Mobile Polaris Discussion

In a nutshell: some reviewers are claiming that RX 480 wants 150W or more when PCI-E 3.0 only provides 75W and therefore AMD must be violating PCI-E specifications, that allowing RX 480 to draw the power it wants will damage their systems, and that AMD ****ed up for not putting an auxiliary power connector on the card.

Their claims are specious. PCI-E 3.0 specs say 300W maximum which is more than enough to power RX 480 without an auxiliary power connector. And if their test rigs can't provide that power, or "burn up" when they try, then it is their rigs that are off spec, not AMD's cards.

 

It seems we have a lot of conflicting reviews.
This is (unfortunately) something to be expected... as it's becoming difficult to ascertain which reviews are legitimate and which ones are full of so early in the launch.
More waiting I guess... but I kinda figured you could undervolt and overclock a GPU while reducing it's power draw and increasing its performance at the same time... did it on my old 9600m GT after all.

 

PCI-Sig still specifies 75W as standard power delivery for PCIe 3.0 slots. It's definitely possible to pull more power as we have seen AMD do before but I wouldn't do that unless you have a robust gaming motherboard, definitely not some cheap budget board.

 

PcPer have done an artical on it measuring how much it pulls from the mobo.

https://www.pcper.com/reviews/Graphics-Cards/Power-Consumption-Concerns-Radeon-RX-480

 

It looks like this is a problem with the RX480, drawing too much power through the PCI-E connector at stock speeds, and especially if you overclock:
https://www.pcper.com/reviews/Graph...ns-Radeon-RX-480/Overclocking-Current-Testing

Witcher 3 RX480 at stock:
"Zooming in a bit on the data we get more detail on the individual power draw from the motherboard and the PCIe 6-pin cable. The white line of the MB +12V power is going over 75 watts, but not dramatically so, while the +3.3V power is hovering just under 5 watts, for a total of ~80 watts. Power over the 6-pin connector goes above 80 watts here as well."

Witcher 3 RX480 Overclocked:
"When we zoom in we find that the motherboard is actually providing more than 95 watts of power over the +12V line and maintains the 5 watts from the +3.3V line, proving that we are indeed getting more than 100 watts through a PCIe connection that is only rated at 75 watts. The 6-pin PCIe power cable is almost crossing that 100 watt barrier too.

Because this is with overclocked settings, AMD is not really responsible for the specification breach in this specific instance, even though the company has talked about the overclocking headroom available on the RX 480 directly."

 

This is very interesting from OC3d (TTL)


sound like the cards proformance is bottlenecked by the 6pin

 

It was already the case for many previous cards, gtx 480 and 580 were already going above pciE specs.

And another thing to take into account :

https://www.computerbase.de/2016-06...el_potenzial_fuer_undervoltage_bei_der_rx_480

By undervolting the powerstates as followed :

P3 0,968 V 0,893 V
P4 1,037 V 0,960 V
P5 1,081 V 1,006 V
P6 1,137 V 1,062 V
P7 1,150 V 1,075 V

You gain 4% perfs, and loose 3.5db, 2 degrees, and most important 33w.

And now that make it the perfect card for its price range imo, wich could almost fit in a laptop with minor adjustments.

 

Anyway, my hype about RX480 is VANISHED.
Claims were that it's max TDP is 150W while it draws way less. In fact it draws way more!

I guess AMD will have to release vBIOS with 0.025V decrease. That should be safe enough and decrease power draw slightly. I think 0.05V would fix problem better but its obvious that such step is impossible hence many cards will become unstable.

And that is exactly why we don't see MXM mobile P10 this day. It would be ~980M performance wise with the same power draw.

 

Was it AMD's intent for 480 to draw more than 150W?
I doubt it.
As i postulated the question before, is it simply possible they ramped up the voltages on the reference cards to a certain level because they expected overclocking?
If so, they simply need to release a modified vbios which will lower the voltages to the point where 30W will be eliminated and raise performance at the same time (win-win), which would likely push the card in the territory of 980 in DX11 for the most part.

