I recently bought an Asus PX13 laptop to replace my Dell 9305 which died in circumstances we don’t need to get into. Let’s just say, I’m better at servicing desktops.
This is, I believe, a now EOL model with a HX370 APU and 4060 GPU and 24GB. With the combination of the APU and the GPU I wanted to get a feel for what settings I should be using when on different power sources. The supplied 200W adapter is large and I wouldn’t want to travel with it and would rather stay within the limits of a compact 65W or 100W USB charger.
I used the Cyberpunk 2077 benchmark for this. Other games/apps will perform differently, but I didn’t want to spend all week on this. Resolution of 1920×1200, medium preset.
I mostly only ran each test once. The power measurements are from a in-line USB monitor and I just eye-balled it. The margin of error on each result is quite high. A few FPS or Watts either way doesn’t count.
APU Only
| APU (W) | FPS | Approx USB draw (W) |
|---|---|---|
| 10 | 10 | 23 |
| 15 | 19 | 28 |
| 20 | 31 | 36 |
| 25 | 34 | 44 |
| 30 | 35 | 51 |
| 40 | 37 | 60 |
| 50 | 37 | 65 |
| 60 | 37 | 62 |
| 75 | 37 | 60 |
The plateauing from 40W wasn’t a surprise as I’ve seen a similar pattern with the Z1E in the Ally. That has big jumps in performance up to 20W, then a little improvement to 25W then completely flat up to the maximum 40W.
With Upscaling
Using FSR 2.1 Performance.
| APU (W) | FPS | Approx USB draw (W) | Notes |
|---|---|---|---|
| 20 | 54 | 36 | Low preset |
| 25 | 56 | 44 | |
| 30 | 58 | 51 | |
| 30 | 63 | 51 | Low preset |
For the first tests I did with FSR the power ratings were coming back in the 70W range even though the TDP limit was 25W. I was setting the TDP with G-helper, with CPU boost off and just setting the main TDP slider. The 2-min and 2-second control were on 75W. It seems that FSR ignores the CPU boost setting. When I lowered the 2-min/2-second levels to match the main TDP limit draw was as expected. Something to be aware of.
APU & GPU
GPU power settings are including the minimum 5W dynamic boost, i.e. 60W here is 55W + 5W boost.
| CPU (W) | GPU (W) | Clock (MHz) | FPS | Approx USB draw (W) | Notes |
|---|---|---|---|---|---|
| 15 | 60 | 400 | 19 | 45 | |
| 15 | 60 | 700 | 34 | 54 | |
| 15 | 60 | 800 | 36 | 54 | |
| 15 | 60 | 900 | 36 | 55 | |
| 15 | 60 | 1000 | 41 | 58 | |
| 15 | 60 | 1500 | 63 | 76 | |
| 15 | 60 | Full | 70 | 80-90 | |
| 15 | 65 | Full | 72 | 80-90 | |
| 15 | 70 | Full | 70 | 80-90 | |
| 15 | 75 | Full | 70 | 80-90 | |
| 15 | 95 | 1500 | 63 | 70-80 | 70W + 25W boost, 87C target |
| 20 | 65 | Full | 70 | 80-90 | |
| 25 | 60 | Full | 72 | 80-90 | |
| 25 | 95 | Full | 70 | 80-90 | 70W + 25W boost, 87C target |
I was surprised at the complete lack of difference between any of the power settings. To the point I’m not entirely convinced G-Helper was actually doing anything. But the clock speed settings were effective.
I’m less confident of the power levels at the higher numbers as I’m using a 100W supply and I don’t think the laptop will actually pull 100W from it. With my 65W Anker Nano charger it seems to top out at 60W. I think this one is topping out at 90W. The last 25/95W test was still measuring at 90W draw but it did seem louder.
The effect of clock speed was very interesting. At a 60W power budget it’s ~10% better than using the APU.
