17 Replies Latest reply on May 14, 2012 3:04 PM by ECBowen

    Xeon benchmarks

    JCschild Level 3

      ok after much fun (sarcasm), delays, too dang busy to get to it. Dual Xeons along with Ivy etc.



      Premiere Pro CS5 Version 5.5.2Testing

      4 WD 1Tb Sata 64 Meg Cache 600 Drives in 2 Raid 0 arrays

      Video material - AVCHD 1080P 24 Frame Each Cut to 30 minutes of material

      Export Codec - H264 HDTV 1080P 24 Preset Default

      4 Effects per Layer - Fast Color Corrector, Brightness & Contrast, Video Limiter, Sharpen

      Each Layer Scaled to 50% for 4 frame PinP view.


      Dual Xeon E5-2760 2.6GHz 8 core x2. 16 cores 32 threads

      64gig  1600 ram

      570 2.5G

      3 Layer - 36:53

      4 Layer - 38:07

      basically on par with a stock 990


      SINGLE <--  E5 2670 @ 2.6 GHz

      32GB  1600

      570GTX 2.5GB

      4  1Tb Sata 32 Meg Cache 600 Drives in 2 Raid 0 arrays

      3 Layer -

      4 Layer - 40:41



      I7 3770 4.7 GHz

      16GB Top Power 1600

      570 2.5G

      4  1Tb Sata 32 Meg Cache 600 Drives in 2 Raid 0 arrays

      3 Layer – 29:45

      4 Layer – 32:58


      X79 3.3 @ 4.5 GHz

      32GB  1333

      580GTX 3GB

      4  1Tb Sata 32 Meg Cache 600 Drives in 2 Raid 0 arrays

      3 Layer - 27:43

      4 Layer - 30:02


      so anyone want to buy a Xeon..?


      will ask Intel for a pair of 3.1GHz but rather pointless




        • 1. Re: Xeon benchmarks
          FloridaG Level 1

          I have a new server I've built here with Dual E5-2640's on a Supermicro X9Dai motherboard running Server 2008RC2, its scoring just over 16 on Cinebench R11.5. Its much faster than my i7 3930k w/mild OC which hits around 11.6 on Cinebench. Maybe the dual E5 setup will work better under CS6, but the E5's are easily more powerful than the previous Xeons so not sure why your scores are so bad with those killer CPU's.

          • 2. Re: Xeon benchmarks
            Chuck A. McIntyre Level 3

            Back around 2004, I went to an Adobe "Roadshow".  At that time, the company I worked with was using the Matrox RT.X100 in the Asus P4C800 Deluxe motherboards.  That was a nightmare experience as a side note. Adobe reps mentioned during the presentation that they used Xeons for system stability.  I recently read on Intel's website that memory errors are very rare.  Or did they say extremely rare?  I don't remember. It's my take that ECC memory is best suited for servers.  Not sure if ECC memory + Xeons would make that much of a stability difference for video editors.  So if they're much more expensive, not considerably more stable and insult to injury... slower, why would any editor spend the money on a dual Xeon editing system?  I suppose there could be an advantage with Xeons for usual editing plus 3d work.  What's your take Scott?

            • 3. Re: Xeon benchmarks
              lasvideo Level 4

              Slightly different evidence from Toms Hardware showing ES- 2687 superiority over i7-3960 for specified function.


              Screen Shot 2012-05-10 at 3.07.53 PM.png


              Im sure with over clocking the 3960s performance would incresse, but when Im running editing sessions and rendering 20 hours a day (5-7 days a week) Im not that comforable running machines over clocked.

              • 4. Re: Xeon benchmarks
                Bill Gehrke Most Valuable Participant

                Scott, great information thanks much.  One minor questions should that dual system processor be an E5-2670?

                • 5. Re: Xeon benchmarks
                  FloridaG Level 1

                  Of course most servers are designed to run using ECC ram because of the Xeon processors, but ECC ram is certainly never used for greater performance, its supposed upside is stability in 24/7 uptime environments but Ive got plenty of mission critical equipment that doesnt run ECC Ram and works just fine.


                  The moral of the story is simple

                  Do.you want the fastest machine? Get the new Xeon E7's, and run the Maximus setup aka Tesla card.


                  A regular i7 with a Gtx580 will beat up a dual Xeon with a videocard unsupported by PP because even all the additional CPU power cant match the effectiveness of the GPU acceleration.


