Asus ROG Ally

Turbo to 45-50 when plugged in is correct. It ramps down after a set period of time normally due to temps (when gaming). The battery usage is normal as whilst the apu is set to 25w the whole device is pulling 40sh as shown by the overlay.

Would love to see D2 settings to get 50-60 at 1080p. Neptune drops into the 40’s for when you get to the main centre hub at 720p with RSR to 1080p.
 
FloppyPoppy said:
i’ve done a lot of de bloating of windows 11 over the last short while and have gotten my idle usage down from 4.2gb to 3.4gb

Doing this has helped me gain noticeable performance and reduced stutter for sure. Can’t wait for atlas os to release for W11
Click to expand...

Could you do a quick list of the major tweaks you have made? Not something I have ever really bothered messing around with but would be interesting to tinker.

@sandys - Finally got around to installing and testing HZD using the built in benchmark:

1080p Ultimate
NHVTBxxh.jpg

1080p Ultimate w/ High Textures
bqwsrYth.jpg

1080p "Original" - which IIRC matches a PS4 Pro
nPPB2S3h.jpg


I also did a run which returned 34 Avg / 28 Low using the original custom settings I had running from my laptop but since changing for benchmarking purposes I can't remember what they were. I think a mixture of High and Ultimate, probably with Motion Blur turned off.

On the small screen the Original settings with 1080p and a bit of FSR should net 50-60FPS in most areas. Not bad little Ally, not bad.

I will run a quick test on my Flow (12900H / 3050Ti) to see what that can achieve. I am expecting it to beat the Ally but not sure by how much. :)
 
Yep so I ran it on the Flow and to no ones surprise it beat the Ally. 55 vs 30FPS at the same settings. Main difference being the Flow using well over double the power to do that so in terms of efficiency the Ally has it. :)

Also as a side note my Flow nearly ended up out of the window after playing a game of lets not connect to the wifi I was just connected to....
 
oweneades said:
@sandys - Finally got around to installing and testing HZD using the built in benchmark:

I also did a run which returned 34 Avg / 28 Low using the original custom settings I had running from my laptop but since changing for benchmarking purposes I can't remember what they were. I think a mixture of High and Ultimate, probably with Motion Blur turned off.

On the small screen the Original settings with 1080p and a bit of FSR should net 50-60FPS in most areas. Not bad little Ally, not bad.

I will run a quick test on my Flow (12900H / 3050Ti) to see what that can achieve. I am expecting it to beat the Ally but not sure by how much. :)
Click to expand...

Those numbers look great for 16Gb, I wouldn't have expected that, they are not too dissimilar to mine though looks like you may have capped the CPU or something, or perhaps that might be where the z1 is doing some cleverer stuff perhaps to save and share power, my CPU runs rampant taking all the power despite turning Turbo off, so the GPU never clocks too high, your GPU driver looks a lot newer than mine too, mines is still from April, I suspect it needs an update as my lows happen at the start, there is a big dip as things start up, happens in everything but then they plays flawless, weird, whereas you are straight in at ~30.


hzd-fps.png
 
sandys said:
Those numbers look great for 16Gb, I wouldn't have expected that, they are not too dissimilar to mine though looks like you may have capped the CPU or something, or perhaps that might be where the z1 is doing some cleverer stuff perhaps to save and share power, my CPU runs rampant taking all the power despite turning Turbo off, so the GPU never clocks too high, your GPU driver looks a lot newer than mine too, mines is still from April, I suspect it needs an update as my lows happen at the start, there is a big dip as things start up, happens in everything but then they plays flawless, weird, whereas you are straight in at ~30.


hzd-fps.png
Click to expand...

The runs were just stock turbo for the Ally albeit plugged in. All runs showed 30w so none of the initial ramp up (maybe where I got the 34FPs result from as that was with the full bore turbo). The Ally does seem to favour GPU clocks over CPU as a default. Nothing changed on my end to limit turbo or anything. I wonder if that is where the main differences of the Z1 come in over the 7840u. We know it’s only a minor “custom” design unlike the SD, so I guess a lot of it is firmware.

