Mobile Polaris Discussion

the mxm spec alows for higher wattage then extra can be added with additional plugs (like on the desktop 980 laptop card)

i don't recall AMD doing this before

 

If the case of some cards being affected remains, then that means some bad apples have gone through... Nvidia has similar issues sometimes, and most GPU's go well over the PCI-E specs anyway.

Undervolt to its lowest possible stable voltage and then see how it behaves... if you are STILL concerned about power draw (although at that point you shouldn't be), your option is to return it and ask for a refund and get a different gpu, or get a new one that doesn't go over the PCI-E specs.

Btw... the power spikes being registered as going over the PCI-E specs are mostly that... power spikes. The gpu won't be consistently trying to guzzle that much power necessarily... and even if it tries, as I said before, no 480 should be operating on such high voltages in the first place at stock clocks.

Do we even know that the voltages of the affected GPU's are even the same for those that do not go over the PCI-E specs?
Some other people have apparently reported that their 480's are drawing about 20W LESS than 150W under load (at 1.29V - but still).

Drop the voltages some more and possibly overclock a bit, and you now have a power efficient and faster than stock 480.

 

the issue is that the average power draw is too high 75w, a lot of other cards go above this but they do only momentarily. its like with a psu they can put out more than what they are rated for but they are not desinged to run a such high of a wattage

if your on a good quality mobo you should be fine but you are causing greater wear you are better off waiting for a after market card with an 8 pin connector & a better cooler. better coolers are always good

multipule websites have done testing showing it pcperfor example
https://www.pcper.com/reviews/Graphics-Cards/Power-Consumption-Concerns-Radeon-RX-480

toms hardware
http://www.tomshardware.co.uk/amd-radeon-rx-480-polaris-10,review-33587-9.html

 

Overhyping is what AMD does regular. If you listen to AMDs prerelease statements you start seeing AMD Bulldozer and Excavator being above Gen i5 and i7 respectfully.

 

AMD posted 8 new videos 4 days ago...
https://www.youtube.com/user/amd/videos

Playlist 8 + Polaris Intro

 

i think its unfair to single AMD out on that nVida, Intel & Apple oh apple, most game companies its just marketing you should never belive to what they clame untill its in the hands of independent reviewers

 

They must be obeyed!

Spoiler

 

As shown in a video a few pages ago, there are nGREEDIAs with higher power spikes on the market for a while now and I don't hear anyone moaning about it. They run entirely out of PCIe power. Again, I'm not saying that AMD is not for blame, but it seems "a bit" unfair, isn't it. Just like dying 7970ms, while in the same time there are quite a handful of green deaths and just occasional mention about that and most of the time it's buried in a topic unrelated to a GPU death, so one has to really look for it. I see a very alarming pattern here. Whatever nGREEDIA does wrong can get away with it, but God forbid if AMD does something wrong all hell breaks loose. Oh and BTW, if you are so worried about the reference GPU, there's always AIB boards - better power delivery (take that for granted, as they would have 8pin at the very least, as well as the usual better MOSFETs, caps and chokes), cooling (do I need to elaborate?), of course and higher price, but people paid $700 for a SCAM Editions, so I don't see how that would be unfair. It would still get better performance for the money, not to mention better cooling.

 

"Expert" said spikes are OK to withstand but constant level of this would kill motherboard at the end.
I hope AMD gets enough backlash on this to lower their share's price to 4$ at least. #deservedit or #almostliterallyaskedforit.

And I don't support arguments like "others are bad too", yeah, that is why I wanted to support opposing side of "others" and I feel betrayed now and I didn't even buy it!

 

This is something that is utterly nonsensical. AMD goes over the PCI-E specs and everyone loses their minds, while Nvidia does it frequently, and people turn a blind eye to their practices.

What kind of motherboards were used with 480 with the spikes? They were likely weak/cheap ones with not a lot of headroom in the first place.
More quality like mobo's should be able to handle 480 just fine even with non-adjusted voltages.

In the cases where power spikes and higher than expected power consumption occurs, just undervolt the reference GPU to it's lowest setting where it will be stable at stock or slightly higher overclock (1333MhZ), and case closed.

AMD can easily enough devise an algorithm to analyse the current voltages and lower them automatically via driver update, or automate the process for the most part, and guide people through possibly finding even lower settings down the line.

 

Please, give at least one single proof of Nvidia's card which consistently takes more than 75W from PCI-E slot.
P.S. It's already proven that 960 doesn't.

