Post your Ryzen Blender Demo Scores! (AMD clarifies Blender Benchmark Confusion, Run @ 150 Samples)

[inf64]

Yep, there's nothing there indicating that you've done calculations to equalize things. Those statements can just as easily mean that you've simply misread the spec

Come on, at the same clock means at the same clock. There is no uncertainty there and nothing to misread.

PS there is no 4.3Ghz stock i5 part on the market which means I manually OCed mine. This in turn means I know what I'm doing .

 

[Valantar]

Come on, at the same clock means at the same clock. There is no uncertainty there and nothing to misread.

PS there is no 4.3Ghz stock i5 part on the market which means I manually OCed mine. This in turn means I know what I'm doing .

You assume that I assume you to be technically competent. On the one hand, that might be a good idea in these forums. On the other hand, some times I get baffled

 

[asendra]

i7-5820k encode 00:01:20
Stock CPU settings but it doesn't even turbo boost, stays at 3.3GHz. Same situation with blender.

Just for fun, my 4770k@4,2Ghz ddr3-1600 does it in 00:01:29

 

[siriq]

2:05.48 on the 8350 (stock), win7 64bit, 16gb ram DDR3 @1600, ran it from hard drive.

How the heck you got the same score with my OCed 8350@4.8 ?

I think i know why, i forgot to turn off the APM

Just tested it. Even with APM off i cannot get better result. That CPU with STOCK clock with that result feels unreal.

 

[nismotigerwvu]

Could be a RAM issue or some other quirks (OS and whatnot). A few pages ago, another member saw dramatically slower results to what I saw on my systems. Can you try monitoring utilization and clocks during the test to make sure all of your cores are pinned and running at 4.8?

 

[rvborgh]

ran it on one of my non AMD machines... i7-5930K (bone stock 3.5 GHz on all cores during the run) at 150 samples.

48.73 seconds

 

[Sven_eng]

76 seconds on my stock 6700k.

 

[superstition]

Windows 8.1 64-bit, 8370E 5 GHz, 1814 speed CAS 9 16 GB RAM, 238 BCLK, UD3P 2.0 board FC BIOS

CMT off (4M/4T, 1 integer and 1 FPU per module, Gigabyte BIOS option)

Blender 2.78.4 — The Stilt "SIMD" build (Blender_278_GIT_SIMD_INTRIN)

100 samples: 0:59.59
150 samples: 1:29.36
200 samples: 1:58.46

Blender 2.78.4 — The Stilt "MSVC" build (Blender_278_1412GIT_AVX2_MSVC)

100 samples: 1:06.21
150 samples: 1:39.53
200 samples: 2:11.45

Blender 2.78a

100 samples: 2:04.66
150 samples: 3:06.01
200 samples: 4:08.26

Blender 2.75a

100 samples: 1:56.83
150 samples: 2:54.58
200 samples: 3:52.77

Removed full 8 core 5 GHz results because they may not be accurate due to having a weak board with VRM issues.

 

[superstition]

Let's do the Time Warp again! (back to 2009)

i5 750 “Lynnfield” 4C/4T, 3.8 GHz, 16 GB (2 x 8) 1600 RAM CAS 9, Windows 7 x64 SP1, clean run

methodology and notes: As with the 8370E tests, I found that running the Blender render twice-in-a-row usually yields the best performance. Running it a third time doesn’t. So, the following scores are the best of two runs. This trend didn’t hold true on the Mac, of course, because of thermals and turbo.

Turbo is disabled on this Lynnfield as it was on the 8370E.

Blender with AMD's Ryzen file results, in order of performance, rounded to tenths

2.75a

100: 1:17.7
150: 1:56.4
200: 2:35.2

The Stilt "SIMD" build, 2.78.4

100: 1:18.6
150: 1:57.1
200: 2:36.7

2.76b

100: 1:19.3
150: 1:58.9
200: 2:38.3

2.77a

100: 1:20.7
150: 2:01.0
200: 2:41.2

2.78a

100 samples: 1:22.0
150 samples: 2:03.1
200 samples: 2:43.2

The Stilt "AVX2" build, 2.78.4

100: 1:22.7
150: 2:03.8
200: 2:44.7

———————————

Cinebench R15

multi: 454
single: 116
mp ratio: 3.90x
Open GL (GTX 460 768 MB at 705/1410): 91.87 fps

