Is there an Undervolting tool for the Ryzen Mobile CPUs?

I've looked around a bit and haven't found jack.

 

Ryzen master iirc

 

IIRC it doesn't support all the options for mobile CPUs

 

Not to my knowledge. I guess you could lower the TDP using the guide below but I imagine lowering the performance is not really what you're looking to do. I bet there is a lot of room for undervolting on these, too. I remember I had a Llano APU like six years ago that was ridiculously poorly optimized. I was able to increase the clock speed by nearly a full 1GHz while lowering the voltages a good bit. I am sure AMD's quality control for setting voltages has improved a bit since then but I bet there could still be some great gains if it was possible.

 

I’m not looking for a way to limit TDP, I kind of want full control like I would get with XTU or TS. I want to undervolt (god knows these things could use it)

 

Sadly, I don't think there is a way to do so yet.

 

Damn, that sucks. Ryzen CPUs would see giant leaps in performance if you could tune them like intel CPUs, the potential boost is game-changing for sure. Maybe they are working on automatic undervolting and whatnot?

 

I had a ryzen 1700 in my gl702zc laptop and I can tell you that cpu was already well optimized.
Yes I was able to drop the stock voltage from roughly 1.17v to about 0.98v on stock... Or I could have overlooked the cpu to 3.6ghz on all cores on stock voltage (or dropping it just a notch to about 1.16v).

And the ryzen mobile parts are ultra low power.
It is possible it could do with voltage modulation, but not sure you'd be able to get massive difference... The uarch was completely revamped after all.

Besides, the biggest problem with 2500u and 2700u are manufacturer limited tdp and inadequate cooling implementation (since they just use pre existing chassis or cooling for Intel and then of course use lousy thermal paste application).

I think AMD would be better off just modifying ryzen master to allow people to undervolt ryzen mobile.

 

I know some of the thinkpads are a bit better under loads than others. I know that undervolting will help, which is always wanted on a mobile system. Even if you could get a tad better performance or just a little more battery life, it would be worth it

 

Undervolting anything on mobile is always recommended (be it AMD or Intel).
It usually (and unfortunately) falls to the end-user (ourselves) to correct the bad design implementations by OEMs.
As for Thinkpads better under load than others... Actually, I think that Acer Nitro 5 with 2500u and 2700u would be better... probably because they both come with RX 560x dGPU... hence the beefed up cooling and no TDP limit on the 2500/2700u APU's (at least the last time I checked).

Thinkpads have a severely TDP limited APU (from 25W down to 15W) - so their CPU clocks tends to drop with prolonged use.

Blame Lenovo (but if you ask me, AMD should not have allowed these APU's to have a configurable TDP - might not have made any difference though, because in the past, OEM's would just shove previous APU's into laptops pre-existing chassis and have them thermally throttle like there's no tomorrow - so, again, they paid 0 attention to cooling).

There are ways to go around the TDP limit of course (as it was posted in that video above), but, that will obviously result in these parts temps going up to 80 deg C... or over that - now, call me crazy, but such parts shouldn't be allowed to reach those temperatures - then again, OEM's thermal application is lackluster to say the least).

 

I think the ThinkPads have pretty beefy coolers and such. The A485 can be modded to take the T480's dual heatpipe dGPU heatsink and TPFC can let you crank up those fans to keep it nice and chilly

 

Yeah, but that's if you go do heavy modding yourself to keep the temps under control.
The ThinkPads actually don't have that good of a cooling if the following picture is any indication for E585:
https://www.notebookcheck.net/fileadmin/_processed_/9/6/csm_internals_31f82501d8.jpg

And here's the inside of Thinkpad A485 innards:
https://www.notebookcheck.net/fileadmin/_processed_/6/0/csm_DSC_0648_watermarked_103b7b124d.jpg

I don't think I've seen any T480 models with AMD APU's.
But quite honestly, the cooling seems atrociously poor.
E585 seems to have a dual heat pipe system, but the T480 is really bad.

 

The dual heatpipe with single fan should be enough for those CPUs. The T480 has a dual heatpipe version you can throw into the A485

 

You would think so... but as I said, Lenovo limits the Ryzen mobile TDP to keep the temperatures under control.
As such, if the TDP limit is removed and the APU is allowed to reach its full TDP value, they will climb up to over 80 degrees C (which suggests a lousy cooling implementation)... that's with the dual heatpipe version.
Another user has E858 which has the dual heat pipe installed, and when he increased the TDP limit to 30W, temps shot up to over 80 deg. C.

Mind you, having over 80 deg C under load is not exactly bad for a laptop (although, many systems simply cannot cope even with this).... but if we can get mid and high-end laptops with temperatures at 75 degrees C or less under maximum load... then that means that the cooling design in these Ryzen mobile units is simply speaking, atrocious at best when they allow the temperatures for such low power units to reach that high up.