And possible reason is failing 6-pin specification too and connecting half of VRMs to PCI-E 12V lane directly. Does it mean #NonStopThrottling as a fix?

 

Deks people keep posting why it's an issue & it's an issue wich will be fixed by 5 or 12 phase cards. We have posted articles & videos that explain it's about constant sustained power draw that is the issue have you read/ watched any?
For example on a motor on start it will pull much higher current than the breaker is rated to trip at but then at load it will run no way near the trip current.

 

Then take a look at this please:
http://www.legitreviews.com/amd-radeon-rx-480-undervolting-performance_183699
http://arstechnica.com/gadgets/2016/07/amd-rx-480-pcie-power-statement/

So, AMD admits that certain (not all) reference 480 GPU's are exceeding the TDP because the power tuning is not optimal.

What does this indicate?
Simply that someone at AMD probably overdid it with the voltages on some of the reference cards.
Also, some people have reportedly found that the 480 memory could be run at HALF the voltage of what it's usually set (and let's not forget that same users at Reddit discovered that the VRAM is chewing away at the power the more its used).

So, here's a suggestion: the problem isn't a big deal and the gpu can probably still run at roughly 120W-130W tops under stock clocks and proper undervolt (possibly less).
Yes, some of the reference cards are affected, but the AIB versions probably won't be.
Also, the issue is easily remedied by software that's released by AMD itself.
AMD gave you the tools needed to lower power consumption while at the same time allow the 480 to keep itself at its boost clocks continuously.

If I was getting myself a desktop, I would actually spend LESS money by getting the reference GPU, undervolt it as low as it can go, and happily game away at 2K and High settings in most recent games (that is if any peaked my interest) at stock clocks.

And again I say: people are blowing this issue out of proportion.

Heck, AMD even went as far to give people the tools to modify their 480's to how they want it.
Admittedly, someone did mess up the voltages in the first place, but again, it's nothing that cannot be resolved by a simple and proper undervolt.
At least this time around, we don't have to go through complicated procedures to undervolt the GPU.

 

That AMD chose the most damage controlled answer in expense of it being the least truthful. It's not the proper answer when all cards have "unoptimal power scheme" and people's motherboards die. The proper one would be "We analyze situation, all who has got issues or want a refund, contact us."
The GPUs potential is not a concern, the company and it's tactics ARE.

Whenever average consumer starts googling about How to lower voltage of RX 480 that will be already after one of his PCI-E slots died or he had shutdowns unless he luckily spots this news or told by friend. Otherwise why would he do? Those who buy reference don't use overclocking tools usually.

Do you REALLY believe that AMD gonna lower voltage from 1137mV to 1050mV? LOL. So that they would get even more complaints about unstable cards? Because some aint binned high, you know?
And at that best case scenario they got - 10-30W which translates to 5-15W reduction from PCI-E - still doesn't fully solve it... and they won't set 1.05V anyways.
P.S. Do you think that undervolting GPU 480 will provide a better max Ocing? Because if not then we have another proble, AMD gave hints that it is overclocking good.

 

Is overdoing it for effect something that people do these days?
Seriously... the premise that most users PCI-E slots will go dead when they start looking into undervolting is an assumption. You base that assumption on the idea that ALL 480 cards will do the same thing and will do so in the shortest amount of time possible

I'm sick to death of people blowing things out of proportion for no good reason.
Is there a reason to be concerned?
Yes, but it is highly unlikely that people's motherboards and/or PCI-E slots will die en-masse just because you say they will.

AMD may have opted for most damage controlled response, but Nvidia would do the same.
Both are corporations out to make money... and in all honesty, we only have a handful of people reporting the problem (usually ones with the means to attract attention), while other bunch of people/posters also seem to be reporting nothing wrong with their GPU's and also not drawing anywhere near as much power as some of the affected cards.

Plus, AMD is already working on a software solution to the problem... my concern here however is what form will the solution take?
Will it be undervolting the GPU to lowest possible levels where performance will go up and power consumption go down, or simply underclock the GPU?

Undervolting would likely affect the root cause of the problem, whereas underclocking would simply side-track the issue.
Corporations aren't the best in attacking root causes though in general though.

As for how fast this problem will be resolved... if AMD's Radeon division is hard at work as usual, then it shouldn't be long at all.

 

Next 2 days tell us how exactly and how effectively AMD tries to solve this issue.
But it's not needed for all or even a half of all cards burn mobos. Not even 40-30, heck, even 10%! It will just be enough of 20-50 well documentally confirmed cases and considering how minimalistic AMD's response was about it and they gonna get a massive backlash.