Cinebench R11.5

multi: 5.12
single: 1.30
mp ratio: 3.94x
Open GL (GTX 460 768 MB at 705/1410): 45.97 fps

CPU-Z 1.77.0.x64

single: 1365 (consistent)
multi: 4777 (best result)

 

[superstition]

0:55.51
i7-4960HQ, late 2013 Macbook Pro, 16 GB DDR3L 1600, 10.9.5 "Mavericks", Blender 2.78a

https://s27.postimg.org/v20uikmqb/time.png

 

[RichUK]

7700k @ 5Ghz / DDR4-2800:

Blender V2.77 - Ryzen Render @ 150 Samples (45.47s)

Blender V2.78a - Ryzen Render @ 150 Samples (46.08s)

 

[biostud]

40.57s with rig in sig.

 

[superstition]

8370E @4.4 4M/4T (CMT off) non-turbo 2133 9-11-10-30-43-1T, Windows 8.1 x64

all benchmark results are the best of two passes

Blender and Ryzen file:

“SIMD” build 2.78.4

100 samples: 1:07.7
150 samples: 1:41.1

AVX2” build 2.78.4

100 samples: 1:15.5
150 samples: 1:53.2

2.75a (7-1-2015)

100 samples: 2:12.7
150 samples: 3:18.1

2.76b

100 samples: 2:15.9
150 samples: 3:23.8

2.77a

100 samples: 2:17.7
150 samples: 3:25.2

2.78a (10-30-2016)

100 samples: 2:21.6
150 samples: 3:32.4

———————— other benches ————————

POV-Ray 3.7.0.msvc10

kernel — .11
user — 990.17
total — 990.28
elapsed — 251.64
CPU vs. elapsed ratio — 3.94
render averaged 1041.74 PPS (264.72 PPS CPU time)
262144 pixels, 4 threads

Cinebench R15

single: 106
multi: 413
mp ratio: 3.90x
OpenGL (7870 GHz): 100.3 fps

Cinebench R11.5

single: 1.17
multi: 4.61
mp ratio: 3.94x
OpenGL (7870 GHz): 74.4 fps

CPU-Z 1.77.0.x64

single: 1295
multi: 4987

CPU-Z 1.78.1.x64

single: 1295
multi: 4991

Once again, we see that something isn't right with the official Blender builds. They scale normally with Lynnfield but even with CMT off, they fail tremendously with Piledriver. The Stilt's "SIMD" build is vastly quicker.


8370E 4.4 GHz 4M/8T non-turbo, 2133 9-11-10-30-43-CR1 RAM 16 GB (2 x 8), Windows 8.1 64, 200 BCLK, 2400 FSB, 2600 HT

The Stilt’s “SIMD” build, 2.78.4:

100: 0:40.0
150: 0:59.7
200: 1:19.3

The Stilt’s “AVX2” build, 2.78.4:

100: 0:45.0
150: 1:07.2
200: 1:29.5

2.75a

100: 1:11.2
150: 1:46.6
200: 2:22.0

2.76b

100: 1:14.1
150: 1:51.3
200: 2:27.2

2.77a

100: 1:14.8
150: 1:52.4
200: 2:28.8

2.78a:

100: 1:16.0
150: 1:53.9
200: 2:31.3

———————————————

CB R15

single: 106
multi: 706
mp ratio: 6.66x
OpenGL (7870 Ghz): 102.72 fps

CB R11.5

single: 1.17
multi: 7.64
mp ratio: 6.51x
OpenGL (7870 Ghz): 75.55 fps

CPU-Z 1.77

single: 1293
multi: 8400

 

[rvborgh]

Perhaps OT... but perhaps this is what Ryzen remedies.

http://www.ilsistemista.net/index.p...n-whats-wrong-with-amd-bulldozer.html?start=6

8370E @4.4 4M/4T (CMT off) non-turbo 2133 9-11-10-30-43-1T, Windows 8.1 x64

all benchmark results are the best of two passes

Blender and Ryzen file:

“SIMD” build 2.78.4

100 samples: 1:07.7
150 samples: 1:41.1

AVX2” build 2.78.4

100 samples: 1:15.5
150 samples: 1:53.2

2.75a (7-1-2015)

