Page 3 of 11 FirstFirst 12345 ... LastLast
Results 31 to 45 of 159

Thread: Cheaper (or faster?) CPUs are coming

  1. #31
    www.Digitawn.co.uk rustythe1's Avatar
    Join Date
    Feb 2006
    Location
    england
    Posts
    1,279
    no, that was the 10 min render, my previous a year and a half ago was 17 then 14, both of which are still almost or more than half the render time.
    Intel i9 7980xe, Asus Rampage Vi extreme, 2x NVIDIA GTX1070ti, 64GB DDR4 3200 corsair vengeance,
    http://digitawn.co.uk https://www.shapeways.com/shops/digi...ction=Cars&s=0

  2. #32

    it is getting tricky to follow you with guesswork.

    - always upload the render
    - always provide the PC specs
    - and always provide other info like scene info / LW version / cinebench version
    - providing even more info is a bonus.

    people cannot follow you if you don't provide good info.


    but i'm out, too much time spent on this.

    and please fix your signature.
    LW vidz   DPont donate   LightWiki   RHiggit   IKBooster   My vidz

  3. #33
    Male Modeler ccclarke's Avatar
    Join Date
    Mar 2006
    Location
    New England
    Posts
    113
    Quote Originally Posted by MichaelBeckwith View Post
    I have always been facinated with the concept of a render farm. If you do have a render farm over a network does each computer work on separate frame of an animation or can the whole render farm work on the exact same image, say 10 computers working on the same 1 image?
    Each CPU in the network renders one frame; the animation frames are rendered in parallel, greatly speeding up the process.

  4. #34
    www.Digitawn.co.uk rustythe1's Avatar
    Join Date
    Feb 2006
    Location
    england
    Posts
    1,279
    Quote Originally Posted by erikals View Post
    it is getting tricky to follow you with guesswork.

    - always upload the render
    - always provide the PC specs
    - and always provide other info like scene info / LW version / cinebench version
    - providing even more info is a bonus.

    people cannot follow you if you don't provide good info.


    but i'm out, too much time spent on this.

    and please fix your signature.
    I did a while ago, maybe you need to refresh or clear browser, should be an i9 7980xe,
    hardly any one has uploaded a render of that scene as people are only showing the stats,
    I understand what you mean but it all just seems to be nitpicking at a larger picture, doesn't matter if it was 17,14,10 its still a massive difference, to be honest what I can do has absolutely nothing to do with what I wanted to know, ill just keep all that to myself, all I want to know is how fast can a 2950 or greater render the marbles scene, then I can decide for my self if they were worth considering if I were to upgrade/add because yes there is a big price difference, but from my view with LW native its far from cost effective , I want to see real world rather than stats as they never seem to be remotely correct.
    Intel i9 7980xe, Asus Rampage Vi extreme, 2x NVIDIA GTX1070ti, 64GB DDR4 3200 corsair vengeance,
    http://digitawn.co.uk https://www.shapeways.com/shops/digi...ction=Cars&s=0

  5. #35
    Quote Originally Posted by rustythe1 View Post
    Where is the 30% from, if both lito and beevee rendered in 33 mins on 1950 and I did it in 14 originally that's over 100% as its over twice as fast, so like having two 1950s?
    I'm talking about the cinebench score which puts your CPU in the mid 9000's while the threadrippers are in the mid 6,000's. The marble scene I don't even see as relevant any longer since LW has changed their shading model since and the industry at large has moved on. Either way, even if somehow you could show that your CPU is twice as fast as the 1950, you still payed 4 times as much for only twice the performance (and again, that's based on your idea that its twice as fast)

  6. #36
    Quote Originally Posted by rustythe1 View Post
    Where is the 30% from, if both lito and beevee rendered in 33 mins on 1950 and I did it in 14 originally that's over 100% as its over twice as fast, so like having two 1950s?
    I still think most the noise is from windows snip, but maybe I have accidently changed a setting that isn't obvious as shaving nearly 30% of render seems a bit excessive, the main settings seem correct as most are listed in the render window and I assume the biggest contributor is the AA passes as all the other settings should be rendered in the first pass, so any setting that didn't affect aa should be negligible.
    Attachment 145916
    noise from windows snip,
    either way even my original post a year and a half ago with no turbo showed it was almost twice as fast, and the cinebench scores back that up, so doesn't that say the intel is more cost effective? at least from a Lightwave point of view.
    I think you might have changed some settings. Here I cut the rendering time by more than half with 4 changes:
    Render:
    Ray Recursion Limit: 12 changed to 6
    Ray Precision: 6.0 changed to 3.0
    Global Illumination:
    ISBG samples: 1024 changed to 512
    Rays Per Evaluation: 4 changed to 2

    Noise in the image has also increased. Though still can't see why the memory usage is so insanely bigger on my system. Still with these settings we are in the area of the IPC difference between Intel and AMD for the first gen Threadripper chips. The third gen TR chips should make things about even to intel's current offerings.
    Attached Thumbnails Attached Thumbnails Click image for larger version. 