                  The new Xeons are awesome, but I went with Supermicro because the their motherboards are solid. For dual 2011 LGA build  I would only have run Supermicro or Asus. If i were going with off the shelf machines, a Z800/820 would be the choice. I run them 24/7 and they never fail.

                  • 6. Re: Xeon benchmarks
                    JCschild Level 3




                    Cinebench (useless and has been for many yrs now but its free thus why sites keep using it and its not about editing its about animation) and some other synthetic benchmarks?

                    really guys? (oh and it also shows Quadros doing better hmmm.. )


                    this benchmark is real workflow with real footage and effects that are commonly used and also MPE driven. its a very common workflow and why we chose to use it as a benchmark. (have red footage #s as well which i forgot to add )

                    at the same time its become far too "lite" so we started adding the lightning effect. (maybe Cepia would be better) seems we didnt run the lightning test so need to re do..


                    i coule easily create a benchmark that would consistantly show Xeons winning and convince people they needed them @ 2x the price or more of a single system.

                    and i would sell far more Xeons and make more money per sale.

                    now anyone that knows us even a little knows we are NOT about that.  (Tom we have been telling you from day one you dont need a Xeon.)


                    not to mention someone with some sense would call BS on it.


                    fact is the avergae editor does not need Dual Xeons and the average work flow is better served with a single higher GHZ system (preferably over clocked)  (read will outperform a xeon) and do it for far less money


                    where they (Xeons) are neeeded is red 4K and heavy 3D modeling/animation.

                    even with this system (Xeon) full picture preview stuttered with Red4K  (at least it actually did play it back) where before 1/2 preview was best we could get with stuttering on older Xeons. (1/4 preview with single systems)


                    so again we will not recommend Xeons to anyone unless they are doing red4K type work or heavy animation/render farm.




                    • 7. Re: Xeon benchmarks
                      JCschild Level 3

                      good catch Bill!

                      • 8. Re: Xeon benchmarks
                        JCschild Level 3


                        you would be correct. memory errors are far less common (assuming all is correct in component choice and ram)  (i am leaving out present issues with X79)

                        it used to be if you wanted Dual Xeon you had to use ecc its only been the last few yrs where you could build xeons and use non-ecc (aside from a few boards here and there by asus/gigabtye)


                        ECC is much slower and adds latency and an extra step in the processing, thinking was better slow than wrong/lost info.. also rarely do you find ECC in a high GHz. 

                        the comment that somehow Xeons were more stable than single systems i have to say BS. but again if you have the wrong components well...



                        • 9. Re: Xeon benchmarks
                          lasvideo Level 4

                          Scott, actually you assume to much.


                          In my extended dialogs with Eric, he understands the scope of the kind of projects I work on (Red, Prores 444, 3D raytracing animation now available in AE CS6, etc.) and concurs with my selection I think you tend to mistakenly assume that all editors here at the site are hobbyists or work strictly on basic projects like weddings or family videos. That is not the case with me and a large influx of folks now gravitating to PrP. And its something you should to understand since the needs of this influx can be different than what you have been used to in the past. Red and Proress 444 (from Alexa) are used a lot in the industry these days and many folks need a system that is diverse enough to cover a wide range of projects, from simple to complex.


                          On a different note - FYI th Quadro 4000 is necessary for folks want to leverage more speed when dealing with the new raytracing mode in AE. And since dynamic tracking is rock solid, lots of folks will want to.


                          Here are examples of the kind of projects I do and clients I edit with that will greatly benefit from the dual xeon system I purchase.   www.hdshotsandcut.com


                          It may be time to be more open minded to the varying needs of new potential customers 

                          • 10. Re: Xeon benchmarks
                            FloridaG Level 1

                            Oh BS,


                            The Xeons are undenably more powerful, and just because Premiere Pro doesn't exploit their strengths doesn't mean much. By any measurable statistic, the E5/E7 upper end  Xeons are easily more powerful than my overclocked 3930k systems (even with my Dual Xeon machine running ECC and peaking at 2.8ghz)


                            I've got people bouncing between AE, PP, Cinema 4D, and I would never give them a mega-clocked i7 for mission critical broadcast installations.


                            A Xeon isn't inherently more stable thats true, but workstation class motherboards are a ton better than the average desktop garbage so the whole sum of the components is relevent to the stability, not whether its a i7 or Xeon.

