I’ve recently been ramming the Framework with benchmarks, and I’ve come across – a perhaps not startling – finding. Whenever the latest firmware is available, it usually gives a great benefit to performance for the hardware that its on.
Yet, when I installed the Framework driver package, everything worked. I ran a benchmark, with the results here and I noticed stellar cpu performance (100th percentile). I might have gotten a golden chip for CPU and garbage for integrated graphics, but I don’t know.
What did concern me, was the fact that anything beyond DX9 was completely unwatchable, with much of it getting below 1FPS.
I upgraded the graphics drivers, and WOW! from 0fps to 27 is quite an achievement (although DX9 might have dropped) Results here
Can anyone run the benchmarks so I have a reference baseline? I’m currently interested to see how much the drivers improve performance on average.
Do not use that website for benchmarking.
Other than echoing that Userbenchmark is fairly useless, a major reason for under-performing integrated graphics could be running RAM in single-channel or flex-mode (2 mismatched capacities).
Could anyone explain as to why it is considered useless? I’d like to know the reason, and what other alternatives are there? Haven’t seen one that tests graphics performance yet.
I’m running 2x8 DDR4 on my framework, and CPU-Z also confirms this with 64 bit.
Run a Passmark benchmark and post up your numbers.
The heaven benchmark is pretty quick to do too.
To be honest lots of GPU benches on the Framework are not going to show you that much. Most will show a certain score ±10% for those with the bigger Iris and faster RAM and then its up to the rest to get near/work out they might have a config optimisation to do or not be bothered.
On a side note for sh*ts and giggles I ran the Userbench again and got these - (same CPU as you)
CPU- 88% Normal - 193 Heavy - 576 Server - 642
Iris - 23.4% 3D DX9 - 27.5fps 3D DX10 - 26.9fps
Ram - 96.6% Multi - 35.2GB/s Single - 25.9GB/s Latency 75.4ns
I think @feesh was speeking mismatch in term of speeds… are they both 3200?
I think several video of HardwareUnboxed / LinusTechTips showed that this is not a good source of performance data. Notably when ryzen was out, and they were still praising Intel…
They are both the same speed, 3200 MHz, straight from FW
From the day I built it two weeks ago.
Have you tried upgrading the Iris integrated drivers? I don’t think I can downgrade, at least easily. (And just 2D and 3D passmark?)
I have the latest as of two weeks ago. I have the Irix set to max performance in all the Intel control panels when on mains power.
Thing is this will all be for nothing as it’s likely we will all be getting a performance crippling BIOS soon fixing the latest Spectre MK2 (or 3) vulnerabilities.
Bang goes 10-20% CPU performance.
This is why we can’t have nice things.