Just running the same test on the SD at 800p and it is turning in around the same 30FPS or so. That aligns with the general Ally delta up lift along with what you are seeing as well.
 
sandys said:
This is just a memory limitation as shown in my post above as I can do 1080p Ultimate in HZD with basically the same chip and 32Gb of RAM, the more you try and do with an APU the more of the sys RAM it needs, as it has no dedicated VRAM, recommended specs call for 8Gb of VRAM, that's a big chunk out of the systems 16Gb.
Click to expand...
Interesting thanks, maybe I will try with the 8gb profile to see if HZD is better and report back.
 
Getting about 30-35fps on hzd on ultimate with 8gb assigned to gfx. Its definitely playable, but not as nice I think as 720 with the extra frames. I think it runs better on 1080 with the extra ram assigned as it wasn’t playable before but occasional glitching with sound and stutters when system is running close to ram limit.

Destiny wise I neglected to mention even though textures are highest, shadows high, I have AA on FXAA and detail distance to medium, motion blur on.

Its a pretty solid 45fps on 15w, and 30w between 50-60.
 
sandys said:
I just best to appreciate its an APU, whack on FSR and enjoy it at 60fps, looks better this way than setting 720p.
Click to expand...
I can’t really see the difference with FSR, at least toggling on and off, its always default to on, when / how does it apply it?

I agree, my main rig is running a 3090ti and twice the system ram. Seeing what this thing can do that is less than a third the size of my gpu, infinitly lighter, and is an entire gaming system in one is the impressive thing! Things have come a long way in a few years to have last gen console power in the palm of your hand.
 
Last edited:
sorry for the spam, I need to clearly read up on rsr / fsr. So I thought enabling in the overlay worked, but it seems I had to enable manually in hzd own settings.

Turning off RSR in the overlay and enabling FSR @ 1080p with Quality setting is getting me extra frames, maybe around 45 now with most settings on ultimate. Vram usage dropped slightly to within 4gb. Defintily smoother and playable at 1080 now!
 
FSR is like DLSS, needs game support but works really well, particularly on these low end chips with small screens, gives you a good quality upscale quality first port of call with these APUs and obviously dropping settings that you might not notice.
 
I noted in a previous post the differences between FSR and RSR. Main difference is with FSR you set native res at 1080p and then set the FSR level in game. It will then render the game internally at a lower res and upscale. RSR uses FSR but is global. With this you set the resolution in game to what you want to scale from whilst leaving the Ally at 1080. It isn’t as good as FSR2 but is great when you need that extra grunt. Works essentially the same as the FSR scaling option baked into the SD.

Sorry typing this on my phone so formatting is a bit whack
 
Cool thanks guys, so by default, I’ve always left RSR on. And kind of been using it right, the only game I’ve ran so far is hzd at 720, and I left RSR on thinking it should be upscaling. The literal game changer was reading previous posts by you guys on FSR - didn’t even realise it was an option in game. This is my first AMD, so it was a bit confusing. When I saw FSR and enabled it the vram usage dropped, so I could use ultimate at 1080p with 4gb. And its given at least 45 frames! Higher textures and 50% nore performance is an absolute winner, well happy I saw this thread :)

For mobile gaming at least on battery, I always set frame cap at 45, as I feel thats the sweet spot for my eyes, its a nice compromise between performance and smoothness, and in a lot of cases 45 is reached in performance mode @15w.

I haven’t had chance to read the entire thread but I also saw elsewhere that the stock fan profile maxes out at about 50%? Not sure if this is true. But if so sounds like there may be some more performance we could get increasing power and clock speeds for longer.
 
steveoz said:
Cool thanks guys, so by default, I’ve always left RSR on. And kind of been using it right, the only game I’ve ran so far is hzd at 720, and I left RSR on thinking it should be upscaling. The literal game changer was reading previous posts by you guys on FSR - didn’t even realise it was an option in game. This is my first AMD, so it was a bit confusing. When I saw FSR and enabled it the vram usage dropped, so I could use ultimate at 1080p with 4gb. And its given at least 45 frames! Higher textures and 50% nore performance is an absolute winner, well happy I saw this thread :)

For mobile gaming at least on battery, I always set frame cap at 45, as I feel thats the sweet spot for my eyes, its a nice compromise between performance and smoothness, and in a lot of cases 45 is reached in performance mode @15w.

I haven’t had chance to read the entire thread but I also saw elsewhere that the stock fan profile maxes out at about 50%? Not sure if this is true. But if so sounds like there may be some more performance we could get increasing power and clock speeds for longer.
Click to expand...

