【Maxon】Cinebench R20 Benchmark Thread

[csbin]

Download: https://www.techpowerup.com/download/maxon-cinebench/

Only support 64bit System



https://www.anandtech.com/show/14053/maxon-releases-cinebench-r20-benchmark

The new benchmark implements a number of new features available for x86 processors. This includes newer AVX instructions, implemented through Intel's Embree raytracing technology, applicable on both AMD and Intel processors. The new benchmark focuses purely on CPU performance, and solves one of the issues with the R15 version - it now scales beyond 64 threads better by having a larger scene. There will be a minimum performance limit to run the new benchmark, which is listed as a memory limit.

Similar to R15, there is a single threaded test and a multithreaded test. The GPU test has been dropped (it hasn't been relevant for a long time).












Went from full sse/sse2 128wide code to AVX2+FMA 256bit.Power consumption nearly the same.I think Maxon has added only a few parts of embree on the renderer(looking at the leftover sse stuff).Embree as a whole is way more demanding than this.





Tweet link removed due to profanity.
I inserted an edited version.

Iron Woode
Super Moderator

 

[TheELF]

But, but,it hasn't even been a decade since the last release...

 

[TheELF]

There will be a minimum performance limit to run the new benchmark, which is listed as a memory limit.

Seems to use 2.2Gb private mem for me.

 

[PotatoWithEarsOnSide]

I guess that AVX is more heavily included to ensure that Intel still "maintain" the single thread crown...?

 

[tamz_msc]

I guess that AVX is more heavily included to ensure that Intel still "maintain" the single thread crown...?

?
Intel already had the ST crown in R15.

 

[inf64]

Th difference in ST test is not that large though . 2990WX gets a monstrous score in both ST and MT subtests.

 

[.vodka]

3.8GHz R7 1700:

It seems it can be looped, so it can be used as a nice realistic stress test, too:

 

[SPBHM]

hey, the number is higher than the R15 result, so I think that's good!

 

[TheGiant]

I guess that AVX is more heavily included to ensure that Intel still "maintain" the single thread crown...?

AVX has some uses and I expect on the contrary that when AMD implements 256bit wide core with ryzen 3xxx cinebench will suddenly use more AVX

/sarcasm off

for more technical members- cinebench/rendering and AVX/2/512=friends or not so much?

 

[jpiniero]

Doesn't look like it uses AVX2 much and AVX-512 at all. Mostly just regular AVX.

Coffee Lake at 5 Ghz gets around 510 or so in single thread it looks like.

 

[Phynaz]

I guess that AVX is more heavily included to ensure that Intel still "maintain" the single thread crown...?

Sure, sure. That's why AMD benchmarks with Cinebench I'm sure.

What's that, Cinebench uses an open-source renderer, that's most likely the best one available? And you can download binaries compiled with ICC 2019, VS 2015 and VS 2017?

Five minutes of Googling can still prevent people from posting really stupid remarks on the Internet.

 

[jpiniero]

Isn't Cinebench based upon Cinema4D? And that's a commercial product.

 

[PotatoWithEarsOnSide]

The whole point was that as the latest gen AMD CPUs were likely to take the ST crown, the goalposts get moved in a manner that would favour Intel.
Whilst it's true that AMD will be improving on the AVX front themselves, they are still likely to lag in any workload that utilizes it.

 

[exquisitechar]

The whole point was that as the latest gen AMD CPUs were likely to take the ST crown, the goalposts get moved in a manner that would favour Intel.
Whilst it's true that AMD will be improving on the AVX front themselves, they are still likely to lag in any workload that utilizes it.

I'm not very familiar with it, but if Cinema 4D uses AVX2 I don't see how it's goalpost moving, it's more like the old version of Cinebench was outdated and not representative of modern rendering workloads, which was actually a common criticism of it.

Zen 2 based CPUs should be equal on that front to Coffee Lake, at least.

 

[Gideon]

The new version uses Intel Embree raytracing library (mentioned in the Anandtech article). It's 100% Open Sourced Software, so everybody can validate it for shenanigans.

It's also one of the most developed and widely used Software Raytracing libraries (just check out this siggraph presentation). Not just by Cinema4D but also Blender, Corona, V-Ray. On top of that it's used for niche stuff by companies like Autodesk, Dreamworks, Ubisoft and Valve.

Yes, it's developed by Intel and yes, it supports AVX2 and AVX-512 (on top of SSE and AVX) but it most certaninly does not cripple the slower code-paths, which is quite visible of how well AMD is still doing, despite it's only 128bit wide vector units (just look at the Threadripper scores).

Are you guys honestly arguing that actual Rendering Engines doing real work (and their benchmarks) should cripple themselves to SSE3 (as Cinebench 15) just that AMD would look more competitive. If they don't, they are somehow favouring Intel?

 

[Gideon]

Just to prove the ponit. If anything, my work 6-core Coffee Lake (i7-8850H in 15" Macbook Pro) actually performs worse in CB 20 than CB 15, relative to my Ryzen 1700X as the render takes longer and it needs to lower the clock. Will run benches to compare in a sec ...

EDIT:
So did 2 quick runs on both @stock, without really closing all apps. The 1700X definitely could be faster if tuned (as seen from the best CB15 results as well in the screenshot)

1700X scored:
CB15: 1563 pts
CB20: 3474 pts

8850H scored:
CB15: 1000 pts (yes really! on this one reference run, screenshots below)
CB20: 2358 pts

That is despite of that 8850H having AVX2 (as it's one the latest Coffee-Lake Laptop parts)

Screenshots here

So the 8850H 's CB20 score is about 67,8% of the Ryzen Score.
while the 8850H 's CB15 score is 63,9% of the Ryzen Score

Running the benchmarks multiple times changes the scores in one or other direction within 1-2%


Bottom line:
I was wrong and the laptop did not end up slower. AVX2, actually makes the Coffee Lake shrink it's gap with the Desktop Ryzen by about 4% in the newer benchmark. Still, this is very far from an unrealistic improvement, considering the CPU can finally use it's 2x wider vector units. It's a totally valid change for a real-world scenario from (For instance, comparing the Handlebars HVEC encoding case with x264, havin wider vector units extract considerably more performance, and that is a valid encoding use-case).

 

[Phynaz]

The whole point was that as the latest gen AMD CPUs were likely to take the ST crown, the goalposts get moved in a manner that would favour Intel.
Whilst it's true that AMD will be improving on the AVX front themselves, they are still likely to lag in any workload that utilizes it.

Read: The industry should continue to use six year old benchmarks so my favorite brand looks better.

 

[Markfw]

My STOCK 2970wx 9548 !!

 

[scannall]

My stock settings and cooler 2600X gets 3018.

 

[IEC]

i7-8700K @ 4.7GHz core/cache, no AVX offset = 3675 cb

 

[hdfxst]

https://imgur.com/4YMe8gx

 

[IEC]

Stock Ryzen 7 2700X, air cooled, 3000 CL16 RAM, X370 board = 3940 cb

 

[bbhaag]

Stock i7-2820QM with 8gb ram. Just clicked run on the test.

 

[virpz]

Still scoring!

 

[IEC]

i7-8700K 5.1GHz core, 4.7GHz cache, no AVX offset = 3980 cb