Mobile Polaris Discussion

At 35W 100% It's BGA, and P11, damn low end.

 

Polaris 10 was also said by AMD it will be available in high-end gaming notebooks.
Polaris 11 is entry level to mid-range.
7970M performance at 35W for Polaris 11 is nothing to sneeze at.
It's rather impressive when you think about it.

Just imagine what Polaris 10 could achieve at say 85W to 100W (we won't know until we get more concrete numbers though).

 

Nope but I will pave the way for more faster mobile versions, let's say a 70watt card that beats the 980m.

I have to agree that I'm not interested in the m480 (even if they had a mxm version, it's way to slow for next gen), but again it's performance per watt is pretty impressive.

 

Indeed... Hopefully there is a higher end version

Well if they do make a 100W part, that will achieve something... That's a massive if....

 

They can't get a higher end part out of P11 or they would have done it so now we wait to see if they put out a P10 part or not.

 

So about GTX 960M performance but using just 35 watts instead of 60 watts [1].



That actually is pretty impressive. It means ok-ish gaming (1080p medium quality) on super lightweight devices, ones that so far could only get 940M (ultrabooks like Asus Zenbook, hybrids like MS Surface).

------

On the other hand, Polaris 10 in RX 470 seems to have less gains vs Nvidia's mobile Maxwell - performance similar to GTX 980M, but using 110 watts instead of 122 watts [3].



------

[1] http://www.notebookcheck.net/NVIDIA-GeForce-GTX-960M.138006.0.html
[2] http://www.notebookcheck.net/NVIDIA-GeForce-GTX-980M.126692.0.html
[3] https://tweakers.net/nieuws/112519/...-van-rx-480-en-rx-470-en-onthult-rx-480m.html

 

Polaris 11 will be used in new MacBooks though (or at least, Apple will use it en-mass for their thin form factors - at of course almost double the cost of an equivalent Windows laptop).


I'd like to wait and see what Polaris 10 in mobile form will do (although, it does make sense that Polaris 11 would be showcased first).
Plus, those early benchmarks shouldn't be taken at face value since they aren't indicative of real world performance. DX12 performance is something completely different... but I guess we will wait and see - and I am slightly disappointed that a 110W version of Polaris 10 is only supposedly on the level of 980M in a synthetic benchmark (which doesn't make sense)

This begs the question what the heck will Nvidia do with a mobile top end solution?

 

nVidia is doing a cut GP104 for their top end. At least for now.

 

nVidia say: they're doing away with the distinction between desktop and mobile graphics processors.

Reality: http://wccftech.com/geforce-gtx-1080-gpu-for-notebooks-specifications-leaked/

 

dutch website broke nda and have result for rx 480m

https://tweakers.net/nieuws/112519/...-van-rx-480-en-rx-470-en-onthult-rx-480m.html

I expected more of that mobile card though, wonder if it's mxm or not

 

http://wccftech.com/amd-rx-480-faster-than-nano-980/
Another benchmark of the RX 480 (4GB and 8 GB).

 

That's pretty cool, still only one benchmark, no game fps comparisons, but we are getting closer.

It's interesting to read the comments on that site. My favorite so far:

Spoiler

Close 2nd:

Not Completely Nvidia Biased • 3 minutes ago
"Quick everyone buy a gfx card weaker then xbox scorpio!!!! Quick!!!!"

That graphic for the Xbox Scorpio showing the cooler really has the "AMD Cooler" look to it

Spoiler

 

I was just browsing through Alienware's desktop portion of their website only to find this little option (that isn't available yet?).

What kind of OEM card would the R9 470X be? A rebranded RX 470 or a rebranded R9 380(X)?

 

Just saw this commented on in the recent wcf article:

Rx 480 Steam VR score: 6.3
GTX 970 Steam VR score: 6.9
R9 390 Steam VR score: 7.1
GTX 980 Steam VR score: 8.9

Spoiler

Can that be correct?

The RX480 is 10% slower than the GTX 970m, which is then 10% below the VR minimum performance requirement?

Spoiler: It looks like it is...

 

Someone mentioned how that Steam VR test was done with an older driver - hence the lower performance.
The actual performance would be in the ballpark of around 9 or a bit higher, since it basically delivers Nano performance at 100W
http://wccftech.com/amd-rx-480-faster-than-nano-980/

I think driver improvements alone might easily push the card in the area of FuryX on those benches, while real life performance could end up higher.

