ThinkPad E485 - (Underrated gem!)

The processors are officially available, but it might take Lenvo quite a few months to add these to their systems.
Looking at the early benchmarks and specs, you can expect the 3500u to be a tad slower than 2700u, while 3700u to pull ahead by about 5%. Not much difference to worry about it. Maybe the 4000 series offers a much better performance jump, but that is at least a year away.

 

hi, the youtube link is down.. able to share yr undervolt settings?

using a e585 with 2700u too!

 

For sure. It depends on the TDP you are trying to reach. First use uProf to check the TDP levels your machine is set to. For reference, E485 stock settings are:

Max STAPM (or sustained power) = 22W
Max Long duration Limit = 25W
Max Short duration Limit = 25W
VRMMAX - ?

Now, let's say all you want is undervolt for lower temps. User Ryzenadj and set vrmmax_current=40000. Then check uProf again (run some GPU stresstest like Furmark) and see if the STAPM power still shows at 22W. If it does, use Ryzenadj and lower vrmmax to say 39000. Check the STAPM again when the GPU is under stress. If you see the STAPM value drop below 22W, - you've found your lowest VRM. If not, - keep lowering it. Rinse and repeat. Should only take a couple min to find your stable VRM to sustain 22W.


Now, if you want to get max performance out of your APU, - this a different story. At stock clocks, both CPU and GPU are crippled by about 15-20%! The most optimal overall performance I was able to achieve so far using the following :

Max STAPM (or sustained power) = 35W
Max Long duration Limit = 35W
Max Short duration Limit = 35W
VRMMAX Current = 59000

You may have to repaste to have the temp under control though. Let me know how it goes.

 

Just pulled the trigger on this and 32gb of memory. Looking forward to having a play around with it. I debated long and hard about spending more (XPS 13/15 and a Lenovo X1 Extreme) but I figure those are overkill for what I need which is really going to come down to running Office/ web browsing and spinning up VM's.
Not to mention I can actually reuse some of the 7mm SSD's I have lying about.

Also, given it's price I'm going to feel fairly comfortable taking it apart and applying decent cooling paste.

Knowing my luck Lenovo will annouce the 3x00 series in the E series tomorrow, but given I have gone from a Ryzen 1700 to a 2600 and didn't notice much difference I'm guessing it'll be pretty similar to what will happen with the mobile chips - I'm not really too fussed about battery life improvements...just hoping that any improvements from a driver perspective get passed down.

 

Welcome to the club!
The 3x00 series won't be much of an improvement.
3500u is about 5% stronger than 2500u, but still a bit weaker than 2700u, while the 3700u is a tad faster than 2700u. I wouldn't worry too much.

Meanwhile another piece of info!
It's possible to manually set CPU frequencies of your Ryzen mobile!
I' was able to run it @3.5ghzx8 sustained @35W while keeping the temps below 85C
If you don't need that much horse power, can set it to 3ghz @25W and keep it cool.
The highest I was able to go was 3.6ghzx8 but it draws around 40W and the temps approach 90C
Once my cooling mod has been finalized, it should be possible to achieve a full 3.8ghzx8 sustained clocks @45W

 

Not really that CB score is Skylake 6700HQ level performance. GB4 scores is impressive too.
Have you tried unlocking TDP in linux like this mentioned in this thread? https://www.reddit.com/r/Amd/comments/ad13w8/undervolting_ryzen_mobile_2500u2700u_in_linux/
https://www.reddit.com/r/Amd/comments/acy2nu/universal_way_to_modify_ryzen_apu_stapm_on_linux/
If Parrot distro supports synaptic pkg manager, try installing tuned-gui and set to balanced on AC and switch to battery saver in battery manually.
Only issue with lenovo is their small battery capacity provided for budget and mainstream devices.
You can also use TLP for controlling charge rate and other thinkpad stuffs. Just comment on intel CPU drivers and leave radeon GPU as it is.

