Return-Path: Received: from mail-vb0-f41.google.com ([209.85.212.41] verified) by media-motion.tv (CommuniGate Pro SMTP 4.2.10) with ESMTP-TLS id 4739780 for AE-List@media-motion.tv; Fri, 08 Jun 2012 19:10:14 +0200 Received: by vbkv13 with SMTP id v13so447984vbk.28 for ; Fri, 08 Jun 2012 10:12:46 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20120113; h=mime-version:in-reply-to:references:date:message-id:subject:from:to :content-type; bh=HnGZp1YjY86Y37gp8jqhTC17MhDnwjudK747V49T538=; b=mnNcGbaQKHK1yhCdro8jENvJsJsFhcnZZZmAP5EVQjng1Tre/hM+B690uMPrhErii9 rLOT9Az1w8az0+3FjCorNMdhs7QOGA7eRwbk6tcMw80lm5kR9neIcjpEXKzwYTKFsyeQ oOhLiRavpcSThLDxC86wazeNcodz+iQbzQeQs3vhwYVTKnq4zhHAS2BL96Bkmbxd9nbu GvHN0K8WyY7IbuLoWG0vDs4WR0PfCV4lTsRyK3mdP8M6QXxNTWPU8mly+hpUk+7vUx9B 3OlK7RZOcFrFyBe15SGPsrf//knPozgPD0ulq9M902micA2YDgdMABpEmfCLNMWPe3bm sTXA== MIME-Version: 1.0 Received: by 10.52.21.177 with SMTP id w17mr5992547vde.98.1339175566258; Fri, 08 Jun 2012 10:12:46 -0700 (PDT) Received: by 10.220.149.70 with HTTP; Fri, 8 Jun 2012 10:12:46 -0700 (PDT) In-Reply-To: References: Date: Fri, 8 Jun 2012 13:12:46 -0400 Message-ID: Subject: Re: [AE] Why a raytraced renderer? From: rendernyc To: After Effects Mail List Content-Type: multipart/alternative; boundary=20cf307abf67eb338004c1f91b57 --20cf307abf67eb338004c1f91b57 Content-Type: text/plain; charset=ISO-8859-1 Im not sure what you mean by explicity testing the GPU. I didnt build this for benchmarking purposes but rather had it laying around so just started testing some cards The comp is extruded shape layers with some lights, shadows, reflections, motion blur and DoF being raytraced. Its not relying on any outside footage or really doing anything else other than rendering the 3d objects its heavier than just some text, but if thats all that "you" will be using the raytracer for than that would be a better benchmark. On Fri, Jun 8, 2012 at 1:00 PM, Teddy Gage wrote: > ah, just to clarify, the robot project took 48 min vs 13 - the AEBENCHCS6 > is closer to 5 minutes on a gtx 580; I have to look up the exact times if > anyone's interested > > > On Fri, Jun 8, 2012 at 12:58 PM, Teddy Gage wrote: > >> Danny, I'm not convinced your scene is explicitly testing the GPU, have >> you tried the benchmark file I posted a while ago? I'd be curious to see >> your results on it. I actually did a migration test and I got the exact >> same render times in two different machines with the same GPU. Not so with >> this robot project... >> >> http://www.teddygage.com/AEBENCHCS6/ >> >> And Greg (Balint), using the CPU engine for raytrace, the above benchmark >> project took 48 minutes vs. 13 minutes with GPU on, on a highly overclocked >> core i7 3930K @ 4.7 ghz >> >> One thing is clear, the CS6 engine is optimized solely for GPUs. This is >> significantly slower than any CPU renderer I use like mental ray or Vray, >> for a scene an order of magnitude less complicated. However with GPU on it >> is certainly comparable / possibly faster for some situations. >> >> >> >> On Fri, Jun 8, 2012 at 12:35 PM, Greg Balint wrote: >> >>> Wow. Alright, thanks! >>> >>> >>> ////Greg Balint >>> ///Art Director / Motion Graphics Designer >>> delRAZOR.com/ >>> >>> On Jun 8, 2012, at 12:30 PM, rendernyc wrote: >>> >>> did a little test with lloyd the other day that had some extuded text >>> with a reflective floor. 