It's a 100W card after all. If AMD releases this version on laptops (which they could easily do, and then also overclocks it while staying in that TDP), I think we can surpass Fury X.

 

That article from wccftech also mentions that in actual games the performance is even higher, plus some cards from the partners will feature the so-called "beast mode", i.e. heavy OC with even more performance increase.

However, I am becoming really skeptical about its mobile form. I mean if RX 480M is at 7970M performance, RX 485M is at M390X (Tonga/970M) performance, then RX 490M won't probably be that much more powerful. Probably, the latter chip will be based on a desktop RX 470 with about 980M performance. I would be absolutely disappointed in this case, because it is SO easy to use the full P10 chip in mobile, considering it consumes only 100W at gaming loads. I will probably stick with my old 7970M before something worthwhile comes from the red team or switch to 1070(M) in case it is properly priced.

 

And I still think that they play games. AMD were targeting affordable VR from get go, not that I care, but they should cover the VR specs at the very least and I think that the actual performance will be around 980Ti. What makes me wonder is whether the supposed $300 GPU is AMD or partner one. As per the article, it says that partner GPUs would go up to $300. What if AMD has a reference GPU at this price point and partners at higher price? The new Scorpio chip is supposed to feature 40CUs, and since it is APU, it's not like it's cannibalizing desktop/mobile chips. What I mean is, hopefully we can expect 40CU chip. As to what we'll get in laptops. The 35W is awesome, but let's not forget that it most likely scales to 50W, this is within budget of MXM-A and would certainly be great for 15" Precisions and ZBooks, and of course all the soldered iterations. The current P10 with supposed TDP of ~100W is perfect for MXM-B, but I still think that it would be Alienware/iMac/soldered only. I hope that I'll be proven wrong, since I REALLY want a P10 MXM, but there's nothing from MXM-B makers, well (hopefully) except the 17" Precisions and ZBooks, at least it's still an option. The 40CUs, if it ever hits a separate chip (not APU) would be desktop exclusive. That's my guess on the matter so far.

 

I've looked at t he whole set of AMD's slides and they spent quite a bit of time discussing the mobile GPUs. I suppose that means they're getting ready to let them loose? It'll be a bit confusing having laptops with a 75W 470, 35W 480 and 100W 485, but OK, people will just have to look at what they're getting and they're probably not going in the same type of laptop anyway.

 

100W and the performance in benchmarks is a little lower than Fury X! Properly binned and lets say running with lower boost (at stock) and they could have a 80W mobile card here that runs as fast as a stock 980 desktop card before even OCing it! Fantastic.

 

Why cannibalize it's performance?
At around 100 to 105W power draw, they could easily slap non underclocked Polaris 10 RX 480 into a laptop.
Properly binned, the power requirements would probably go down further.
Then they can further try to undervolt and overclock the gpu as much as it can go and probably end up delivering Fury X level performance at around 100 W.

The desktop version probably has 45W headroom for potential overclocking capabilities (which apparently AMD mentioned).
So, theoretically, if a desktop 480 oc's well, it could probably deliver and surpass 1070 in benchmarks - however, we also don't know how much driver enhancements will increase the GPU's overall performance in not just synthetic benchmarks (which is what these mostly are), but also gaming performance (which already seems to be higher than Fury X and 980Ti in 1440p at DX12).

 

cannibalize its performance so that the m15x doesnt suffer from thermal problems with the new card

Sent from my Huawei Mate 8 NXT-AL10

 

Indeed... while putting a 180W Nvidia GPU in a laptop = no problem.

 

https://tweakers.net/nieuws/112519/...-van-rx-480-en-rx-470-en-onthult-rx-480m.html

 

Something is off with those specs.
The 470 is said to draw 110W, but that's not exactly realistic considering that 480 is supposed to be in this range of power draw (the 150W is just what the interface can provide at maximum, AMD intentionally said that 480 will consume much less power than that and there have been talks about highly overclocked versions of Polaris 10).

In that sense, we also need to keep in mind that old drivers might have been used for those benchmarks, and that benchmarks are not indicative of gaming performance.

The 480 appears to be in the range of Fury and close to Fury X actually.
I guess we will have to wait and see.

