Mobile Polaris Discussion

They have done it again; The 980M is such a card when it runs at "normal" 1.065V. You need an unofficial unlocked 1V vbios from Prema to have any chance keeping it anywhere close to a 100W tdp card. The 680M and 7970M were 100W cards and they run way cooler when pushed hard. I know cause I have run them all in the same laptop with same cooling.

 

So AMD is holding "Polaris Tech days" before Computex starting from Saturday! Since the title specifically mentions Polaris, we can hope for a lot of information on the upcoming gpus. Considering Nvidia is apparently not releasing Pascal based mobile gpus soon, I really think this is a golden opportunity for AMD to shine in mobile. From the looks of it, desktop R9 480X is consuming only 130W while having, at least, Fury level of performance, so stuffing that card with negligible (3-5%) loss in performance in mobile shouldn't be a problem.

http://videocardz.com/60316/amd-radeon-gets-a-new-logo

 

Or properly bin the thing and stick it on MXM unaltered

 

Power consumption has gotten out of hand with nVidia. 680M was 100W like you said, 780M was 122W, 980M when pushed to that unlocked vBIOS on 1.065V is around 130W TDP. Too much. I want an efficient board, 100W should be the limit for reference design (I feel at least)

Anything more is a ***** to cool in a laptop, and this is coming from a M18x/AW18 with beefy heatsinks and high CFM fans.

 

I agree! My old machine cant even handle the 980M at 1.065 at an OC at full load. Shuts off. Power delivery was never designed for cards needing so much pwr.

I am really hoping they can return to the 100w standard of before. Would be nice if 1070M would be 70W too.

Sent from my SM-A500FU using Tapatalk

 

Speaking of cooling in laptops... we're talking about an outdated design method of cooling for the sole reason that OEM's are lazy/greedy to the point where they don't want to bother innovating in the area.
Whatever happened to using superior synthetic materials that would achieve same or better cooling via passive methods, and for instance water cooling in laptops?
Oh right... they don't want to spend any money on that.

Btw... on Polaris 10 performance... some numbers are a tad... iffy.

Polaris 10 was able to achieve in the following leaked benchmarks Fury (non X) level performance:
http://www.tweaktown.com/news/52253...0-cf-beats-titan-leaked-benchmarks/index.html

However, in some DX12 titles, we see that Fury (non X) already beats Titan X in DX12, such as here:
http://en.yibada.com/articles/12662...-unveiled-computex-2016-crossfire-version.htm

Also, AMD demonstrated Polaris 10 running Hitman Pro at 1440p in DX12 at highest settings... achieving 60 FPS (locked, so we don't know if this is it's fullest potential) and surpassing FuryX/980Ti at same settings.

Even if the benchmarks turn out accurate (I somewhat doubt it because most of these benches are done with programs incapable of properly reading new hardware, not to mention that testing is done on units that are for the most part intentionally disabled/crippled to skew the results and not get accurate representation prior to release - and at 14nm, Polaris 10 should be able to clock to 1400 MhZ - and that benchmarks are not indicative of real world performance), it seems like Polaris 10 might be sitting right in between 1070 and 1080 as far as DX12 goes.
And if the talks about reworked hardware for better performance under DX11 turn out accurate, then there's performance gains there too.

So... under which circumstances could Polaris 10 CF not beat 1080?
Was 3dMark 11 also properly modified so it can take advantage of Polaris (familiar enough with it - because Pascal is for the most part shrunk down Maxwell that's relatively improved in compute).

This is what doesn't make sense.
One set of leaked benches said that Polaris 10 achieves Fury (non X) level performance which already beats TitanX comfortably in some DX12 titles (which would make it as fast or faster than 1070) and with a small difference in FPS between it and 1080.

Are we seeing an incorrect picture here or mere wishful thinking?

 

Ok, so seems like AMD will announce Polaris during Computex, NDA for reviews will end on 29th of June (so damn late!!!!!!) and hard launch even later....... I was so much hoping to get something new before my so long awaited game is released (No Man's Sky).

http://videocardz.com/60373/amd-polaris-tech-day-nda-ends-on-june-29th

 

We have been patient until now. Shall we see how this tech event pans out before we go out and panic buy ngreedia cards!?

