Page 19 of 44 FirstFirst ... 9171819202129 ... LastLast
Results 271 to 285 of 655

Thread: 11.5's BenchmarkMarbles.lws - share your machine's render time here

  1. #271
    Quote Originally Posted by eV1Te View Post
    I haven't done benchmarks in Lightwave, but Hyper-threading tends to work slower while doing serious simulations etc.
    NOT in rendering, HT in LW will give you 25-35% better render times. Actually in any rendering/calculating software not just LW.

    Try it if you don't believe , turn on "automatic" mode in LW threads and render scene and then switch to 4 Threads only and compare rendertimes.
    RAM-Studio
    WS - Dual Xeon E5-2698v4/128GB/Win10x64/4xRTX 2080Ti
    My LWM Video Car Modeling Tutorial

  2. #272
    Super Member JonW's Avatar
    Join Date
    Jul 2007
    Location
    Sydney Australia
    Posts
    2,235
    Quote Originally Posted by eV1Te View Post
    My CPU only has 4 cores, i.e. it can only calculate 4 things simultaneously.I haven't done benchmarks in Lightwave, but Hyper-threading tends to work slower while doing serious simulations etc.
    Quote Originally Posted by JonW View Post
    W5580 x 2

    1h 13m 17s (4397.7) Hyperthreading on (8 real + 8 HT cores)

    1h 33m 57s (5637.1) HT off (8 real cores only)

    With HT on the render is done in 78% of the time.

    The benefits of HT off with single core work might outweight having HT on. It would be worth a test with Bullet.
    Quote Originally Posted by JonW View Post
    I tried 2 instances of LW, only 20 seconds quicker on W5580 x 2 box.

    1:13:17 1 x LW
    1:12:57 2 x LW average
    My dual W5580 CUP box (similar to 2 x 920 CPUs at 3.2 GHz)
    Procrastination, mankind's greatest labour saving device!

    W5580 x 2 24GB, Mac Mini, Spyder3Elite, Dulux 30gg 83/006 72/008 grey room,
    XeroxC2255, UPS EvolutionS 3kw+2xEXB

  3. #273
    Semiconductor node? eV1Te's Avatar
    Join Date
    Jan 2007
    Location
    Sweden
    Posts
    49
    Quote Originally Posted by Lewis View Post
    NOT in rendering, HT in LW will give you 25-35% better render times. Actually in any rendering/calculating software not just LW.

    Try it if you don't believe , turn on "automatic" mode in LW threads and render scene and then switch to 4 Threads only and compare rendertimes.
    Maybe you loose that much if you only use 4 threads while you have HT enabled in BIOS as you suggested me to test. I found this in a 1 year old article were they used only as many threads as the physical cores even with HT turned on, and you see a significant drop when HT is enabled:
    Click image for larger version. 

Name:	LW96-HTComp1.png 
Views:	106 
Size:	8.6 KB 
ID:	116426

    They never tested it with double the threads while HT was enabled unfortunately. Anyone has any good benchmarks on this in Lightwave?

    Next time I do a reboot I will try it myself (but my uptimes are quit long as you see from the screenshot below, I'm glad LW is so stable now days)

    Click image for larger version. 

Name:	cpu.png 
Views:	89 
Size:	23.2 KB 
ID:	116427

  4. #274
    eV1Te - trust me it's noticably faster with HT ON, Whatever they did on that test on your screen has no sense or they swaped out results by mistake. HT ON is faster esp. in rendering. You are crippling your machine by 25-35% on renderign if you are using HT OFF . Go and reboot now but first open up few scenes and hit F9 and type in render times so you can compar eon your machine and not some online graph which is wrong .
    RAM-Studio
    WS - Dual Xeon E5-2698v4/128GB/Win10x64/4xRTX 2080Ti
    My LWM Video Car Modeling Tutorial

  5. #275
    Semiconductor node? eV1Te's Avatar
    Join Date
    Jan 2007
    Location
    Sweden
    Posts
    49
    Quote Originally Posted by JonW View Post
    My dual W5580 CUP box (similar to 2 x 920 CPUs at 3.2 GHz)
    Interesting, that is a huge difference! I must try it myself in Lightwave before I believe it!

