| | | | | | |

Ryzen 9 3900 vs Engineering Sample 906ED 2.1GHz


Description
The 3900 is based on Zen 2 architecture while the 906ED 2.1GHz is based on Comet Lake.

Using the multithread performance as a reference, the 3900 gets a score of 687.5 k points while the 906ED 2.1GHz gets 657.2 k points.

Summarizing, the 3900 is 1 times faster than the 906ED 2.1GHz. To get a proper comparison between both models, take a look to the data shown below.

Specs
CPUID
870f10
906ed
Core
Matisse
Comet Lake-H
Architecture
Base frecuency
3.1 GHz
3.6 GHz
Boost frecuency
4.3 GHz
5 GHz
Socket
AM4
LGA 1151
Cores/Threads
12/24
8/16
TDP
65 W
- W
Cache L1 (d+i)
12x32+12x32 kB
8x32+8x32 kB
Cache L2
12x512 kB
8x256 kB
Cache L3
4x16384 kB
16384 kB
Date
September 2019
March 2020
Mean monothread perf.
74.97k points
86.65k points
Mean multithread perf.
687.5k points
657.19k points

Non-optimized benchmark
The benchmark in Mode 0 (FPU) measures cpu performance with non-optimized software. It uses the basic µinstructions from the i386 architecture with the i387 floating point unit. This mode is compatible with all CPUs so it's practical to compare very different CPUs
Monothread
3900
906ED 2.1GHz
Test#1 (Integers)
4.39k
5.2k (x1.18)
Test#2 (FP)
16.99k
21.19k (x1.25)
Test#3 (Generic, ZIP)
7.74k
6.85k (x0.88)
Test#1 (Memory)
24.57k
15.48k (x0.63)
TOTAL
53.69k
48.71k (x0.91)

Multithread

3900

906ED 2.1GHz
Test#1 (Integers)
48.45k
42.31k (x0.87)
Test#2 (FP)
230.81k
198.07k (x0.86)
Test#3 (Generic, ZIP)
122.82k
68.37k (x0.56)
Test#1 (Memory)
54.73k
7.38k (x0.13)
TOTAL
456.8k
316.12k (x0.69)

SSE3 optimized benchmark
The benchmark in mode I (SSE) is optimized for the use of SIMD instructions with 128 bits register and the SSE set up to version 3. Nearly every modern CPU has support for this mode.
Monothread
3900
906ED 2.1GHz
Test#1 (Integers)
16.59k
18.76k (x1.13)
Test#2 (FP)
24.55k
27.15k (x1.11)
Test#3 (Generic, ZIP)
8.55k
7.15k (x0.84)
Test#1 (Memory)
26.28k
15.31k (x0.58)
TOTAL
75.97k
68.37k (x0.9)

Multithread

3900

906ED 2.1GHz
Test#1 (Integers)
181.33k
156.95k (x0.87)
Test#2 (FP)
261.01k
250.7k (x0.96)
Test#3 (Generic, ZIP)
126.62k
70.02k (x0.55)
Test#1 (Memory)
66.64k
7.43k (x0.11)
TOTAL
635.6k
485.11k (x0.76)

AVX optimized benchmark
The benchmark in mode II (AVX) is optimized to used 256 bits registers beside the first version of the Advanced Vector Extensions (AVX). The first AVX compatible CPU was released in 2011.
Monothread
3900
906ED 2.1GHz
Test#1 (Integers)
16.16k
18.88k (x1.17)
Test#2 (FP)
23.75k
28.5k (x1.2)
Test#3 (Generic, ZIP)
9.32k
6.91k (x0.74)
Test#1 (Memory)
23.17k
14.86k (x0.64)
TOTAL
72.4k
69.14k (x0.95)

Multithread

3900

906ED 2.1GHz
Test#1 (Integers)
174.73k
157.47k (x0.9)
Test#2 (FP)
275.63k
265.75k (x0.96)
Test#3 (Generic, ZIP)
125.65k
68.86k (x0.55)
Test#1 (Memory)
44.17k
7.38k (x0.17)
TOTAL
620.18k
499.47k (x0.81)

AVX2 optimized benchmark
The benchmark in mode III (AVX2), like AVX1, is optimized to used 256 bits registers beside the second version of the Advanced Vector Extensions (AVX). The first AVX2 compatible CPU was released in 2013.
Monothread
3900
906ED 2.1GHz
Test#1 (Integers)
16.85k
33.92k (x2.01)
Test#2 (FP)
26.03k
30k (x1.15)
Test#3 (Generic, ZIP)
9.54k
7.01k (x0.73)
Test#1 (Memory)
22.55k
15.73k (x0.7)
TOTAL
74.97k
86.65k (x1.16)

Multithread

3900

906ED 2.1GHz
Test#1 (Integers)
229.41k
302.78k (x1.32)
Test#2 (FP)
292.81k
278.98k (x0.95)
Test#3 (Generic, ZIP)
128.48k
67.89k (x0.53)
Test#1 (Memory)
36.8k
7.53k (x0.2)
TOTAL
687.5k
657.19k (x0.96)

Performance/W
3900
906ED 2.1GHz
Test#1 (Integers)
3529 points/W