2008 macpro CPU render was abotu 15mins. he turned >>> on his 8800GT that is not supported and it rendered in about 7 mins. on a >>> q4000 the same scene took 37 seconds >>> >>> >>> >>> On Fri, Jun 8, 2012 at 12:26 PM, rendernyc wrote: >>> >>>> CPU based render? >>>> >>>> >>>> On Fri, Jun 8, 2012 at 12:24 PM, Greg Balint wrote: >>>> >>>>> Danny, could you do a test for me on the same project file with the >>>>> 580 turned off? Just to see how much the Nvidia cards are helping the >>>>> situation? >>>>> >>>>> ////Greg Balint >>>>> ///Art Director / Motion Graphics Designer >>>>> delRAZOR.com/ >>>>> >>>>> On Jun 8, 2012, at 12:16 PM, rendernyc wrote: >>>>> >>>>> if ANYONE has one of the cards tested already or a different one and >>>>> wouldnt mind popping off a render please email me so we can get more results >>>>> >>>>> danny >>>>> >>>>> On Fri, Jun 8, 2012 at 12:06 PM, Andrew Embury wrote: >>>>> >>>>>> Fantastic! >>>>>> >>>>>> Thank you ever so much for doing this. This is truly phenomenal of >>>>>> the user group to come together to make this all possible. >>>>>> >>>>>> Cheers. >>>>>> >>>>>> - Andrew >>>>>> >>>>>> >>>>>> On Fri, Jun 8, 2012 at 11:59 AM, rendernyc wrote: >>>>>> >>>>>>> The 480 uses more power than the 5XX >>>>>>> im expecting the 570 to be about the same speed as the 480 while >>>>>>> using less power >>>>>>> >>>>>>> waiting for some tests on that one >>>>>>> >>>>>>> >>>>>>> On Fri, Jun 8, 2012 at 11:57 AM, rendernyc wrote: >>>>>>> >>>>>>>> we had the 580 at 15:07 >>>>>>>> so 480 was about 8% slower >>>>>>>> >>>>>>>> >>>>>>>> >>>>>>>> On Fri, Jun 8, 2012 at 11:39 AM, Andrew Embury wrote: >>>>>>>> >>>>>>>>> For the performance to cost ratio....WOW! That's phenomenal! >>>>>>>>> >>>>>>>>> Now, the 580 is cutting that time by 3 minutes is that correct? >>>>>>>>> >>>>>>>>> Cheers. >>>>>>>>> >>>>>>>>> - Andrew >>>>>>>>> >>>>>>>>> >>>>>>>>> On Fri, Jun 8, 2012 at 11:36 AM, rendernyc wrote: >>>>>>>>> >>>>>>>>>> 480 in a 12 core macpro was 16:21 >>>>>>>>>> >>>>>>>>>> >>>>>>>>>> On Fri, Jun 8, 2012 at 11:31 AM, Andrew Embury >>>>>>>>> > wrote: >>>>>>>>>> >>>>>>>>>>> What about the 480? Did that not score between the 580 and 680? >>>>>>>>>>> >>>>>>>>>>> I'd be very interested to see that. >>>>>>>>>>> >>>>>>>>>>> Cheers. >>>>>>>>>>> >>>>>>>>>>> - Andrew >>>>>>>>>>> >>>>>>>>>>> >>>>>>>>>>> >>>>>>>>>>> On Fri, Jun 8, 2012 at 11:28 AM, rendernyc wrote: >>>>>>>>>>> >>>>>>>>>>>> interesting >>>>>>>>>>>> thats why we need more results... >>>>>>>>>>>> >>>>>>>>>>>> the 680 tested in a z800 rendered in 17:34 >>>>>>>>>>>> a 580 in an 2008 mac pro was 15:07 >>>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>>> On Fri, Jun 8, 2012 at 11:20 AM, Teddy Gage < >>>>>>>>>>>> teddygage@gmail.com> wrote: >>>>>>>>>>>> >>>>>>>>>>>>> GTX 680 - 2 GB VRAM - clock speed 1096 MHZ >>>>>>>>>>>>> >>>>>>>>>>>>> core i7 3930K (6 logical cores) overclocked to 4.7 gHZ >>>>>>>>>>>>> >>>>>>>>>>>>> render time : 13 min 5 seconds = 780 seconds >>>>>>>>>>>>> >>>>>>>>>>>>> >>>>>>>>>>>>> GTX 580 - 3 GB VRAM - clock speed 866 MHZ >>>>>>>>>>>>> >>>>>>>>>>>>> core i7 940 (4 cores) oc'd to 3.7 gHZ >>>>>>>>>>>>> >>>>>>>>>>>>> render time 13 min 50 seconds = 830 seconds >>>>>>>>>>>>> >>>>>>>>>>>>> >>>>>>>>>>>>> So these results seem to imply there is a disk and / or CPU >>>>>>>>>>>>> bottleneck in your benchmark project, or there is something else going on. >>>>>>>>>>>>> Maybe RAM plays a factor. Because in my own benchmark tests with a >>>>>>>>>>>>> different project, I get different results, where the GTX 580 is faster, >>>>>>>>>>>>> not the 680 in the faster machine. Interesting... >>>>>>>>>>>>> >>>>>>>>>>>>> >>>>>>>>>>>>> On Fri, Jun 8, 2012 at 11:07 AM, rendernyc < >>>>>>>>>>>>> rendernyc@gmail.com> wrote: >>>>>>>>>>>>> >>>>>>>>>>>>>> email sent >>>>>>>>>>>>>> >>>>>>>>>>>>>> thanks >>>>>>>>>>>>>> >>>>>>>>>>>>>> >>>>>>>>>>>>>> On Fri, Jun 8, 2012 at 10:26 AM, Dave Bittner < >>>>>>>>>>>>>> dave@pixelworkshop.com> wrote: >>>>>>>>>>>>>> >>>>>>>>>>>>>>> I'm running a flashed 470 from eBay. I'll run the test and >>>>>>>>>>>>>>> send you the results. >>>>>>>>>>>>>>> >>>>>>>>>>>>>>> On Jun 7, 2012, at 5:41PM, rendernyc wrote: >>>>>>>>>>>>>>> >>>>>>>>>>>>>>> > yup, been testing lots of diff cards with the raytracer. >>>>>>>>>>>>>>> > want to run a test on your 570 for the chart? >>>>>>>>>>>>>>> > >>>>>>>>>>>>>>> > have the 285, 480, 580, 680, q4000 and q4800 so far >>>>>>>>>>>>>>> > >>>>>>>>>>>>>>> > was just saying that the only card someone from adobe >>>>>>>>>>>>>>> could "officially" talk about is the currently shipping q4000 >>>>>>>>>>>>>>> >>>>>>>>>>>>>>> >>>>>>>>>>>>>>> +---End of message---+ >>>>>>>>>>>>>>> To unsubscribe send any message to < >>>>>>>>>>>>>>> ae-list-off@media-motion.tv> >>>>>>>>>>>>>>> >>>>>>>>>>>>>> >>>>>>>>>>>>>> >>>>>>>>>>>>>> >>>>>>>>>>>>>> -- >>>>>>>>>>>>>> danny princz >>>>>>>>>>>>>> >>>>>>>>>>>>>> exposedideas.com >>>>>>>>>>>>>> >>>>>>>>>>>>> >>>>>>>>>>>>> >>>>>>>>>>>>> >>>>>>>>>>>>> -- >>>>>>>>>>>>> Animator & Editor >>>>>>>>>>>>> www.teddygage.com >>>>>>>>>>>>> Brooklyn >>>>>>>>>>>>> >>>>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>>> -- >>>>>>>>>>>> danny princz >>>>>>>>>>>> >>>>>>>>>>>> exposedideas.com >>>>>>>>>>>> >>>>>>>>>>> >>>>>>>>>>> >>>>>>>>>> >>>>>>>>>> >>>>>>>>>> -- >>>>>>>>>> danny princz >>>>>>>>>> >>>>>>>>>> exposedideas.com >>>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>> >>>>>>>> >>>>>>>> -- >>>>>>>> danny princz >>>>>>>> >>>>>>>> exposedideas.com >>>>>>>> >>>>>>> >>>>>>> >>>>>>> >>>>>>> -- >>>>>>> danny princz >>>>>>> >>>>>>> exposedideas.com >>>>>>> >>>>>> >>>>>> >>>>> >>>>> >>>>> -- >>>>> danny princz >>>>> >>>>> exposedideas.com >>>>> >>>>> >>>> >>>> >>>> -- >>>> danny princz >>>> >>>> exposedideas.com >>>> >>> >>> >>> >>> -- >>> danny princz >>> >>> exposedideas.com >>> >>> >> >> >> -- >> Animator & Editor >> www.teddygage.com >> Brooklyn >> >> > > > -- > Animator & Editor > www.teddygage.com > Brooklyn > > -- danny princz exposedideas.com --20cf307abf67eb338004c1f91b57 Content-Type: text/html; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable
Im not sure what you mean by explicity testing the GPU.
I didnt build this for benchmarking purposes but rather had it laying arou= nd so just started testing some cards