Name:	LW2015BenchmarkMarbles1950x3.7Ghz-changedsettings4.jpg 
Views:	17 
Size:	136.2 KB 
ID:	145921  

  7. #37
    www.Digitawn.co.uk rustythe1's Avatar
    Join Date
    Feb 2006
    Location
    england
    Posts
    1,279
    I don't live in the US so maybe prices are different there, when I got my 7980 it was £1400, the TR was £999 at the time, so I get 25% more from my point of view, yes you are right about the marbles, but its the easiest way for me to see a comparison without much bother,
    how is it an "Idea"? I know what my times are, and its just against the 2 scores that were shown in the marbles thread. im not out to prove one is better than the other, im just asking some one to prove the other is better than what I see it to be,
    Intel i9 7980xe, Asus Rampage Vi extreme, 2x NVIDIA GTX1070ti, 64GB DDR4 3200 corsair vengeance,
    http://digitawn.co.uk https://www.shapeways.com/shops/digi...ction=Cars&s=0

  8. #38
    Quote Originally Posted by rustythe1 View Post
    I don't live in the US so maybe prices are different there, when I got my 7980 it was £1400, the TR was £999 at the time, so I get 25% more from my point of view,
    Well I got Zbrush for $250 through a deal with DAZ about 10 years ago, but that's not really relevant to someone looking to buy Zbrush now which is 900. Right now that 7980 is about 2000 and that 1950x is 500.

    I'm simply saying your idea because all the benchmarks I'm seing don't peg your 7980 to be twice as fast. If you want a comparison, you can see this page: https://www.cpubenchmark.net/compare...50X/3092vs3058

    What really stands out there is the much higher value you're getting performance vs $.

    I'm using the 2950X in my rig and I paid about 700 and its quite a bit closer matched to the 7980 then the 1950 is. https://www.cpubenchmark.net/compare...50X/3092vs3316
    Last edited by hrgiger; 09-08-2019 at 10:40 AM.

  9. #39
    www.Digitawn.co.uk rustythe1's Avatar
    Join Date
    Feb 2006
    Location
    england
    Posts
    1,279
    yea and I agree with you that the numbers etc are all much better toward AMD, but the point of my argument is Lightwave specific, this is a Lightwave forum and it seems to me there is an issue between certain renderers and AMD (its been pointed out in a couple of others on another benchmarking forum too but cant remember the link) so id rather see a real world comparison rather than a by the numbers,
    Intel i9 7980xe, Asus Rampage Vi extreme, 2x NVIDIA GTX1070ti, 64GB DDR4 3200 corsair vengeance,
    http://digitawn.co.uk https://www.shapeways.com/shops/digi...ction=Cars&s=0

  10. #40
    www.Digitawn.co.uk rustythe1's Avatar
    Join Date
    Feb 2006
    Location
    england
    Posts
    1,279
    https://www.tomshardware.com/reviews...ores,6312.html
    and if you read Forbes there were a couple of articles where they thought results are either false (systems they tested were a few hundred points below what cinebench results showed) or that there were rumors that the cinebench scores were done with prototype 5000 MHz ram,
    maybe some one else can do a score as well as marbles?
    Intel i9 7980xe, Asus Rampage Vi extreme, 2x NVIDIA GTX1070ti, 64GB DDR4 3200 corsair vengeance,
    http://digitawn.co.uk https://www.shapeways.com/shops/digi...ction=Cars&s=0

  11. #41
    Registered User
    Join Date
    Nov 2018
    Location
    Germany
    Posts
    38
    Quote Originally Posted by rustythe1 View Post
    as you can see even that one is way higher than what they list on their site, 9616 when it should be down near 5000, like I said something is off
    Your 7980XE isnt running at stock speeds, you are overclocking it. No wonder it has a much higher score than anticipated.
    Quote Originally Posted by rustythe1 View Post
    that's getting close to the 2990 score according to your link
    You compare a highly overclocked CPU with one at stock settings.

    Btw my result:


    And this is a list you can find at Computerbase.
    According to the list your 7980XE must be running at about 4,4 Ghz all core.
    Oh and look at the stock result of the Ryzen 9 3900X 12 Core CPU.
    Two of them would be just as fast as my overclocked 2990WX with 32 Cores.
    That should give you an idea of how good the Ryzen 3000 Series is.