 

Actually, that info is outdated, need to upload new pics.
Cinebench15 - 730 @3.38Ghzx8. Just waiting on a bigger power brick and cooling mod supplied to bump it to 3.8ghz. Expecting to break 800pts. That's way above my AW 13 R3 with 6700HQ (stock 45W)
Geekbench4 - 16.5k was the best score so far, but again only @3.4ghz.
Yes, the TDP in linux is just as easy to unlock, but for some reason the temp seems quite a bit higher. I've seen it shoot up from 50 to 85 in one second. Others report up to 100c spikes. Something is definitely off, need to get to the bottom of it. Though, I don't need the full 3.8ghz power in linux, so can keep it locked at 20W and still have a very nice performance.
Battery could be bigger for sure, 4-5 hours is all you get at stock settings. You can, however, downclock the CPU to 400mhz! So, it should be possible to tweak the battery life to reach a full day computing.

 

My A485 comes in tomorrow, I'm excited to see what the little guy can do. I hope it works out and has good battery life...

 

Curious about it too as I haven't seen anyone with a A485 on our ryzen discord server.
For starters, I'd use uProf to check the existing settings, then stresstest and check the stock temps before repasting and modding.

 

What's the discord server name? Be good to jump on there when I get the thing to more easily share info.

 

Here you go - https://discord.gg/qEAfkuA

 

That's impossible. I hate Ryzen, it made my 45W CPU look like POS.
There's another trick for increasing CB and benching scores by putting the process to real time priority to eek out max points. Have you disabled Spectre protection in Windows 10?
AMD calculate the temps by querying max/worse temps unlike Intel.
How the boost clock works? Is it 38x on 1 core,36x on 2/3 cores and 35x on all cores? I refuse to believe that 15W pack so much power to beat a full blown 45W-55W HQ Chip but its a fact.

 

I hate when people tweak the settings to get higher scores. I don't! Not only I didn't touch any priorities, but also didn't optimize anything, running all the same processes in the background, including 20 tabs in Firefox and all the monitoring tools, photo viewer, snipping, discord, steam, etc. I do this to test the real life performance and to compare the results to those from notebookcheck, as we know they don't optimize anything either.
This is @3.4ghz on all cores.
This chip CAN run 3.5, 3.6, 3.7 and even 3.8ghz on ALL cores. But it draws as much as 45W at 3.8ghz

 

Pretty good score.

What's the fan like when it's doing the benching? That's one thing I've been a bit disappointed with my Yoga X1, whenever you throw anything at it that stresses the CPU the fans go ballistic.

It's obviously a much thinner laptop so that's to be expected, but interested to see how the thicker chassis deals with it

 

This isn't cheating, real time priority ensures max scores and if the median between stock and real time priority is more than 100-200 pts then you must optimise your OS esp. Win 10. I scored mere 630 pts and real time pushed it to 688 and after optimising it, I get consistent of 696 and on real time priority I get 699.
I dare you to run CB R15 at Nice value of 0 in WINE or Q4Wine and see the scores, you'll feel Windows is POS.

 

Normal use - I don't hear the fan at all. I've repasted though, so the temps are lower. But regardless, at full boar 100% CPU @3.4fgz+ there's no escaping the noise. It's definitely not as loud as those gaming machines, but I do hear it. Not annoying me at all, so take it with a grain of salt as I'm biased here.

If we want to know the absolute max score for each chip - yes, I agree. We would need to run it real time, fully optimized OS. But since I've been comparing my results to those from notebookcheck, had to do exactly what they did - no special optimizations. It's also not a good comparison vs HWBOT top scores as different motherboards and BIOS settings may allow certain chips to be overclocked and overvolted.