 

http://wccftech.com/amd-rx-480-1500mhz-overclocking-tool-voltage-control/

http://wccftech.com/amd-rx-480-faster-than-nano-980/

With stock clocks of 1266mhz it consumes 100w so if they get that on an MXM I'll be super stoked... also wccftech is saying its super overclockable 1500mhz. With stock performance close to nano/980 OC performance MIGHT be closer to 980ti 1070 stock which is pretty exiting for a 200-300 dollar card. Please AMD make a 100w Crossfireable MXM form factor without cut down cores that will work in an alienware 18........ My fingers and toes are crossed.

 

Good news on RX 480 OC potential and a new tool for OC and even voltage control:
http://wccftech.com/amd-rx-480-1500mhz-overclocking-tool-voltage-control/

Not included in the article, most probably fake, but OCed RX 480 should perform close to that:

 

lol I stole your thunder by one minute XD

 

So now, we at least know what Lisa was implying when saying cards will cost up to $300. We hoped for RX 490, but it will just be "beastly" clocked 1500mhz+ variants of RX 480. Well, if it brings performance of 1070 for 250-300 USD, then I'm all for it.

 

this is very good news i already thought to myself that a healthy 20% OC should indeed be possible, bringing polaris to very respectable 980 Ti / Titan X levels

edit: whoa, WAIT A SECOND! that overclocking / voltage tool is actually by AMD THEMSELVES?!?! holy benchmark!

 

Niiiice!!! There IS hope at the end of the (AMD)tunnel!

 

RX 460 3DMark11 slightly below GTX 950:

http://wccftech.com/radeon-rx-460-c...utm_campaign=Feed:+Wccftechcom+(WCCFtech.com)

 

Awesome news, it's just that I can't let myself be all excited just yet (for the eventual mobile/MXMs that is). Did any of you took a second glimpse on the cooler? It's the same as the reference R9-380 (190W)*. Go check it out. I guess the thermal density of these 14nm chips doesn't play nice with current cooling technology, so how they would perform in the confines of a laptop is even bigger question. We might not see 100W mobile GPUs after all. At least not with current cooling technology. ASUS' 3D vapor chamber (there was a less fancy name in some old papers, but I forgot it) might be the way to go.

*Can't find a single review with this cooler, in order to make a guess where things are heading.

 

http://wccftech.com/radeon-rx-480-reference-model-leak/
Reference models for the RX 480 look to be coming to the market, which is different than the aftermarket coolers-only for the R9 380(x).

 

I wonder why, maybe the more expensive (higher clockable) versions would come with some bells and whistles cooling solutions (would they be the new X models, like same in every way with the non-X (clocks and stuff), bit different coolers and 8pin ). Of course this one is supposed to cool 190W (although 28nm), so it would be interesting to see what temps and clocks one can pull out of it, but again there's no base to compare it to (lack of R9-380 reviews with this cooler).

 

From what i read in this article, seems like 480m is a polaris 11/R470. Cold shower for me -_-

 

Same here for now.. Not seen any Polaris 10 mobile GPUs... Looks like AMD will be a let down again..

Sent from my LG-H850 using Tapatalk

 

Yes and no.

The top end sucks unless there mobile variants of P10, but I'd imagine two 480m in crossfiire could potentially be the next 750m SLI ~$1100 bang4buck gaming laptop.

 

There will definitely be P10 based mobile card, the question is whether it will be based on 470 or 480 and how much will it be gimped. Considering Nvidia 1080m and 1070m cards will be of titan x and 980 performance respectively, I really doubt AMD can counter that anyhow.

So far the lineup looks like this:
Rx 480m === 965m performance
Rx 485m === 970m performance
Rx 490m === 980m performance

And if we're lucky there will be Rx 495m with desktop 980 (1070m) performance. If we're SUPER lucky.

 

Correction: rx 470 is polaris 10, so 480m is based on a pathetic desktop 460.

 

So what laptops can we expect with the mobile Polaris cause I want to atleast play aots on the laptop?

 

Correct. However, keep in mind that the TDP of the 965M is 60W whereas the RX480M is only 35W. Thus, the latter is actually a substantial improvement coupled with a rather strange change in naming convention. It'll be interesting to see if Nvidia follows suit -- otherwise we'll be comparing a 35W RX480M to a 100W 1080M.

 

Just to know how you see your table, specifically based on what chips your 485m and 490m are if they equal to 970m and 980m performance respectively?

rx 490m if exists should be P10.

 

I've been reading the 490 and 495 are Polaris 10..are you guys just speculating they're gonna be 11 and/or 10 or has anything been confirmed as of now??

