Mobile Polaris Discussion

3DMark11 Performance Preset

I dont know the name of this GPU but have seen benchmarks. Could be R9 M490 too for all I know.
This might be the specs for one upcoming Polaris 10 GPU considering the clock is only 800MHz
http://ranker.sisoftware.net/show_d...a09dac8ae2dfeaccb489b89efb9ea393b5c6fbc2&l=en

 

Now that certainly looks promising.
If AMD clocks that at 1000 mhz though (as opposed to 800 - which seems like an engineering sample and not a finished product), the score could jump up nicely - but I'd be more interested to see in-game performance (as opposed to benchmark scores).

 

There are benchmarks of one Polaris chip running at ~800MHz and the same one at ~1200MHz. I would guess the one at 800MHz is mobile although it could still be testing of ES chips like you say of course.

 

3DMark11 score of 13000-14000 for Polaris 10 seems a bit low since that's 380X territory but if it's a low clocked ES and a cut down variant at that with early drivers it's possible. Or it might just be a highly clocked Polaris 11. In any case the difference between clocks of desktop Polaris and a mobile one should only be about 10-15% this time around.

 

Time for some updated news on Polaris 10:
http://www.itechpost.com/articles/1...is-10-provides-impressive-gpu-performance.htm

It would appear some websites are claiming that Polaris 10 will be performing on par with R9 390/390X.

If the artificial benchmarks are any indication, then the scores themselves (as far as FutureMark/3dMark11 goes) will be close to FuryX... however, actual gaming performance is something else (seeing how it would appear that a demo-ed Polaris 10 surpassed FuryX in Hitman Pro at DX12 - granted at a specific resolution, but still...).

Also, some posters online stated something about Polaris having revamped hardware that will enhance DX11 performance (because the older GCN architecture wasn't particularly well oriented towards it - or so the explanation went if I'm not mistaken - something to do with schedulers?). Is there anything to substantiate these claims?

All in all, when one takes into account the transition to 14nm, the ability to operate at higher clocks without as many instabilities, etc... could we see 980ti level of performance (or close to it) in actual games?
Pro software is another story, though I wouldn't mind AMD's computational power being more used in Autodesk programs such as 3d Studio Max.

AMD needs to give us an uncut Polaris 10 in mobile space. At those performance levels it would essentially surpass 980M by a good margin while being in the same TDP range, and would effectively offer desktop 980 and beyond (most likely beyond in actual games and pro software) at a lower TDP (and if 1080m offers mostly 20-30% performance increase at same TDP, then AMD could still be in the lead with Polaris 10, or potentially exactly the same as far as DX11 goes... while bringing modern features and far better support for DX12).

Still need to wait for an actual release as this is nothing more than speculation right now.
I'd love to see Vega with HBM2 in mobile though... and I wouldn't be surprised if they give us something like Vega Nano with HBM2 next year in laptops.

 

I really hope AMD put something out that makes me want to switch from nvidia to amd especially after having to deal with the 980m & its throtling issues

 

Some event coverage by Jayztwocents and LinusTechTips:





Also, I love how for a couple hours we all completely forgot how evil NGreedia was and just succumbed in awe to Pascal graphs.

So, is the Polaris launch event pretty much confirmed for late June?

 

i just hope nvida dont do what they did that last few gens & go o look your to card is faster than ours and suddenly put out a ti version or amd need to do the same but i wouldn't want them to do such ****ty things to there competion

 

Every. Single. Time. Fanboys are fanboys, are fanboys, but you let you think that they are not, until they confirm it with their next purchase.

 

http://www.fudzilla.com/news/graphics/40638-amd-could-announce-polaris-graphics-cards-in-late-may
So potentially late May for some news. Shortly before, there'll probably be more leaks.

 

27-29-th May to be specific. 26-th at very best as I heard.

 

Ok.. I see some websites already claiming Nvidia's superiority over AMD given what was released on 1080 and 1070 thus far.
1080 was said to have around 20 - 30% higher performance than Titan X at half the TDP (125W).

Polaris was (thus far) mentioned to have a theoretical TDP of around 150W, though that it would consume 'far less' (likely around 100 to 110W?).
So... given what we know of the demo-ed Polaris capabilities... AMD essentially clocked it relatively low (but this could be for mere demonstration purposes).

Seems to me that AMD could easily (if they wanted to) release a fully unlocked Polaris 10 GPU with say same TDP and clocks like 1080... in which case, any performance differential might be nullified or be just within a few % of each other.

So, why the doom and gloom beforehand?
It seems Polaris is more than capable of going up against Pascal at same/similar TDP and clocks... and its (thus far) demonstrations point towards a very capable GPU that moves previous generation high-end performance into the mainstream (Nvidia though seems to be targeting the 'high end', but at relatively low gains).