With Ryzen 2x00u, we are still just trying to reach its reference clocks for CPU and GPU.
What Lenovo and other vendors did is unacceptable. Most of them locked the TDP at 11-15W total draw for the chip. Even Intel's i7-8x50u, while also rated as 15W, can boost all the way to 45W+ when needed. And their iGP is nowhere near as demanding as Vega. E485 fairs a bit better after recent BIOS updates, sustained TDP=22W, boosts up to 25W. Still, it can only get to about 2.8Ghzx8 for the CPU and 1000Mhz out of 1300 for the GPU.
At 30W, you get - [email protected], GPU@1183Mhz
At 35W, - [email protected], GPU@1220Mhz
At 40W - [email protected], GPU@1300Mhz - Maxed out
At 45W - both CPU & GPU maxed out. [email protected]

So why would they advertise something that is only possible at higher TDP? And why would Lenovo ignore this and cripple the APU with their BIOS?

Regardless, we are now able to fully control the TDP, voltages and temp thresholds, as well as manually set CPU frequencies on the fly both in W10 and Linux.

Currently working on RAM timings. Should be able to get another 15-20% GPU boost once I can fully utilize my 3200MHz memory, which is currently downclocked to a mere 2400MHz.

And I totally agree about W10, Linux is my OS of choice even though still lacking the proficiency here.

Didn't know you could run Cinebench R15 in WINE! Will certainly give it a try! Thanks

 

You're not reviewing your laptop or showing it off how good it is, rather you're using for security/forensics and people won't want to hear anything saying CPU is unoptimised or slow. They'll suggest you to buy ThrottleBook Pro from Apple.
Sadly my CPU can max out only at 40W and higher than will throttle it down for some odd reason.
Use Q4Wine with Win 8 compatibility layer. Allow CB.exe to run as exe permission. The rendering will be quick so keep your eyes on screen. Sadly, CPUz won't run and so does any Winring0 based apps.

 

Well so far I'm both impressed and disappointed in this A485. Build quality is pretty good but they've done a few silly things (like not have a dedicated NVMe slot, it goes through a cable to the MB at 2x). I'm benching it "stock" (well I already added my 16GB HyperX CL15 ram to it and SX8200 480GB SSD). I'll make a dedicated thread and not derail yours any more @Kuro Kensei, thanks for making me look into this little cpu/gpu that hopefully can, lol...

 

@Kuro Kensei Give tuned-adm a try. Choose your version of power saving or performance poison. I chose balanced profile. Tuned-gui is a hit or miss and constantly needs gtkkmm modules to reinstalled to work. Cmd lines are easy to work with.

Code:

tuned-adm list
tuned-adm profile balanced

You will see a minimum of 1-3W of drop.

 

cmd is fine. 3W drop would be epic, as I currently have about 8-9W discharge rate in Linux. 5W would be ideal but that may be impossible.

 

5W drain is becoming greedy and you won't get it unless you break pico meter lithos.
Maybe 1-2W minus your current wattage reading excluding the screen. I never thought Ryzen was this much power efficient. I was holding off the upgrade to complete AMD system incl. laptops and desktops. I didn't get true DTR like AMD based laptop that gave greater driver flexibility like Intel and not to rely on OEMs for updates.

 

So I've been playing around with RyzenAdj and I'm not convinced that changing the VRM current is "undervolting" the processor. I can change the VRM current enough to underclock the cpu/gpu but it doesn't actually change the temp of the cpu when running prime/furmark. Sadly until they figure out how to do actual voltage control I don't think undervolting is a reality with mobile ryzen (really lame since I can undervolt my desktop with ryzen master).

 

You need to play with both STAPM, Long and Short Limits in addition to VRMMAX current.
Lower VRMMAX current means lower voltage and lower power draw. If you drop it too low, it affects the clocks for sure. Think of your 2700u as a highly optimized Intel chip, that's been undervolted already. If you set Intel CPU @ 50W and 1.4v (just giving random numbers) while it's able to run at max clocks @ 33W @ 1.3v, then by undervolting you will effectively drop it down to those values, your CPU power draw will be reduced to 33W @1.3v.