I think that they could release the option with a new driver release, offering people the option to undervolt the gpu and increase it's performance.
The process could technically be automated.

 

I won't go as far to say that it draws "way more" than 150W (stock that is), but yes, it draws more, not less. Then again, there are reviews that claim ~110W, go figure. Maybe @Deks is onto something.

There's still some hope for mobile. As you said yourself, they could decrease the voltage and as the post above yours says - performance improves, fans gear down, and the power consumption drops significantly. As for performance it would be well above 980m you can be sure about that and BTW 980m is NOT a 100W GPU, just so everything is clear.

 

Still cheaper than a 970 and with driver improvements will always be a step ahead. In DX12 it will stomp both 970 and 980. All for $200. Let's not lose sight of what AMD is trying to do.

They want marketshare and if successful who knows perhaps down the line we benefit as the company starts to make money and become more competitive with Ngreedia.

The bigger broader and brighter picture no more doom and gloom..

 

Stomp? Ehhh I think you're overstating things quite a bit there. The 480 gets stomped by stock 980s most of the time and 980s overclock like crazy too. I really wanted to see something good from AMD but Polaris isn't it. And with the power draw issue, I definitely won't be putting one in my machine. Replacing that motherboard would be too expensive if it killed my PCIe slots. I doubt my Maximus VI Formula would fail from the amp increase but I'm not willing to risk it either.

 

Is there a power draw issue? Seems very early days.

Perhaps we should wait to see more user reviews and enthusiast use before making final conclusions about it no?

Sent from my SM-A500FU using Tapatalk

 

triturbo Maybe my pessimistic calculations need to be calibrated a little bit but I won't recommend anyone a RX 480 anymore.

Real deal may be the RX 470 if it is priced properly... I would likely to buy one for 150$. Because it shouldn't have power overdrawing and is cheaper. Maybe I really should buy a PC with RX470 and sell my notebook... But only if 150. 160$ pricetag will make me think longer.

 

I think it's fair to say that we need to wait for the RX 480 to unleash its full potential with proper driver and good overclocking, then only we can compare it to gtx 980
as of now, RX 480 does actually performs better in DX12 compared to GTX 9xx series. that also with significantly smaller power draw and lower price when compared to 970 or 980
there is really no point degrading it this far

 

Except for the fact that currently overclocking is guaranteed to put the RX 480 consistently over the 75W PCIe power limit which will definitely fry some motherboards. It needs more power and AMD doesn't want to admit they screwed up. A card limited to 150W power draw is obviously defective if it's pulling 165W. AMD should have stuck on an 8 pin connector for 225W total instead of a 6 pin for 150W total, end of story.

 

Which is precisely what AMD promised: a VR-capable video card at a budget price. It was never going to be competition for the top end and, as I've been ranting about, even trying would be suicidal. What RX 480 does, it does well, and does it for a lot less money than nVidia.

And there's this: Async compute is one area where AMD absolutely wrecks nVidia, and Vulkan was designed to take full advantage of async compute capabilities. I expect that as more games take advantage of Vulkan's features, and as Vulkan drivers themselves are optimized and fine-tuned for various workloads, we're going to see a lot more of RX 480's real capabilities.

 

pay game devs money to delay/gimp vulkan and include gameworks!

 

Using that ASync won't just help them, the DX11 perf sucks on Fury vs Maxwell, and I don't expect any improvements in that area with Polaris, I think I read the Dx12 is fine but not by a huge stomping margins, except that Oxide's Ashes which is optimized for Radeon, they couldn't even beat Maxwell, very unfortunate, no one will pick this RX480 for laptop now(pcie powerover draw on Internet rekts faster than cyanide), and MXM spec I highly doubt with all that shiny number pushing false marketing ASUS corp rushing towards that GTX N cards or SLI M cards and proprietary MXM MSI push. It's over for AMD on mobile I'm afraid, we are screwed with the monopoly from the OEMs / Greed green and BGA filth.

