Upgrading mono Xeon to Dual. Ram Management & Benefit Dual CPU for After Effects?

[xavierFr06]

Hi,

I use AE

I have mono Xeon E5-2687w (8 core, 16 thread) CPU running at 3.4Ghz, with 64 Gb Ram (8 x 8 Gb) and 1xTitan GTX, with Windows 10 on SSD.

In AE i have chose 3 Gb per thread (16x3=48Gb) for AE and keep 12 Gb Ram for Windows.

I make footage with Sony action Cam Sony FDR-X1000VR, in 4K 100Mbits/s. Xavcs, 30p, MP4,H264.

The computer need 53 minuts,for 5minuts and 33seconds of video, with effect for keep out Fisheye effect, as like it :

https://www.youtube.com/watch?v=stmzNSCRI_M

and i keep 4K from Effect by this way :

https://www.youtube.com/watch?v=BPE8jWJg-Os

But i begin more long video than 5:33 and the computer need few hours.

I think to will upgrade my computer with a second CPU Xeon E5-2687w ( 8 core, 16 Threads), then, 2 x Xeon E5-2687w ( 16 core, 32 Threads).

The next motherboard : Z9PE-D8 WS

-My first problem, this motherboard have only 8 (2x4) Ram slot, 4 slot for CPU1 and 4 slot for CPU2. Then, i was thinking to put 4x8 Gb per CPU in the motherboard . 32Gb for CPU1 and 32 Gb for CPU2. In AE i can chose the number of CPU thread for AE/System and to give 0.75; 1; 1.5; 2; 3 Gb per thread for AE and keep Ram for Windows. i can select 1.5 Gb per thread in AE, 32thread x 1.5Gb=48Gb for AE, and i keep 16Gb for Windows.

From yours past experience of multi CPU, i don’t know if the système need so much 24Gb, and AE so less 1.5Gb per thread. Do you see any problem in my project ? it will run well ?

-My second problem, AE, for 4K video and few function, use well the 12 at 16 core maximum. Arround 16 core, after the benefit fall. I can keep few core for Windows and i will be under 16 core for AE, then i can have :

-- 14 core for AE, 28Thread is 28x2Gb = 56 Gb for AE and 8 Gb for Windows

-- or -- 13 core for AE, 26Thread is 26x2Gb = 52 Gb for AE and 12 Gb for Windows

-- or --12 core for AE, 24Thread is 24x2Gb = 48 Gb for AE and 16 Gb for Windows

By this way for Ram management, I make a mistake or how run Ram for AE and the 2 CPU ?

for example, if i needed 10Gb for Windows with my past mono Xeon E5-2687w, then, with dual Xeon E5-2687w,  i will need 5Gb for CPU1 and 5 Gb for CPU2, or i will need 10 Gb for Cpu1 and 10 Gb for CPU2 ?

Du you know some thing about it?

From yours past experience, what will be the best Ram per thread for AE, maybe for Windows it will be half or double per CPU ?

-My third problem, i have AE CS6, and some test show to need to use 2014 CC 13.2 for multi CPU, and not after with 2015 CC. I have estimate with dual CPU, 37% minimum of benefit, but possible more benefit. Do you estimate the % benefit value i can hope? or from type of job in AE?

Today, from i see about AE need between 12/16 core maximum, from the incredible price for Xeon V2, V3, V4 upgrade for make dual CPU, i can upgrade my computer for cheap and to be near of the maximum performance it is possible to obtain from AE. I keep some high frequency CPU and dual CPU. I don’t hope from Intel 16 Core CPU with much more than 3.5 Ghz… not soon and not for cheap. I use my computer for lot of thing, and the benefit fall after 12-16 CPU core.

I think in futur to begin effect and some 3D.

If you have any council over my first point of view for this project…

Thank

Xavier

P.S.: Sorry for my English level...

[Mylenium]

I'm really not sure what you are asking. Throwing around these theoretical considerations is of no use to anyone. You won't know anything until you actually try and given, how limited AE's multi-threading is and that there are other things to consider like file I/O, rendering to a clip-based format and so on, another CPU may have zero effect. Honestly, and no offense, you are not thinking straight here and obsessing about technical minutia that quite likely don't have any practical relevance, especially since we don't know what your projects will actually look like. It's one thing to unwarp some footage or render a text slate, but a completely different thing to create complex projects with lots of pre-comps and heavy effects...

Mylenium

[xavierFr06]

Hi,

Thank for your answer

in First, I am interresting by the experience of peolple from here. I am not looking for the best answer and for to be sure at 100%. I think with simple example, the people have some experience from one of her past projet using in 2 different performance hrardware (in same type of her project). I am not looking for point of view with big projet or effect to much complex.

i have CPU, mother board for 1 CPU, and 64 Gb Ram.

If i upgrade i will need second CPU, a mother board for 2 CPU and second air cooling, and it will be a cheap upgrade performance, but, before to buy, i wanted to konw the upgrade memory i will need and perfomance i will obtain.

After my first experience and message and your answer (Thanks for your answer), i was looking for other experience in internet for confirm my experience and all i was writing in my first message.

For it, i needed obtain the confirmation about the more performance i can hope. I was thinking to obtain arround 37% with a second CPU. The next article show possible 50% at 60%. But i am interresting by the experienc of people and if they can confirm ?

For to know if 64 Gb will be enought with second CPU (because second CPU will need more RAM for windows system, not Adobe, for same job than 1 CPU), the next articles show my 64Gb will be OK (for the more Ram will need Adobe and will need Windows System in Dual Xeon), but 3D it will be not enougth with 3D. But i am interresting by the experience of people and if they can confirm ?

For the Ram management from Windows, i not have find fully answer.

From i have find, if (for example) Windows needed 10 Gb with mono CPU, then i will not need 10 Gb for CPU1 and 10 Gb for CPU2 with dual CPU for Windows system. But i will need minimum 10 Gb. Then i will need between 10 Gb and less than 20Gb for Windows system in Dual CPU. But i am interresting about the experience of people.

For my project to obtain more performance for cheap, i don’t know if i will have satisfaction to hope more performance with old component  and to not upgrade to new more recent component. For it, t need to change CPU for more expensive. For me, i think my cheap solution will be better than to expensive new component. Old component will be to stay in the past performance. The new componant wil be to see the futur but the price are more important. But the people with her experience can have more council about to chose between old technology component and new. I am interresting by her point of view. With past mother board, i will have today 64 Gb only with Asus Z9PE-D8 WS (8 slot ram only).

From my first experience, my cheap projet look like possible for nice more performance with only 64 Gb… and the information i have find from internet, show it to. Else, i make mistake ?

I am not looking for the best answer and for to be sure at 100%. I think for simple example, the people have some experience for the same type of her project in different hardware solution. I not have the answer for all and i not have confirmation if my information are more real than impossible.

If the people have some experience or council, i am interresting.

Thanks

Xavier

1. P.S. : I understand, it need to not create doubt. I am not looking for it.

for my fist post, i had some idea from my experience, but i have find article can explain the information i was looking for. The next informations are only as like reference with , i think, from simple example for give comparison with out to use big complex projet… but a first level of point of view, as like i was looking for here :

-------------------------------------------------------------------

-------------------------------------------------------------------

Here, i understand After Effect can manage less than 64 thread (and for 4K):

http://www.intel.com/content/dam/www/public/us/en/documents/guides/workstation-adobe-4k-guide.pdf

For Large-Frame Video Editor, it need For 1 CPU 8 core, it need 21 Gb, 1Gb video Ram. For 2 CPU 8 core, it need 42 Gb, 4Gb video Ram.

i have 64 Gb and 6 Gb video Ram. Then it is ok for 1 or 2 CPU 8 core each.

Then i will need 2 CPU 2687W (8 core) with minimum 42 Gb, 4 Gb video Ram.

But For Motion Graphics/3D Animations and Compositing, it need For 1 CPU 12 core, it need 64 Gb, 2Gb video Ram. For 2 CPU 24 core, it need 256Gb Gb, 12Gb video Ram.

i will have 2 CPU 8 core each, it need arround 85 Gb and 12 Gb vidéo Ram. I will have 64 Gb and 6 Gb video Ram.

Then it is ok for 1 CPU, but not for the 2 CPU and video.

But, it is writing « A minimum of 3 GB per available core is recommended, and it can be set as high as 6 GB. » and after in the array it is wrtiting other information « 3 GB per thread for Adobe After Effects* multi-processing support. »… then, i don’t know about minimum, 3Gb per Core or Thread?

3 Gb * 16 core = 48 Gb, then i am ok.

3Gb * 32 thread = 96 Gb, i am bad…

--------------------------------------------

Here :

Deciding between a single CPU or dual CPU

http://ppbm7.com/index.php/tweakers-page/95-single-or-dual-cpu/109-single-or-dual-cpu

i understand Adobe can use 64 thread maximum. Then 2 CPU with 16 each core maximum. Then it is ok for me because i have only 2x 8core. And i will obtain 50 et 60% more power, It is ok and for cheap.

