Rendertime-Considerations using Superfly (Cycles)



  • I plan to make a Video on Rendertimes-Considerations with Poser 11 soon. For this i am doing some tests about what works and what not.

    Anyway, i thought sharing my Rendertimes-Results with you all here.
    Then you can make your own considerations if new hardware will help you and how much.

    I have rendered two Samples, one was the default Scene, with default settings at 1920x1080, second is a more complex Scene at 2048x2048 resolution and still below medium settings.
    As this article gets already too long, I'll keep the other results for the video. CPU is a 12 Core i7.

    This first table shows the Rendering Time Tolerances for the CPU, same picture, same device. Different Render-Times.

    "Field-Size" -> "Used Time"
    Have to add that in these tests Poser-Cycles was stable and did not crash also had no memory leaks (Simple Scene).

    Render: Intel Render: Intel
    4 -> 57.8689999580383 4 -> 57.1500000953674
    5 -> 60.0460000038147 5 -> 57.1220002174377
    6 -> 56.201000213623 6 -> 57.816999912262
    7 -> 56.5390000343323 7 -> 56.1370000839233
    8 -> 56.1889991760254 8 -> 55.9029994010925
    9 -> 56.2449998855591 9 -> 60.0009999275208
    10 -> 59.0640001296997 10 -> 56.1750001907349
    11 -> 58.7619996070862 11 -> 57.8020005226135
    12 -> 55.2620005607605 12 -> 57.143000125885
    13 -> 55.1630001068115 13 -> 57.219000339508
    14 -> 54.9109997749329 14 -> 58.0549998283386
    15 -> 55.1519999504089 15 -> 57.0430002212524
    16 -> 55.1189999580383 16 -> 57.1820006370544
    17 -> 55.2849998474121 17 -> 58.0419998168945
    18 -> 55.2840003967285 18 -> 59.9649996757507
    19 -> 58.6979999542236 19 -> 57.2090001106262
    20 -> 55.2560000419617 20 -> 56.2000002861023
    22 -> 57.1770000457764 22 -> 57.2069997787476
    24 -> 55.1619992256164 24 -> 56.2469997406006
    28 -> 57.0110001564026 28 -> 56.2220001220703
    30 -> 55.2259998321533 30 -> 55.9370002746582
    32 -> 57.1089997291565 32 -> 59.2740001678467
    34 -> 56.8410000801086 34 -> 55.9219999313354
    36 -> 57.0770001411438 36 -> 56.188000202179
    38 -> 57.0309996604919 38 -> 56.1720004081726
    40 -> 55.2720003128052 40 -> 56.1849999427795
    45 -> 56.8100004196167 45 -> 56.1930003166199
    50 -> 57.8640003204346 50 -> 57.1609997749329
    55 -> 56.139000415802 55 -> 57.1340003013611
    60 -> 56.1399998664856 60 -> 57.793999671936
    64 -> 55.9640002250671 64 -> 57.018000125885
    70 -> 60.0389995574951 70 -> 60.0859999656677
    80 -> 57.1110000610352 80 -> 58.1709995269775
    90 -> 56.9570007324219 90 -> 58.018000125885
    100 -> 58.1649994850159 100 -> 61.03600025177
    110 -> 58.1089997291565 110 -> 58.0819997787476
    120 -> 58.800000667572 120 -> 58.960000038147
    128 -> 60.0790004730225 128 -> 61.8449997901916
    130 -> 59.9790005683899 130 -> 58.183000087738
    150 -> 59.128999710083 150 -> 59.9819993972778
    180 -> 65.1230006217956 180 -> 63.1380000114441
    200 -> 64.1329998970032 200 -> 66.0010004043579
    225 -> 64.960000038147 225 -> 61.1650004386902
    256 -> 66.7920002937317 256 -> 66.1290001869202
    280 -> 85.0019993782043 280 -> 85.2329998016357
    300 -> 83.0359997749329 300 -> 84.978000164032
    320 -> 81.9220004081726 320 -> 88.0699996948242
    400 -> 94.4679999351501 400 -> 95.2150001525879
    512 -> 116.455999851227 512 -> 121.24599981308
    768 -> 243.521999835968 768 -> 254.354000091553
    1024 -> 379.680000305176 1024 -> 394.803999900818

    So what we see is that about CPU-Rendering, the field size does not play a significant role.

    Now comes the interesting part: GPU and Field-Size.