 

I'm not over doing it i am repeatedly tying to explain the difference between peak & sustained power draw

Most people aren’t blowing it out of proportion 75w as an sustained average is to high, No one is saying that people's motherboards and/or PCI-E slots will die en-masse, people are saying that it can cause damage & personally I & I’m sure plenty of others don’t really want to risk their £150+ mobo also how can system builders guarantee systems were there is this issue atm, it doesn't matter if its 50% or 10% they should of picked it up in the factory AMD realy needed to pull this off perfectly but their making so meany simple mistakes

The a good number of people who buy low to mid-tier cards will be building a gaming system on a budget & the best place to save on would be a motherboard / cpu their target market are the ones most likely to see the issue. As mentioned earlier when we were talking about OCing a lot of people don’t even want to touch this stuff (disappointing to hear I know) I’ve a build I’m doing for someone I’m working on have told them about the card & now they want one & I can’t in all good consciousness sell them a computer with a reference card so I’m waiting to see what happens with the custom cards

Unfortunately a under volt may not be best in all circumstances as with cpu under volting due to the silicon lottery their will be a lot of diffrence in what is required, but hopefully it will fix it, would need testing across a lot of cards thou

don't want to leave it on a downward not but has anyone seen this?

you only need to watch from 6:30-13:30mins


I personal really want AMD to do well I’m quite vocal about my opinion of nGreeida & would like to switch over to their cards for my personal use it’s just they keep messing up

 

If it is possible to undervolt every GPU, it means that somebody at AMD made a mistake in setting the base voltage. This would be quite fortuitous for them, but usually, the voltage is set to the default value because it is possible to undervolt some, perhaps even most, GPUs -- but not all of them. It wouldn't do AMD any good to issue an update which makes 95% of the cards operate perfectly within the advertised power budget if the remaining 5% start experiencing random crashes because the voltage is too low. Of course, it is possible that they either set the voltage too high or had left themselves some margin and are now going to cut into that margin, but it sounds more plausible that something (perhaps the memory?) will have to be downclocked together with the voltage decrease.

 

Per the reviewers tests themselves, all of the affected cards seem to have voltages set to too high levels.
I'm not saying every 480 will undervolt the same, however, every 480 will likely undervolt low enough where most, if not all cards would experience much lower power draw and better sustained stock boost clocks.

The differences will likely start cropping up in the area of which GPU can undervolt the LOWEST (not whether it can undervolt in the first place)... for instance, my P9600 CPU is a very old one (from a technological point of view) and it can undervolt to roughly 1.0V or just under that, however, some people were able to push their own P9600's even lower than mine - which means they got a relatively good silicon batch.

So, it's not a question if a GPU can undervolt... pretty much every GPU is capable of it, because the manufacturers set a high enough voltage to ensure stability of their products across the board... however, in the case of AMD, it would appear that someone ended up trigger happy with the voltage and it was set too high for some of the reference cards.

 

AMD said something about auto aging managing which means it need higher voltage with time. I don't know how that statement now fit if they lower voltage to the limit... if not now then later even if it happens to 5% only. #timebomb?

 

https://www.techpowerup.com/223893/amd-4gb-radeon-rx-480-can-be-flashed-into-8gb
If you want to save $30 to $50, this may be of help. Of course, check to see if it physically has 8 GB.
http://www.tomshardware.com/reviews/amd-radeon-rx-480-power-measurements,4622.html
Also, Tom's Hardware re-did their power measurement tests to get a second round of results.

 

Oh AMD, Please never change

 

Choice quotes:

Specifications covering just how much power is allowed to flow through the pins of a 16x PCIe motherboard slot were established a long time ago. However, saying that they include a maximum wattage isn’t just misleading, but flat out false.​

and

4. Current hardware should be able to handle this amount of current without taking any damage, as long as the motherboard’s slots are clean and not corroded. It’s also advisable to make sure that the graphics card sits precisely in its slot. This should always be the case, though, even with significantly lower amounts of power.​

Summary of the summary: Yes, there is a problem with at least some of the reference cards out there or the drivers for same. No, this problem is not anywhere near as big a deal as the trade press or this thread have made it out to be. No, your mobo isn't going catch fire or otherwise be destroyed by this problem unless it's already so far off spec that it would fail, catch fire, or whatever regardless of the GPU you put in it.