Here, i am not fully bad :

http://international.download.nvidia.com/adobe/pdf/adobe-hardware-performance-white-paper.pdf

and my titan , and a second, will save me.

And next confirm about maximum number of 2x 12 core max CPU, then i am no bad with 2x 8 cores = 16 core, with only 14 or 12 for after effect

Adobe After Effects CC 2014 Multi Core Performance :

https://www.pugetsystems.com/labs/articles/Adobe-After-Effects-CC-2014-Multi-Core-Performance-716/

Adobe After Effects CC 2015.3 CPU Comparison :

https://www.pugetsystems.com/labs/articles/Adobe-After-Effects-CC-2015-3-CPU-Comparison-847/

Adobe After Effects CC 2015.3 Multi Core Performance :

https://www.pugetsystems.com/labs/articles/Adobe-After-Effects-CC-2015-3-Multi-Core-Performance-843/

Adobe After Effects CC 2015 Multi Core Performance :

https://www.pugetsystems.com/labs/articles/Adobe-After-Effects-CC-2015-Multi-Core-Performance-714/

and After Effects CC 2014 (13.2) maximum are better for multi core.

features not available in After Effects CC 2015 :

https://blogs.adobe.com/creativecloud/features-not-available-in-after-effects-cc-2015-13-5/?segment=...

And Premiere

https://www.pugetsystems.com/labs/articles/Adobe-Premiere-Pro-CC-2015-Multi-Core-Performance-Update1...

https://www.pugetsystems.com/labs/articles/Should-you-use-a-Dual-Xeon-for-Premiere-Pro-CC-2017-932/

they suggest 2 CPU E5-2697W V3, 10 core, more hight Ghz. I understand i will have more performance with second CPU than I7-6950X, but no so much, but i will have more cheap solution becaus i need only other mother borad with air cooler and the I7-6950X will be very much more expensive… with mother borad + new 64 Gb ram DDR4.

------------------------------------

Deciding between a single CPU or dual CPU

http://ppbm7.com/index.php/tweakers-page/95-single-or-dual-cpu/109-single-or-dual-cpu

it need for 2 CPU 8 core, 85 Gb… and i will have only 64 Gb

Building a 32-Thread Xeon Monster PC for Less Than the Price of a Flagship Core i7

http://www.techspot.com/review/1155-affordable-dual-xeon-pc/page8.html

same question than me

http://www.tomshardware.co.uk/answers/id-3196211/2670-xeon-render-machin.html

Asus Z9PE-D8 Owner's thread

http://www.overclock.net/t/1261060/asus-z9pe-d8-owners-thread/180

mother board Supermicro X10DAL-i

http://www.supermicro.com/manu"als/motherboard/C600/MNL-1701.pdf

see "System Block Diagram 1-10"

FUJITSU PRIMERGY SERVERS MEMORY PERFORMANCE OF XEON E5-2600/4600 (SANDY BRIDGE-EP) BASED SYSTEMS

https://sp.ts.fujitsu.com/dmsp/publications/public/wp-sandy-bridge-ep-memory-performance-ww-en.pdf

-Memory Management in NUMA Multicore Systems: Trapped between Cache Contention and Interconnect Overhead

http://people.inf.ethz.ch/zmajo/publications/11-ismm.pdf

In a NUMA system a process’s data can be allocated in the memory of any processor in the system. We say that a process p is homed on Processor i of the system if the process’s data was allocated only on Processor i. If a process runs on its home processor, it is executed locally. Similarly, if a process runs on a processor different from its home processor, it is executed remotely. F

From i understand, we not have double ram Windows quantity needed in dual CPU.

- Dual CPU vs Single CPU

https://www.youtube.com/watch?v=L18Ck_tH-uA

- CPU Benchmarks

Multiple CPU Systems

http://www.cpubenchmark.net/multi_cpu.html

Single Thread Performance

http://www.cpubenchmark.net/singleThread.html​

mono to dual CPU make more performance 42.5% for 2687w

-for between i7 and Xeon

the i7 extreme cpus are faster than some xeons, but 2 xeons together are more powerful than a single i7. With dual xeons you get more threads and similar speeds to the i7.

[Kevin-Monahan]

Hi XavierFr06,

I have moved your post to the Hardware forum. I hope your post gets a bit more traction here.