So many people upset saying they won't wait a month and just buy the 1070M looking at the comments from that link. Amazing that the world is full of cretins!

No we wait and compare the products and choose wisely no matter the wait. 980M already nearly 2 years old we can wait a month or 2 more

 

Amd did say they wil have both polaris 11 and 10 available for purchase on desktops and laptops before back to school season. So if availability happens early July, they would be on track, though really cutting it close.

I am also dissatisfied that nothing new will appear in June though.

 

All the first 1080/1070 cards are designed to make you want a newer better 1080/1070 card later, built in obsolescence.

The only 1080 worth getting long term is a water-cooling, dual 8-pin powered, totally OC'able card.

I don't think that's coming out in the first month

If it is, then wowsa, that's great, but even then... the big GPU is still coming, fit that into the same OC'ing framework, and you will want to drop the first wowsa for the second.

Even if you don't want to wait for Polaris/Vega, maybe do it to see how the first 1080/1070 cards OC and perform?

It's gonna be fun

 

At this point, forget Polaris. Vega is where the big cards are gonna be shown, and I'm hoping they get passed to mobile space too.

That being said, if there will be a high performance Polaris MXM, you bet I will be the first to get them and stuff them in the M18x R1 and R2

 

I will be closely behind you doing this for the M15x. Yes it gets the 8th generation (can you believe that!!) GPU since release but this time not top end I hope to cut down power requirements

..and unlike the M18x it is likely to work plug and play

 

I might just break and buy a top end laptop with best polaris 10 part to replace my ageing acer in signature . Though, I might end up waiting for mobile zen and possibly mobile vega. Apparently AMD leaked some impressive performance gains for Zen , and given HSA/huma amd dx12 readdiness in amd hardware , I might want to get an a AMD laptop instead.

 

Here ya go.

https://www.ekwb.com/news/ek-unveils-new-nvidia-geforce-gtx-1080-water-blocks/

I'm waiting for Polaris as well. I'm tired of paying nVidia prices when AMD comes close to the performance at a lower bracket most of the time. My 780 Ti cost me almost 700 bucks just to have Maxwell come crush it 5 months later.

 

We fixed the M18x R2 once, we can fix it again lol.

 

I think polaris will br worth the wait. And if amd makes more optimisations and mods to both polaris 11 and 10 in the meantime, all the better .

Finally, I would not take those leaked benches at face value. Given what I wrote in my previous post, I think we might not be getting a proper picture (especially with NDA still in effect)

 

At least tomorrow we'll get to know what (mobile) parts they have planned... or so I hope. It's hard to believe that they'll keep the performance behind the curtain until (I would guess) the very launch of the GPUs/laptops(hopefully), but it's not that of a surprise. They'll hold the valuable (again hopefully) advantage (or parity if you wish) as long as possible. After all is there even a rumor about competing mobile parts? I mean, I know that we all take them for granted, but with recent news about new Maxwell machines is there any chance that mobile Pascal would hit the market before Polaris? I doubt that to be honest, and quite likely there will be a quarter difference.

 

I'm definitely not jumping to Nvidia anytime soon with their ridiculously overpriced cards. When I first saw those prices for mobile 980 it made me laugh so much, but I almost started crying when I realized a lot of people are actually buying machines with a single 980 for >3000$, when they can get 980 Ti in SLI for much cheaper price. I'm not AMD fan or something, but earning money in my country is one hell of a task, so literally each cent counts for me and AMD provided me with the best price-performance ratio when I bought my clevo 170em with 7970m for like 800 bucks back in 2013. Let's be patient and wait what AMD has to offer!

 

I've heard 4Q for Pascal, 3Q for AMD but who knows, these things change. Nothing is ever ready on time anymore and if it is, it was rushed and comes out broken.