 

I caught that bit about auto aging managing as well.
It raises the potential possibility of this being a software problem in that case (considering that AMD stated how Wattman is still in relatively early stages).
The AIB versions of 480 will likely end up with lower voltages by default, not to mention custom cooling and of course more power cabling.

In regards to reference GPU's, they weren't meant for high OC-ing.
They are meant for predominant use at stock clocks and slight OC most likely that could bring you to 150W.
Actually, considering that GDDR seems to be quite the power hog, I wonder what happens when you not only undervolt the core, but also drop the voltage on the memory as well by half (some people reportedly stated that the VRAM is a massive power hog) - which could explain some wild fluctuations we've seen in the reviews with the affected cards.

Now, dropping the voltage on all clocks and the VRAM as low as they will go might result in 480 that won't even exceed 110W (I think this claim might be valid in that case).
Also, with carefully raising the voltages bit by bit from their lowest for stock and also OC-ing, you could theoretically ramp up the GPU to about 140W all together and still gain a nice boost in performance.

Sort of like manually optimizing your own GPU's power potential at stock first, then raise the clocks and voltage as needed to reach (but not exceed) for example 140W.

Granted, that could be just wishful thinking, but at the same time, it could also mean there's a potential in the GPU we might not have noticed.
Would be interesting to find someone with the GPU to test this out and see what happens when they drop the VRAM voltage (as well as core voltage) to as low as it will go on stock first, then redo all tests, and after that slowly raise the voltage bit by bit and increasing clocks.

 

What I see here is a product that is forbidden legal requirements and will be working without a problem only if there is some stamina prepared for out of spec products like this...

I don't think that it's motherboards fault that it doesn't have safe reserve over the legal specification it is validated.

Sorry but that is pure speculation, such of kind which predicted RX 480 to be like FuryX.
THG wrote that 110 is what die takes while averages at 165. Even if you take away 20W from die and lets say HALF of memory powerdrain and lower 25W (other components+powerloss) to 15 you still end up at 90+15+15 and that is still like a fairy-tail case.
Real-case scenario is that AMD "fixes" card to average at 150W.

 

Of course it is speculation... I even said as much.
But it's not that far fetched when you think that most people mainly undervolted the core clocks, but left the VRAM more or less untouched (or didn't try undervolting any lower), and STILL got to about 135W of power consumption.

It's a cumulative effect. If the GPU is undervolted on clocks alone, that brings the affected 480 into about 135 W range of power consumption.
Undervolt the VRAM as low as it will go and you might shave off another 10 to 20W (depending).

VRAM power consumption (especially on GDDR5) jumps up quite a bit when a lot of it is used... especially on 2K and 4k (not that 4k is necessarily too usable for most people on this GPU, but 2k should be more than adequate).

 

What are these "legal requirements" of which you write?

 

If you write PCI-E label on the box of GPU then you know... you shouldn't refrain from it's standards. At least it's what I read somewhere

Guys, what do you think about the fix AMD provided today for its graphics cards which may damage some motherboards? Oh yeah, it still hasn't. There is still some time though...

 

Again, what are these "legal requirements" of which you previously asserted? Last I knew, the PCI-SIG is not an agency of any government and therefore is not capable of issuing regulations with the authority of law.

Never mind that the "standards" you've been asserting appear nowhere in PCI-SIG's documents.

 

http://wccftech.com/amd-rx-480-overclocked-air-water/
Custom RX 480s coming soon. While they seem promising, I wonder how much of a price markup they'll have.

 

Re: Radeon released RX 480. (Card is Released)
July 01, 2016, 09:52:08 PM
#503
Well it blew my motherboard. Returning them!!

https://bitcointalk.org/index.php?topic=1494859.msg15437574#msg15437574

Re: Radeon released RX 480. (Card is Released)
July 01, 2016, 10:10:22 PM
#506

When I came home, no fans on the card, no video, no boot.
I had 3 cards hooked up with non powered risers. Sending them back and I'll wait for an 8 pin version.
There are scorch marks on the main power to the motherboard.



Re: Radeon released RX 480. (Card is Released)
July 01, 2016, 11:34:31 PM
#511

Even if it was any other GPU, you should never use more than 2 GPUs without powered risers. It sucks that you destroyed a motherboard and PSU...

Re: Radeon released RX 480. (Card is Released)
July 01, 2016, 11:38:41 PM
#512

I have plenty of power supplies, have to get a new motherboard. I even had the clock speed on the GPU's clocked at 1000. Oh well, gives me an excuse to get better cards in a month or so. Until then, get a new MB and put the 280s back in.

