Announcement

Collapse
No announcement yet.

Curious results with 1950x and different memory

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Curious results with 1950x and different memory

    Hi there,
    I'm just testing a new system - 1950x (at stock settings) + Asrock Taichi and I had some unexpected results while running V-Ray Benchmark, so I hoped someone can tell me if this is normal.

    1st test:
    with 2 stick DDR4 @2666 C15
    V-Ray bench (1.0.6) result: 00:46
    Cinebench R15 result: 2835

    2nd test:
    with 4 sticks DDR4 @3200 C16
    V-Ray bench (1.0.6) result: 00:46
    Cinebench R15 result: 2990

    The strange thing for me is that there is no performance gain in V-ray Bench when using 2-channel DDR4 @2666 vs. 4-Channel @3200, but there is such in Cinebench.
    Is that some limitation of the benchmark itself or it will be the same with V-Ray inside 3dsMax?
    Last edited by Dimiter.Enev; 08-12-2017, 01:05 AM.

  • #2
    Interesting test. Did you run cinebench several times? I've had wildly different scores on a dual 2696v3 across successive runs, from 4100 to 4450. The tests takes so little time that it is very sensitive to any hiccup or background process.

    While there is no performance gain in vray, you could also argue that CB is hindered by dual channel RAM and vray is not!
    www.dpict3d.com - "That's a very nice rendering, Dave. I think you've improved a great deal." - HAL9000... At least I have one fan.

    Comment


    • #3
      For the 2nd test I had made 2 tries in Cinebench - gave the exact same score, and 3 tries in V-Ray bench - one of the results was even lower 00:47.

      (By the way, I had wrote the timings of the 3200 memory wrong on my first post - it was not C18, but C16 - edited it)

      Comment


      • #4
        One more test, just for comparison:
        1 stick DDR4 @2666 C15
        V-Ray bench (1.0.6) result: 00:51
        Cinebench R15 result: 2776

        Comment


        • #5
          I'm not sure about Vray Benchmark but I'm definitely getting faster render times with faster/overclocked RAM in both Cinebench and in real-world projects. Here are my findings with an interior scene I rendered:
          • 1950X 3.9 Ghz, RAM 2400 Mhz: 3270 CB in Cinebench R15 / 213 sec in interior scene Max 2016 + Vray 3.6
          • 1950X 3.9 Ghz, RAM 3600 Mhz 3414 CB in Cinebench R15 / 202 sec in interior scene Max 2016 + Vray 3.6
          I want to stress that all overclocking was done from BIOS.
          Max 2023.2.2 + Vray 6 Update 2.1 ( 6.20.06 )
          AMD Ryzen 7950X 16-core | 64GB DDR5 RAM 6400 Mbps | MSI GeForce RTX 3090 Suprim X 24GB (rendering) | GeForce GTX 1080 Ti FE 11GB (display) | GPU Driver 546.01 | NVMe SSD Samsung 980 Pro 1TB | Win 10 Pro x64 22H2

          Comment


          • #6
            Thanks for the share Alex!
            I presume you are using the same sticks of RAM with different overclocks right? Probably 4x16GB kit?
            Can you run the VrayBencmark so we can have some comparable data if possible?

            Comment


            • #7
              The first test was with done with 2 kits of 4 x 8 GB sticks and the 2nd test was done with 1 kit of 4 x 8 GB sticks (same model RAM in both cases). Testing again would be kind of hard to do since I need to remove half my RAM again in order to achieve 3600 Mhz (with all 8 slots populated, the memory controller is very stressed and can't boot higher than 2933 Mhz). The CPU fan (Noctua U14S) is blocking the RAM sticks so I have to detach it first to reach them and it's kind of a pain to do all over again honestly.
              Max 2023.2.2 + Vray 6 Update 2.1 ( 6.20.06 )
              AMD Ryzen 7950X 16-core | 64GB DDR5 RAM 6400 Mbps | MSI GeForce RTX 3090 Suprim X 24GB (rendering) | GeForce GTX 1080 Ti FE 11GB (display) | GPU Driver 546.01 | NVMe SSD Samsung 980 Pro 1TB | Win 10 Pro x64 22H2

              Comment


              • #8
                Oh, definitely don't bother removing sticks, I just meant to run VrayBenchmark with the config that you have now.
                Thanks for the additional info!
                I'm with the same cooler right now, but hesitate to overclock until I get a proper water block.

                Comment


                • #9
                  I complete Vray benchmark in 41 sec. with the system in my signature. My RAM is clocked at 2933 Mhz. Click image for larger version  Name:	vraybench_results_AMD_1950x_RAM_2933mhz.jpg Views:	1 Size:	91.4 KB ID:	979060






                  Actually, I've watched some youtube videos and having an AIO is not really going to give you better temps and overclocks than the trusty U14S. It's a surprisingly good air cooler that beats most AIOs except for maybe Enermax Liqtech 360. It's not loud too while most AIOs at higher RPMs would sound like a jet turbine. A plus of the Liqtech 360 might be the better RAM clearance, 2-3 C lower temps and prettier looks but apart from that it would cost more, sound worse at high RPMs and have more things prone to failing like the pump for example or liquid spillage. Here's an interesting YT video. You wouldn't be able to overclock Threadripper more than 4.1-4.2 Ghz with the Liqtech 360 anyways so why really bother with expensive AIOs when you can get 3.9 Ghz easily on an U14S that's a lot cheaper and will last you basically forever and doesn't need maintenance apart from blowing the dust off from time to time. Just my 2 cents.
                  Last edited by Alex_M; 29-12-2017, 06:32 PM.
                  Max 2023.2.2 + Vray 6 Update 2.1 ( 6.20.06 )
                  AMD Ryzen 7950X 16-core | 64GB DDR5 RAM 6400 Mbps | MSI GeForce RTX 3090 Suprim X 24GB (rendering) | GeForce GTX 1080 Ti FE 11GB (display) | GPU Driver 546.01 | NVMe SSD Samsung 980 Pro 1TB | Win 10 Pro x64 22H2

                  Comment


                  • #10
                    Thanks Alex.
                    We are going a bit off-topic, but
                    Yes I know about the AIOs.
                    I already have a custom water loop with 420x6.5cm radiator from my previous config and just waiting for the new EKWB TR4 water block (which should be released very soon).
                    The plan is to keep the Noctua as a back up option in case the pump fails in the middle of a project.

                    Comment

                    Working...
                    X