                            • 11. Re: Xeon benchmarks
                              JCschild Level 3


                              actually you assume too much.. our clients range from hobbyiest to full broadcast and sever farm animators.


                              as i said the "AVERAGE" editor..

                              • 12. Re: Xeon benchmarks
                                lasvideo Level 4

                                You missed my point Scott...you assumed to much about ME and my needs. Without knowing the kind of projects I worked on (or listening to my needs for my work flow), you did me the disservice of making repeated recommendations that were ill informed...


                                (Tom we have been telling you from day one you dont need a Xeon.)


                                Listen more... talk less....  


                                • 13. Re: Xeon benchmarks
                                  JCschild Level 3



                                  you pretty much just said it.. "" because Premiere Pro doesn't exploit their strengths doesn't mean much"""


                                  last i checked this was the Adobe forum. and the benchmarks we use are for adobe users and the average editor..


                                  i would also say AE would not benefit either..


                                  and i would very much debate the Xeons vs single board quality. as they jump all over the place.

                                  in fact i would not touch a supermicro board for nothing.

                                  they have an extremely high fail rate and the worst warranty in the biz. this is fact not fiction. i have been burned too many times by SM.


                                  the warrnty is SO bad: warranty date starts the day it ships from SM to the disty. then from disty to reseller, then reseller to end user. there could be months between the date shiped until you get it.

                                  and they only have 1 yr warranty..

                                  Tyan is only marginally better with date based on sale date (like everyone else) but usually is a 2yr warranty.

                                  Intel: 3 yr next day advance replace..


                                  based on that who do you think has more confidence in their products?


                                  an OCed system when done correctly is no more or less stable than 1 not OCed.. to assume it is well.




                                  • 14. Re: Xeon benchmarks
                                    JCschild Level 3

                                    Tom i got it, just not going to do personal comments in the forum

                                    • 15. Re: Xeon benchmarks
                                      JCschild Level 3

                                      ok here is red 4k #s which shows where the Xeons shine..


                                      Dual Xeon E5 2670 @ 2.6 GHz

                                      64GB  1600

                                      570GTX 2.5GB

                                      4  1Tb Sata 64 Meg Cache 600 Drives in Raid 0 array


                                      Red R3D 4K 24P 30 seconds long for each layer

                                      Export 2048 X 1152 24P Standard

                                      3 Layer - 2:02

                                      4 Layer - 2:13


                                      I7 3770 4.7 GHz

                                      16GB Top Power

                                      680GTX 2GB

                                      4  1Tb Sata 32 Meg Cache 600 Drives in Raid 0 array

                                      Red R3D 4K 24P 30 seconds long for each layer

                                      Export 2048 X 1152 24P Standard

                                      3 Layer –  4:11

                                      4 Layer – 5:05

                                      • 16. Re: Xeon benchmarks
                                        FloridaG Level 1

                                        PLENTY of mission critical servers are built with Supermicro hardware, I could care less that you think they are horrible.


                                        An OC'd system is an OC'd system. Thats purely your opinion that its no less stable that running a CPU outside its design parameters, which is pure ignorance.



                                        And you would discredit any kind of mathematical benchmark , but are content to ONLY include PP as a litmus test for the performance value of a Xeon?





                                        So now the Xeon's shine with 4k? Who would have guessed given such a horsepower advantage over an i7.....

                                        • 17. Re: Xeon benchmarks
                                          ECBowen Most Valuable Participant

                                          The Dual Xeons in testing give far greater performance headroom ie Red material. It's not that Premiere does not utilize all the cores. It's simply as stated before, once you have enough threads to handle encoding/decoding then GHz is the deciding factor with the MPE acceleration. So with the hardware acceleration GHz is the performance factor once you exceed 8 to 12 threads with most codecs. However AE is different. Dual Xeons perform far better in AE then single cpu systems because threading and ram ceiling are the primary factors. The difference in AE renders between Single and Dual CPU system is significant 2 to 1 or greater provided you have the ram to support that many threads. So with AE or other compositors then the Dual Xeon systems would be the ideal choice. Just make sure you have the ram to support 32 threads.


                                          On a side note, a client of ours found the file to add cards to the acceleration in AE with ray tracing. So you do not need a Quadro card/Open GL for acceleration for that.