Re: Radeon released RX 480. (Card is Released)
July 02, 2016, 02:07:24 AM
#515
Quote from: QuintLeo on July 02, 2016, 01:48:55 AM
Quote from: madmartyk on July 01, 2016, 11:32:47 PM
This is what I came home to today.


And that's WHY you use powered risers.


I know, but funny thing, I have ran 3 280x's on that system for years with non powered risers and have never had a problem. These draw less power but fried the board. Powered risers (and a new motherboard) or on the way. I'll wait to see what comes out with the 2nd gen of the 480s

Spoiler: problem root cause...

https://bitcointalk.org/index.php?topic=1494859.msg15463590#msg15463590
Re: Radeon released RX 480. (Card is Released)
July 04, 2016, 09:36:03 AM
Reply with quote #566

The 6 pin grounds are all grounded (one should be sensing, not grounded) on the reference rx480. So in practice it can deliver the same power as the 8 pin.

The problem is two fold:

1- AMD soldered the RAM vrm's to pull its power exclusively from the pci-e, without touching the 6 pin.

2- AMD hard soldered two pci-e power lanes to the GPU. So instead of the gpu pulling all its power from the 6 pin, its also pulling from pci-e.

As such, when both the GPU and the full 8GB of ram are being pushed, it will overload the pci-e


This being a physical circuitry problem, all the driver can do is to throttle the card (decreasing performance). Or, if there is a margin, to throttle the gpu Voltage.

 

http://www.anandtech.com/show/10469/amds-tuesday-radeon-rx-480-update-new-driver-by-late-thursday
AMD releases their solutions (on Thursday's driver).

 

AMD's second solution is not really an optimal one considering it can impacts performance (though the new driver should also bring some performance enhancements, which might balance things out).

Better way would be to simply undervolt the GPU on the core as low as it can go, and do the same on the VRAM.
That will solve the power draw problem by quite a bit on its own and increase performance at the same time.
Apparently, one user was able to undervolt the OC-ed boosts (1360 MhZ) down to 1.056V.
I asked him to try and undervolt the VRAM as well to see if that brings additional power savings.

 

Reviewer has power off's on AMD budget build when used with RX480 and not with 980ti.

Spoiler: The $350 Budget Build using the RX480

Seems like a good system to try the AMD driver fix. I will watch his channel for an update.

 

Since we are after "specs" and "as advertized", what's your take on DisplayPort than can't display or VR GPU that can't VR? I don't see any of you mention it. Oh and it's been a month since the initial reports and there's still no solution. Let's see the same fierce for rightougness

 

I'll just put it here:
Nliedia (note "much faster"):

Nrealdia:

 

Because it is Polaris topic, not Pascal
AMD's fix is to disbalance VRMs so 3 of them would draw more because they are connected to 6-pin. it will push 6-pin even further way from specification but this is not a big issue. Yet they still have to propose alternative fix which lowers performance to fully fix the problem... on stock only.
The line about driver 3% enhance is just a salty candy. You would get those anyways.

SUMMS UP: Don't buy reference RX 480 no matter what.

 

I don't see it in the Pascal thread either.

Some reviewers say (the ones that took time to note it) that it's a pretty overbuilt power delivery, so let's see how this dis-balance would affect it. As for the 6pin, they really can pull in excess of 150W, it's up to the PSU, but most PSU would handle even this "excessive" load (actually some say more than that). Makes me wonder why they didn't do that in first place. I mean, I said it a couple of times already, I hoped that it would sip power, but it doesn't, so if they intentionally ran out of spec, the safest bet would've been the 6pin. All motherboards are not equal that's for sure, but (not to say all) PSUs would handle more than 75W load though the 6pin. BTW these wires are thicker than needed for a reason, the same reason I said that motherboards should be able to handle more than the PCIe spec - over-engineering. As for the performance, let's see how much it drops, if any, as the official quote mentions some optimizations as well.

Well I said that a while ago - just wait for the custom solutions. Ever since the first leaks of the GPU and the cooler (just an aluminum radiator), I was actually hoping that it would be so efficient that it doesn't need a fancy cooling. Obviously it needs a fancy cooling and more than 6pin, so custom solution it is.