You can create custom fan profiles (for both fans) within Armoury crate, but by default it depends what bios are you on which dictates the fan profile. Bios 3.22 brought a much more aggressive fan profile for the turbo profile which helps keep temps in check.

Performance wise we are mostly limited by power rather than temps. The Ally only really gets too hot when running plugged in during the initial turbo boost window. Funnily enough, with the XG mobile taking the load of the GPU portion of the APU, it allows the CPU to turbo hard.
 
I tried RSR in GTA V but changing res in game just changed global res?

Also it's impressive so far but it's a shame I can't get it to run cities skylines with mods and assets. :(

Also I've observed the API wattage can be pulling 40w on its own??

Also, VRAM to auto seems the best way it'll just adjust on the fly as needed?
 
Last edited:
willhub said:
I tried RSR in GTA V but changing res in game just changed global res?

Also it's impressive so far but it's a shame I can't get it to run cities skylines with mods and assets. :(

Also I've observed the API wattage can be pulling 40w on its own??

Also, VRAM to auto seems the best way it'll just adjust on the fly as needed?
Click to expand...

For RSR you may have to play around with Fullscreen and Windowed Fullscreen depending on the game. The key thing is ensuring the resolution per the Ally overlap is set to 1080p. Setting a lower resolution than this in game forces RSR to upscale to the 1080 res of the Ally screen.

APU can pull up to 45w (or 50w) briefly when plugged in to its external power brick (or 65w compatible charger). The boost window doesn't last very long before settling in at around 30w sustained. The window duration is generally constrained by temperature but even if it wasn't it still drops. Similar to Turbo on Intel CPU's (PL1/PL2 settings).

VRAM on Auto is what I have set as so far it performs better / more consistently in the games I play. You may find some games prefer the fixed amount so its a little trial and error. As @sandys has noted above, if the Ally had 32GB of RAM you could just set it to 8GB and leave it there.
 
Yup once you fix an amount for VRAM it is taken away from the system entirely so generally better to leave it at Auto when you are resource limited, fixing an amount of VRAM does tend to makes thing smother as assets can be cached but when you are fighting with the system and game for resources better to let it manage it as you don't have the luxury RAM abundance on the Ally.

This shows the difference on a 5700g whilst the overall averages are much the same, its the lows that show the hit, he only has 16Gb though so far from optimal if a big game, so you can see the odd title being worse as not enough left of game. Its always going to be system dependant though, you can probably do a build on the Ally with tiny11 etc that'd free up a lot of waste. Will no doubt try that when I get a new SSD.

I should do a re-test on my NAS it has an APU and 64Gb, it was years ago when I tested previously and AMD has improved drivers quite a bit over the years, of course its last gen 5xxx though. Would take to long on the Zoe its drive is too small.

 
Last edited:
Need to do more thorough testing by D2 at 1080p (80 FOV) and a mixture of Medium / low settings returns around 45FPS average running around the main Neptune patrol area. Goes into the 30's once you have a lot of particle effects and lots of enemies so likely memory bandwidth limited in these scenarios. Plugged in it would likely go into the 50's thanks to the higher TDP unlocking more CPU and GPU clock speed but not as representative of on the go performance. I can certainly see lesser demanding areas (in strikes for example due to their closed instance thus less CPU overhead) closer to 60 being achieved.

Kind of interesting tbh as the performance isn't that different to what I was seeing at 720p Med / Low with RSR up to 1080p. I can only gues that I am actually CPU limited at 720p, likely due to the netcode instancing overhead causing peak frequencies to drop. The D2 engine also runs noticeably better from my own experience on Intel / Nvidia hardware - perfect example being my Legion vs the Flow. The Legion has a 5800H ~4.1Ghz boost in game) paired with a very strong 3080M GPU. The Flow on the other hand pairs its slightly weaker 3080M (in terms of sustained boost clocks) with a 12900H which happily hums along at 4.6Ghz in D2. The Flow is outright faster, and by a good 10-15 FPS in the same gameplay scenarios.

One thing I would love to see on these little handheld PC's would be a dedicated VRAM pool or, if that isn't easily possible, double memory bus akin to what we see on the console APU's. Maybe not practical but I would love to see a 7840U class APU run unconstrained by its memory bandwidth. I reckon we could see Series S performance quite easily (less CPU but higher IPC and higher clocks combined with much faster CPU cores).
 
You must log in or register to reply here.
Back
Top Bottom