 

Depending on which country you are in... I can assure you that even here in the UK, earning money is not a simple task, given that majority of one's income goes on basic necessities of life (which realistically no one should be paying for in this day and age of technological automation and abundance - at this point, its all artificially induced scarcity, and has been for decades, controlling access to goods and services via 'money' that is a simple piece of paper that has no 'real value') and that most jobs are paid minimum wage (not living wage), while on another end of the spectrum you have moderate increase in wages that require expertise and qualifications that even most college or university graduates cannot produce (that's not even going to the 'high end' - the entire scheme has become ridiculous).

Granted, if you don't drink, smoke, drive, or eat out (but you can still go out for walks, runs, and talk with people for instance which doesn't necessarily cost anything), you can easily save up decent amount of money over the longer term (months or a year)... but many people are swayed into the consumption mentality, so they spend a lot of cash on things they don't really need.
I just spend on basic necessities, and shove everything else into savings which I rarely tap into for something that I might need. Incidentally, I was able to save up money like that on a part time job no less... but I have it a bit easier what with sharing the flat and some expenses with someone.

What I find problematic with new technology is that a lot of it is severely overpriced, even though there is no justifiable reason for it to be.
Manufacturing costs have gone down to almost nothing, and yet this has NOT translated to lower prices for consumers.
The older or weaker tech gets showed into the lower pricing bracket eventually, but the more powerful ones still remain relatively expensive.

Plus, let's not forget the shoddiness of products themselves and built in planned obsolescence which guarantees that your product will almost certainly break down right after the warranty expires. Quality is not assured even if you are buying a 'top end' product.
People have stated on more than one occasion that 'you get what you pay for', as if they are indicating that higher price means better quality... but this is not the case because I had numerous 'cheap' products being of higher quality and even modularity that outlasted various overpriced hardware.

AMD however seems to offer cheaper products at same or comparable performance levels.
But as I said before, I doubt that Polaris 10 will be under-powered or not worth the wait. Given all the changes Polaris architecture has, along with the advantage of being on a 14nm process, it stands a good chance of equalizing things with Nvidia easily enough in the mobile segment (in the desktop, it doesn't need to compete with 1080... why would it when the difference between Fury (non X) and 1080 in DX12 is only a few FPS... and if Polaris 10 achieves AT LEAST Fury (non X) level of performance according to that 3dMark 11 benchmark [which is NOT to be taken literally as its probably not suited towards taking advantage of Polaris that well which was also likely running underclocked - and let's not forget that 3dMark11 doesn't even test DX12 performance], it is already very efficient and plenty fast, not to mention cheaper.

 

Well, now that No Man's Sky is delayed to August 9, even more reasons for me to be patient and save up some money in the meantime Personally for me, it would suffice even if mobile Polaris is just a bit faster than 980M, since more important for me is the price. If I can get 980M level of performance for $1K, that's fine, if I can get Fury (non X) for that amount, then it's just awesome, because I'm not living in US and additional 250-300 USD would be needed for shipping and customs.

Btw, any news from Polaris Tech days?

 

I don't think AMD would allow Polairs to be 'just a bit faster' than 980M, because 980M is an ageing card on an old manuf. process, and they are probably aware that Nvidia will not sit on their rear ends when it comes to releasing mobile Pascal cards.
Plus it would downright criminal to NOT take advantage of 14nm as much as possible and clock the GPU's as high as they can go without raising TDP.

Given what's happened with Pascal, it would seem there have been 0 news on it and then 'suddenly' (or so it seemed), Nvidia made a paper launch of 1070 and 1080 (reportedly 'high end cards' - even though they don't really fit that bracket).
So, there's a worrying possibility that Nvidia does have mobile Pascal cards ready (or in the works) to be launched as well... we just know nothing about them (like we knew next to nothing about desktop Pascal).

I still don't think Nvidia will necessarily have anything in terms of mobile (least of all mid-range) Pascal cards ready for release before AMD releases Polaris.
In this sense, I think that particular market segment is 'temporarily safe' for AMD to reclaim TAM that Raja spoke about... though in all honesty, it would be encouraging for them to release Polaris a bit faster (maybe mid-end of June as opposed to early/mid July).

