Return-Path: Received: from mail-qe0-f46.google.com ([209.85.128.46] verified) by media-motion.tv (CommuniGate Pro SMTP 4.2.10) with ESMTP-TLS id 5082115 for AE-List@media-motion.tv; Thu, 23 May 2013 21:33:07 +0200 Received: by mail-qe0-f46.google.com with SMTP id 1so2014686qee.19 for ; Thu, 23 May 2013 12:42:23 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20120113; h=mime-version:in-reply-to:references:from:date:message-id:subject:to :content-type; bh=QZ4QRjZOWqzY3iokLHXrGu6DHy4ytltecfPUL/3j0fY=; b=RTUUQ8LIIw/33yMsPGxKoh0V3qD0CrXmQH7wsXRqMQU5mNxGVe7tjo/MkO4x05DrJb 6BeopD3624qXZbymqpW08oaHea6CCFLVjDP0nsm+O+FJJoG62NkW/On+ps+79sKc7gCt T1vsDpiwt2kwx0/Wgjv9j5YXHOFUClxI2L8ur2aKHzeTwTI6/8eNJ+GAPDaFI7I63ow+ O6H7nOcr50MFyURztMWtMQcmllemMMNcSn1z+HsJS0iX3bw3PJsSCqz6DNdU+OX7Dxkx FU6fRH5x1VRamOJd7kxwtdE5lFShtM06GTnlhdb9GkUbvM1ZFrwMNzph2gSowntONuGd 7Plg== X-Received: by 10.224.179.77 with SMTP id bp13mr13177894qab.34.1369338143615; Thu, 23 May 2013 12:42:23 -0700 (PDT) MIME-Version: 1.0 Received: by 10.49.88.102 with HTTP; Thu, 23 May 2013 12:41:43 -0700 (PDT) In-Reply-To: References: From: Teddy Gage Date: Thu, 23 May 2013 15:41:43 -0400 Message-ID: Subject: Re: [AE] [OT] yet another graphics cards thread - Titan raytracing benchmarks To: After Effects Mail List Content-Type: multipart/alternative; boundary=485b397dcc97a0cedc04dd67e144 --485b397dcc97a0cedc04dd67e144 Content-Type: text/plain; charset=ISO-8859-1 The titan is in the machine with a 3930K @ 4.6 ghz, 32 gb ram, asus p9 mobo The sli GTX 580 is in a new machine with an MSI board, the quad core 3770K, 16 gb 2133 mhz RAM and 128gb RAID 0 SSD drives. On Thu, May 23, 2013 at 3:32 PM, rendernyc wrote: > is that the same system that you ran the 680 in? > > i7 3820@3.6 > > > On Thu, May 23, 2013 at 3:17 PM, Teddy Gage wrote: > >> also, @Danny aka rendernyc I ran your robot bench on both an OC titan @ >> 1120 mhz and an OC sli GTX 580 x2 @ 860mhz >> >> titan results were here, last render item http://i.imgur.com/royBZpj.jpg >> >> 8 min 12 sec >> >> sli 580 GTX x2 rendered in 9 min 12 sec >> >> which is interesting because in the other benchmarks the gtx 580x2 was >> faster >> >> >> On Thu, May 23, 2013 at 12:51 PM, Teddy Gage wrote: >> >>> With double-precision on, the benchmark renders slower. Just tried it. >>> not by much, maybe 5-10%? Honestly I have no idea what double-precision is >>> useful for. Anyone can fill me in? >>> >>> I was also wondering whether it was possible to edit a preferences file >>> allowing AE to use more available VRAM. I understand limiting cards with >>> less memory, but I'm only getting 4.5 GB of usable VRAM with the titan in >>> AE. Granted, that is a lot but it would be great to get another GB and I >>> doubt the OS needs more than 500 mb of overhead. >>> >>> >>> On Mon, Apr 22, 2013 at 10:26 AM, Darby Edelen wrote: >>> >>>> I'd be curious to find out how Titan performs in these tests with >>>> double precision enabled vs. disabled from the nvidia control panel. I >>>> expect that the ray-traced renderer in AE is probably not double precision >>>> but I'd like to find out :) >>>> >>>> >>>> On Sun, Apr 21, 2013 at 1:40 PM, Teddy Gage wrote: >>>> >>>>> So a while ago I posted an AE graphic card benchmark at creative cow >>>>> that has gotten a lot of responses, so I threw together a quick, readable >>>>> spreadsheet of the results. I was mainly curious about compute performance >>>>> of the new titan card vs. last gen 5xx cards and the kepler 6xx series. I >>>>> wanted to see if the titan was worth the extra $1,000, because it is in >>>>> fact a consumer version of the nvidia k20 tesla which costs over $3k. The >>>>> results were very interesting. The original thread is here >>>>> >>>>> http://forums.creativecow.net/thread/2/1019120#1035979 >>>>> >>>>> And I have posted the spreadsheets (and benchmark file) on my server >>>>> at >>>>> >>>>> http://www.teddygage.com/AEBENCHCS6/ >>>>> >>>>> What I would love to add are any results with the newer quadros (5000 >>>>> / 6000) or any tesla card. >>>>> >>>>> Furthermore I'd be interested to hear any thoughts from 3D users about >>>>> their experiences using an AMD 7950 vs nvidia consumer card for viewport 3D >>>>> (using zbrush / maya / or C4D especially) >>>>> >>>>> The amount of C4D work I've been doing has only increased in the last >>>>> year and I can imagine when CS7 comes out that's only going to increase. >>>>> viewport accuracy is going to be of a larger concern and nvidia is using a >>>>> poor strategy to differentiate the quadros by gimping compute performance >>>>> in the 6xx consumer cards. >>>>> >>>>> Would also love to hear any thoughts on these results, I'll refrain >>>>> from saying what I think until people get a chance to check it out. >>>>> -TG >>>>> >>>>> >>>>> -- >>>>> Animator & Editor >>>>> www.teddygage.com >>>>> Brooklyn >>>>> >>>> >>>> >>> >>> >>> -- >>> Animator & Editor >>> www.teddygage.com >>> Brooklyn >>> >> >> >> >> -- >> Animator & Editor >> www.teddygage.com >> Brooklyn >> > > > > -- > danny princz > > vimeo.com/rendernyc > -- Animator & Editor www.teddygage.com Brooklyn --485b397dcc97a0cedc04dd67e144 Content-Type: text/html; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable
The titan is in the machine with a 3930K @ 4.6 ghz, 3= 2 gb ram, asus p9 mobo