 

485M is M485X, which is a rebrand of M390X based on Tonga. It was announced more than a month ago.
490M is a speculation, but in regard to gradation in performance, it cannot be much faster than 485, so I objectively expect it to be at about 980M performance.
Above chips exist or have a high probability to be. Now 495 is not even a speculation, but a guess. It has never been mentioned, but if it exists, it should be above 980M performance, at best perform like 980.

480M was confirmed to be based on Polaris 11 (desktop RX 460). Everything below that is confirmed to be rebrands. Polaris 10 mobile chips are yet to be seen.

 

If AMD is shooting for VR specs (and if they are aiming to bring it to laptops as well)...the 980m level of performance simply won't do.
Besides, benchmarks are not representative of real world performance... plus, every so-called 'source' is shoddy at best that came up with very low (sub 980) and relatively high (Fury X) level of performance... and we also need to keep in mind that drivers for Polaris are far from mature.

Also, AMD knows nvidia will release new Pascal for mobile too... so... either the Polaris mobile line is not supposed to go up against those chips because they have something else up their sleve, or something else.

 

As far as I understood, AMD is only shooting for VR specs on desktops. Their marketing since a while ago was heavily pushing only 2 things for Polaris:

1) cheap VR min-specs for desktop
2) "console-level" performance for notebooks with low TDP (so something that would fit in slim notebooks)

Re 1): RX 480 is roughly GTX 980 but for just $199, so exactly what they promised.
Re 2): console-level performance is ~GTX 960M, so RX 480M at 35 watts is right about there.

Everything else beyond that was just people projecting their wishes but AMD was not promising more.

 

Are we talking N16E-GS or N16E-GR for the GTX 965M?

 

Former, but the difference is like 10% between them, so doesn't make much difference.

 

But doesn't that strike you as odd?
That AMD's latest GPU at 100W is no better than 125W 980M?
This strikes me as very odd when you compare the performance pew watt ratios - meaning that the 470 will likely not be using 100W to begin with... it will probably be in the ballpark of about 70-75W.
Also keep in mind that 480 was said that it will NOT consume 150W at maximum (at best, it would be roughly 100 to 110W)... 150W is the amount of power the interface itself can supply... hence the mentioning of overclocked cards (which would realistically need the headroom).

As we already saw, benchmarks do not tell the whole story. And all reviews have 480 performance fluctuating across the board (in synthetic benchmarks no less).
Plus, if 480 is drawing 105 to 110 W at maximum, how can the 470's 100W be 980M performance when 480 outclasses it in the demo's we've seen by using about 10 W more (surpassing the level of FuryX and 980Ti by a marginal amount at DX12)? Granted, AMD is far better optimized for DX12 (which is one of the reasons I don't buy into the benchmarks and 'around 980 performance' considering that most new games are coming out with DX12 support anyway (and the older games will probably run more than adequate on 480 - especially if the inclusion of primitive discard accelerator brings performance enhancements (for which it is likely that driver updates will be needed).

So... we might not be seeing the whole picture (emphasis on 'might').
We know relatively little of the mobile lineup when it comes to Polaris, so we don't know what kind of performance the top end mobile will have (it might very well be the desktop 480 in it's full form - while the overclocked models that can make use of the rest of the thermal headroom [up to 150W] will likely be left for desktop - which realistically would be pretty good).
The 480M looks to be Polaris 11... but Polaris 10 is coming to notebooks, although we know very little of it's top end performance in this segment - the 470 might end up in mobile, but at around 75W TDP, and the actual 'high-end polaris 10' might be a desktop grade/stock 480 at 1266MhZ (which if the benchmarks are actually accurate would meant 980 level performance in DX11, or possibly more over time, and FuryX/980Ti level performance in DX12 under specific resolutions and high settings - 1080p might result in 60FPS, while 2k resolution would yield about 35 to 40 FPS - more than playable if you ask me).

Of course, I have no idea what the high-end Polaris 10 line-up will be and the above are all hypothetical best case scenarios, but if AMD will provide a 100-110W mobile GPU, then that could be the full fledged Polaris 10 from desktop at stock clocks), while the 470 would be the bridge between the low and mid range - akin to what 970m is for Nvidia).

 

The HD7870 was a 175W TDP (actual around 125W) card that we saw a full implementation in a 100W form factor with very similar performance.