not to fast, but it much better than what my old DualCore would have done with it so I am happy.

I did not get very far using this. I was curious, but for me it's not an issue.

Just load the scene and press the Render button on the top toolbar. That uses the "default" renderer (yes, it's the Full Render).

As for your problem of not being able to see the full text of the OS choices, that sounds like an example of an Internet Explorer browser issue, rather than a problem with the site. IE does not expand dropdown boxes to show full text, unlike pretty much every other browser around - Firefox, Chrome, Opera, Safari...

- Oshyan

I'll try it again. I have a couple days off coming up (need them, it's been a weird week) Let's see what this P4 HT can do. Only in the last few months has it shown any 'I'm giving it all she's got Captain!' on some new games.

my render takes 17m44s.
i have an intel core 2 duo E8400 on 3600 MHZ
windows xp home sp3
2 GB Ram DDR2 on 1066+ MHZ

Make sure to report your render time on the site. Posting here in the forum is not necessary.

- Oshyan

Quote from: matrix2003 on March 05, 2009, 06:08:48 PM
Just read that Vista SP2 RC is leaking around on the internets, and the final will be out by April first-ish.
    Wonder if that will change any times? ::)

By April Fools' Day? It must be a joke!  :D

I ran the benchmark last night. My poor old p4 HT! It gets the job done, but the comparison...
I do not think I will post the result. 

Please do - for the benefit of others! :)

- Oshyan

Okay. Pentium 4 HT, 4 gigabytes ram. Radeon 1300, 512 megabytes. Used both logical processors. 1 hour, 15 minutes, 22 seconds.

I'll post at the benchmark site also.

Core i7 920 at OC to 2.8 GHz 6gb DDR3 vista x64 ultimate with Geforce 9800 GT
Rendertime was 9:50. I was actually impressed that an image like that could be rendered so fast.

Ambico, be sure to submit your results to the site. :)

- Oshyan

I only see the logo for Terragen 2.  I don't see any picture to download.  When I click on the logo, I'm brought right back to the same page.  What's going on???

The download is the benchmark file in a ZIP. When you click Terragen 2 from the main benchmark list it will open a sub-list of versions. Click the most recent version (it will be updated with 2.1 shortly) and then a new page will load. On the upper-left of that page is the download link, under the picture.

- Oshyan

ok, on my new box dual xeon 5590 on a tyan s7025 with ssd256 gig pny 1800fx I got 4 min 42 seconds ;D ;D

Quote from: JHD on March 28, 2010, 05:33:26 PM
ok, on my new box dual xeon 5590 on a tyan s7025 with ssd256 gig pny 1800fx I got 4 min 42 seconds ;D ;D

Nice result! I'll be updating the version numbers on there shortly to reflect the latest available version. In the meantime go ahead and enter your result under 2.0.2.1 version.

- Oshyan

I am a bit puzzled with my benchmark results.  Perhaps someone can shed some light on my observations.  I have an AMD Phenom II X4 955 (3.2 GHz) with 8GB RAM.  First, I made sure that my pre-render was on "Pause" so that it didn't take cycles away from the render under test.  I am guessing that preview rendering only uses one thread because it never seems to take more than about 25% (one core) of the total CPU usage.  This system should have come in under 16 minutes on this benchmark but it took longer (17:10) than the only other Phenom II (by cyphyr @ 3.0 GHz) in the chart.  Starting the render brought my CPU utilization from about 1% to 99%.  At about 08:56, one of the cores seemed to drop out and the CPU utilization went to 76%.  At 10:12 a second core dropped out and utilization went to 51%.  Finally, at 11:09 a third core dropped out and utilization went down to 26% and stayed there until the render was finished.  In the final stages it seemed to be working only on the vegetation in the lower right quadrant of the render window.  It appears to me that mutli-threating is not used in all parts of the render pipeline.  Pausing the render at 08:01 had little effect on the outcome.  Does anyone have an explanation for this behavior?   ???

A difference of 1 minute could come from a number of factors, including background idle processes, different OSs and handling of multithreading, etc. The behavior of the CPU usage is normal. What TG2 does is spawn render threads to handle titles across the image sequentially until rendering is done. When one tile is done rendering, that CPU thread will be assigned to a new tile. When it's out of rendering to do, that thread will become idle, even if others are still working. TG2 does not do tile subdivision nor can it assign multiple cores to a single thread. So what happens is the last tile of the image is always rendered using just 1 thread. All renders will end that way since there is inevitably one tile that takes longer than the others. This is an inefficiency in the renderer, but it's difficult to dynamically split or reassign work tasks. It's something that can hopefully be addressed in the future.