    Did you properly disable HT in bios or did you just reduce the number of threads?

  6. #276
    Super Member JonW's Avatar
    Join Date
    Jul 2007
    Location
    Sydney Australia
    Posts
    2,235
    Quote Originally Posted by eV1Te View Post
    Did you properly disable HT in bios or did you just reduce the number of threads?
    In BIOS.


    I also changed the cores to fixed speed (forgot the exact terminology) & on some other test got a tiny bit more performance. The Xeons can't be overclocked as they are locked. But you can OC the 920 & 940 easily.

    So:
    I just leave Threads on Auto. So it will render with the same number of threads as the CPU the scene is rendering on. There is basically no difference having threads on manual. Sometimes it's a touch quicker, sometimes it is a touch slower. But if you are sending scenes off to your render farm Auto is the way to go if you have different number of cores on the computers & you are using 1 render node per computer.

    So with 2 instances of LW with HT on, threads on Auto, I can get an average frame rate 1:12:57. Not bad on a box that is over 4 years old!

    So if you OC your CPU to 3.2 GHz you will render this scene in 2.5 hours!
    Procrastination, mankind's greatest labour saving device!

    W5580 x 2 24GB, Mac Mini, Spyder3Elite, Dulux 30gg 83/006 72/008 grey room,
    XeroxC2255, UPS EvolutionS 3kw+2xEXB

  7. #277
    Not doing nothing.... Tranimatronic's Avatar
    Join Date
    Dec 2003
    Location
    Vancouver BC
    Posts
    496
    The Hyperthreading argument was one I had with a sysAdmin at the last studio I worked at. For some reason, he WOULD NOT believe hyperthreading made a difference until I sat him down and rendered the same scene, once with HT on and once with it off. The difference on a medium complexity frame was 12 minutes for no HT and 10 with it enabled.
    I have a feeling this is different depending on which OS you use. I believe Windows has a better implementation, but this is just a hunch.

    Curiously the nerd at my local computer parts store swore hyperthreading makes no difference for 3d rendering also. The same guy swore MSI motherboards were the best too, but when asked, couldn't explain why.

    Its like telling people to turn off their fuel injection systems in their cars because the fuel will get there eventually anyway.

  8. #278
    Semiconductor node? eV1Te's Avatar
    Join Date
    Jan 2007
    Location
    Sweden
    Posts
    49
    I'm glad I now tried to enable HT once again, there must have been some new driver update which allows the dynamic Intel Turbo-boost over-clocking even if only single thread process is running at 100% while HT is on (previously my CPU stayed at 1.9 GHz even if I was running Prime95 on one of the 8 virtual cores), and it also works with Bullet Dynamics. The other option is to disable any dynamic clocking, and always running at full frequency with a higher power consumption.

    For rendering it surly gives me a 5-25 % speed increase, where; radiosity, transparency and difficult shaders tend to favor HT turned ON it seems. A noticed only very small difference in performance for scenes without radiosity and fancy materials.

    So I retract my previous statement, HT does give you a boost and at least for now I can't find anything negative about it!

  9. #279
    Super Member JonW's Avatar
    Join Date
    Jul 2007
    Location
    Sydney Australia
    Posts
    2,235
    Scenes without radiosity there is not a lot of difference. My very old Dual E5335 Xeon box performs reasonably well (4 cores each CPU 2.0GHz no HT). So if I bake radiosity on my W5580 then use the E5335 box on the farm it is ok. On the other hand, if I am not baking, I don't even bother with this old box with radiosity on the farm, it is just far too slow. There was & bit of an improvement with my E5450 box.

    CPUs from the i7 onwards handle radiosity extremely efficiently. & with HT on are a world of difference over previous CPUs.
    Procrastination, mankind's greatest labour saving device!