Thanks,
Kevin

[Bill Gehrke]

Take a look at my PPBM test data on the CPU page for some results of the Premiere Pro CPU intensive scores

[JEShort01]

Bill,

Back in December 2015 I had run PPBM on a dual e5-2687w v2 and the CPU score was 173, much faster than you show in the PPBM test data CPU page. You should probably update it. Here's my really old thread with this test score:

Dell T7600 Super Titan & CC = fast, fast, fast!

Xavier,

Bottom line - don't do it! (the upgrade) I'll add some more specific comments below which are based on running dual Xeon PCs for Premiere Pro and other Adobe CC applications for more than a few years.

I'll expand on some of my various specific thoughts regarding where you are and what you are considering:

- your current system is very balanced - that's good for Premiere Pro and After Effects

- And with a balanced system, it is rather difficult and expensive to add parts that will increase the whole balance

- Adobe CC in 2017, as others have pointed out, including the links that you share above, is able to utilize both CPU and GPU

- And interestingly, adding more CPU cores than 6 or 8 cores (+ associated Hyperthreads) really doesn't start becoming really that useful unless you are cutting on RED 6k or RED 8K media (your Sony 4k media is really not that taxing)

[note: read the following thread carefully and my "correct answer" for it:

Premiere 2017 not using all CPU when exporting H.264

- This means that if you really want to achieve a significant improvement in performance for your specific media and workflows, you would want to go with a fast 8 or 10 core (i.e. i7-6950x overclocked so all cores are running at 4.2 GHz); I know this would be much more expensive than the upgrade that you are considering, but: you could sell your current build and recoup some value from that, a 2017 build with DDR4 memory, faster USB, etc., you would have a build that could be better incrementally added to in the future

- e5-2687w is 2013 technology; it was great in its day, but it is still 2013 technology - PCs, drivers, etc. doesn't change that much year to year, but face it, it's been 4 years now

- I've used systems with 32GB, 64GB, 128GB, and even 256GB and Adobe CC only rarely benefits from anything past 64GB

- Reviews on the Asus Z9PE-D8 WS motherboard are really pretty poor (I was reading from Newegg's site for this discontinued motherboard)

Cheers,

Jim

[xavierFr06]

Hi ,

Thank Bill Gehrke.

Test 1 Disk Test : 1:15 (75sec)

Test 2 MPEG2-DVD : 7:52 (472sec)

Test 3 MPEG2-DVD : 7:48 (468sec)

Test 4 H264 : 12:34 (754sec)

It is the same test here with PPBM6 ?

Have I a problem performance with my computer?

https://forums.adobe.com/thread/1273580

Cuda and GPU render test exist ?

Thanks JEShort01...

Sorry , I am french. I hope to not make a mistake of understanding...

I have read all you show me and i think it is the same conclusion than my links, about high frequency, number of core...

it is sure, E5-2687w is 2013 technology. mine run at 3.4Ghz every core, and some time at 3.8Ghz with 1 core.

I understand about i7-6950X.

About render with GPU, more GPU way SLI Titan i have same preformance, or more, than 2 GTX 1080 for render) and new video card (GTX 10xx) permit only 2 SLI (Maybe, only Titan X Pascal permit 4 SLI), GTX Titan have double precision, 6Gb for permit 4K render... At the end, First Titan by Cuda compatibility (only Titan Black have 10% more powerfull, but the price are more higer than 10%, and first Titan can increase her performance by 19% ),  double precision, 4 SLI, low price... for me, Titan are the best.

GTX 1080 (since GTX 9xx) , not have double precision, no full Cuda compatibility, only 2 SLI, expensive...

But, 4 Titan SLI need 4x16 Pcie lanes = 64 lanes. Only 2 Xeon permit it.

From i have find, there are 2 motherboard in the wrold permit to obtain true 4 x 16 Pcie:

-Asus Z9pe-d8 for Xeon E5 V1 and V2.

-Supermicro X10DRG-Q for Xeon E5 V3 and V4.

Few other motherboard permit 3 SLI.

All other motherboard use 32 Pcie lanes, ou 40 Pcie Lanes, and it is the why we obtain 2.5 x GPU performance with 3 or 4 way SLI. A fake 4 SLI permit only to use all performance of bandwidth from 32 or 40 lanes (SSD M2 and other use Pcie performance bandwidth). Fake 4 way SLI run as like it,.. if one GPU are not at 100% of 16 Pcie lanes Bandwidth ( clock delay or other...) , then the other GPU will use more Bandwidth but with a maximum of 40 lanes bandwidth. in any way, we have 2.5x GPU maximum and less. The Pcie can modulate the bandwidth between all Pcie lanes. It is the why, there are no différence with some test on internet with Pcie 2.0 and 3.0.