    GTX 1050 Ti Render: GTX 1080
    5 -> 823.87099981308 5 -> 676.327000141144
    6 -> 615.606999874115 6 -> 501.124000549316
    7 -> 478.88300037384 7 -> 388.960000038147
    8 -> 382.940999984741 8 -> 315.248999595642
    9 -> 325.803999900818 9 -> 265.851999759674
    10 -> 282.519999980926 10 -> 229.250999927521
    11 -> 244.586999893188 11 -> 200.178999900818
    12 -> 215.552000522614 12 -> 176.236999988556
    13 -> 191.452000141144 13 -> 157.408999443054
    14 -> 172.385999679565 14 -> 140.213000297546
    15 -> 155.28600025177 15 -> 129.093999862671
    16 -> 140.291999816894 16 -> 116.255000114441
    17 -> 130.473999977112 17 -> 108.290999889374
    18 -> 116.258999347687 18 -> 97.1579999923706
    19 -> 107.207999229431 19 -> 89.210000038147
    20 -> 99.1340003013611 20 -> 84.9819993972778
    22 -> 85.0060005187988 22 -> 70.210000038147
    24 -> 72.840000629425 24 -> 61.1360001564026
    28 -> 58.9159994125366 28 -> 51.9269995689392
    30 -> 53.0319995880127 30 -> 44.0349998474121
    32 -> 48.797999382019 32 -> 42.9190001487732
    34 -> 46.889000415802 34 -> 37.8459997177124
    36 -> 42.079999923706 36 -> 34.0869994163513
    38 -> 39.9910001754761 38 -> 31.9980001449585
    40 -> 38.0569996833801 40 -> 29.9920001029968
    45 -> 33.0309996604919 45 -> 25.9969997406006
    50 -> 30.8180003166199 50 -> 22.742000579834
    55 -> 25.9970002174377 55 -> 19.9610004425049
    60 -> 23.9950003623962 60 -> 19.746000289917
    64 -> 23.835000038147 64 -> 16.6619997024536
    70 -> 23.8739995956421 70 -> 15.7519993782043
    80 -> 23.9640002250671 80 -> 15.2779998779297
    90 -> 21.7470002174377 90 -> 12.6809992790222
    100 -> 24.8220000267029 100 -> 20.621000289917
    110 -> 20.9089999198914 110 -> 11.835000038147
    120 -> 20.8359999656677 120 -> 11.9969997406006
    128 -> 19.9579997062683 128 -> 10.9219999313354
    130 -> 24.5160002708435 130 -> 12.8250002861023
    150 -> 22.7569999694824 150 -> 11.1889996528625
    180 -> 21.6929998397827 180 -> 10.9629998207092
    200 -> 19.9199995994568 200 -> 11.0110001564026
    225 -> 22.6639995574951 225 -> 10.9899997711182
    256 -> 23.5289998054504 256 -> 11.5699996948242
    280 -> 20.6700000762939 280 -> 11.704999923706
    300 -> 19.6610007286072 300 -> 9.8769998550415
    320 -> 20.0419998168945 320 -> 9.7
    400 -> 23.8680000305176 400 -> 9.70700025558472
    512 -> 25.8030004501343 512 -> 9.67799949645996
    768 -> 24.7579998970032 768 -> 9.9079999923706
    1024 -> failed 1024 -> 11.7990002632141

    Note that in large field sizes, Cards with lower Memory fail.
    The only "safe" field size - even for complex scenes was in my tests:

    Where the 1080 with more memory allows generally larger field sizes.

    Finally a table where both cards work together.

    Render: Geforce GTX 1050 Ti + 1080
    5 -> 380.261999607086
    6 -> 281.544000148773
    7 -> 217.598999977112
    8 -> 176.411999702454
    9 -> 149.237999916077
    10 -> 129.208999633789
    11 -> 129.892999649048
    12 -> 130.081000328064
    13 -> 88.191999912262
    14 -> 80.2239999771118
    15 -> 71.844000339508
    16 -> 65.896999835968
    17 -> 62.8359999656677
    18 -> 55.1339998245239
    19 -> 51.0159997940063
    20 -> 48.0599994659424
    22 -> 41.8580002784729
    24 -> 36.8850002288818
    28 -> 28.9190001487732
    30 -> 29.7109994888306
    32 -> 25.8039999008179
    34 -> 22.7020001411438
    36 -> 20.9449996948242
    38 -> 20.0569996833801
    40 -> 19.1870007514954
    45 -> 17.0499997138977
    50 -> 14.6539998054504
    55 -> 14.7750000953674
    60 -> 13.0159993171692
    64 -> 11.9679999351501
    70 -> 11.9530005455017
    80 -> 10.7019996643066
    90 -> 9.98600006103516
    100 -> 9.9060001373291
    110 -> 9.77500009536743
    120 -> 9.92000007629394
    128 -> 10.8359999656677
    130 -> 10.5720000267029
    150 -> 9.92199993133545
    180 -> 8.77500009536743
    200 -> 8.96900033950806
    225 -> 10.8109998703003
    256 -> 9.98999977111816
    280 -> 8.72899961471558
    300 -> 11.7469997406006
    320 -> 12.0069994926453
    400 -> 12.8439998626709
    512 -> 9.03499984741211 (possibly failed?)
    768 -> 9.80600023269653 (possibly failed?)
    1024 -> 6.87899971008301 (possibly failed?)

    Have to say that all tests run automated and i was not there, but from what i see, i expect that the render failed at the last three results.
    Also there is a large random factor if two cards with different speed work together. It depends on which card gets the most difficult blocks at the end of the render. If its the slower card, this card may need significant time at the end of the render, when the other card is long time idle.