- Oshyan

Thanks Oshyan.  So if I understand correctly, the main thread divides up the target screen area evenly between the cores (i.e. threads) at the start.  Each thread then works on its part until all are done.  I did some calculations to find that my system took a total of 4120 Core*Seconds to do a frame.  Of this, there were a total of 1273 Idle Core*Seconds which accounts for almost 31% of the total Core*Time.  This means that 1.24 Cores is Idle on average during this render (might be true for all quad-core systems).  So my system should (theoretically) be able to complete the render in 11:51 if all cores were kept busy the whole time.  I would agree that some optimization could speed things up substantially.  Of course, it would only benefit systems with more than one core.  Do any of you with i7 processors make TG2 "think" that it is running on an 8-or-more-core system?  It might not make a difference in the end. 

Your analysis is generally correct, and the ability to subdivide render tiles would certainly improve render times in many situations.

I use 8 threads on my i7 920 and i7 860 systems, with generally good results.

- Oshyan

Wow man, real cool. I like the usefulness of your ambition! Nice site.

Just tryed the benchmark on my AMD Athlon II X4 630 2.8GHz, 4 Gig ram, win7.
24:27s 980196 micro-triangles.

What dos threds do in tg?

Each additional thread splits the rendering of the image. You always want the number of threads to equal to or less than the number of actual CPU cores (or execution threads) you have available on your system. Never use more threads than the number of cores/cpu threads you have. The more threads you are able to use, the faster your renders will be, generally speaking. Most systems these days are dual core or quad core, so you can use 2-4 threads, depending on the system. Your X4 should I think have 4 cores.

- Oshyan

I've added systems in the past but have been unable to add my new system to the site. When I click submit, I get a window that says "Your system was not saved."

Also, when I choose my CPU model, it puts this under speed:
<br /> <b>Warning</b>:  include_once(/home/thrdspee/public_html/forum/SSI.php) [<a href='function.include-
once'>function.include-once</a>]: failed to open stream: No such file or directory in
<b>/home/benchweb/public_html/ajax/processor_ajax.php</b> on line <b>2</b><br /> <br /> <b>Warning</b>: 
include_once() [<a href='function.include'>function.include</a>]: Failed opening
'/home/thrdspee/public_html/forum/SSI.php' for inclusion (include_path='.:/usr/share/php:/usr/share/pear') in
<b>/home/benchweb/public_html/ajax/processor_ajax.php</b> on line <b>2</b><br /> 3.33

I've changed that to my actual CPU speed but still cannot submit my machine.

I can't reproduce this issue at the moment so I'm really not sure what's going wrong. But we're moving the site to a completely new content management system in the near future, so it should resolve all this kind of stuff. If you can, keep an eye out for the new site (I'll probably mass mail the members to notify them of the update), and then enter your benchmark there. It will manage benchmark results much better than the current one anyway.

- Oshyan

New version of the benchmark site is up!
http://3dspeedmachine.com/benchmarks/render-times/terragen-2/default-renderer/
Now based on an all-new sweet CMS, new bench filtering tech, integrated forums, etc, etc. Perhaps most importantly, when you enter a benchmark result, you're given the option of specifying which version you used for the benchmark and then viewers can filter on that. In the near future we'll also be adding columns so you can view which version a benchmark was performed with, and thus compare across versions and systems.

More big plans for the future! :)

- Oshyan

Great, I do like the benchmark site, it helps see peoples renders in some kind of perspective. I dont remember too well but did not the last version also have a column for a user name?
cheers and thankyou
Richard

Quote from: cyphyr on February 17, 2011, 04:21:24 AM
Great, I do like the benchmark site, it helps see peoples renders in some kind of perspective. I dont remember too well but did not the last version also have a column for a user name?
cheers and thankyou
Richard

Yes, it did. We'll be making some tweaks to the column display upcoming as there are other potentially useful data points to put on there.

- Oshyan

The site somehow doesn't work well here, whatever browser I use.
Could someone post a direct link to the benchmark results page?
Thanks in advance :)

All of sudden it showed up! :)

Cheers,
Martin

I must be doing something wrong.  I check my TG2 version and it says "2.2.23.1".  I try a check for updates, and it says there are none.  The drop-down entry for the tg2 version on the benchmark page has three choices; 2.0.2.1, 2.3.6.0, and beta 1?  I can't seem to enter anything other than the three choices.

Am I missing something?

The version numbers are entered by users. Someone probably entered the wrong version (or is possibly using an alpha version). You should be able to add your own when you click Add/Edit Benchmarks (you must be registered and logged-in to do this). Simply fill in the right version number when you enter the benchmark results. I have gone ahead and added 2.2.23.1 already.

- Oshyan

Still having problems.  I log in and have the Terragen 2 Default Renderer page displayed.  I click the "Add/Edit Benchmarks" button and I get a "Save Benchmark" window but I can't type in the system field. The drop-down for that field offers up a blank, green filled line.  I can enter the time and software version fields, but a "Save" gives me a "System - required" message. ???