Fake 4 way SLI permit 2.5 GPU performance maximum, but we lose arround 14%, as like it. then, we have 2.2x GPU performance...maybe, it is the why the news GPU permit only 2 SLI...but 2 SLI permit to hope 2/1.14= 175% GPU performance.

With true 4 SLI x16 Pcie, we obtain 4x GPU - 14%, ( 4 /1.14) then 3.5x GPU performance.

we obtain a difference of performance between true and fake 4 way SLI x16 Pcie :

3.5/2.2 = 160%

New next Xeon and Pcie 4.5 , 5.4 will not permit more... maybe in 2020 with very expensive new Xeon.

i think the chose are between CPU (+ 140% of performance, possible more) or Cuda (+ 160% of performance) performance or CPU and Cuda if we not have price limit:

1 x CPU I7 , the most high frequency and arround 8 core, with 2 SLI

or

2 x CPU XEON V1 or V2 , the most high frequency and arround 8 core, with 4 SLI

or

2 x CPU XEON V3 or V4 , the most high frequency and arround 8 core, with 4 SLI

... and after it is the price/performance.

I had and have Intel Xeon E5-2687W V1. if i need to change, i need to think to video render and other software.

I wanted CPU: Intel Xeon E5-2697 v3 or Intel Xeon E5-2687W V4, for me, betwen 8 and 14, is ok for other software.

from single core, Xeon V3 and V4 architecture, some time, with same core and frequency, we obtain 20% more performance per core.

By frequency and core, with E5-2687w Vs E5-2697 V3, E5-2687w V4, in theorical with  benchmark software test , there are more power full, 190%.

But, for true test, i can see only 140%. Lot of bechmark not give the real performance, but a proportional result at frequency and core number.

I am not sure to obtain very much more performance with other software.

for the price

in any way, for build it or update, i need same PSU, cooling... but for dual Xeon V1, i need E5-2687w ... it will coste me 300$ x1 .

for E5-2697 V3 , i will need 1000$ x 2 + 64Gb DDR4 ecc... 2800 $

then, 140% of CPU performance for a more bigger price...

http://www.cpu-monkey.com/en/cpu_benchmark-cinebench_r11.5_single_core-2

https://asteroidsathome.net/boinc/cpu_list.php

http://cpu.userbenchmark.com/SpeedTest/19642/IntelR-XeonR-CPU-E5-2697-v3---260GHz

http://cpu.userbenchmark.com/SpeedTest/8860/IntelR-XeonR-CPU-E5-2687W-0---310GHz

From "Bottom line - don't do it! (the upgrade) I'll add some more specific comments below which are based on running dual Xeon PCs for Premiere Pro and other Adobe CC applications for more than a few years.".

from i had understand, the best are 2014cc 13.2 for multi CPU core, not after.

but not next after 13.2.

you means over 13.2.... CC2015 ... CC2017 ?

About "- And with a balanced system, it is rather difficult and expensive to add parts that will increase the whole balance".

If i not make a mistake of understanding, i think the best CPU are between E5-2697-v3, E5-2687W-v4. we can decrease the number of CPU and to keep high frequency. and for the other soft ware, we have lot of performance.

But, there are other software, and betwen E5-2697-v3, E5-2687W-v4 or to not change E5-2687w for cheap... it is not easy.

Today, for GPU Cuda performance, it will be 4 true way SLI. for it, it will be Xeon E5-26xx.

For other soft ware it will be Dual Xeon.

For performance it will be maximum 16 core/CPU, the minimum will be 8 core/CPU.

About Ram, yes, 64Gb look like ok.

From my information, 3Gb/core are enough. the Win10 system need 8Gb maximum. 64-8= 56 Gb. 56/3=18 core.

About "Reviews on the Asus Z9PE-D8 WS motherboard are really pretty poor".

With the forum, I have find lot of problem about Asus dual xeon and few true SLI.

Z10PE-D8 WS, have only 4x SLI Pcie 8 lanes.

Z10PE-D16, have only 3x SLI Pcie 16 lanes.

Z9PE-D16 , not make SLI

and this 3 mother board have important issue.

