Maxwell Render | Official public forum

Since the benchwell site is abandoned again I thought I'd post some benchmarks on modern hardware I've been getting in:

Intel i9-7940X CPU @ 4.6 Ghz, 128GB RAM, Asus Rampage VI x299

• Benchwell Maxwell 4.1.1.1: 2:32 time, 1843
• Cinebench: 3350
• V-Ray: :39

Intel i9-7980xe CPU @ 4.2 Ghz, 32GB RAM, EVGA x299 FTW

• Benchwell Maxwell 4.1.1.1: 1:55 time, 2419
• Cinebench: 4180
• V-Ray: :30

Intel i9-7980xe CPU @ stock, 32GB RAM, EVGA x299 FTW

• Benchwell Maxwell 4.1.1.1: 2:21 time, 1974
• Cinebench: 3340
• V-Ray: :38

Intel 2x Xeon E5-2699v4 CPU @ 2.2Ghz, 256GB RAM, Supermicro X10DRG-Q

• Benchwell Maxwell 4.2.0.2: 1:21 time
• Cinebench: 5402
• V-Ray: :25

Intel 2x Xeon Gold 6154 CPU @ 3.0Ghz, 384GB RAM, Supermicro X11DPG-QT

• Benchwell Maxwell 4.2.0.2: 1:08 time
• Cinebench: 6431
• V-Ray: :20

AMD Threadripper 1950x @ stock, 32 GB RAM, Gigabyte x399 DESIGNARE EX

• Benchwell Maxwell 4.2.0.2: 2:47 time
• Cinebench: 2889
• V-Ray: :46

Apple iMac Pro, 8 core 3.2 GHz Intel Xeon W, 32 GB RAM,Vega 56

• Benchwell Maxwell 4.1.1.1: 4:41 time, 993
• Cinebench: 1662
• V-Ray:

Apple iMac Pro, 10 core 3.02 GHz Intel Xeon W, 128 GB RAM,Vega 64

• Benchwell Maxwell 4.1.1.1: 3:57 time, 1176
• Cinebench: 1881
• V-Ray: 1:03

AMD Threadripper 2990x @ stock, 64 GB RAM

• Benchwell Maxwell 4.2.0.3: 1:45 time
• Cinebench: 5119
• V-Ray: :27

Will update as tests are completed.

I can add my own home PC:

Ryzen 7 1700 CPU @ 3.4 Ghz, 16GB RAM, Maxwell 4.1.1.1: 739.85

Good idea, thank you,
pretty interested by the bench of the 7980XE


The benchwell needs filters to sort the processors...
It is unusable in its current form.
(with an average of all same processors, among others)

Another reference, my Mac Pro from 2009, running High Sierra:

Dual Xeon X5690 @ 3.46 GHz, 48GB RAM:

• Benchwell Maxwell 4.1.1.1: 785.63

• Cinebench 15: 1602

update above numbers with the 18 core i9. It can be run higher, I ran at 4.6 for some tests. What I'm noticing is that the VRM cooling issue on i9 overclocks can be a problem. Choose a motherboard with a good VRM design with dual 8-pin CPU power connections, such as the EVGA x299 FTW or similar.

I've also got some Threadripper 1950x CPU sitting here that will be benched next week.

Hi, below mine (both ran in MR v3.2.1.5):

single CPU E5-1650v3 @3.50ghz, RAM 32gb,
Benchmark: 567 Time: 8m12s

dual E5-2620v2 @2.10ghz, RAM 40gb
Benchmark: 315 Time: 14m44s

Question: should the machine with 567 points render in general (almost) 2x faster than the one with 315?
Because in reality, based on day to day usage, rendering some heavy SSS & heavy displacement, they feel almost similar...
i.e. how accurate is Benchwell, is it safe to use the scores posted in this thread while specifying up a new Workstation?

thanks

D

fi3er wrote: ↑

Wed Dec 06, 2017 5:30 pm

Hi, below mine (both ran in MR v3.2.1.5):

single CPU E5-1650v3 @3.50ghz, RAM 32gb,
Benchmark: 567

dual E5-2620v2 @2.10ghz, RAM 40gb
Benchmark: 315

Question: should the machine with 567 points render in general (almost) 2x faster than the one with 315?
Because in reality, based on day to day usage, rendering some heavy SSS & heavy displacement, they feel almost similar...
i.e. how accurate is Benchwell, is it safe to use the scores posted in this thread while specifying up a new Workstation?

thanks

D

Its better to focus on the completion time rather than the benchwell number. It seems to be more consistent.

gotcha, thx dmeyer. I've added the times into my previous post.

So my question is: If I get a i9-7980xe machine (like yours, that completed in 2:21), its gonna render (approx) 4x faster than my current E5-1650v3 (that completed in 8m12s), across all different renders in day to day usage (obviously with slight differences depending on type of scene etc)?

thanks

D

fi3er wrote: ↑

Wed Dec 06, 2017 7:21 pm

gotcha, thx dmeyer. I've added the times into my previous post.

So my question is: If I get a i9-7980xe machine (like yours, that completed in 2:21), its gonna render (approx) 4x faster than my current E5-1650v3 (that completed in 8m12s), across all different renders in day to day usage (obviously with slight differences depending on type of scene etc)?

thanks

D

I can't think of why it wouldn't be a pretty consistent performance difference. The only exception would be in single threaded parts of the render - which in Maxwell is very little.

bump - added AMD 1950x scores

iMac

i7 6700K @ 4.00 GHz

11m 03 Seconds

431.34

Operating System: Win64
Processor: AMD Ryzen 5 1600 Six-Core Processor (No OC)
Total Physical Memory: 6.4 GB
Number of Cores: 12
Engine version: 4.1.1.1
Time: 8 m 10 s
Benchmark: 569.40

Bump for updated Xeon 6154 numbers. Cinebench and V Ray running fine but Maxwell doesn't like something about the Xeons...it takes way longer than it should even though the cpu's are at 100% the whole time. Will investigate.

Thanks for doing this, that threadripper is a bit dissapointing, although it is a cheaper option, I think it will slow down more with large scenes relative to the intels.