This example scene uses voxel preview which does quite well with auto reduction and should not lag the interface too much. If you put it into mesh or gpu preview mode where the auto reduction is not implemented, and forcefully use no reduction, it might make a bigger difference between the GUI and the batch simulation.
Thank you. That makes sense, off course.
I would still love to know how to get the best possible rig for simulation. I guess that a Threadripper system is not a bad choice when considering simulation AND rendering. A similar system would cost much more if Intel based.
Finally, the good choice is to multiply the simulation rigs to be able to launch several simulations over night 
I have 3 AMD TR 2990WX’s and an i9-9900K, Ill download the scene and try it out on the intel and AMD in the next day or so…
-dave
Wow that was a bit of a shock. I had some time so I tried to load up the scene in the OP at the top. Until this afternoon I was actually thinking of switching back to the Threadripper as the 9900k just “felt” slower but apparently that is not the case. Ill drop in a screen shot of the two machines, I’ve also cropped the screenshots purely for size but not hiding anything and can post the full size shots if asked lol. I think I still have access to the nightlies, I might grab one of the latest of those as well and give it a shot too. Anyway here are the results:
Both machines are configured the same with Win 10 | 3dsmax 2020.1 | V-Ray 42001 - Update 2.1 (4.20.01.00001) | Phoenix 3.14.00 Build:20190427 | Forest 621
------------------------------------------------
i9-9900k - Finished in: 12m 22s
------------------------------------------------
TR4 2990wx Finished in: 14m 06s
------------------------------------------------
its not a huge difference just looking at it but if you have or do a lot of simulations then I think the Intel is the way to go.
Any word yet on how Phoenix does comparing the latest AMD 3rd Gen Ryzen CPU’s?
-dave
yeah threadrippers are shithouse, even worse with 128gb RAM
Tho, in H, tr feels very solid and my xeon is slower, like it should theoretically. Don’t know, why phx reacts so differently. Gonna test example liquids too a bit later today.
In phoenix the TR4’s cant compete but in rendering its another story as my TR4’s are substantially faster at rendering than my 9900k. Everything has its use and in most industries you can use two or more similar tools to do the same job, and generally one tool will be more efficient than the other at certain jobs…
Here you will find chart and a scene. My xeon is like 10core at 4.7. Compare to phx results Maybe it is very bad max architecture, that limits phx. Dunno.
yes, that’s why I found that the TR 2950X was good choice (high clock speed). It might not be the fastest in sim, but it does a good job when rendering. So, when you have to do both, it’s a good alternative for a good price.
Might as well revive this thread with the launch of the AMD 5000 series!
CPU AMD RYZEN 5950x 4.875Ghz [32 threads]
Total elapsed
9m 47s
Performance
4.45 vox/s
RAM is 4x32GB (128GB) Corsair LPX 3600c18 running at 3600c16, I think there is more headroom in this RAM for overclocking but haven’t got around to it yet, I wanted to confirm this was stable first. Manually setting the 1.8v PLL to 1.8v dropped my time and performance by a lot (15m 43s, 2.67M vox/s) but bumping it up incrementally gave significant performance uplift, right now it is at PPL 2.05v for this score. PBO is enabled on the CPU, I see single core hitting 5ghz briefly. FCLK 1800mhz for 1:1.
Saving to an SSD with compression disabled.
Interesting. Which version u used ? Gonna retest, since u hit me with performance, but more total time… Gonna retest Thos l3 speed is quite boosted. Kewl. And what temp was ?
5950x looks pretty decent for a workstation / sim box with a fast gpu
I used 4.20. After a bios update that came out a few days ago I don’t have to mess with the 1.8 PLL and push it beyond 1.8v anymore and I’ve been working on tightening up my DRAM timings to reduce the latency too but its not as easy with 4 sticks of duel rank high capacity memory. Here is my new score:
CPU AMD RYZEN 5950x 4.875Ghz [32 threads]
Total elapsed
9m 30s
Performance
4.57 vox/s
Temps didn’t go above 77c (I have a waterblock on there).
Todays nightly. 4.30
CPU AMD RYZEN 3960x 4.2Ghz [48 threads]
Total elapsed
9m 27s
Performance
4.76 vox/s
Hi Guys,
Did any of you test already a 3970X or 3990X with Phoenix (simulation and rendering) ?
Tested with the Car tire burnout scene for 3ds Max and Maya with the Preview disabled during simulation.
Link to the 3ds Max scene - Phoenix for 3ds Max
Link to the Maya scene - Chaos Docs
AMD 3990x 128 threads @ 4.0 GHz
Memory 3400 mhz
SSD Samsung 970 EVO Plus
3ds Max - 4.30.01 Nightly, Build ID: 2020112730466
[B]Total Elapsed 10 min 22 s[/B]
Maya - 4.30.01 Nightly, Build ID: 2020112730466
[B]Total Elapsed 22 min 10 s[/B]
And the next one
AMD 3970x 64 threads @ 4.23 GHz
Memory 3200 mhz
3ds Max - 4.30.01 Nightly, Build ID: 2020112730466
[B]Total Elapsed 9 min 36 s[/B]
Maya - 4.30.01 Nightly, Build ID: 2020112730466
[B]Total Elapsed 19 min 46 s[/B]
Errr… reassure me… maya and max scenes don’t have exactly the same settings, right ? :shock: Or do I have to ditch Maya ?
We might have discussed this already, but I’m not sure…
hehe, no worries, whoever ported the scene just did something for Maya without taking care of equalizing the settings.
As an addition - there is no reason for simulations in maya to be any different in speed versus Max. There are some additional thing that might differ between Max and Maya and be faster in either - texture sampling, talking with particles and with forces, and the previews, which are often faster in Maya.