- Joined
- Mar 10, 2010
- Messages
- 11,878 (2.25/day)
- Location
- Manchester uk
System Name | RyzenGtEvo/ Asus strix scar II |
---|---|
Processor | Amd R5 5900X/ Intel 8750H |
Motherboard | Crosshair hero8 impact/Asus |
Cooling | 360EK extreme rad+ 360$EK slim all push, cpu ek suprim Gpu full cover all EK |
Memory | Corsair Vengeance Rgb pro 3600cas14 16Gb in four sticks./16Gb/16GB |
Video Card(s) | Powercolour RX7900XT Reference/Rtx 2060 |
Storage | Silicon power 2TB nvme/8Tb external/1Tb samsung Evo nvme 2Tb sata ssd/1Tb nvme |
Display(s) | Samsung UAE28"850R 4k freesync.dell shiter |
Case | Lianli 011 dynamic/strix scar2 |
Audio Device(s) | Xfi creative 7.1 on board ,Yamaha dts av setup, corsair void pro headset |
Power Supply | corsair 1200Hxi/Asus stock |
Mouse | Roccat Kova/ Logitech G wireless |
Keyboard | Roccat Aimo 120 |
VR HMD | Oculus rift |
Software | Win 10 Pro |
Benchmark Scores | 8726 vega 3dmark timespy/ laptop Timespy 6506 |
You realise all modern processor's are designed for Turbo, and dash to rest operation.Lets actually talk Geekbench for a sec. I know Geekbench3 was highly flawed, but why does everyone think that Geekbench4 is bad?
Here's Geekbench4's workload: https://www.geekbench.com/doc/geekbench4-cpu-workloads.pdf
Now I recognize that a lot of Geekbench4's benchmarks fit inside of L1 cache, but that's more of a testament to how big L1 caches have gotten. (128kB on the iPhone). Lets be frank: if 128kB L1 cache is what's needed for the modern consumer, then we should be blaming AMD / Intel for failing to grow their L1 to 128kB (AMD / Intel still have 32kB L1 data caches).
Lets really look at Geekbench4's benchmarks. Unlike Geekbench3, AES is downgraded to be just another test instead of its own category. (And mind you, AMD Zen2 and Intel Xeons have doubled their AES pipelines recently: AES remains an important workload). There's JPEG compression (emulating a camera), HTML5 parse, LUA scripting, SQLite database, and PDF rendering. Lots of good workloads here. Very similar to a wide variety of workloads of the modern, average consumer. Even an LLVM compile (3900 lines of code).
There's a bunch of "synthetics" too: 450kB LZMA compression, Djikestra, Canny (Computer-vision), a 300x300 Raytracer, etc. etc. A bunch of tiny synthetics.
--------------
Geekbench4 is what it is: a small test for testing L1 cache and Turbos of modern processors. Its probably closer to the average phone-user or even desktop-user's workflow than SPEC, LINPACK, or HCPG.
But yes, the iPhone crushes Geekbench. Because the iPhone has 128kB L1 cache. But is that a legitimate reason to call the test inaccurate? We can't just hate a test because we disagree with the results. You should instead attack the fundamental setup of the test, and tell us why its inaccurate.
Its pretty insane that the iPhone has a 128kB L1 cache per core. Yeah, that's its secret to crushing Geekbench4 and its pretty obvious. But Intel Skylake's L2 cache is only 256kB and AMD Zen2's L2 is 512kB. Having such a large L1 cache is a testament to the A12 design (larger caches are usually slower. Having such a large cache as L1 must have been difficult to make).
Any bench shorter than the Tau value isn't worth shit regardless IMHO, not really, you can gauge performance to a degree but it's not the whole picture , and that's geek bench for you , short bursts , a test designed for phones and light use cases.