The comp is = extruded shape layers with some lights, shadows, reflections, motion blur a= nd DoF being raytraced.
Its not relying on any outside footage or really doing anything e= lse other than rendering the 3d objects
its heavier than just som= e text, but if thats all that "you" will be using the raytracer f= or than that would be a better benchmark.





=


On Fri, Jun 8, 201= 2 at 1:00 PM, Teddy Gage <teddygage@gmail.com> wrote:
ah, just to clarify, the robot project took = 48 min vs 13 - the AEBENCHCS6 is closer to 5 minutes on a gtx 580; I have t= o look up the exact times if anyone's interested


On Fri, Jun 8, 2012 at= 12:58 PM, Teddy Gage <teddygage@gmail.com> wrote:
Danny, I'm not convinced your scene is e= xplicitly testing the GPU, have you tried the benchmark file I posted a whi= le ago? I'd be curious to see your results on it. I actually did a migr= ation test and I got the exact same render times in two different machines = with the same GPU. Not so with this robot project...

http= ://www.teddygage.com/AEBENCHCS6/

And Greg (Balint), using the CP= U engine for raytrace, the above benchmark project took 48 minutes vs. 13 m= inutes with GPU on, on a highly overclocked core i7 3930K @ 4.7 ghz

One thing is clear, the CS6 engine is optimized solely for GPUs. This i= s significantly slower than any CPU renderer I use like mental ray or Vray,= for a scene an order of magnitude less complicated. However with GPU on it= is certainly comparable / possibly faster for some situations.



On Fri, Jun 8, 2012 at 12:35 PM, Gre= g Balint <greg@delrazor.com> wrote:
Wow. Alright, thanks!


////Greg Ba= lint
///Art Director / Motion Graphics Designer

On Jun 8, 2012, at 12:30 PM, rendernyc <rendernyc@gmail.com&g= t; wrote:

did a litt= le test with lloyd the other day that had some extuded text with a reflecti= ve floor. 2008 macpro CPU render was abotu 15mins. he turned on his 8800GT = that is not supported and it rendered in about 7 mins. on a q4000 the same = scene took 37 seconds



On Fri, Jun 8, 2012 = at 12:26 PM, rendernyc <rendernyc@gmail.com> wrote:
CPU based render?


On Fri, = Jun 8, 2012 at 12:24 PM, Greg Balint <greg@delrazor.com> wro= te:
Danny, could you do a test for me on the same= project file with the 580 turned off? Just to see how much the Nvidia card= s are helping the situation?

////Greg Balint
///Art Director / Motion Graphics Designer
delRAZOR.com/

On Jun 8, 2012, at 12:16 PM, render= nyc <rendernyc@= gmail.com> wrote:

if ANYONE = has one of the cards tested already or a different one and wouldnt mind pop= ping off a render please email me so we can get more results

danny

On Fri, Jun 8, 2012 at 12:06 P= M, Andrew Embury <aembury@gmail.com> wrote:
Fantastic!