AMD are happy with the console margins, they settled for mediocre level no race for crown anymore, the CEO is a failure. The GP102 will blow the Vega for sure at this rate, even if they manage (maybe get close) the GP100 is there as a trump card to bash (moreover this Pascal is just a die shrink with clock bump and high OC with a vBIOS lock B$, Founders scam, milking at best, gen 2 pascal anyone ?), they need to step up their game else very very bad future for us consumers.

No idea about Zen vs Intel's monster CPUs on the other hand.

 

As the pricepoint is talked about alot, in Sweden both the cheapest 970 and 480 is ~240€. Cheapest 480 8GB is 270€.

Sidenote: Similarly specced a 1070 desktop is only ~100€ above a 480 desktop. No-brainer for most.

 

Considering they start at around £183 for the 4GB version here that's gouging right there.

 

Food for thought.

 

More from PcPer helping to explain

 

I think this over current problem is probably occurring due to the 480 having the stock voltage set to high.. Seems people are actually getting better performance and not having this issue when doing a small undervolt..

Sent from my LG-H850 using Tapatalk

 

AMD shouldn't expect their customers to undervolt to get the cards in spec. On top of that, the overclocking capabilities of the RX 480 have been touted but any overclocking will further stress the power situation. AMD really made a mess this time.

 

Anandtech just got this: http://www.anandtech.com/show/10465/amd-releases-statement-on-radeon-rx-480-power-consumption

(Bolding and commentary for snarkiness is mine)

I'm being rather hard on them, I know problems happen, they really should have caught this though. Rushing to market to beat the 1060 my guess.

One question, is the software fix going to affect performance?

 

They're probably going to undervolt all the chips to compensate. That should make average boost actually better but of course it changes nothing if the cards are overclocked.

 

Exactly... f..ed up.
All was done perfect... except overvolting 0.025V up. and 25-35% more of hype.

 

The Async compute thing is very important, since AMD's solution is a proper hardware based one. WHen we see actual proper games released built for DX12, I guarantee you that the story will be different. And if we do see widespread Vulkan adoption, I'll bet you my best rendition of El Risitas' laugh that AMD will beat them in that department as well.

Eh...not exactly. They did do some weird trickery with the 390X and 390, so they're not direct rebrands as they do outperform the 290X and 290 respectively. Keep in mind, AMD won every single performance bracket against Maxwell except for the high end. I have to keep reminding people that the 390 kicks the 970's ass in benchmarks. AMD should have been seen as the winner of the Maxwell/Fiji generation, but they weren't because of idiots going around saying "hurr-durr rebrands bad" and Nvidia's enormous marketing budget.

 

If we are talking only performance wise then yeah, but if you take into consideration power consumption, efficiency, and heat generation it is a different story, so even if Fiji had a small perf advantage, Maxwell were arguably the better cards overall (and had considerably more OC headroom), though if you don't look at these things yes, you can say Fiji outperformed Maxwell .

 

You wanna know a secret? As much as you might think you care about the power consumption, you don't. The difference was small enough that it wouldn't cost you more than a few dollars a year. The heat argument is a valid one, but only if you're building a mini-ITX rig. In which case, you should get a Nano anyway (unless it's above your price bracket). But let's face it, mini-ITX rigs make up a very small percentage of gaming computers, so that barely matters. Yeah the Maxwell cards had more OC headroom, but the 390 and 390X were no slouches either in terms of OCing (another thing AMD pulled off with their trickery when 'rebranding' them), and the performance difference between your average 970 at max OC against your average 390 at max OC would probably still favor the 390 (unless you really struck it rich in the silicone lottery). And Fiji did not outperform Maxwell. I'm more talking about the Hawaii rebrands. The Fiji cards that were worth getting were only worth getting because they won by default; the Nano because Nvidia had no competing product, and the Fury because Nvidia had no product in the same price range. The 980 Ti outperformed the Fury X even at base clocks (even though the Fury X will age much better and scaled way better in crossfire, with a Fury X crossfire setup handily beating SLI Titan Xs in about 90% of scenarios), but that was it. AMD won everywhere else. Literally the only reasons I'd ever recommend a Maxwell card to anyone would be:

• If they were building a mini-ITX rig but could not afford a Nano
• If they were limited by their PSU (which happens far less often than you would think)
• If they wanted a high-end card but never planned to go dual-GPU in the future and were going to upgrade next generation (since they would not get to harness the crossfire potential of the Fury X and the fact that the Fury X will age much better would not be relevant to them since they will be upgrading soon anyway, I would recommend the 980 Ti)
• If they really wanted ShadowPlay

AMD is the way to go in any other situation.

 

Well I do. When I switched from laptop to desktop, electricity bill grew noticeably. Imagine if I bought a 290X...

 

That's not a fair comparison. It's not the difference from laptop to desktop. It's the difference between a 390 and a 970. A desktop is typically left on longer and always draws power from one source (your house), whereas your laptop is often put to sleep, charged places that in places that are not your house, and has power modes specifically designed to limit consumption to increase efficiency several times over (where it can get 4-6 hours of extra battery life with clever software).

 

Let's dismantle this, shall we?

Consumers don't care about the difference between 75W and 150W. That's one incandescent light bulb. That's literally pennies on their monthly electric bills.

Consumers don't care how hot their computers run, either. Only people operating server rooms and nerds like us care about heat.

Consumers don't care about overclocking capabilities. They don't overclock their gear.

Each thing you say is wrong about RX 480 is a thing that consumers don't care about. What consumers care about is the best bangs for their bucks when they walk into Best Buy or Microcenter. Regardless of any perceived faults in RX 480, it is unquestionably the best bang for the buck.

 

The one point of contention I might have with that is the one about OCing, as I think a significant percentage of gaming-grade GPU owners do that (since it is very accessible nowadays), but then again, that percentage will be much lower with the $200 GPU market.

 

OC'ing is niche.

Sent from a 128th Legion Stormtrooper 6P

 

yeah unfortunately with overclocking most people just want to plug & play with there stuff or are scared they'll brake it, i most of the people i know with k series CPU's & black editions IRL wont even touch the multiplier, & even fewer oc the gpu. its disappointing really

cost to run wise (@ Full load)
my laptop (speck in my Sig) - 230w - 0.45p p/h would take aprox 9.2 days to cost me £1 (of running at full load constantly)
the GF's Desktop oc'd i7 & 980ti SLi - 840w - 1.66p p/h would take aprox 02.8 days to cost me £1 (of running at full load constantly)

given figures are rounded & estimated slightly (my watt meter died) desktop includes monitor (90w) and estimated on addition wattage drawn for oc's

if power draw was to come down it would be nice but the change would be slight

 

I think it is bold tactics at best to promote and offer new Overclocking Software for the product which is likely to show signs of whole PC's failure due to this.
By signs of failure I mean how PC shuts off for example because its easy to notice but it's not limited by this.
What customers will likely feel is a Bang on the Butt.

 

I rarely put my laptop to sleep, and it ran from wall socket in the house.
So anyway, my M60J consumes about 55W on load, whereas my desktop, what, 250-300W?
So say you have a 6h gaming session (which I did during holidays), 55W*6 vs 250W*6 is quite the difference. I imagine the 390 would have another 100W over 970.
The difference in electricity bill when I switched from lappy to desktop was around 20EUR. For some, mere peanuts, for other, money.

 

100w = £0.002 per hour

also the load wouldnt be constant only when every thing id at full load

 

Yeah, I don't live in UK.