The sli GTX 580 is in a new machine wit= h an MSI board, the quad core 3770K, 16 gb 2133 mhz RAM and 128gb RAID 0 SS= D drives.


On Thu,= May 23, 2013 at 3:32 PM, rendernyc <rendernyc@gmail.com> = wrote:
is that the same system tha= t you ran the 680 in?

i7 3820@3.6


On Thu, May 23, 2= 013 at 3:17 PM, Teddy Gage <teddygage@gmail.com> wrote:
also, @= Danny aka rendernyc I ran your robot bench on both an OC titan @ 1120 mhz a= nd an OC sli GTX 580 x2 @ 860mhz

titan results were here, last render item http://i.imgur.com/royBZpj.jpg
8 min 12 sec

sli 580 GTX x2 rendered in 9 min 12 sec=

which is interesting because in the other benchmarks the gtx = 580x2 was faster


On Thu, May 23, 2013 at 12:51 PM, Teddy Gage <teddygage@gmail.com&g= t; wrote:
With double-precision on, the benchmark ren= ders slower. Just tried it. not by much, maybe 5-10%? Honestly I have no id= ea what double-precision is useful for. Anyone can fill me in?

I was also wondering whether it was possible to edit a preference= s file allowing AE to use more available VRAM. I understand limiting cards = with less memory, but I'm only getting 4.5 GB of usable VRAM with the t= itan in AE. Granted, that is a lot but it would be great to get another GB = and I doubt the OS needs more than 500 mb of overhead.


On Mon,= Apr 22, 2013 at 10:26 AM, Darby Edelen <dedelen@gmail.com> = wrote:
I'd be curious to find = out how Titan performs in these tests with double precision enabled vs. dis= abled from the nvidia control panel.=A0 I expect that the ray-traced render= er in AE is probably not double precision but I'd like to find out :)


On Sun, Apr 21, 2013 at 1:40 PM, Teddy Gage <teddygage@gmail.com&= gt; wrote:
So a while ago I poste= d an AE graphic card benchmark at creative cow that has gotten a lot of res= ponses, so I threw together a quick, readable spreadsheet of the results. I= was mainly curious about compute performance of the new titan card vs. las= t gen 5xx cards and the kepler 6xx series. I wanted to see if the titan was= worth the extra $1,000, because it is in fact a consumer version of the nv= idia k20 tesla which costs over $3k. The results were very interesting. The= original thread is here

http://forums.creativecow.net/thread/2/1019120#1035979
And I have posted the spreadsheets (and benchmark file) on my se= rver at

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

What I would love to add are any results with the newer qu= adros (5000 / 6000) or any tesla card.

Furthermore I'd be interested to hear any thoughts from = 3D users about their experiences using an AMD 7950 vs nvidia consumer card = for viewport 3D (using zbrush / maya / or C4D especially)

The amount of C4D work I've been doing has only increased in the l= ast year and I can imagine when CS7 comes out that's only going to incr= ease. viewport accuracy is going to be of a larger concern and nvidia is us= ing a poor strategy to differentiate the quadros by gimping compute perform= ance in the 6xx consumer cards.

Would also love to hear any thoughts on these results, I'= ;ll refrain from saying what I think until people get a chance to check it = out.
-TG


--
A= nimator & Editor
www.teddygage.= com
Brooklyn




--
Anima= tor & Editor
www.teddygage.com
Brooklyn



--
Anima= tor & Editor
www.teddygage.com
Brooklyn



<= /div>--



--
Animator & Editor
www.teddygage.com
Brooklyn
--485b397dcc97a0cedc04dd67e144--