Also, just because Polaris is targeting the mainstream on desktops, doesn't necessarily mean it would be worse than Nvidia (who by itself is moving previous high-end performance to what they call 'high end').

This might/will come down to architecture's capabilities/features, seeing how synthetic benches are not representative of in-game performance.

Finally... I think performances might actually end up nearly equal between the two in DX11, with a potentially widening gap going into AMD favor given its features and DX12 readiness - though this might not be the case.

As for release dates... do we know of any concrete release dates for both Polaris and Pascal?
Isn't Nvidia's launch date earlier but also just a paper launch? Whereas AMD's would be actual launch of both Polaris 10 and 11 that will be released together for desktops and laptops within a month or two later.

 

1080 is 180W at 1600MHz, it uses closer to 200W when in boost.

 

Ah... tnx for the correction.
So... doesn't that even give more headway for AMD?

 

It would be stupid for AMD to try to compete head to head with nVidia right now. As I wrote previously, Intel have defacto dominance in the low end and nVidia are focused on the very high end where supercomputers and benchmark junkies live. There is a gap between where affordable consumer VR lies, where Intel can't yet compete and nVidia treat as an afterthought. Targeting this market is smart. It gives AMD a first mover advantage in the mainstream consumer VR space.

So what if Polaris won't be winning any benchmark contests. It doesn't need to win benchmark contests. It's going to be better than almost anything you can put in a desktop or tower today for less than half the money. That's a big win for consumers and for us, especially since nVidia are pricing GTX 1000-series upwards of $50 more than GTX 900-series.

 

I was actually thinking something along those lines as well.
Thing is, I wasn't necessarily thinking that Polaris HAS to butt heads with 1080... I'm just saying that the architecture could certainly pull it off (and quite possibly at a lower TDP compared to Nvidia) if AMD decided to put out a Polaris derivative with comparable TDP and clocks.

Besides, Polaris 10 is essentially mid-range... so the kind of performance Nvidia touts for 1080 would essentially be in the mid-range for this generation of GPU's.

It's a shame Zen won't be released until next year... I was hoping to have an all AMD system (Zen APU with Polaris or Vega for discrete gpu and HBM2)... guess I could wait until Zen is released.

Though I'll certainly be looking at Polaris with great interest.

 

We do know that Polaris 10 was already demonstrated in Hitman Pro at 1440p resolution, maxed out settings and DX12... and it achieved 60FPS (which is higher than FuryX/980Ti... and, the game was locked to this FPS, so we don't know Polaris' maximum achievable FPS potential as of yet).

I guess some people and websites are going into 'uproar' because Nvidia is bigger than AMD.
Will be fun to see how things unfold.

 

Polaris might be awhile according to this.

http://www.nordichardware.se/nyhete...osakerhet-hos-amd-partners-infor-polaris.html

 

Polaris will be released this month according to these....

http://wccftech.com/amd-polaris-10-polaris-11-launch-event/

http://www.digitaltrends.com/computing/amd-polaris-macau/

 

If there is two things consistent about AMD, is their ability to create seemingly crappy products that somehow becomes great products after time and their ability to utterly completely screw up a product launch. We will see what happens.

 

We will indeed see what happens, however, AMD was plagued before wih lack of cash to release more efficient architecture on preexisting manuf. process. As such, it would not be accurate to project these notions on an entirely new product and manuf process. Besides, amd did release fury which did come close to Maxwell in terms of efficiency after all, and performance. All without introducing a new architecture. I think that says something about AMDs ability to squeeze out more efficiency on an older product.

 

I am not sure if this has been posted here. But it seems that, to compete with Nvidia, AMD has decided to launch Vega in October this year instead of next year.

http://www.techpowerup.com/222403/amd-pulls-radeon-vega-launch-to-october

 

Uhm... that's only speculation for a few reasons:
1. We have 0 clue if this is actually happening, because the whole premise is based on 'allegedly'.
2. Might have nothing to do with 1080 and 1070 - it could simply mean that if Vega has already been taped out (which someone confirmed if I'm not mistaken), they might just want to jump the gun early now that they know what Nvidia is launching with... or the earlier release (if its happening) might have been prompted by something else... perhaps AMD's contract with China, or other factors?

Again though, we have 0 clue if this is actually happening or not as of yet.

All in all, I wouldn't discount Polaris just yet.
That thing was demo-ed in Hitman Pro in DX12 and 1440p, getting consistent 60FSP and besting both FuryX/980ti as a result... with locked FPS (meaning we do not know if this is the maximum what Polaris 10 is capable of - for all we know, this could be the notebook high-end version... or at least one version/step below the high end.... if AMD releases more than one Polaris 10 for laptops that is).

It was just a demo after all to give us a preview of what is to come... so this might indicate (might) that further improvements are on the way.
And considering that Polaris 10 still has yet to be released, AMD could still play with it to improve performance by raising the TDP and clocks if they deem it necessary.

