Cheaper (or faster?) CPUs are coming

[Ernest]

https://www.techpowerup.com/258951/i...cascade-lake-x

The first indication of this comes from an Intel slide that claims a 1.74 to 2.09x increase in performance-per-Dollar over "Skylake-X."

[Free4Ever]

Or you can buy a 16core AMD for 550$ now and render away, while Intel pushes more PR BS.
https://www.amazon.com/AMD-Ryzen-Thr...dp/B07GFN6CVF/

The new 3950x 16 core is about to be released in Sept for 750$ while intel 16 core costs twice.

If you like the waiting game, the upcoming Threadripper 3 in Q3, will slaughter any intel cpu in price/perf, like the server Epyc already does.

[erikals]

Or you can buy a 16core AMD for 550$ now and render away, while Intel pushes more PR BS.

Yes, unfortunately it looks like intel is far behind. (cost vs performance)

[rustythe1]

if AMD are so good please can some one show the marbles test scene results, on stock turbo i.e. around 4ghz or whatever (so not stupidly overclocked) I still have not even slightly taken back by any results ive seen in technical comparison to intel, and the amount of people I see complaining their threadrippers are not playing ball concerns me, anything I see on PCmark or CB just seems rubbish to me as none of the results seem anything like they should be, and I don't think any of those tests translate into how modern software actually utilises tools that intel provide,
I think I can guarantee no one with a TR2 on stock will beat my render time even though I have almost half the cores. and in that respect I guess the TR3 might struggle too

I know this sounds like im being very negative, I just still waiting for someone to prove me wrong with out BS

[TheLexx]

Maybe worth mentioning is there was some sort of Windows bug slowing AMD performance, but I'm not techy enough to know how relevant that is overall..

[Ryan Roye]

10 years ago it would have been a completely different story, all my machines used to be intel, now AMDs are populating my render farm.

[MichaelBeckwith]

10 years ago it would have been a completely different story, all my machines used to be intel, now AMDs are populating my render farm.

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?

[MichaelBeckwith]

Maybe worth mentioning is there was some sort of Windows bug slowing AMD performance, but I'm not techy enough to know how relevant that is overall..

I remember that. Level1 tech person found out it was the way that windows schedules threads and something to do with NUMA memory allocation. As far as I am aware (not 100% sure) Windows has fixed that with the latest creators update patch.

https://www.youtube.com/watch?v=M2LOMTpCtLA

[erikals]

I just still waiting for someone to prove me wrong with out BS

check >
https://www.cpu-monkey.com/en/cpu_be...5_multi_core-8

and then do the math of a complete PC versus performance.
https://pcpartpicker.com


but i discussed this with you before, so this is it as far as what i can provide you of info.

[Free4Ever]

You can get a preview of the Threadripper 3 and Ryzen 3950x 16 core in the benchmarks of the new 3900x 12 core. At 500$, Ryzen beats Intels 12 core i9-7920 at 1200$.
Includes Corona, Luxrender, POVRay and CineBench:

https://www.anandtech.com/show/14605...sing-the-bar/9

Price/Perf is more than 2x in favor for AMD.
Power of 105w for AMD and 140W for Intel will limit Intel even more in high core CPUs when cooling is the limit.

[Free4Ever]

Render farms are less interesting, now that you can buy a single rig with 2x64 CPUs or 256 threads using AMD EPYC 7742 Rome server CPU.
Cinebench R20:
2xAMD 7742 (128 cores): 31179
1xIntel 9980XE (18 core): 8791
1xIntel 8180 (28 core): 8585

So 4 high end Intel desktops or two 2P Xeon workstation/server to match one AMD 2P workstation/server.

https://m.hexus.net/tech/reviews/cpu...server/?page=7

https://www.cgdirector.com/cinebench...dated-results/

https://www.servethehome.com/crushin...ecord-edition/

[erikals]

Render farms are less interesting,

agree, most won't need it. unless one needs to quadruple the power.

[Lito]

if AMD are so good please can some one show the marbles test scene results, on stock turbo i.e. around 4ghz or whatever (so not stupidly overclocked) I still have not even slightly taken back by any results ive seen in technical comparison to intel, and the amount of people I see complaining their threadrippers are not playing ball concerns me, anything I see on PCmark or CB just seems rubbish to me as none of the results seem anything like they should be, and I don't think any of those tests translate into how modern software actually utilises tools that intel provide,
I think I can guarantee no one with a TR2 on stock will beat my render time even though I have almost half the cores. and in that respect I guess the TR3 might struggle too

I know this sounds like im being very negative, I just still waiting for someone to prove me wrong with out BS

I have done that benchmark previously in another thread, so I added it here and a new one I did this morning:

LW2015.3 x64 - BenchmarkMarbles.lws (included w/ LW 2015 content file and untouched)
System: Win 10 x64 1803 ; AMD Threadripper 1950x OC to 3.7ghz ; 64GB RAM @3066mhz ; Nvidia 1080Ti

The second image was done previously when I was still running a AIO cooler on my threadripper. Those coolers ended up being defective so I switched to a air cooler so I dropped my OC from 4Ghz to 3.7Ghz. With LW2015 the difference was about a minute per frame. To be fair though, I was downloading files and browsing the web while I was benchmarking but that should have negligible impact on the results. Also this is a first gen Threadripper, so results will be faster with a 2950x and probably significantly faster with the upcoming 3rd gen threadripper with the much improved IPC (which I most likely will upgrade to since I can just drop it into my current motherboard if AMD doesn't change their socket )

BTW, I don't think I have seen any of your benchmark times for your particular rig. What are your render times with this benchmark?

[Hail]

if AMD are so good please can some one show the marbles test scene results, on stock turbo i.e. around 4ghz or whatever (so not stupidly overclocked) I still have not even slightly taken back by any results ive seen in technical comparison to intel, and the amount of people I see complaining their threadrippers are not playing ball concerns me, anything I see on PCmark or CB just seems rubbish to me as none of the results seem anything like they should be, and I don't think any of those tests translate into how modern software actually utilises tools that intel provide,
I think I can guarantee no one with a TR2 on stock will beat my render time even though I have almost half the cores. and in that respect I guess the TR3 might struggle too

I know this sounds like im being very negative, I just still waiting for someone to prove me wrong with out BS

My mate's 3900x is about 40-44 percent faster than my i7 8700k, in rendering in c4d with the physical renderer.
That's his main workstation, so it'd be difficult to find time to run the marbles scene.
Will try to see if I can get him to run that.

[rustythe1]

check >
https://www.cpu-monkey.com/en/cpu_be...5_multi_core-8

and then do the math of a complete PC versus performance.
https://pcpartpicker.com


but i discussed this with you before, so this is it as far as what i can provide you of info.

and this is my point, because as far as lightwave is concerned all these scores appear to be meaningless, don't know if its a bug or an intel specific feature, or maybe they are just running the tests at stock 2.6 but the scores are way off what I get out of mine, and the render times I have seen don't seen in other areas that spectacular when they should going by scores and numbers be twice what I can do, so no body here will do the marbles test scene? (not the 2019 update, just the original 11-15)