As for the heat, if you set your 2700u @ 50W and 1.4v it will run very hot @3.8ghz. It will run at the same 3.8ghz when VRM current is set to 67500 or essentially dropping the voltage to 1.3v and TDP to 42-43W. This will produce less heat.
I think we are looking at the same picture but in different ways. What I'm saying is, Ryzen mobile is already highly efficient and optimized to not need much tweaking. The problem, however, is that it's been restricted to run below its reference clocks. Now, when we unlock the TDP, voltage and temp thresholds, we need to find out the max stable state when both the CPU and GPU are able to run at their max frequencies, while using the least amount of power and producing the least amount of heat.

Last night I cycled through about 20 Fire Strike runs and found that anything above 37W produces lower scores and unstable clocks because my 65W power supply keeps failing! The total draw from the wall reaches 70-75W and the internal battery starts draining. Which brings me to conclusion that the effective limit of the power charger has been reached. So, I've ordered a 90W brick. Will know the truth in a few days!

 

I kept STAMP+Long+Short limits at the default (22/30/25) and adjusted VRMMAX. I adjusted it to the point where it would start running at slower speeds (STAMP of 15) on the CPU/GPU and then back up to where it would run at the better values. There was no difference in temperatures between the stock VRMMAX and the lowest VRMMAX values. I even raised it to 60000 and the temps were still the same.

Perhaps I am doing something wrong, but while playing with the current values of VRMMAX I had prime95 large FFTs and Furmark running in the background. I was logging temps/frequencies/power package with uProf to get my results.

I think VRMMAX is more similar to IccMaxx on an Intel CPU. It basically says "you can use this much total current regardless of voltage." Lowering IccMaxx causes the CPU to run at a lower frequency but does not play a hand in the voltage that it uses. I'll play around with it more though.

I am running with a 95W USB-C charger so running full tilt shouldn't be an issue.

 

You should see voltages change in HWinfo as you adjust VRM values. If it's not happening, then please join the discord server and ask FlyGoat to help getting a custom config matching your machine's DSDT table. I've seen some models that are "immune" to ryzenadj's manipulation, but Lenovo's generally respond well.

As for the temps, have you tried forcing the fan @100% using TPFC (ThinkPad Fan Control)? Could be that your fan simply doesn't get properly engaged till the temp reaches 80C+. Stock thermal paste is also not very good. BTW, does your machine have 2 or 1 heat pipe?

 

I'll do some more testing with it to see if voltages are changing or not. I've already repasted the laptop (after taking a few temp logs stock), but sadly it's just one big heatpipe. Temps are low but then again it's keeping the CPU/GPU at a pretty low wattage.

*on edit*
I forgot to ask, do the values in ryzenadj remain after a system restart?

 

Maybe you need ACPI OSI trick mentioned here http://iam.tj/prototype/enhancements/Windows-acpi_osi.html
I used it on my old intel laptop yesterday and I'm getting 20 mins-30mins of extra battery for no reason. Battery drain is minimal.
Can both CPU and GPU be loaded perfectly w/o clocks being throttled on either subsystems?

 

Ryzenadj settings will reset on restart but there are ways to make it run on boot like XTU. I wouldn't do it until you find the golden combo, then create a startup process and apply those settings.
Please take a look at RyzenTest as well, it lets you set the CPU frequencies manually and is another amazing tool to fine tune the system performance. Just make sure to only manipulate the P1 state.

What are your temps btw? One heatpipe could be ok if it's big enough.

Will check the link, looks pretty neat!
It seems that models with 2500u can be fully loaded at about 35W with both CPU and GPU reaching their stable max clocks. The 2700u needs more power. Once I receive my new power brick will be able to confirm. At the moment I'm sitting at TDP 37.8W as the absolute max without tripping the charger and draining the battery. This means a stable 3.55GHz x8 + 1300/1300 GPU clocks. This gives the best [at the moment] scores in benches. 3DMark11 P GPU 5150, CB15 - 740.
We are trying to get the bios unlocked, to see if the RAM timings can be changed. I can only imagine what performance boost can be achieved when going from 2400 to 3200 mhz. Based on the desktop Vega 11 tests it should be around 15-20% which is insane. Guess more fun and exciting news coming soon!