I have find some issue with Asus Z9PE-D8 WS. But less than the Z10PE-D8 WS, Z10PE-D16, Z9PE-D16 and with less important problem. But, it need to take last Asus Z9PE-D8 WS version.

i have find from Z9PE-D8 WS users lot of mistake, usuali with Ram model (problem of Boot and other), use only 1 Cpu...

what is the problem you know about Z9PE-D8 WS ?

What you means about "really pretty poor"?

I have find it "Since CS6 does not support multiple video cards," here :

http://ppbm7.com/index.php/choosing-cpu-mobo-memory

From few Cuda test, SLI run...no ?

Thank

Best regards

[xavierFr06]

With this link from  Bill Gehrke, i have check few performance:

CPU Ranking

there are 43% performance between E5-2687w V1 and 2697 V3. and other CPU permit to obtain same conclusion.

the result confirm the increase performance with more high clock and number of CPU.

[Bill Gehrke]

Jim,

I will add your v2 to the list I suspect the posted score was for a v1

[Bill Gehrke]

Jim Thanks much I posted your results see if it looks right now,

Clock speed is important

[xavierFr06]

JEShort01 , i have read again your council.

I'm no video expert and Adobe or other.

About old and new CPU technology…

Since 2 yaers, I was thinking to wait the next 2018/2020 CPU.

In first, it was for PCIe 4.0, PCIe 5.0, More CPU performance, More PCIe lanes per CPU, 4K H265 10 bit encoding…

But, for obtain  more PCIex16 it need Xeon.

With next Xeon V5 and V6, Intel add more Core, but not more frequency. Only the new architecture permits more performance. I can see the performance betwen CPU with Cinebench , but the performance are no so much between CPU in some application. And next Xeon will not have high frequency.

http://wccftech.com/intel-broadwell-ep-xeon-e5-2698-v4-processor/#ixzz44y2WyJSg

end 2017/2018, we can hope 48 PCIe lanes with Xeon V5 and V6 with PCIe 3.0, not 4.0 and not 5.0. But it will be possible to have 3 way SLI per CPU. But I don’t think we will have, with dual Xeon, a motherboard with 6 x16PCIe slot. The new CPU will be very expensive and it will need to wait.

From I have seen, CPU H265 10 bit encoding, will be not 2018, maybe 2020.

PCIe 4.0 are very late… I not speak about PCIe 5.0.

Today, we can obtain nice Xeon for cheap but there are only Supermicro x10drg-q and Asus Z9pe-d8 WS for 4 true way SLI. The ultimate chose come from Xeon CPU price and performance. The difference between this 2 motherboard with Xeon V1 V2 V3 V4 or i7-6950X… there are the same technology, but we have some more USB3, Sata3… but, I not have seen so much more benefit. Possible I make a mistake… about NVME between Xeon V1 V2 and V3 V4 i7-6950X, I have create my Bios for AHCI as like NVME, there are some little difference.

For i7-6950X and difference with Dual 2687w V1:

http://ppbm8.com/CPU/CPU.html

I see only a difference with price of i7-6950X. and I will not have 4 way SLI. And some test not show  i7-6950X at the best and not so often and no so much. Else, I make a mistake.

With out to speak about CC 2015, but, only performance between CPU:

https://www.pugetsystems.com/labs/articles/Adobe-After-Effects-CC-2015-3-CPU-Comparison-847/

From the price, the performance of i7-6950X are not so much, but the I7-6700K… Waow!

But without 4 true way SLI

About Adobe CC in 2017…need to wait (again) more years…But i have learn a understanding mistake from me.

From Szalam :

The old CS6 version of After Effects that you're using doesn't use CUDA or the GPU for any of this.

The newest version that came out last month can use the GPU to accelerate the Sharpen, Gaussian Blur, and Lumetri effects (details here) and that's about it. But it doesn't have to use CUDA for these; it can use Open CL too. (And Metal on Macs.)

Thanks Szalam

https://helpx.adobe.com/after-effects/system-requirements.html

For me GTX Titan are the best Cuda compatibility with lot of render. But, there are lot of render application with GPU only in CC 2017. I was hope to use CC2014, but there are not so much render by GPU. The minimum are CC 2015…

https://blogs.adobe.com/creativecloud/after-effects-cc-2015-3-in-depth-gpu-accelerated-effects/

https://en.wikipedia.org/wiki/Adobe_After_Effects

About Balanced system… the chose, if i not make mistake, are betwen 4 way SLI and to not brake the Balanced system. with unactivate core (without Hyperthreading) for keep 4 way SLI, better clock speed, it will permit better the performance with software as like Adobe AE and premiere.