On a final note... if releasing Polaris by early/mid July means AMD might make further improvements to all Polaris gpu's, then so be it.
I'd rather they release quality product than rush it out the door.

 

I don't know if anyone mentioned this because this is in opened source since 17-th of May.
R9 M485X has 256 bit memory interface and up to 8GB.
http://www.amd.com/en-us/products/graphics/notebook/r9-m200#

 

the m485x is a variant of the m395x

hopefully they will put out a good new mxm & desktop cards not just re brands i would like to see them give nvida a run for there money / just straight up better, the GPU market needs some competition, nvida's strangle hold on the market is worrying

 

Finally something from Polaris Tech Days! AMD was obviously demoing R9 480 or 480X (it was shown as RX 480) running Doom at 1440p on ultra. Though fps was not shown, we can safely assume it was not below 60 or there wouldn't be any reason to demo it. I'm not sure if Doom is Nvidia optimized but according to the benchmarks to achieve 60 fps at 1440p it would require 390X or 970 (whereas normally 390X is rivaling 980). Now it all depends whether it was 480 or 480X, because if latter, it seems somehow underwhelming.

http://videocardz.com/60561/amd-shows-off-radeon-rx-480-running-doom-at-1440p144hz

 

Maybe RX = R10 and this is just the R10 480. Everyone seems to be replacing the 10 with X these days. Last gen was R9, would make sense to have this one R10.

I agree if its 480X its kind of underwhelming but if its priced between 250-300, that's strong competition for the 1070 no matter what its called. We don't know the actual framerate it was running either. I mean they decided to upscale it to 1440p on a 1080p 144hz monitor so its got to be over 60. Wouldn't make any sense otherwise.

 

Nope. AMD just has garbage for OpenGL drivers.

 

And if it can do this with garbage drivers, ask yourself what if can do with in other games with proper ones

Sent from my LG-H850 using Tapatalk

 

1440p 144Hz without no FPS counter... using virtual super resolution on a 1080p screen (if actually running at 1440p). Using Freesync, no less.

What I think is they couldn't even hit 144fps nearly constantly; something 980Tis have been reported to do in that game at that resolution while maxed out. I don't trust it. And I'm quite frankly annoyed that AMD who is in no position to pull any kind of stunts, would do some crap like that.

 

Indeed, these benchmarks do not really impress in desktop space, but in mobile this could very well be a game changer, because, for instance, 980M can barely sustain 50 fps even at 1080p (according to notebookcheck). At least AMD has always been saying they want to target mainstream segment first and these results are not that bad. Pricing will be crucial and of course whether they bring this level of performance to mobile.

 

In what? DOOM? That's just plain untrue, people with 970s clear well over 100fps with ease in Doom; 980Ms cannot be getting 50fps. Did they use some old broken drivers or something?

 

Here it is:
http://www.notebookcheck.com/Doom-2016-Notebook-und-Desktop-Benchmarks.165882.0.html

Though there is no 970, but 980 is getting about 97 fps at 1080p, so there might have been driver problems with 980M, since that is too big of a difference...

 

I only played the free multiplayer demo of doom but the 980M was certainly pushing easily 60fps+ and that tethered to an old nehelem chip so newer laptops with the 980M should fare even better.

Sent from my SM-A500FU using Tapatalk

 

Let's wait and see before passing judgements on rumours.. There is only a few more days till they announce it at least, well 2 days..

 

I've seen 980s get FAR more than 97fps. That is weird as all moo.

I suppose. I truly hope Polaris 10 will do better than we thought, but still.

 

Expect less performance or up to a 1070, for less money. That's going to be the highest priced polaris. These new GPUs are underwhelming, to say the least.

 

for midrange mainstream gpus 1070 performance would be fine, especially at a low price point

Sent from my Huawei Mate 8 NXT-AL10

 

It's almost June 1st, the date a few sites said Pascal would launch.
Hopefully we get some new info soon, I remember when Bristol Ridge was Pre-announced for those HP products a while back they said shipping early June.