 

Instead of power brick, check the specs of cable, specifically the one you connect to Mains. I just checked mine and Dell gave 9A/240v ~50/60Hz cable for AGA and laptop. I was cleaning my cupboard and got hold of 13A cable from my old radeon 5750 and improvements were only minor mainly on demanding tasks.

 

Mine is 100-240v 1.8A 50-60Hz for input (cable CSA 152192 SPT-2 300v, can't find any detailed specs) The brick is rated max 65W, but drawing 75W+ when benching. Even with plus platinum efficiency (I wish) it would still struggle and seeing my battery drain when setting TDP 40W+ kinda points towards the brick itself?

 

I still don't think that adjusting the VRM MAX does what you are thinking it does.

I just did a 44 minute prime95 small FFT bench (well 54 minutes if you count the 10 minute warmup) and here are the results of adjusting the VRMMAX from 40,000 to 30,000 in 1,000 increments every 4-5 minutes. You can see it has no effect on cpu core voltage or average SoC voltage. It also doesn't effect CPU power package. If I go below 30,000 I do get package power state reductions... And you can also see my temps are not very good, lol...


*on edit*
I am in the process of doing another test and this time I'm going all the way down to 25,000 at one minute increments to see what the clockspeed does.

 

Wow, something is clearly messed up! It does look like the settings have no effect on your APU. Are you using the ryzenadj applet with sliders or the batch file instead. Try making the changes via the batch file. I'll generate some graphs from my machine tonight. But this is totally different from my experience. Voltages change immediately after I set a vrmmax value, provided the resulting voltage is lower than the current one. If you simply rise the TDP (STAPM, Long and short values), it will only go as high as the vrmmax allows it. In other words, vrmmax current dictates the maximum package voltage which restricts the TDP. If you are not seeing any dependency and dynamics during your tests, - it's definitely not working as intended.

The temps aren't bad. I think it's the fan profile which isn't perfect and lets the CPU heat up to 70C before kicking in.

 

I put a more accurate temp reading in the chart.

I am using ryzenadj from an elevated command prompt with the following command example:
ryzenadj -a 22000 -b 30000 -c 25000 -k 40000

If I go low enough it restricts the package wattage (compiling the data for you now).

VRM Max is a current and not a voltage though. Yes V=IR but it is not actively altering the current nor voltage on my system. Perhaps it is because I am on windows...

Here is the latest test where I went from VRMMAX of 40,000 to 25,000 in 1,000 increments every 1 min. You can see that the package starts to drop power...

 

Now this makes more sense. You can see the voltages go down along with the clocks and wattage once vrmmax is below 30k. Apparently, 40k is high enough [max] current for the V=IR=1.Xv or high enough [max] Voltage to keep the power draw at 22W. I think it's working as intended and you now have a baseline for the stock CPU performance stats. I'd run some GPU and combined CPU+GPU tests before pushing higher.

And give ryzentest a try too, as you can force the CPU to run at higher frequencies. Don't mess with P0, just bump the clocks on P1.

 

But the voltage isn't dropping for a good reason. It's throttling the package because it can no longer draw enough current. Because of this the clocks are dropping so voltage drops. This is exactly like IccMax on the Intel platform. It's not a way to undervolt, but really a way to limit package power. CPU temps were not helped by this method (only when the power package/clocks were lowered).

This is not undervolting my friend, I hate to say it. This is throttling.


*on edit*
I also ran Cinebench R15 extreme at stock settings and then at 30,000 VRM Max. The CPU score went from 146 to 130. My X1E does 331 as a reference point.