With Upscaling
Using the GPU over the APU has the benefit of allowing DLSS over the inferior FSR. These tests were all with DLSS Transformer model.
| CPU (W) | GPU (W) | Clock (MHz) | DLSS | FPS | Approx USB draw (W) |
|---|---|---|---|---|---|
| 15 | 60 | 600 | Performance | 55 | 59 |
| 15 | 60 | 600 | Balanced | 43 | 53 |
| 15 | 60 | 800 | Performance | 67 | 62 |
| 15 | 60 | 800 | Balanced | 62 | 63 |
| 15 | 60 | 1000 | Performance | 71 | 64 |
| 15 | 60 | 1000 | Balanced | 60 | 63 |
| 15 | 60 | 1500 | Performance | 80 | 70 |
| 15 | 60 | 1500 | Balanced | 74 | 67 |
| 15 | 60 | Full | Performance | 79 | 76 |
| 15 | 60 | Full | Balanced | 79 | 79 |
| 15 | 75 | Full | Performance | 79 | 76 |
| 15 | 75 | Full | Balanced | 79 | 81 |
| 25 | 75 | Full | Performance | 80 | 81 |
A few thoughts:
- DLSS users more power than unscaled at lower clock speeds, but less at higher ones
- Under DLSS the power never got into the mid-80s to 90, so I don’t think I was capping the supply
- There’s some bottleneck at 80fps which all of the higher settings are hitting
Cyberpunk Best Choices
| Power Budget | APU | GPU | Upscaler | FPS |
|---|---|---|---|---|
| 60W | 25W | – | FSR Performance | 56 |
| 60W | 15W | 55W, 600MHz | DLSS Performance | 55 |
| 100W | 15W | 55W, Full speed | DLSS Balanced | 80 |
Using a 100W charger gives a lot more flexibility, you can use 15W/55W as the headline settings with DLSS and you’re fine.
At 60w things are trickier. There’s a few better options just out of range at ~65W, but rules are rules. In practice using those would just mean 5W drains out of the battery which would be an issue after 12 hours.
But sticking to 60W, there’s a straight choice between FSR performance and DLSS performance, and of those I’d prefer DLSS on quality grounds. Plus, it has the advantage of using the GPUs own memory and not taking system RAM. But this could be very application specific, so I did another test.
Satisfactory Mini Test
Satisfactory doesn’t have a proper benchmark so I just hovered in front of some buildings for a bit. Run with full 2880×1800 resolution, medium preset.
| APU (W) | GPU (W) | Clock Speed (MHz) | Upscaler | FPS | USB Draw (W) |
|---|---|---|---|---|---|
| 25 | FSR Balanced | 30 | 44 | ||
| 15 | 55 | 600 | DLSS Performance | 29 | 54 |
| 15 | 55 | 800 | None | 27 | 59 |
| 15 | 55 | 800 | DLSS Balanced | 29 | 58 |
| 15 | 55 | 800 | DLSS Performance | 35 | 57 |
| 15 | 55 | 1000 | None | 33 | 63 |
| 15 | 55 | Full | DLSS Performance | 70 | 83 |
There’s some movement here above and below the 60W line, but 600Mhz DLSS performance is safe on both games and as good as, or better than FSR.
Final thoughts
This was actually quite enlightening. I have three presets on G-helper, which I can now set as:
- Silent – 10W APU only, can run fanless and good enough for general use
- Balanced – 15W APU, 55W GPU @ 600Mhz, good enough for 65W sources
- Turbo – 15W APU, 55W GPU @ full speed, good enough for 100W sources
There’s no performance beyond 100W that’s worth worrying about for the sake of a bigger charger and all the extra heat and noise.
Lastly, it’s worth remembering that the APU-only profile for 60W is only 44W draw, it’s just that going up to 60W on the APU doesn’t gain anything. This comparison is just about running off a mains adapter, so that efficiency doesn’t matter, but if you were to actually run on battery then the APU is far superior. Drop it down to 20W (36W total) for most of the performance and adjust game settings to suit. With a healthy battery that’s 2 hours of playtime.