15 Replies Latest reply on Jul 8, 2011 10:32 AM by Jim_Simon

    2500K vs 2600K

    Jim_Simon Level 9

      I'm in the market to upgrade my personal system, which I use for Internet, gaming and such.  Some minor video stuff, but nothing on the level of what I do for work.  In looking things over, I came to the conclusion that the 2500K was a better deal than the 2600K.  Yes, the 2600K offers Huperthreading and better performance, but the improvement just didn't seem worth the extra money - at least on the gaming side of things.

       

      I've been wondering if anyone had tested the two processors against each other for PP and other Adobe apps, again wanting to know is it was worth it to spend the extra $100 for the 2600K and it's Hyperthreading.

       

      Well, Tom's has run a few tests, and I'm seeing the same kind of performance scores for media apps that I saw for games.  Very small performance improvement for that extra $100.

       

      Right now, I think the i7 2500K may be the best deal in processors, even for the professioanl Adobe user.

       

      http://www.tomshardware.com/charts/x86-core-performance-comparison/benchmarks,128.html

        • 1. Re: 2500K vs 2600K
          Harm Millaard Level 7

          I know you are not a regular visitor of PPBM5 Benchmark

           

          It seems you are afraid of the results.

           

          Best ranked i5-2500K at # 128 with a RPI of 3.12 times slower than a fast system with dual raid0's,

          Best ranked i7-2600K at #   56 with a RPI of 1.85 times slower than a fast system with dual raid0's plus 1 OS disk,

           

          both measured at stock speed.

           

          Draw your own conclusions. Look at the beta page here: http://ppbm5.com/DB-PPBM5-2.php

           

          BFTB wise € 171 or € 250 for a 70%+ performance gain seems to negate your argument.

          • 2. Re: 2500K vs 2600K
            Scott Chichelli Level 3

            HI Jim,

             

            the 2600k performs vastly better than the 2500k

            while we could OC the 2500 to 4.5GHz the lack of HT killed it performance ability.

            it performed about as good as the older lynnfield 760/870ish...

             

            this was true for both audio and video applications.

            for another $100 more you are buying the right processor.

             

            Scott

            ADK

            • 3. Re: 2500K vs 2600K
              Jim_Simon Level 9
              It seems you are afraid of the results.

               

              Not afraid.  It's just that PPBM5 doesn't perform a proper "comparison" benchmark, with all things identical except the items being tested.  With so many variables between systems, it can be difficult to read anything definitive into those results.

              • 4. Re: 2500K vs 2600K
                Jim_Simon Level 9
                the 2600k performs vastly better than the 2500k

                 

                Well, I'm just not seeing that in any published benchmarks.  The difference is always quite minimal, regardless of the test being run.

                 

                If you guys have run proper comparison benchmarks (one rig, changing out only the processor), I'd be interested to hear your results, both at stock speeds and overclocked.

                • 5. Re: 2500K vs 2600K
                  Scott Chichelli Level 3

                  Feel free to believe Toms hardware if you will, we tested it doing our normal 30 minutes AVCHD to H264

                  That’s a real benchmark not some silly synthetic test.

                  PPBM is a real test albeit a tad light

                  Rarely do I see in real world tests what I see published online at hardware review sites.

                   

                  I also find it hard to believe for an NLE editing system that $100 is even a question

                  • 6. Re: 2500K vs 2600K
                    RjL190365 Level 4

                    Jum,

                     

                    You've been seeing only minimal improvements with the 2600K over the 2500K because you are comparing its performance in applications that do not take advantage of more than four threads and also do not make any use whatsoever of HyperThreading. However, Premiere Pro does take advantage of more than four threads - and the four-thread maximum of the 2500K results in that processor not performing as fast as it could have under Premiere Pro. In fact, with 16GB of RAM, a stock-speed i5-2500K performs only about as well as a stock-speed i7-930 with 24GB of RAM (with all other components being equal in both systems).

                     

                    And expanding on Harm's observations, the two systems under comparison have different amounts of RAM: The top-scoring stock-speed i5-2500K had only 8GB of RAM while the top-scoring stock-speed i7-2600K had 16GB of RAM. When both systems had only 8GB of RAM, that same i7-2600K system with half of its RAM removed (making it 8GB) ranked at # 108  with an RPI of 2.64 times slower than a fast system with dual RAID 0's plus 1 OS disk. (For the record, the best ranked i7-2600K with 8GB of RAM ranked # 98 with an RPI of 2.19 times slower than a fast system with two SSD's - but then again, that system also has a GTX 580 rather than the older, slower, reference-clocked GTX 470 in the system cited by Harm.)

                     

                    Message was edited by: RjL190365

                    • 7. Re: 2500K vs 2600K
                      Jim_Simon Level 9
                      we tested it doing our normal 30 minutes AVCHD to H264

                       

                      OK.  Same rig?  Changing only the CPU for each test?

                       

                      Have you the results handy?  What were the export times for the two models?  Were either overclocked at all?  Were they also tested at stock speeds?