 

Anybody figure out how to undervolt these puppies yet? AMD APUs would benefit greatly from undervolting, more so than intel CPUs due to that integrated GPU being as powerful as it is

 

146 and 130 for multicore? Or is it single and your X1E scores 331 in the single core bench? Then I rest my case, as I've never seen anything like this before. You should be seeing 180-190 in single core and 1100-1200 in multi if it doesn't throttle and holds its clocks. Not talking about HWbot world records here.
My 2700u scores 735 at 36W @3.4GHz, and you will see it break 750 if I can control the APU at 45W.

Well, if you say it's not undervolting - then it's not. Just for kicks, I've undervolted my 6700HQ in AW 13 using Intel XTU by -0.080 and the load temps dropped along with the total power draw of the package, which in itself dropped by at least a few watt. This is just simple physics. If you lower the voltage - you lower the power draw. Unless, the original voltage is configured too high, with a hefty safety buffer. Intel lets you play with it both in desktops and laptops. Maybe AMD will offer something like that too. But for now, this all we got. Which isn't bad at all. You can, just by using ryzenadj and ryzentest, bump make it run @3.5ghz instead of 2.6ghz in your graphs, while keeping pretty similar temps and maxing out the GPU clocks. Which again, isn't bad at all. It's just a different beast with its own hurdles to jump.

But hey, we aren't done with these bad boys yet. Some of us are working to unlock the BIOS, and when that happens, will see if we can offer better controls of voltages, GPU clocks and RAM timings. It's all just a matter of time, as we have access to the code already and figured out a way to flash it even with the silly digital signature.

Edit:
Take a look at the screenshot. The HWinfo column shows the max values. Look at the voltages and tell me this is not undervolting. I can provide a stock run with both CPU and GPU hitting 1.36v. But hey, what do I know about it, probably just a fluke.

Edit2:
I know what's going on. You are going down on VRMMAX and thus indeed throttling the chip. This would work if the chip was at max performance with a voltage headroom. Instead, it's already crippled by the bios and you are gimping it even more. Try going higher on all your TDPs and VRM instead. Get it to 35W across all TDPs and VRMMAX=55-65k. Set the CPU P1 state to 3.5ghz in Ryzentest0. Then run a few quick tests to see if the clocks are holding up. Then slowly reduce vrmmax to the lowest that can still hold the clocks - this IS fine-tuning.

Well, apparently what I regard as undervolting, is in reality a throttling. Dunno, look at the benchmarks in here, I'm getting a pretty solid performance at low temps.
To max out the iGPU all you need is about 36-37W. The CPU can be manually set to whatever max frequency you want. I currently have it at 3.4ghz while doing some gaming benchmarks and this is what I'm getting. Look at the GPU clocks and max temps. This is all stock air cooling. Nothing special. I can game at these settings all day.

 

You need to invest in 90W+ PSU to get best performance. These days OEM skimp on these looking at all core multipliers and TDP. My sister's lenovo had same issue, they provided with 45W crap PSU and I had used mine(65W) and I saw instant boost in processing and did you know that Lenovo crippled the dGPU by downclocking it by 100MHz on core and memory. I OC'ed it and supplied 20% extra power limit. No complains now. After updating MEIFW the PC is a rocket now even on HDD aside from Windows 10, Linux works fine.
@custom90gt Do you review any taptops or desktops using Linux? Have you tried Cinebench in Linux with Q4Wine? Is the VRM Max same as Intel's DC Loadline or IA Slope that reports less Watts consumed.

 

No sorry I shouldn't have added confusion to this. Cinebench Extreme is a MUCH more difficult cinebench program. Those are multicore scores.

I can promise you it's not undervolting the CPU in anyway. There is a big difference between undervolting (which will lower the package power while keeping clocks at the same speed) and limiting package power which reduces clocks and causes a lower voltage. Check my clockspeeds if you don't believe me. Sadly I know a thing about physics as a mechanical engineer... My hope is AMD will allow for an adjustable voltage offset, until then they are going to be pretty far behind in regards to mobile tweaking. I agree that it's nice you can unlock the TDP and I'm testing that now. That's something you can do on Intel also though.

It's hard to tell from your screenshot, but without seeing your logs I can't tell you what is going on. You have to take into account what your clocks are doing (likely hitting higher max clocks on stock without package power limiting). I'm not trying to be a negative nancy, I just don't think that limiting package power is undervolting. I agree that your scores are good for what the laptop is. Sadly this A485 is really limited temp and TDP wise and I don't think it would ever be close to what yours can hit.

Sadly I don't use any Linux outside of what I used to use as an engineer. I think that VRM Max is the same as Icc Max (max current that can be supplied to the CPU via the VRM).

 

I agree that TDP limiting is NOT undervolting. The very definition is LOWER voltage for the SAME clock speeds. Limiting TDP lowers clock speeds and voltage as a result of lower clock speeds. There is NO offset of actual voltage, just a drop corresponding to the lowered clock speeds.

 

In essence here is what limiting the package power allows the GPU to do:

 

I thought VRM max was the limit until the mobo blows up.

 

I'm confused as to what you are referring to perhaps?

With the RyzenAdj VRM Max sets the maximum current that the VRM is allowed to supply to the cpu under full load.

 

Well, similar to ICC max VRM Max is the max current the PSU can deliver for CPU operating at high clocks and if PSU supplies more current than requested value the mobo or CPU will fail in due time.

 

It has nothing to do with the power supply itself. It's the power regulation circuitry (VRM) for CPU which determines the max wattage that your CPU could use (if requested). The CPU won't/shouldn't draw more current than it needs. The only way the CPU will be damaged is if for some reason you added too much voltage.

 

Alright, let's try to figure it out. Are you guys familiar with AMD [desktop] undervolting? What are we looking for exactly? The bios will eventually get unlocked same as SMU, we just need to know what to look for. Are these make any sense?

Settings
-a, --stapm-limit=<u32> Sustained power limit (10e-3 W)
-b, --fast-limit=<u32> Fast PPT power limit (10e-3 W)
-c, --slow-limit=<u32> Slow PPT power limit (10e-3 W)
-d, --slow-time=<u32> Slow PPT constant time (ms)
-e, --stapm-time=<u32> STAMP constant time (ms)

-g, --vrm-current=<u32> Vrm Current Limit (mA)
-j, --vrmsoc-current=<u32> Vrm SoC Current Limit (mA)

-l, --vrmsocmax-current=<u32> Vrm SoC Maximum Current Limit (mA)
-m, --psi0-current=<u32> PSI0 Current Limit (mA)
-n, --psi0soc-current=<u32> PSI0 SoC Current Limit (mA)

 

Sadly the extent of my Ryzen Desktop undervolting/overclocking experience is through the bios and ryzen master.

With my motherboard I have a decent amount of voltage control. I can run with a standard voltage, use a custom base vcore, a vcore offset, or a combination of any of those:



Adjusting the VRM current limits would only allow for a potentially increased (or decreased if you lower it) max power load to the CPU.

In Ryzen master the voltage control is much more limited, but you can see it directly effects the vcore of the CPU itself:


Playing around with current limits to the CPU doesn't allow for proper voltage control IMHO. Really we need a proper voltage control to the CPU itself. How to achieve this, I have no clue. I would love it if eventually we had a Throttle Stop implementation on Ryzen, but that may never happen.

 

It would be very different from throttlestop, probably a different program entirely

 

Right, but ultimately something like throttle stop would be the ultimate undervolting utility IMHO. I consider Uncle's work in it to be the best when it comes to undervolting/increasing performance. You can even do different power profiles. Ryzen has a long way to go...

 

Part of the problem is how new Ryzen is. Given some time some guys will figure it out. I might even try myself if I can get my hands on ryzen.

Throttlestop is my favorite tool for any tuning on laptops, as long as they are intel