    Quote Originally Posted by rustythe1 View Post
    https://www.tomshardware.com/reviews...ores,6312.html
    and if you read Forbes there were a couple of articles where they thought results are either false (systems they tested were a few hundred points below what cinebench results showed) or that there were rumors that the cinebench scores were done with prototype 5000 MHz ram
    No serious reviewer would test a system with an altered BCLK.
    Prototype 5000 Mhz RAM? Can you shed some light on this?
    Btw RAM speed doesnt affect CB results. If there is a difference at all its within the margin of error.

  12. #42
    www.Digitawn.co.uk rustythe1's Avatar
    Join Date
    Feb 2006
    Location
    england
    Posts
    1,279
    im not overclocking beyond the standard turbo in these tests, its at about 4.2 max, so again those scores are wrong, so if you have a 2990 please run the marbles on lw11 - lw2015, as I said, I don't care about scores as cinebench scores, they just don't add up anywhere
    Intel i9 7980xe, Asus Rampage Vi extreme, 2x NVIDIA GTX1070ti, 64GB DDR4 3200 corsair vengeance,
    http://digitawn.co.uk https://www.shapeways.com/shops/digi...ction=Cars&s=0

  13. #43
    Registered User
    Join Date
    Nov 2018
    Location
    Germany
    Posts
    38
    Quote Originally Posted by rustythe1 View Post
    im not overclocking beyond the standard turbo in these tests, its at about 4.2 max, so again those scores are wrong, so if you have a 2990 please run the marbles on lw11 - lw2015, as I said, I don't care about scores as cinebench scores, they just don't add up anywhere
    I am sorry but those scores are most certainly not wrong and CB results do add up.
    Especially the CB20 results as this version was designed for CPUs with many cores. It has a more intense workload than the older r15 version.

    No i will not run the Marbles Bench as its completely outdated and cherry picking on your side.
    CB 20 is an established benchmark with real world performance measurements.

    There are so many render benchmarks out there that do represent real world performance like blender and vray with up to date render engines.
    Marbles is an antique relic.

    A few examples.
    It doesnt include the 7980XE but the 12 Core Intel CPU 7920X and the also 12 Core part 3900X from AMD.
    The amount of performance you get for your money is just really great.







    Last edited by Lemon Wolf; 09-15-2019 at 05:52 PM.

  14. #44
    www.Digitawn.co.uk rustythe1's Avatar
    Join Date
    Feb 2006
    Location
    england
    Posts
    1,279
    but again I use any of those tests and it comes out higher, so without over clocking with blender benchmark, my 7980 is actually above the 9980
    Click image for larger version. 

Name:	blender.JPG 
Views:	7 
Size:	299.1 KB 
ID:	145989
    and means its faster than the 2970wx that has 12 more threads and a higher clock speed,
    Intel i9 7980xe, Asus Rampage Vi extreme, 2x NVIDIA GTX1070ti, 64GB DDR4 3200 corsair vengeance,
    http://digitawn.co.uk https://www.shapeways.com/shops/digi...ction=Cars&s=0

  15. #45
    Registered User
    Join Date
    Nov 2018
    Location
    Germany
    Posts
    38
    Quote Originally Posted by rustythe1 View Post
    but again I use any of those tests and it comes out higher
    Well then you are not running at stock settings.
    Can you provide images of your bios settings?
    What is your CPUs frequency during these tests?
    Do you really think that all the benchmarks are wrong and you must be right?

    Quote Originally Posted by rustythe1 View Post
    so without over clocking with blender benchmark, my 7980 is actually above the 9980
    Ok.
    Do you even know the difference between the 7980XE and the 9980XE?
    Its the same thing just with slightly better boost frequencies.
    There has been no real improvement between the 7980XE and the 9980XE.
    Go ahead look at the reviews. The 9980XE is even worse in some cases as it has really bad thermals because the heat spreader isnt soldered to the die like with the 7980XE.
    The 7980XE has a better OC headroom as well.
    And i do not believe that your system is not overclocked in some fashion.

    Quote Originally Posted by rustythe1 View Post
    and means its faster than the 2970wx that has 12 more threads and a higher clock speed,
    The 2970WX has most certainly not a higher all core or single core frequency than your 7980XE.
    And your list in the screenshot needs more context. Where is it from? Is it normal user results? What is it?
    Besides its about the new 3000 Series not the older 2000.
    Last edited by Lemon Wolf; 09-16-2019 at 06:10 AM.

Page 3 of 11 FirstFirst 12345 ... 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
  •