 

The wording is a key though. I hate marketing to start with, especially in recent years, but the AMD's claim is "The World's fastest notebook GPU", which is true and they don't mention by how much ("by a HUGE margin" for example), while in the nGREEDIA's it says "Much faster" where the actual scale shows 15% at best. Is that "much faster" to you? I think that's why @Hurik said "(note "much faster")". If it was "faster" that would be true, as in AMD's case, but then that's not catchy enough. It's all marketing after all, but the wording here is a key.

 

Going to test my newly arrived RX 480 in my desktop. If I come back mad, you know what happened.
Since the new driver is also coming out today, I can test both as well. You know, to act as another source.
EDIT: And I'm back from doing a 3DMark11 benchmark. Given that I haven't used my desktop in almost half a year, take the CPU data with a pinch of salt.

Spoiler: Screenshot

I wanted to show off a comparison with my old R9 285, but I remembered I used an older driver with it. I did an undervolt of 80mv, and upped the fan curve on it to be more aggressive. In an open-air "case", the max it reached was about 75 C with my undervolting and different fan curve. Voltage was around 1.02, I think. However, according to GPU-Z, it did consume a maximum of about 119 watts on the GPU core (as the rest isn't measured in GPU-Z) during the benchmark.

 

Make sure to undervolt the core as low as it will go, and also to do the same on the VRAM.
One or two people have claimed that the voltage for 480 memory can be undervolted by half (which seems a bit far fetched) - but you never know.
Undervolt the VRAM as well on stock and then measure the power draw and performance.
It would also be interesting if you then progressively raised the voltages and clocks on both the core and vram bit by bit until you do reach say 140W consumption... should result in an ok performance boost.

AMD's driver solution... I don't think it will be as effective as proper undervolting.

 

I don't have any power/wattage measuring tools (I should get some), so I have to rely on GPU-Z's core wattage report for the time being, but I didn't know I could undervolt by that much (in the best case scenario). When I do more runs later, I'll be sure to undervolt more.

 

Don't quote me on halving the vram voltage though. As I said it seems a bit far fetched even to me. Though one person managed to oc the core to about 1360MHz and undervolt to about 1.1v if I'm not mistaken. He has not tried undervolting the vram.

It would seem that gddr5 is causing the power spikes and higher than usual power draw (amd also stated something to that effect as well). So the vram should definitely be undervolted.

I also meant you should use gpu-z for wattage reporting.

Keep us posted.

 

Wow man, your GPU score is like 1000 points more than stock 980, right at 970M sli! Btw, here we go:
http://videocardz.com/62009/amd-rel...mson-16-7-1-drivers-fixing-power-distribution

 

This is great hopefully this issue gets debunked.

Seriously though why would AMD make such a mistake at this point in time. Not good timing at all!



Sent from my SM-A500FU using Tapatalk

 

I think I may have undervolted a bit too much in my second run since it crashed during the benchmark run.
Doesn't help hat my RX 480 has an ASIC quality of 89.3%, so I may not be able to undervolt much to begin with. Also found out that I undervolted my VRAM by 80mv and the core by 31 mv from 1062 mv Not sure, but given how Wattman, as buggy as it currently is, works, when the RX 480 reaches state 7 (I gave it a clock speed of 1266MHz at the state, which should be the maximum stock boost clock), the undervolt I gave it was the same amount of voltage applied for state 6 (which operates at 1235 MHz).
EDIT: Undervolted the VRAM by an additional 20mv. Total score didn't change, but the graphics score dropped by 70 points. As well, GPU core wattage was the same (since I didn't change the undervolt).
EDIT 2: I did do a stock run to see if there were any improvements to the undervolt, and there was. Compared to the stock graphics score, my 1st undervolt performed better by 370 points in the graphics score. Same core wattage, though.
Also did a core undervolt by 11mv (total of 42mv), and kept the VRAM undervolt the same 80 mv. Performance dropped, and the wattage didn't. I think I'm at a limit.

 

Only thing that matters to me from AMD is if their mobile GPU is used dedicated w/o the worst thing that ever happened to mobile gaming, Enduro/Optimus...

That is the only way AMD can prove to me they are serious about mobile gaming, if they can convince MSI, ASUS, Alienware or Sager to use a Polaris fully dedicated GPU w/o support for Enduro.

 

So, what was the wattage according to GPU-Z with the undervolt on both core and VRAM?
Btw... if you start noticing performance drops due to undervolting (which seems quite odd to begin with), then just raise the voltage just a tiny bit to not give you any trouble.

 

Around 121 after the additional core undervolt. That alone strikes me as odd.