Ah, yeah you need to have added at least one System (computer profile/details) to your profile. This points out that it's not very clear at that point, so I'll make some changes. It's easy to add a system though. Just click the My Profile tab at top and then click Add System below, enter the details and save. Then you can add benchmarks for any app for that system. If you have multiple systems enter each one as a new system and then you can pick which system the benchmark was run on when you add a result.

- Oshyan

Thanks! That did it.  I thought I had already set up a system profile a while back, but I guess not.

Everything should have been migrated from the old site, but it's possible something was missed. In any case glad you got it working now and I'll update the site to let people know that they need to enter a system in order to add a benchmark.

- Oshyan

0:11:12s mins on my i5 2500K stock 4gb ram

Can't seem to enter a system (combo drop down is empty) into the benchmarking site right now...  :(
But nevertheless I am very pleased with my new Z8NA-D6 mobo with two Xeon E5620 @2.4GHz (8 cores) and 12MB RAM + 4 1TB drives in raid10. I guess its not stacking up that well against the new i5's for cost/speed but I use it for more than just TG2.

Cheap to buy from Newegg (1740=260 + 2*390 + 2*130 + 4*90 + 80 +case+graphics) and did the benchmark (in 2.3.20.1) in 8:53. A big improvement over my previous(single old school Xeon) system at 2:25:34. In fact 533 vs 8734 seconds. Almost 16.4 times faster.

and almost twice as fast as the 8core Mac it is most equivalent to. :-)

Are you logged-in? To enter a new system you go to My Profile then click Add System and fill out the details and click Save. Then go enter your benchmark and select the system you entered. Should work fine, just tested here.

- Oshyan

Hi Oshyan - You are right. Somehow completely missed that. Obvious when I see you can benchmark other pieces of s/w also... :-(

No problem. Could probably use some FAQs.

- Oshyan

Hi,

Quote from: neon22 on May 14, 2011, 05:38:18 PM
and almost twice as fast as the 8core Mac it is most equivalent to. :-)

Previous to v2.3 the Mac version had problems scaling and could slow down a lot when it started to use hyperthreads, as an 8 core Mac would with the benchmark. My dual quad core 2.26 GHz Mac Pro renders the benchmark in 11:16 with v2.3, so only a couple of minutes slower than your faster machine really :-). With v2.3 the Mac and Windows version have pretty much the same rendering performance.

Regards,

Jo

Hmm - yes looks like the clock speed would account for the difference. Your 2.26GHz and my 2.4GHz.

On the surface of it the best price/performance looks like an overclocked Core i7 with D0 stepping
But many of these benchmarks are not on the latest release.
That would be a good column to have visible - I think.. Hint :)

But the newegg price of a high end i7 is still between 1500 and 2000. So there's not much price diff if you start with a 3.4GHz system. But then homebuilt is always a bit cheaper at that top end. :-\
Of course the dual 6core E5660 is 4600USD...(2.8GHz) Hmmm If that really got you to 4 minutes then maybe...

blah blah blah .. CPU speed .. blah blah blah .. nephalem .. blah blah  ::) LOL

There's another thread around here discussing some recent price research I did which found that a 6 core i7 extreme (previous generation, not Sandy Bridge) is probably the best bet for max performance at this point. It's not cheap, but it should equal a dual CPU quad core Xeon system due to higher clock rate and fewer threads. The best price/performance is probably the i7-2600 Sandy Bridge at 3.4Ghz. It's really remarkably cheap considering the clock rate increase from the past generation.

- Oshyan

Quote from: Oshyan on May 16, 2011, 03:13:01 PM
There's another thread around here discussing some recent price research I did which found that a 6 core i7 extreme (previous generation, not Sandy Bridge) is probably the best bet for max performance at this point. It's not cheap, but it should equal a dual CPU quad core Xeon system due to higher clock rate and fewer threads. The best price/performance is probably the i7-2600 Sandy Bridge at 3.4Ghz. It's really remarkably cheap considering the clock rate increase from the past generation.

- Oshyan

and the i7-2600K chip goes to 4.4GHz with a basic fan (for $23 more). So maybe it takes the lead.
http://www.anandtech.com/show/4083/the-sandy-bridge-review-intel-core-i7-2600k-i5-2500k-core-i3-2100-tested
But I think I should stop hijacking this thread... This is going to look so old later....

btw did someone execute this benchmark with the i7 2600K ? I did not find it on http://www.3dspeedmachine.com/

luc

I should be able to run my new build sometime late October.

Surprising we have no i7 2600 benchmarks on there yet. I think there are a fair number of people who have them here now.

Time to update the benchmark with the latest features anyway...