 

My laptop pulls ~350W from the wall which depending on the time of the year costs different rates so it's between 0.0161c per hour to 0.0315c per hour while my desktop runs ~550W or 0.0253c to 0.0495c per hour. Let's say I was gaming 7 hours per 30 days at the maximum rate. My laptop would cost me $6.62 and my desktop would cost me $10.40. Take that out over a year and the laptop is $79.44 and the desktop is $124.80 - it definitely adds up.

 

I recalled that we are discussing mobile Polaris afterall. So the question to those who defend ref RX 480s: Would you defend a laptop which supposed to come with 120W AC and does use all 120W at stock clocks?

Before you answer that question bear in mind that I have MSI GT780 which came with Fermi 570M supported by 150W AC. "Fun" fact was that after ~15% overclock my laptop shutdown itself in the middle of any game... and it didn't matter if i7 or dual i5 was inside! Frustrating wasn't the correct word, pi$$ed off was. Yes, I was overclocking a notebook and perhaps my motherboard became too old too but I don't care and consider it's the GPU's issue for bringing mobo and AC to the limit and above before reaching max stable overclock. And here we have GPU at stock doing this.

 

Don't forget the laws of thermodynamics and the energy consumed also ends up as heat. If you live in a cold country the cost is essentially zero compared to putting out that heat from a heater or your computer. Essentially free!!! gas heating is cheeper tho...

 

that is not a problem with the hardware that is company's that are making laptops not supplying the products with a mains adaptor capable of running the hardware sounds like MSI should of sold the GT780 with a 230w psu

 

oO
GT780 was sold with 180w psu, and you can put a 240w alienware psu without any troubles.

 

You do realize that AMD came out and said that only a few RX 480's were supposedly affected and NOT the entire lineup?
Also, in the cases of affected cards, the issue with the power consumption seems to be what I proposed as a potential problem on here before: too high voltage.
The 480 doesn't need to operate at 1.3V as it appears to do so (hence the high temperatures and above 150W power draw).
By undervolting the GPU to it's lowest sustainable setting at base clocks (1266MhZ boost), power consumption will go down radically and won't present a problem at all... heck, power consumption can still be slashed by about 33W while overclocking the 480 (reference) to 1330 MhZ.
Laptop GPU's usually undergo a lot more aggressive binning methods than desktop ones due to smaller surface area and inferior cooling.

As for your laptop issue... that has more to do with how laptops are designed. They don't really have a lot of thermal headroom, and they are specifically designed to accommodate certain power draws.
Overclocking goes beyond those limits and you are essentially pushing an already constrained system that'd designed to save power (not expand on it).

If you want to play with an overclock in a laptop, it's your call... just as it's your call to do so on a desktop.
Laptops have inferior cooling to desktop computers, and manufacturers have a tendency of cutting corners in laptops.

Plus, it's not the first time that manufacturers do not really automatically change the setup to include a more powerful power brick so it can handle a more demanding GPU... the same issue happens all the time (I'm not saying its justified, I'm just saying that before you go on saying negative things about 480, you should really recheck if you have all the relevant data)

So, I wouldn't be surprised that overclocking could result in the above mentioned issues - which are basically self-inflicted ones by users themselves.

This whole issue about power draw is blown out of proportion.

 

its seems all RX480's do it they seem to be tuned to clock speed wise very well to the wattage the reference cards that come overclocked preform worse then stock until you increase the power slider see TTL's video


Also some laptops OC very well depends how well the cooling has been engineered

 

A valid point often overlooked in these forums. Electrical consumption provides heating to your home with a coefficient of performance (efficiency measure of a heat pump) of exactly 1. This pales in comparison to an actual heat pump in moderate cold weather or combustion-based heating where the equivalent COP can be more than 3 but as a pure by-product I'll take it with a smile.

Having said that, the waste heat makes your AC work harder in the summer so the benefits or lack thereof will vary depending on individual circumstances.

 

You are right, I have 180W AC. Aging as well as laptop itself but still.
Deks
yeah, but untill I read that other cards with the same voltage don't go over PCI-E spec I don't buy this. Fooled by you once - your fault, fooled twice - mine.