                      • 8. Re: 2500K vs 2600K
                        Jim_Simon Level 9
                        you are comparing its performance in applications that do not take advantage of more than four threads

                         

                        Looking over the Tom's test, I see that you're right.  Tom's has equalized the testing to a single clock speed using only one core, in order to test the architecture itself.

                         

                        I'd still be interested in Scott's results, if has still has them.

                        • 9. Re: 2500K vs 2600K
                          Jim_Simon Level 9
                          PPBM is a real test albeit a tad light

                           

                          If performed on the same rig, changing only one thing at a time (whichever thing is being tested), then I agree.

                           

                          I wasn't knocking the test, so much.  It was the published results that can be less than effective as a true benchmark, given all the variables between systems.

                          • 10. Re: 2500K vs 2600K
                            Scott Chichelli Level 3

                            Have to check with Eric as this way back when they first came out to see if he saved the numbers (intel sent us a 2400, 2500k and 2600 (no k))

                            And beta boards a few months before release.. We had to wait for stock to actually show up before we could get the 2600k

                            I looked in the benchmarks folder on the server and didn’t see them in there.

                             

                            But yes obviously the correct way to test is an identical system and just change the CPU.. and yes that’s how we test.

                             

                            But as we have all said the lack of HT is why it performs so poorly for Adobe and honestly for every other app we tried both audio and video.

                            HT and Turbo are more effective with sandy than any other gen thus far and therefore is more important now.

                            And yes they would be OCed numbers and possible stock #s as well assuming we have them.

                             

                            If not I can run it again, I am sure I have a 2500 sitting around

                             

                            Scott

                            • 11. Re: 2500K vs 2600K
                              Scott Chichelli Level 3

                              I agree this is one issue with PPBM (well 2 as I said its light) user submitted data is difficult to weed thru and the potential variables make it difficult.

                              Even with identical specs but windows/adobe setting different it can be a pretty big change

                              • 12. Re: 2500K vs 2600K
                                Jim_Simon Level 9
                                If not I can run it again, I am sure I have a 2500 sitting around

                                 

                                That's be awesome.  Stock and OC'd, if you're so inclined.

                                • 13. Re: 2500K vs 2600K
                                  JEShort01 Level 4

                                  Randall,

                                   

                                  You've done a lot with PPBM5; I'd be curious to see what your 2600k rig does with HT enabled/disabled in the bios.

                                   

                                  That would be really quick to do, and I would think interesting to this thread. As Jim Simon points out there are a lot of "variables" from system to system regarding drives, configuration, tweaks, etc.

                                   

                                  Regards,

                                   

                                  Jim

                                  • 14. Re: 2500K vs 2600K
                                    Harm Millaard Level 7

                                    Jim,

                                     

                                    I have a hard time understanding you. On another thread you are talking about a new Panny of $ 6K+, here you are talking about a measly $ 100 difference. If you have the funds to invest in such a camera, then what is $ 100 in CPU? It does not even buy you a decent pola filter, but will give you better performance all around, every day.

                                     

                                    I agree that the weaknesses of the PPBM test are twofold:

                                     

                                    1. The widely different setups,

                                    2. The short tests.

                                     

                                    1 has only two solutions to resolve that weakness and that is that you all send me your hardware, I test it with the same OS/programs/Adobe setup and return your system or we create a disk that contains the same setup and you replace that as the boot disk and then run the test. Neither of which is sensible, apart from the legal issues about activation and license agreements.

                                     

                                    2 is easily remedied by extending the benchmark but has the drawback of making the download larger, the testing time much longer and for some that testing time is already close to an hour in the current version.

                                     

                                    However, instead of having only a very few systems in a test, like Tom's hardware, we have over 500 systems in the test and then, despite different setups, individual differences will tend to grow less important because of the large numbers involved.

                                     

                                    If you have two runners, one aged 20 and one aged 25 that both run the 100 meters and the younger one does it in 10 seconds flat and the 25 year old does it in 11 seconds, you can't say that 20 year olds have a 10% better performance. That is lying with statistics. Now if you had 100 runners evenly divided over both age categories and the results (on average) were 10.49 for the 20 years old ones and 10.81 for the 25 year old ones, then I would agree there might be a statistically significant difference of 3% if the results were corrected for outliers, but if that is really true if the sample were 200 runners? It may be reversed, depending on your sample or it may not show any difference at all.

                                     

                                    The problem with benchmarks like Tom's or ADK's is that they can not be repeated outside of their offices and you can't know the testing circumstances. With our benchmark you can replicate results, try it out yourself, and with the large number of entries even different setups can be compared with a large degree of confidence.

                                     

                                    Sample size is extremely important to give statistical validity to any test and both Tom's and ADK's tests lack the number of measurements to give it any semblance of statistical validity. Those tests are limited to only a few number of observations, from which one cannot derive - with statistical accuracy - differences between systems, even though they were tested under the 'same' conditions.

                                     

                                    Statistics is a difficult profession, and easy to manipulate if you have the wrong intentions.

                                    • 15. Re: 2500K vs 2600K
                                      Jim_Simon Level 9

                                      Let me just say that because I want a particular camera, does not mean I can afford that particular camera.