100 samples: 2:12.7
150 samples: 3:18.1

2.76b

100 samples: 2:15.9
150 samples: 3:23.8

2.77a

100 samples: 2:17.7
150 samples: 3:25.2

2.78a (10-30-2016)

100 samples: 2:21.6
150 samples: 3:32.4

———————— other benches ————————

POV-Ray 3.7.0.msvc10

kernel — .11
user — 990.17
total — 990.28
elapsed — 251.64
CPU vs. elapsed ratio — 3.94
render averaged 1041.74 PPS (264.72 PPS CPU time)
262144 pixels, 4 threads

Cinebench R15

single: 106
multi: 413
mp ratio: 3.90x
OpenGL (7870 GHz): 100.3 fps

Cinebench R11.5

single: 1.17
multi: 4.61
mp ratio: 3.94x
OpenGL (7870 GHz): 74.4 fps

CPU-Z 1.77.0.x64

single: 1295
multi: 4987

CPU-Z 1.78.1.x64

single: 1295
multi: 4991

Once again, we see that something isn't right with the official Blender builds. They scale normally with Lynnfield but even with CMT off, they fail tremendously with Piledriver. The Stilt's "SIMD" build is vastly quicker.


8370E 4.4 GHz 4M/8T non-turbo, 2133 9-11-10-30-43-CR1 RAM 16 GB (2 x 8), Windows 8.1 64, 200 BCLK, 2400 FSB, 2600 HT

The Stilt’s “SIMD” build, 2.78.4:

100: 0:40.0
150: 0:59.7
200: 1:19.3

The Stilt’s “AVX2” build, 2.78.4:

100: 0:45.0
150: 1:07.2
200: 1:29.5

2.75a

100: 1:11.2
150: 1:46.6
200: 2:22.0

2.76b

100: 1:14.1
150: 1:51.3
200: 2:27.2

2.77a

100: 1:14.8
150: 1:52.4
200: 2:28.8

2.78a:

100: 1:16.0
150: 1:53.9
200: 2:31.3

———————————————

CB R15

single: 106
multi: 706
mp ratio: 6.66x
OpenGL (7870 Ghz): 102.72 fps

CB R11.5

single: 1.17
multi: 7.64
mp ratio: 6.51x
OpenGL (7870 Ghz): 75.55 fps

CPU-Z 1.77

single: 1293
multi: 8400

 

[superstition]

The only thing Ryzen appears to remedy with the official Blender builds is that it's not being hobbled unnecessarily like the construction cores appear to be.

2009 Lynnfield = normal scaling. Tiny benefit from the "SIMD" build, tiny hampering from the "AVX2" build

Piledriver = terrible performance from official Blender builds, massive performance increase from "SIMD" build and "AVX2" build, with the former being faster than the latter

Phenom and even older AMD architectures beating the construction cores, even when CMT is off? A CPU running DDR2 800 is more efficient? Sure...

This data progression doesn't look like it fits the claim that the reason we're seeing it is because of Piledriver's design inadequacies. I don't buy it as being useful for measuring Ryzen's performance versus the construction parts:

23,437 02:08 min : AMD FX-8350 „Vishera” (4M/8T, 4.0 GHz, 4.2 GHz Turbo, DDR3-1866 DC) von Nero24
25,518 09:30 min : AMD Athlon II X3 400e „Rena” (3C/3T, 2.2 GHz, kein Turbo, DDR3-1600 DC) von donmartin3000
25,805.06 FX-8370E (@ 5Ghz) single thread per module 25,805.06
31,968 11:55 min : AMD Athlon II M320 „Caspian” (2C/2T, 2.1 GHz, kein Turbo, DDR3-1066 DC) von Nero24
37,137 08:37 min : AMD Athlon X2 7550 „Kuma” (2C/2T, 2.5 GHz, kein Turbo, DDR2-800 SC) von Nero24
38,784 03:58 min : AMD Phenom II X4 810 „Deneb” (4C/4T, 2.6 GHz, kein Turbo, DDR3-1066 DC) von Nero24
51,282.05 Sandy Bridge 2500K @ 4.5 Ghz
51,306.17 i5 750 (“Lynnfield” 4C/4T, 3.8 GHz)

Let's see The Stilt's builds progressed out like this, eh?