 

Won't be surprised about 1200 MHz clocks for example. And hope they make it even higher.

 

Based on an original article that was speculative.

 

Polaris should be out next month but don't expect miracles. As for Vega, how is AMD going to rush Vega with an HBM2 shortage? Everything is going to nVidia Tesla P100s right now.

 

Just use HBM1 then.. Does the job and AMD have full dibs on it anyways

Sent from my LG-H811 using Tapatalk

 

Not going to happen on Vega. It would be a stupid move to do so when Vega will most likely be used in workstations too. Tesla wins out.

 

Comments say that Tesla gets ECC HBM while consumer-class one goes mass on September.

 

even if polaris matches 980m in performance , how many oems are going to build laptops with polaris?

 

It shouldn't be a problem for laptops offering MXM solutions.
If the price is right and it performs well enough with good efficiency I don't see why soldered laptops would avoid it either.

 

Doesnt AMD have early access agreement for HBM2?

 

That's what I thought too.. Just bought a Acer XF270HU in preparation for Polaris.. Price was unbeatable..

Sent from my LG-H811 using Tapatalk

 

It's pretty valid question. I hope that we'll see offerings from various OEMs. So far only DELL is pretty much confirmed (they never skipped a revision), but the Alienwares are soldered and Precision's MXMs are rather expensive, since they are FirePro variants. Other OEMs (especially Clevo) is what bothers me. If they skip this generation as well, then I see no other explanation than massive bribing.

At least from hynix, so I don't think that it would be a problem to source the needed chips.

 

nVidia buys pre-production HBM2 from Samsung that's why the HBM clocks are low. AMD has a contract with SK Hynix so they will get most of the HBM2 stock from Hynix and by October it should be in mass production too.

 

Polaris will exceed GTX 980m by a significant margin.

 

That still doesn't change the fact that Vega doesn't seem to be near ready to market. AMD's first priority is Polaris and for good reason with the mainstream and custom chips for the new consoles being the focus. PS4K is coming in October I believe which means AMD has to be producing the chips pretty quickly.

 

Possible Polaris 10 and Polaris 11 specifications:



http://www.techpowerup.com/222450/more-polaris10-and-polaris11-specifications-revealed
http://videocardz.com/59903/possible-polaris-10-and-polaris-11-specifications-emerge

So this would put mobile Polaris 11 performance to somewhere between GTX 970M and 980M, though using only 50 watts (vs 70-100 watts).

 

If that TDP is real and AMD delivers and not going by TFlops only & if they really perform then BGA laptops will outperform the MXM 980Ms, that's embarrassing !!?, Well this time I'm sad about the whole node transition, they could've gone for the 20nm in between instead of milking the whole 4 Years of 28nm....People who had invested in this gen will be ripped off, considering the gimping drivers.

 

You know mxm is just a BGA chip on a board, right?

Sent from a 128th Legion Stormtrooper 6P

 

you know a computer board is just a bunch of silicon, gold, silver, and plastic, right?

 

Mind blown.

Sent from a 128th Legion Stormtrooper 6P

 

http://wccftech.com/amd-radeon-400-mobility-gpus-confirmed/

 

So if R9 M485X is just a rebrand of R9 M395X and it's positioned above two new Polaris 11 GPUs (R9 M480X and R9 M480), it would mean Polaris 11 is actually below GTX 970M performance, making its 50 watts TDP less impressive.

 

How depressing! They cant be rebranding again! AMD be damned

Sent from my SM-A500FU using Tapatalk

 

Wow AMD. Really.

 

We should get mobile Polaris 11 and 10 as well... so, those won't be the rebrands.
Though to be honest, I really don't understand why are AMD even releasing rebrands of their older mobile gpu's to begin with when Polaris is supposed to take place in the mid/high mobile range.

 

R9 M490 and 490X are Polaris 10 GPUs with 1250MHz clock rate / 256 Bit, only 2 cards the Polaris 11 is a stupid 50 TDP / 128 Bit which may end as BGA trash imo. And Since the Alienware now gone full BGA might not feature these & the Macs would be powered by that low end Polaris 11 and maybe a high end 10 even though Apple or AW - BGA trash as usual.

Only time will unfold if there will be any MXM based M490X...

Seems like they are trying hard to even making slightest margins or slacking off over those console trash. They should sell themselves to Samsung if it's the former case. The Pascal Hypebeast seems like unstoppable now until Vega...

 

I don't know whether to LOL or not... Although the AMD failure bug hit my old desktop last night so maybe I'll just laugh. Nothing like a red screen and a reboot followed by "Generic VGA" to brighten your day!

 

Your Sager packed up and now your desktop.. Gee how unlucky can you be?

BTW, when did you have anything AMD in your desktop?