 

I think that actual launch date will be when/after the nda expires (29th june).

 

RX 480 specs leaked:
http://videocardz.com/60752/amd-radeon-rx-480-specifications-leaked

Also, quoting WhyCry:
WhyCry
BTW I heard AMD demonstrated RX 480 in CF beating GTX 1080 at Polaris Tech Day

 

Specs seem misleading.
All of the demo's to this point of Polaris 10 indicate at least Fury (non X) level of performance in DX12 (actually, they point to faster than Fury X level performance in DX12 and 1440p).
DX11 might be a different story with some performance enhancements due to primitive discard accelerator incorporation, but overall, the Teraflops alone don't give us a full story, especially because we have relatively little clue how much faster the new architecture is clock per clock, and what the final clocks (or at which clocks the demo's ran) will be.

Still need to wait until Polaris 10 NDA is lifted and the cards themselves properly tested.
These websites and people indicating 970 and 980 level performance are not getting the whole picture.

The 1080 is barely 10% faster than FuryX in DX12... what happens if Polaris 10 ends up only 10% to 20% less in DX12 compared to 1080 at 1440p? Although, Polaris 10 also has 8GB of GDDR5, so it might far similarly to FuryX at 4K even due to higher VRAM (though Fury has a massive bandwidth advantage)

 

How is this supposed to be an accomplishment. If it takes 2 cards to beat 1 single card, that's losing terribly in my book.

Also that R9 M485X...rebadge of the M395X, which is a rebadge of the R9 M295X, which is a cut down from a R9 280X core from 2014.

FFS AMD, get your crap together.

 

Actually the R9-M295X is the full Tonga (the supposed 285X) but I think there never was 285X, anyway, not cut-down chip. There was a 380X though, which is this very chip (full Tonga). Rebranding aside, I think a few pages ago there was a graph that 2x P10 were unable to beat 1080, so I guess this comes just to disperse this rumor, and no one said by how much it is over it. The true performance is still a big question mark.

And something that most of you probably don't care, but I AM MORE THAN HAPPY to see once again (since the very first Polaris anouncement) - HDR Ready When was the last time where most displays had to catch-up to graphics, it's usually the other way around. Why all the numbers if you can't enjoy the colors and dynamic range?

 

http://videocardz.com/60773/amd-radeon-rx-480-to-cost-199-usd

If this comes true, AMD have broken the midrange market and owned it completely!

 

I don't see how they are making any money pricing so low. 250 would make more sense.

 

Maybe. But they undercutting nvidia. If they capitalise on mid range. They can ask sell the 490 bit more. At least 980 performance pro early cheaper than a 980. All those that are waiting for 980s might just get that instead. Newer tech. Amd claws some money back.
And wouldn't sell at a loss. When you go two a shop and buy any special offer its still a profit. If amd sold this at a loss. It's because it's an introductory price. Or they want rid of stick because they failed and want to release better stuff sooner.
Nobody complains buying stuff cheaper. You almost make it sound like you would be too guilty to gutsy have card for cheap lol. I love cheap. Not tat cheap but just cheap prices. Unfortunately there's no real cheap laptop with great performance unless it's performance is out of date lol.


Sent from my iPhone using Tapatalk

 

Keep in mind Polaris 10 is expected to compete more with GTX 1060 (which should have performance between GTX 970 and 980) than with GTX 1070 or 1080 (performance of GTX 980Ti and more).

At least that's what AMD kept saying - it'll definitely fulfill min VR specs, but no more promises beyond that. Everything else is just speculations.

GTX 960 was exactly in this price range, so GTX 1060 should have similar price. AMD has to price Polaris that low to stay competitive (e.g. $200 vs maybe $250).

 

Isn't the m295x Tonga based? It's just a heavily lower-clocked version of the R9 380x.
Doesn't mean much, though.

 

Computex Web cast time

http://www.amd.com/en-us/who-we-are/corporate-information/events/computex

 

AMD Computex 2016
Previously Live, recorded playback, Starts @ 15:57