Return-Path: Received: from mail-ob0-f169.google.com ([209.85.214.169] verified) by media-motion.tv (CommuniGate Pro SMTP 4.2.10) with ESMTP-TLS id 4920087 for AE-List@media-motion.tv; Sat, 08 Dec 2012 20:55:07 +0100 Received: by mail-ob0-f169.google.com with SMTP id v19so922448obq.28 for ; Sat, 08 Dec 2012 11:59:09 -0800 (PST) X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=20120113; h=mime-version:in-reply-to:references:date:message-id:subject:from:to :content-type:x-gm-message-state; bh=/ulhSnRJPYChtCSM4ni0eC1x7vBBYI/0sdTlwwyYiFc=; b=lpwUukRqPrXmrNBbxvmKeSdBR+e6HFPdhjKHfkhskvHVwd1tuZG+/y2w9vNp/obFcJ 5Y3Q521pYk0ubTXRW2tQhmcgy4bdQvQv8efzcYOnIFFfr74+MdhKX3QcFOjxft7mlay0 xVrSiIocB87KvfoxC3qxRAKbd/9O7Wa9mjepyUO47zoquvWpVJrFicl49TK/UJemlrjj E4+pY89i/7mMddvGV/5t7oULU23j8QCnMHIqgM2GSf7Gi4TDeUMWTwHTOiyKFYIeA12f ZUbP+GhgV78E/vVgoZIy1ZzZFEqVah+irPGl+e0+89/3VtlO9bExA0B0L+fKb/TZUrvg fraw== MIME-Version: 1.0 Received: by 10.182.98.19 with SMTP id ee19mr5106855obb.90.1354996748819; Sat, 08 Dec 2012 11:59:08 -0800 (PST) Received: by 10.76.125.34 with HTTP; Sat, 8 Dec 2012 11:59:08 -0800 (PST) Received: by 10.76.125.34 with HTTP; Sat, 8 Dec 2012 11:59:08 -0800 (PST) In-Reply-To: References: Date: Sat, 8 Dec 2012 14:59:08 -0500 Message-ID: Subject: Re: [AE] 32core benchmarks From: Walter Soyka To: After Effects Mail List Content-Type: multipart/alternative; boundary=14dae93b6014e29d5204d05cc3dd X-Gm-Message-State: ALoCoQnF2sQv69YEh3TbyP803CTlLXBnreqJwYshFnzVhuraihPEjfkzHvRhdMMBNzgDsoORw04m --14dae93b6014e29d5204d05cc3dd Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable More RAM could help, but absolute performance on AMD historically trails absolute performance on Intel. Performance/cost or performance/power might be a different story, though. Also, I wouldn't be surprised if CINEBENCH scales better over ever-more cores than Ae does -- I'd think there's often much more raw computation and less I/O or inter-process communication with CINEBENCH or most 3D renderers than Ae. Walter Soyka On Dec 8, 2012 2:19 PM, "Teddy Gage" wrote: > Yes it is a bit mysterious which is why I wanted him to try a real world > AE project. > > My overclocked sandy bridge i7 3930 K, with the same graphics card gets 5= 5 > FPS and a 12.74 CPU score in cinebench, and cost around $2,200... I feel > like something is missing? It should be at least 2x faster, unless AMD ha= s > really fallen far behind. > > Did you run the 64 bit version of cinebench? are all cores firing 100% if > you open the task manager performance view? is it 16 physical cores or 32 > physical? > > Could be possible the cores are starved for RAM.... > > On Sat, Dec 8, 2012 at 2:00 PM, S=F8ren Christensen = wrote: > >> >> That was a bit low cpu score for 32 cores,no? Maybe cinebench isn't >> scaling well with so many cores? >> >> ( My 4 core MBP from mid 2011 clocks in at OpenGL 42,61 fps and CPU 5.53= ) >> >> >> Cheers, >> S=F8ren >> >> >> |||| MOTION & INTERACTIVITY >> |||| dslnc studio - www.dslnc.com >> |||| +34 93 268 0953 / +34 61 555 9963 (mobile) >> >> |||| SOCIAL: >> |||| facebook.dslnc.com >> |||| twitter.dslnc.com >> >> On 08/12/2012, at 18:15, "lsslss" wrote: >> >> hi All,**** >> >> just got my 32core in, and out of the box, I've tested cinebench 11.5.**= * >> * >> >> ** ** >> >> I've got a Asus GTX680 DC2G 4GD5 (with 4Gb DDR5) in the machine, and hav= e >> installed the NVidia 306.97 driver.**** >> >> I'm running Win7 x64 ultimate.**** >> >> ** ** >> >> The results are:**** >> >> ** ** >> >> OpenGL: 32.61 fps**** >> >> CPU: 13.36 pts**** >> >> ** ** >> >> I'll run other benchmarks after the weekend.**** >> >> ** ** >> >> kind regards,**** >> >> Marcel van Berk**** >> >> ** ** >> >> *Hoge Bomen/LSSLSS* >> >> ** ** >> >> info@lsslss.nl**** >> >> ** ** >> >> www.lsslss.nl**** >> >> > > > -- > Animator & Editor > www.teddygage.com > Brooklyn > > --14dae93b6014e29d5204d05cc3dd Content-Type: text/html; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable

More RAM could help, but absolute performance on AMD historically trails= absolute performance on Intel. Performance/cost or performance/power might= be a different story, though.

Also, I wouldn't be surprised if CINEBENCH scales better over ever-m= ore cores than Ae does -- I'd think there's often much more raw com= putation and less I/O=A0 or inter-process communication with CINEBENCH or m= ost 3D renderers than Ae.

Walter Soyka

On Dec 8, 2012 2:19 PM, "Teddy Gage" &= lt;teddygage@gmail.com> wrote= :
Yes it is a bit mysterious which is why I wanted him to try a real world AE= project.

My overclocked sandy bridge i7 3930 K, with the same graph= ics card gets 55 FPS and a 12.74 CPU score in cinebench, and cost around $2= ,200... I feel like something is missing? It should be at least 2x faster, = unless AMD has really fallen far behind.

Did you run the 64 bit version of cinebench? are all cores firing 100% = if you open the task manager performance view? is it 16 physical cores or 3= 2 physical?

Could be possible the cores are starved for RAM....

On Sat, Dec 8, 2012 at 2:00 PM, S=F8ren Chri= stensen <soren@desilence.net> wrote:

That w= as a bit low cpu score for 32 cores,no? Maybe cinebench isn't scaling w= ell with so many cores?

( My 4 core MBP from mid 2011 clocks in at OpenGL 42,61= fps and CPU 5.53 )


Cheers,
S=F8ren


|||| =A0MOTION & INTERACTIVITY
|||| =A0dslnc studio = - www.dslnc.com
|||| =A0+34 93 268 0953 =A0/ =A0+34 61 555 9963 (m= obile)

|||| SOCIAL:
=

On 08/12/2012, at 18:15, "lsslss" <info@lsslss.nl> wrote:

hi All,

just got m= y 32core in, and out of the box, I've tested cinebench 11.5.<= /u>

=A0=

I've got a Asus GTX680 DC2G 4GD5 (with 4Gb DDR5) in the machine= , and have installed the NVidia 306.97 driver.

I'm ru= nning Win7 x64 ultimate.

=A0

The result= s are:

=A0

OpenGL: 32= .61 fps

CPU: 13.36 pts

=A0=

I'll run other benchmarks after the weekend.

=A0=

kind re= gards,

Marcel van Berk=

=A0=

Hoge Bomen/LSSLSS

<= u>=A0

= info@lsslss.nl=

=A0<= /p>

www.lsslss.nl<= u>




--
Animator & Editor
www.teddygage.com
Brooklyn

--14dae93b6014e29d5204d05cc3dd--