I am ready for have council about to take Xeon V3 or V4 and not V1 and V2. But what I will obtain by V3 and V4?

… 40% minimum…90% maximum and not all time. I will like to buy dual Xeon 2697V3 or  dual Xeon 2697A V4. But with next Xeon, Intel add core, no more frequency, and the price go up.

I can hope 5% performance from new DDR4 memory or other Bandwidth. But it is no so much for expensive evolution.

When I see your performance with “Dell T7600 Super Titan & CC = fast, fast, fast!

“, I not understand the problem of my system with the half CPU and GPU, and so much I am far from your performance.

Do you have built new system, and what it is your new performance?

You have a same system I want to build. i can not understand, why to not build same?

Do you have build one other system and obtain 100% more performance?

From i have seen to build 4 true Sli 16xPcie betwen to build cheap with E5-2687W and most powerfull with E5-2697 V3 for most powerfull utilisation, i have 2 solutions with Asus Z9pe-d8 WS or Supermicro x10drg-q.

From E5-2687w V4 (Freq all core 3.2)  , or  2697V3 (Freq all core 3.1). by number of core, the 2697V3 will be more powerfull with 13% with lot of thing , by frequency 2687W V4 will be more powerfull with 3%. And the 2687V4 cost more... In any way we can not have the most power full every where with 1 CPU type. But it is possible to find a high middel powerfull.

My projet are to build

2 x CPU XEON V1 or V2 , the most high frequency and arround 8 core, with dual E5-2687W

and 4 Titan SLI

or

2 x CPU XEON V3 or V4 , the most high frequency and more than 8 core, with dual E5-2697 V3

and 4 Titan SLI

For the 2 projet, i need to buy:

1 PSU, 1500W for the 4 way SLI and Fan.

1 tower, i will like to keep the motherboard horizontaly

2 Noctua NHD15

For

2 x CPU XEON E5-2687W V1.

I need:

1x E5-2687W

For

2 x CPU XEON V3 E5-2697 V3

I need:

2x E5-2697 V3

64 Gb Ram

1x mother board Supermicro x10drg-q

My problem with Supermicro, i will not find some much bigger communauty for help and some update driver and Pionner with this motherboard... as like...

http://forums.guru3d.com/showthread.php?t=409468

If I try to sale all i have... E5-2687 arround 200€, Asus Z9pe-d8 WS 400€, 64 Gb DDR3 600€ ?

=1200€

buy 2 E5-2697 V3 2000€,  Supermicro x10drg-q 700€, 64 Gb DDR4 1100€?

= 3800€

3800-1200= 2600 €... for obtain betwen 140% and 190% more powerfull?

A mono core give best performance, as like high frequency.

From Cinebench (but I think they not use Turbo)

-Dual 2687w V1 obtain 2263 cb 2x8 core--> 141 cb/core … Most cheap CPU

(but I think they not use Turbo, because I have 1200 mono CPU, and arround 2400 in Dual  150 cb/core)

https://www.youtube.com/watch?v=e8sHvYY33tc

-Dual 2687w V2 obtain 2427 cb 2x8core-->151.6 cb/core … more than double price than the most cheap CPU

https://www.youtube.com/watch?v=GLCX5k3LTUc

-Dual 2687w V3 obtain 2970 cb 2x10 core--> 148.5 cb/core … more expensive than the most cheap CPU

https://www.youtube.com/watch?v=aVu1UZpyCyY

-Dual 2687w V4 obtain 4086 cb 2x12 core--> 170 cb/core… Winner but very expensive CPU

https://www.youtube.com/watch?v=5L5BHCW8XHc

-Dual 2697 V3 obtain 3871 cb 2x14 core-->102 cb/core… Loser expensive CPU

https://www.youtube.com/watch?v=fA5ZWPLtfTc

           2687w V1 are no so bad ? No?

From i can see, the best benefit are to put dual Xeon, it will add arround 60% of CPU performance and 4 true way SLI, it will add arround 60% of GPU performance. Xeon 2697 V3, maybe will add 40% or more.

Note :

ASUS-Z10PE-D8-WS Quad SLI multi-graphics card setups only support x8/x8/x8/x8 configuration

http://news.softpedia.com/news/The-Problems-With-ASUS-Z10PE-D8-WS-Dual-Socket-Motherboard-459168.sht...