Thank you ever so much for= doing this. This is truly phenomenal of the user group to come together to= make this all possible.

Cheers.

- Andrew


On Fri, Jun 8, 2012 at 11:59 AM, r= endernyc <rendernyc@gmail.com> wrote:
The 480 uses more power than the 5XX
im = expecting the 570 to be about the same speed as the 480 while using less po= wer

waiting for some tests on that one


On Fri, Jun 8, 2012 at 11:57 AM, rendernyc <rend= ernyc@gmail.com> wrote:
we had the 580 at 15:07=A0
so 480 was ab= out 8% slower



On Fri, Jun 8, 2012 at 11:39 AM, Andrew Embury <aembury@gmail.com><= /span> wrote:
For the performance to cost ratio....WOW! Th= at's phenomenal!

Now, the 580 is cutting that time by 3 minutes = is that correct?

Cheers.

- Andrew


On Fri, Jun 8, 2012 at 11:36 AM, r= endernyc <rendernyc@gmail.com> wrote:
480 in a 12 core macpro was 16:21
<= br>
On Fri, Jun 8, 2012 at 11:31 AM, Andrew E= mbury <aembury@gmail.com> wrote:
What about the 480? Did that not score betwe= en the 580 and 680?

I'd be very interested to see that.

Cheers.

- Andrew



= On Fri, Jun 8, 2012 at 11:28 AM, rendernyc <rendernyc@gmail.com><= /span> wrote:
interesting
thats why we need more resul= ts...

the 680 tested in a z800 rendered in 17:34 a 580 in an 2008 mac pro was 15:07


On Fri, Jun 8, 2012 at 11:20 AM, Teddy Gage <teddygage@gmail.com> wrote:
GTX 680 - 2 GB VRAM - clock speed 1096 MHZ
core i7 3930K (6 logical cores) overclocked to 4.7 gHZ

render = time : 13 min 5 seconds =3D 780 seconds


GTX 580 - 3 GB VRAM - clock speed 866 MHZ

core i7 940 (4 cor= es) oc'd to 3.7 gHZ

render time 13 min 50 seconds =3D 830 seconds


So these=20 results seem to imply there is a disk and / or CPU bottleneck in your=20 benchmark project, or there is something else going on. Maybe RAM plays a factor. Because in my own benchmark tests with a different project, I=20 get different results, where the GTX 580 is faster, not the 680 in the=20 faster machine. Interesting...


= On Fri, Jun 8, 2012 at 11:07 AM, rendernyc <rendernyc@gmail.com><= /span> wrote:
email sent

thanks


On Fri, Jun 8, 2012 at 10:26 AM, Dave = Bittner <dave@pixelworkshop.com> wrote:
I'm running a flashed 470 from eBay. I&#= 39;ll run the test and send you the results.

On Jun 7, 2012, at 5:41PM, rendernyc wrote:

> yup, been testing lots of diff cards with the raytracer.
> want to run a test on your 570 for the chart?
>
> have the 285, 480, 580, 680, q4000 and q4800 so far
>
> was just saying that the only card someone from adobe could "offi= cially" talk about is the currently shipping q4000


+---End of message---+
To unsubscribe send any message to <ae-list-off@media-motion.tv>



<= font color=3D"#888888">--
danny princz

exposedideas.com



=
--
Animator & Editor
www.teddygage.= com
Brooklyn




--
danny princz

exposedideas.com




<= /div>--
danny princz

exposedideas.com




<= /div>--
danny princz

exposedideas.com



<= font color=3D"#888888">--
danny princz

exposedideas.com




--
= danny princz

e= xposedideas.com



--
danny princz

exposedideas.com



--
= danny princz

e= xposedideas.com



--
Animator & Editor
www.teddygage.com
Brooklyn




--
Animator & Editor
www.teddygage.com
Brooklyn




--
= danny princz

exposedideas.com
--20cf307abf67eb338004c1f91b57--