 

[FlanK3r]

i7-5960X 3.5 GHz (locked) with 2400 MHz DDR4 = 39.07s

 

[dfk7677]

i5 3570k (stock): 121.72s
i5 4590: 107.89s

 

[superstition]

6700K Skylake, stock, turbo on, static undervolt, 3200 16 GB 16-18-18-2T, Windows 10 AE

as usual, each score is best of two runs (except Cinebench single thread as it’s slow)

in order of speed, rounded to tenths:

“SIMD”, The Stilt

100: 0:23.9
150: 0:35.7
200: 0:47.6

“AVX2”, The Stilt

100: 0:24.9
150: 0:37.2
200: 0:49.7

2.75a

100: 0:34.8
150: 0:52.2
200: 1:09.4

2.76b

150: 0:52.7

2.78a

100: 0:36.0
150: 0:53.7
200: 1:11.7

———————— other benches ————————

POV-Ray 3.7.0.msvc10

kernel: 0.02
user: 1030.84
total: 1030.86
elapsed: 132.15
CPU vs. elapsed ratio: 7.80
render averaged 1983.75 PPS (254.30 PPS CPU time)
262144 pixels, 8 threads

Cinebench R15

single: 183
multi: 929
ratio: 5.09x
OpenGL (Sapphire 1070 stock): 157.6 fps

Cinebench R11.5

single: 2.06
multi: 10.22
ratio: 4.96x
OpenGL (same GPU): 89.3 fps

CPU-Z 1.78.1x64

single: 2109
multi: 9304

CPU-Z 1.77.0.x64

single: 2109
multi: 9202

---------------------------------- what these results mean ----------------------------------

The official Blender builds are a great test of 2009's Lynnfield processor and very poor for looking at anything starting with the 2011 Bulldozer 8150 and Sandy.

 

[siriq]

Blender 2.78.4 — The Stilt "SIMD" build
AMD FX 8350@4.8
150 samples : 1:08.02
Lot faster with this build compare to the official one. 2:05 or 2:06 .

 

[siriq]

I just did one more test:
150 samples AMD FX 8350@4.8

01:07:03

 

[Shivansps]

1:48:88 with a 4.2Ghz 2500K

 

[siriq]

I think i will keep my rig. Zen with 16 threads(stock) does 36 sec and FX 8350@4.8 does 67 sec. Despite the fact FX is on OC, still holding up , compare to stock Zen(8/16 threads). I would say, Zen 4/8 will be around 20% faster in most scenarios compare to my rig. Better to just save some money for myself, for REV 2.0 Zen. Let me know about your odds

 

[William Murdock]

I7 6800K Asus X99-A Stock 3.4GHz

100 - 32.70
150 - 48.71
200 - 64.91

 

[superstition]

I think i will keep my rig. Zen with 16 threads(stock) does 36 sec and FX 8350@4.8 does 67 sec. Despite the fact FX is on OC, still holding up , compare to stock Zen(8/16 threads). I would say, Zen 4/8 will be around 20% faster in most scenarios compare to my rig. Better to just save some money for myself, for REV 2.0 Zen. Let me know about your odds

Here are my results with stock Blender, for reference:

Skylake 6700K, 2.78a

100 samples: 0:36.0
150 samples: 0:53.7
200 samples: 1:11.7

Your FX result is significantly better at 4.8 than mine was at 5. I guess I'll need to retest at 4.8. The 5 GHz testing wasn't on a tested-as-Prime-stable clock because my board's VRMs run too hot for Prime testing at that level of voltage so it's possible that the scores are lower than they should be because of less than optimal voltage. It should be able to handle 4.8 much easier. I can run Prime at 4.7 on the most strenuous settings.

update.... It does look like the 5 GHz results are bogged down by inadequate voltage and/or excessive VRM heat. My guess is that the CPU NB voltage was too low. I managed to get a much better CB multi score when I bumped that up when I did 4.6 testing. Here are the results at 4.8:

4.8 GHz

100: 1:09.5
150: 144:5

4.8 GHz “SIMD”

100: 36.8
150: 55

CB 15 multi: 767 (same score, two runs)

CPU-Z 1.78

single: 1414
multi: 9212

If someone with a high-grade board could test at 5 GHz that would be good.

 

[Iron Woode]

Blender 2.78a

stock I5 4690K 8GB DDR3 1333 win10 x64

1.42.64 @ 150 sample