Expansion slots : 4 x PCI Express® 3.0 x16 slots (dual at x16/x16; quad at x8/x8/x8/x8)

SPECIFICATIONS[1]

[1] Specifications, content and product availability are all subject to change without notice and may differ from country to country. Actual performance may vary depending on applications, usage, environment and other factors.

https://www.asus.com/News/ChCMUvu3bWfJTIfq/

quad at x8/x8/x8/x8 = 32 Pcie lanes only

ASUS-Z10PE-D8-WS less performance than mono CPU with X99-E WS

-ASUS-Z10PE-D8-WS Three SLI multi-graphics card setups only support x16/x16/x16

Correspond to 3 GPU / 1.14= 2.63 GPU performance with 48 Pcie lanes.

Any other motherboard single CPU with Fake Three Quad SLI multi-graphics card setups only support x16/x16/x8 Pcie

Correspond to 2.5 GPU / 1.14= 2.19 GPU performance with 40 Pcie lanes.

ASUS-Z10PE-D8-WS , with 2 Xeon, add 2.63/2.19= 1.20… add 20% SLI performance and need 2 expensive Xeon V3 or V4.

ASUS-Z10PE-D8-WS useless for 3 SLI, same problem from Gygabyte.

Thanks

Xavier

[JEShort01]

Xavier,

It's a bit cryptic, but it seems that you are still running CS6 and for the time being not interested in upgrading to CC. In my opinion, that is a big mistake - re-read the Super Titan Dell T7600 thread yet again for my vintage 2013 reflections on CS 5.5 vs. Adobe CC vintage 2013. I don't have any specific proof or benchmarks in the last 4 years regarding CS vs. CC, but while Adobe development is not perfect (sometimes it's 4 steps forwards and 1 step back), the current CC stuff is really great software and I personally would NOT want to go back to CS for 4k projects. Top-drawer dual-Xeons, SSDs, and powerful video cards can't change that either.

As you seem to gravitate to old, but powerful hardware, you may want to seriously look into Black Magic's DaVinci Resolve if you remain opposed to paying CC's subscription pricing. I'm not a Resolve user, but hear great things about it for simple cuts and extensive color correction. I have no knowledge whatsoever regarding whether it can do things like GoPro lens correction and other things you mention in your workflows. Or, if you must remain with CS6, simply convert all of your media to a format that CS6 likes (Sony's 4k consumer CODEC and GoPro's 4k did not even exist back when CS6 was developed, so it is no surprise that it does not excel when using those formats).

You continue (in this thread, and likely in your own mind) to dwell on all kinds of unrelated benchmarks, quad-SLI (how's this relate to running CS6), and more. Mylenium said this in his initial post to your thread and I'm stating it again.

Suggestions:

- Get your current rig working and performing like it should; your PPBM test figures are poor; start a new thread if need help with that goal, don't add that here (this thread is too confusing already)

- Re-consider moving forwards from vintage 2013 software to CC

- Condense your future posts on this Adobe forum to a paragraph, or two, or three; your "blog" style of writing is VERY time-consuming to wade through; anyone helping here is doing it to be nice (nobody's being paid to help you here)

I do have newer systems. I'm choosing not to mention specifics here since I think it would contribute very little to the subject of this thread. I would enjoy building another benchmark scorcher for education and for fun, but work and life are busy and it seems that I never have the time. If I do another "super" build in 2017, it will start with either an overclocked x99 or a dual-Xeon e5-2687w v2.

This will be my final post to this thread.

Good luck!

Jim

[xavierFr06]

Hi JEShort01

Thanks for answer.

As like i was writing, i think to use 2014 CC 13.2. But it will come with new server.

for 2015 cc, and more, today, i don't know. i need other server and to begin to try 2014 CC 13.2.

"I do have newer systems. I'm choosing not to mention specifics here since I think it would contribute very little to the subject of this thread. I would enjoy building another benchmark scorcher for education and for fun, but work and life are busy and it seems that I never have the time. If I do another "super" build in 2017, it will start with either an overclocked x99 or a dual-Xeon e5-2687w v2."

I understand, old Dual  Xeon are no so bad but need to compare with mono overclocked CPU.

I understand, it is possible to overclock I7-6950X. there are some better performance, but some.

For me, i need to run other appliation/software and SLI. it is complicate to not use Xeon.

I have understand about more small, only 1 subject, no a lot, in next thread.

I will continu to find answer in other way. after, i will try to make more small thread.

Thanks

Best regards