- Oshyan

I've just added a benchmark for my i7 2600K. 8m43s.

Matt

Thanks  :)

Couldn't get the right settings in the drop downs, so I didn't add my results.

I've just put together a Sandybridge 2600 3.4Ghz (although the bios says its running at 3.8?? and Win7 Pro 64bit is saying 3.7. I haven't done anything to the bios, so maybe it was upped at the warehouse I bought it from). 8gig of Ram (should be 16, but had a miss matched set of pairs so returned).

Completely clean install, only TG, I got 7:41.

When I decreased the bucket size to 64 (although it looked like it was rendering 128) I got 7:12.

To upgrade form my previous system which had a decent graphics card that I carried over, the cost was £700.

You can enter in your own values on the benchmark site and it should save them as new.

The 2600 will clock up to 3.8Ghz when only running on a single core and it can vary the clock dynamically. So 3.8 might be reported in BIOS because only 1 core (or less, really) is in use. When you viewed the Windows system info it might have been running a minor load and been clocked at 3.7 or something. Would be interesting to see if it reports 3.7Ghz while rendering in TG though.

A more full-featured app like CPU-Z (http://www.cpuid.com/softwares/cpu-z.html) would tell you what your CPU was clocked at with greater detail and certainty.

- Oshyan

this not be working, cannot make register. bad site info?
say this: Fatal error: Allowed memory size of 67108864 bytes exhausted (tried to allocate 94 bytes) in /home/benchweb/public_html/sapphire/core/model/DataObject.php on line 2758

Hmm, that's strange. Can you try again? Anyone else seeing this problem?

- Oshyan

hi, my name is christian. i'm new here and in the world of terragen.  ;)

my question: is the bechmarktest released immediately when i  enter it, or released the admin the data after saving the results?

It should be immediate unless the system or another user flags it as suspicious.

- Oshyan

Just to add something: 6:38 on my new Desktop

specs:
CPU: Core i7 2600K @ 4.53 GHz (OC)
MB: Asus P8Z68-V LE
RAM: 16GB Corsair DDR3 1600
HDD: 120 GB Corsair Force 3 SSD
Graphics: Geforce GTX580

Very similar machine to me Goms, congrats on the new machine! It's a joy to work with TG on something like that. :)

- Oshyan

Thx!
Its quite amazing to see things render in no time compared to my old one. Any experience on how high you can go with the 2600K? I now have a base clock of 103 and the multiplier is set to 44. So far the maximum temperature stays at ~47C, the watercooling is quite effective and I'm concidering to go a little bit higher... :D

I am at 4.6Ghz, probably wouldn't go higher than that myself, but I've heard of people getting beyond there (4.8 at least).

- Oshyan

Good to know, i'll see how stable this clock is with a longer render. :)

Cant seem to add a system or save anything, tried IE and Chrome.. anyways

P9X79 PRO
i7 3930k OC @ 4.7 Ghz (Running below 60 degrees with Corsair H80)
32GB DDR 1866 Mhz
2x 120 GB SSD Raid0

benchmark time 00:05:34

Benchmarked too, just for fun, a hour ago or so. All @ default, tweaking would be cheating, right?  ;)

Got a warning at the standard subdiv cache at 400 MB but meh..

- Xeon E3-1240v2 @ 3744 MHz (bit OC´ed via BCLK at 104), temps ~41 max. (realtemp, + cpu-z)
- 16 GB RAM, 1333 9-9-9-24
- SSD HD (i think it´s all about CPU skills TG wise so..)

Benchmark time was 7:09 minutes.

I just bought recently a new Shuttle SH67H3 with an LGA 1155 MB for less then 200 € (v.2.0 MB is needed for Ivy CPUs) and will just throw that low energy consuming and low heat producing Xeon single server Ivy CPU into that Shuttle + some adds from what was left at home from my old system.
Curious then, what my cheap renderslave, named Franz, will achieve then with less, but faster RAM (7-7-7-21)..

If i can set up a complete remote control, im rly happy then.
There is nothing more annoying than a/your main PC system is hindered because of long time background renderings.. (http://smiliestation.de/smileys/Kotzen/28.gif)

I am going to upgrade by PC soon. I thought that before I do that, it would be nice to run some Terragen benchmark with my old system so that I can compare how much improvement there is with new system.

I did Google search and found this thread. However, when I go to the address http://www.3dspeedmachine.com/, it only gives a login page. But I do not have account there, so I can not enter the site. There is no option to register a new account. How can I register?

--
Pauli

The site is currently down. Plans are to bring it back later this year, hopefully with all the old results intact, although they won't be as relevant anymore. I would suggest in the meantime to just pick a couple of your own scenes and render them on your current system, then re-render on your new one.

- Oshyan