    W5580 x 2 24GB, Mac Mini, Spyder3Elite, Dulux 30gg 83/006 72/008 grey room,
    XeroxC2255, UPS EvolutionS 3kw+2xEXB

  10. #280
    eye kan kode gud jrandom's Avatar
    Join Date
    Dec 2009
    Location
    Seattle, WA
    Posts
    1,143
    On modern CPUs, using hyperthreads with Lightwave gives a performance boost where each hyperthread is worth about 25% of a full core. So while 8 hyperthreads won't halve your render times, it will be measurably faster than just using four cores.

  11. #281
    Not so newbie member lardbros's Avatar
    Join Date
    Apr 2003
    Location
    England
    Posts
    6,056
    Single-threaded applications or simulations will run slower though as they will be running on a single HT Core... effectively halving the speed of the simulation/single-threaded appliation. I've found this with my work machine anyway... have they now got dynamic HT nowadays?
    LairdSquared | 3D Design & Animation

    Desk Work:
    HP Z840, Dual Xeon E5-2690 v2, 32GB RAM, Quadro K5000 4GB
    Desk Home:
    HP Z620, Dual Xeon E5-2680, 80GB RAM, Geforce 1050 Ti 4GB

  12. #282
    Super Member JonW's Avatar
    Join Date
    Jul 2007
    Location
    Sydney Australia
    Posts
    2,235
    Quote Originally Posted by lardbros View Post
    Single-threaded applications or simulations will run slower though as they will be running on a single HT Core... effectively halving the speed of the simulation/single-threaded appliation. I've found this with my work machine anyway... have they now got dynamic HT nowadays?
    Yes, I found Modeler is a bit slower with HT on, if it is a massive job I turn HT off. An SSD helps disc tasks to some degree.
    Procrastination, mankind's greatest labour saving device!

    W5580 x 2 24GB, Mac Mini, Spyder3Elite, Dulux 30gg 83/006 72/008 grey room,
    XeroxC2255, UPS EvolutionS 3kw+2xEXB

  13. #283
    obfuscated SDK hacker Lightwolf's Avatar
    Join Date
    Feb 2003
    Location
    Stuttgart, Germany
    Posts
    13,711
    Quote Originally Posted by lardbros View Post
    Single-threaded applications or simulations will run slower though as they will be running on a single HT Core... effectively halving the speed of the simulation/single-threaded appliation. I've found this with my work machine anyway... have they now got dynamic HT nowadays?
    How did you find this? If you look at how HT works, that certainly isn't the case (and shouldn't be either).
    Did you time it or look at the CPU load (which, with HT enabled, needs to be read differently!).

    Especially the "halving" comment is one that I'm not buying at all. I suspect that is by just looking at the CPU usage, which is misleading in this case.

    Cheers,
    Mike

  14. #284
    obfuscated SDK hacker Lightwolf's Avatar
    Join Date
    Feb 2003
    Location
    Stuttgart, Germany
    Posts
    13,711
    One more, 36m 43s - just unboxed.
    Click image for larger version. 

Name:	Coen_Marbles.png 
Views:	126 
Size:	492.6 KB 
ID:	116692

    Lenovo Thinkstation C30
    2 x Xeon 2680 (8 Core, 2.7Ghz base, HT)
    128GB Ram
    128GB SSD
    I won't mention the current GPU because it's embarrassing and will be swapped out anyhow.

    And if you wonder about me ranting earlier that dual CPU machines aren't really worth it - this one was a (refurbished) bargain.

    Cheers,
    Mike

  15. #285
    Super Member COBRASoft's Avatar
    Join Date
    Dec 2005
    Location
    Oudenburg, BEL
    Posts
    3,184
    Oh my, that's FAST mike! Congrats on the new machine.

    Where can I buy one of these and how much is it?

Page 19 of 44 FirstFirst ... 9171819202129 ... LastLast

Bookmarks

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •