So the testing begins.
I've dialed in stable overclocks for both the 1950x and 7900x. The 1950x is set to 4 GHz on all cores and the 7900x is set to 4.9 on all cores with 4 boosting to 5 GHz. The 1950x is a bit more shaky at this speed than the 7900x.
Temps on the 1950x are better than those of the 7900x despite the 7900x being delidded and the 1950x having 6 more cores.
Before starting, I wanted to make sure that both chips were outputting the expected performance at their given overclocks. With that said, in Cinebench R15, the overclocked 1950x easily punches the overclocked 7900x in the face when it comes to multi-threading, but the 7900x completely decimates the 1950x in single-threaded performance.
1950x - 3416 multi / 164 single
7900x - 2743 multi / 224 single
I've started comparing Handbrake and I can say that the multi-threaded Cinebench results of the 1950x don't necessarily translate into better multi-threaded performance in this application.
-
Handbrake with H.264, the overclocked 1950x is somewhat close to the the overclocked 7900x. In H.265, which is even better with AVX, the 7900x just shreds the 1950x despite a 6 core deficit. Cores weren't being completely utilized by 1950x in either H.264 or H.265.
Handbrake (H.264 Medium - Medium Quality) 1 min remaining
1950x
7900x
Handbrake (H.264 Ultrafast - Low Quality) 1 min remaining
1950x
7900x
Handbrake (H.265 Veryfast - Higher Quality) seconds remaining
1950x
7900x
-
Nice, I’m hoping to get my hands on an e5 16xx 10 or 12 core variant, one of the beastly unlocked OEM ones. I’d love to see how it stacks up vs your 5960x, 1950x and 7900x
- done12many2 and PCGuy_5960
- 1
- 1
-
@Damascus Nice! I didn't know that unlocked 10/12 core Xeons existed!
-
They are super rare, the E5 1681 v3 is what I’m aiming for. 10 cores, 20 threads, a baseclock of 2.9ghz and it’s unlocked to OC. I’m hoping to get my hands on one when all of the v3 machines are replaced by the oncoming x299, TR4 and equivalent Naples/Xeon chips hit the market.
-
I'm using G.Skill Trident Z 4 x 8 3200 c14 kits on both setups. Exact same kits.
As I'm sure you are aware, Ryzen does not exactly overclock memory as well as Skylake-X. I could get 3600 and 3733 MHz to load on the TR 1950x, but they weren't stable at all so I ran with the DOCP of 3200 c14, which is definitely not slow by Ryzen standards.
I only run my CPUs / Memory overclocked so stock performance means very little to me. If my Skylake-X chip can run the exact same kit at 4000 MHz, then that's what I'm testing with. If the Ryzen TR chip could do the same, I would be running it.
Hope that makes sense.
Sounds like you have a lot of fun ahead bud! Let me know if I can help.
-
Thanks, I’m curious to see how a poor mans i9 performs (1681 10c/20t, 1685 12c/24t or 1691 14c/28t) in comparison to the 7900x, 7920x or 7940x.
-
Looks like Blender is where the 1950x really shines!
Blender v2.79 running the BMW Benchmark from here.
1950x - 2:20
7900x - 2:46
1950x
7900x
-
Firestrike and Time Spy Extreme CPU tests.
1950x Firestrike CPU test: 24,922
7900x Firestrick CPU test: 26,705
1950x Time Spy Extreme (AVX2): 8,025
7900x Time Spy Extreme (AVX2): 9,038
1950x Time Spy Extreme (AVX-512): 7,992 **When AVX-512 is selected the 3DMark software defaults to AVX2 since Zen doesn't support AVX-512.
7900x Time Spy Extreme (AVX-512): 11,726 **Ran at 4.8 GHz as apposed to previously mentioned 4.9 GHz do to much heavier load from AVX-512.
1950x FS
7900x FS
1950x TSE AVX2
7900x TSE AVX2
1950x TSE AVX-512
7900x TSE AVX-512
-
It would be interesting to test gaming in Wolfenstein 2: The new colossus. I'd try downloading the demo and check the FPS, because that game utilizes all of the available threads. I think they should perform equally.