specst
This commit is contained in:
@@ -3941,3 +3941,30 @@ _**Agent (model default, mode Agent)**_
|
||||
|
||||
---
|
||||
|
||||
|
||||
The **7800X3D should win** for this script.
|
||||
|
||||
**7800X3D:**
|
||||
- 8 cores, 16 threads → script uses 16 worker processes
|
||||
- Zen 4 (2023)
|
||||
- Strong single-threaded performance
|
||||
- Higher IPC
|
||||
|
||||
**Threadripper 1950X:**
|
||||
- 16 cores, 32 threads → script uses 32 worker processes
|
||||
- Zen 1 (2017)
|
||||
- Lower single-threaded performance
|
||||
- Lower IPC
|
||||
|
||||
**Why the 7800X3D wins:**
|
||||
1. Per-core performance: Each worker finishes faster, so 16 fast workers can outperform 32 slower ones.
|
||||
2. Architecture gap: Zen 4 vs Zen 1 is a large IPC improvement.
|
||||
3. PNG compression: PIL operations benefit from faster cores; the 7800X3D’s cores are roughly 2–3x faster per core.
|
||||
|
||||
**Expected performance:**
|
||||
- The 7800X3D should be roughly 30–50% faster, depending on I/O.
|
||||
|
||||
**Bottom line:** The 1950X’s extra threads don’t offset the 7800X3D’s much faster cores. The 7800X3D is the better choice for this workload.
|
||||
|
||||
---
|
||||
|
||||
|
||||
Reference in New Issue
Block a user