 

To be perfectly honest, I think any temperature would be just fine. The CPU will never reach a temperature that is critical on its own due to some form of throttling. I would not consider a laptop that gets 85 degrees C during max loads a "bad implementation" of cooling, AS LONG AS the clocks speeds and performance is high. Hell, I wouldn't bat an eye if it hit 99 degrees C during double benchmarks and performance stays strong.

As it is, you can throttle a CPU if you want it cool, but pushing a cooling system to it's limits shows a lot about how well designed it was.

IMO a bad implementation has less to do with how hot/cool it runs, but how high performance is during max loads.

As an example, a laptop with i9 hits high 90s during benchmarks, cpu speeds drop BELOW base clocks, would be considered bad. But if you have another that is around the same size, even if it still hits the same high 90s, as long as performance is better I would consider the cooling to be "better"

 

That's the thing.
TDP limited APU's tend to drop their performance with consecutive Cinebench runs in Lenovo units.
You can achieve a good score on the first run, but run it 5 times in a row and the results will drop like a stone.

 

That is true, but this SHOULD be true for most machines. If this doesn't happen, that means either the cooling is overkill, or the CPU is too weak to take advantage of cooling

 

I think my point is, that SHOULD NOT be happening in the first place.
If a hw component inside a laptop is advertised to run at specific speeds, it should be able to reach and maintain those clocks for as long as you need them (which means OEM's need to stop messing about and design proper cooling that would fulfill this task)... anything else is just false advertising and/or subject to legal issues (because they would be in violation of consumer rights laws - which leaves them open to lawsuits, lost business and reputation damage).

 

I feel like it SHOULD be happening. The whole principle behind their throttling is to have very high clock speeds for short bursts, while maintaining a lower clock speed for sustained loads. Getting an i7 hexa core CPU to run at full turbo speeds on all cores in a laptop is a pretty challenging task, but the majority could live with lower sustained clocks if it meant they could get a thinner and lighter package

 

No, it shouldn't be happening.
The 2700U throttles when you subject it to stresses over long period of time which lowers its overall performance compared to where it should be (ergo, doesn't behave as it should).
For instance, we know that high single core clocks won't matter in multi-threaded software because all cores will be running at their designated boost clocks when stressed (which is usually lower than when using just 1 core)... the point is to have cooling which allows hw to boost and maintain their designated speeds for as long as you need them.
So if you want to do rendering for instance on 2500u/2700u using Blender or 3dsMax, the APU should be capable of reaching and maintaining its boost clocks for as long as the render lasts... not for as long as the OEM dictates due to them implementing garbage cooling.

Same methodology applies to games... performance usually drops during games because of the APU being TDP limited (as both the CPU and iGP are fighting to get enough resources - so, in games, CPU clocks usually go down substantially because the TDP limit is preventing both it and the iGP getting enough juice at the same time).

This can't be happening... its a faulty design plain and simple.

Imagine using a the laptop with 2700u for a business presentation, and at first everything is running smoothly... but then, a few mins into the presentation, it starts responding less and less.
That's a design flaw from the manufacturer who created the laptop...

Same thing with my GL702ZC.
It broke down twice due to an identical issue: blown motherboard circuitry for the fans which eventually leads to motherboard burning itself out (probably linked to poor cooling on the VRM's).
Its a £1600 piece of hardware... that kind of an issue shouldn't be happening. Asus skimped on the cooling design in the early units and I was the unfortunate recipient of that problem which is still ongoing thanks to ridiculous bureaucracy.

Don't justify bad design choices on behalf of manufacturers who don't care about consumers.
Demand that they bring you higher quality control.

Desktop CPU's and GPU's are capable of maintaining their advertised boost clocks indefinitely (or as long as you need them for).
Laptop grade hardware should be able to do the same... if it can't, then its an OEM design flaw and shouldn't be compromised on.

 

That is especially true for Ryzen APUs, but throttling over long periods of time is pretty much to be expected. The main point I am trying to get to is that IF you indeed have a CPU that will not throttle uder the maximum possible loads, then the implementation is just plain wrong.

For example, I have an XPS 15 with a hexa core i7. It is certainly a fast machine for sure, but the CPU throttles. Now, if I were to put in a quad core i7 into it, it would not throttle at all. Not even a CPU benchmark would give it any trouble. Which is a "Better" implementation from what I think is your point. But here is the problem, while it can indeed maintain full turbo over any loads, you are not really getting the most out of your CPU. If you had a hexa core i7, even if it throttles to use just as much power (TDP), you would be better off for the vast majority of tasks as that short windows before it throttles will be much, much faster than the quad core. Even if you cannot get the full 100% out of the CPU under loads, that doesn't mean it isn't a bad implmentation at all.

If you had a CPU that doesn't throttle, that would mean your cooling is overkill (12 pounds of cooling for an i9) or your CPU is too weak for the cooling system (45W CPU in a 60W chassis). This isn't a design flaw, that is just how it works.

 

I will just put this here

https://gitlab.com/ryzen-controller-team/ryzen-controller

might help someone...

 

So Ryzen undervolting is now a thing? Consider me intrigued

 

Unless I missed it, it's still only limiting wattage, not undervolting.