So, few weeks ago the Dell G3 3579 review was published for the variant with an I5-8300H and GTX 1050 Ti. I was pretty harsh towards the G3, finding it to be disappointing in multiple areas, including the keyboard, screen, thermals, connection ports. Thermals and CPU clocks behavior were indeed not good, but were also strange – the I5-8300H downclocked to around 1.2-1.4GHZ under Prime95 + Furmark workload and to around 2.5GHZ under Metro Last Light workload, without any obvious reason, since TDP and temperatures were in the ok range.
Few days ago Dell released a new bios, version 1.1.0 (vs 1.0.2 used for the review) and I ran again few tests:
- Metro Last Light, highest settings@1080p
- Aida64 stability test (CPU, FPU, cache) + Unigine Heaven extreme benchmark
- Prime 95 + Furmark@768p.
These are the stress test results:
1. The first graph shows the difference in CPU core clocks when running the Metro Last Light built-in benchmark, set on highest settings available, running at 1920×1080 resolution. This benchmark is quite tasking for a DX11 game. As you can see, with the 1.0.2 bios, CPU clocks go down to around 2.5-2.7GHZ, while with the 1.1.0 bios, clocks remain on 3.8-3.9GHZ, that’s a huge improvement. It also costs in CPU temperatures. I’m pasting here the temperatures table from the full review. As you can see, with bios 1.1.0, CPU stable temps are higher.
|CPU: stable temps [C]||CPU: max temps [C]||CPU: stable clocks||CPU: average utilization||CPU: max utilization||GPU : stable core temps [C[||GPU: utilization|
|Prime95 (bios 1.0.2)||82||85||2.9||100|
|Prime95 CPU@-0.12V (bios 1.0.2)||69||99||3.2||100||70|
|Prime95 + Furmark (bios 1.0.2)||81||86||1.4||100|
|Prime95 + Furmark CPU@-0.12V (bios 1.0.2)||70||99||1.7||100||73|
|Metro Last Light, Highest@1080p (bios 1.0.2)||72||93||2.7||60||72||98|
|Metro Last Light, (bios 1.1.0)||90||94||3.88||31||50||75||99|
2. Another change is that in Prime95 + Furmark stress test, it takes longer time to reach throttle clocks. It took around 4:11 minutes to reach 1.2GHZ clocks, while with bios 1.1.0 it tool around 3:20 minutes, so there is no improvement really. I don’t why. Could be some kind of electrical limits (real or calculated). I don’t know, but although there is no TDP limitations.
3. Although not conducted in the original review, I ran the Aida64 stability test (CPU, FPU, cache) with Unigine Heaven benchmark in the background. The CPU downclocks to around 1.8-1.9GHZ for sometime, while in other periods it maintains 3.1-3.3GHZ. No conclusion here, seems like the strange G3 behavior.
It seems that the 1.1.0 new Dell bios does improve CPU clocks behavior under some workloads and most importantly, under typical gaming workload. It might not be the case in more CPU intensive games, however. Moreover, the higher clocks result in higher CPU temperatures, which were already a problem before.
Bottom line, a small improvement, but Dell has yet some work to do.