Mobile Polaris Discussion

The difference between 980 TI and 980 is absolutely huge. Where the difference between 970 vs 980 is acceptable. Wherea GTX 670 is almost right on 680.....

I think nvidia learned the lesson AMD may be learning, disabling a single set of CU is not enough to properly differentiate the products.

Look like the way they segmented the 980m, 980, and 970m, its absolutely brilliant. Each one has a place and none of them are cannibalizing the other's sale.

 

Except, the naming is off and it would probably be a one off. You have a desktop part, but in a mobile form, and the other two are mobiles... Like they aren't derived from desktop parts as well. I think that 980MTi would've been better, even if mouthful, as an added bonus it sounds like a sports car.

 

Are they really going to call it GTX 1080? because a lot of clueless people will think it is something to do with 1080p resolution

 

I hope they drop the whole R9 R7 bullcrap naming scheme 290X 290 390 390X bs. The ATi naming scheme was perfect ex: 77xx, 78xx, 79xx, just adjust the last two numbers to indicate which is higher up.... and stick an M to indicated mobile 77xxM, 78xxM, 79xxM

Don't know why they dropped that naming scheme. It was clean and easily understandable

 

What exactly does that "R" stand for, anyway - Radeon?

 

No because they already say it in the name...AMD Radeon R9 390X

 

The only way AMD makes a comeback, is if they combat the GTX 960M and GTX 970M equivalents. I hope these rebrands don't mean they're conceding the mainstream and mid-range again.

 

I doubt it. It's the OEMs apparently that want rebrands, and they seem to want to use them in Mac's (though Polaris was stated to be used in them too)... but AMD's Polaris 11 at 50W could easily go up against both 960m and 970m and win (if I'm not mistaken).
The 960m alone is using 75W.

I don't get it. AMD has Polaris 11 and 10 for mobile, so why mess with the lineup using rebrands in the first place?
The entire 4xxx series should be Polaris gpu's.

 

The demoed, super raw, prototype was consuming ~37-38W and already went head to head with desktop 950 (965M equivalent), while this one consumes 50W, most probably with much more performance, so beating everything in current mid to high end range is very much achievable Even if these cards won't be able to compete with mobile Pascal in high-end, they would at least attract more OEMs to use AMD GPUs, hopefully such grands as ASUS and MSI. AMD needs more money to be able to afford making less rebrands.

 

The way some search engines work, I can see a lot of confusion even for someone who knows what (s)he is looking for.

I agree. Why the leading two symbols, since you clearly distinguish them with 420, 470 or 490 afterwards. It could have something with the APUs, since they say "R7 level of graphics" and one could look it up and get a ballpark figures of the performance. That's the only logical explanation in my head.

 

Pascal was stated to have twice perf. per watt of Maxwell.
AMD managed to squeeze out 2.5 perf. per watt vs the best of Fury line (Fury X)... meaning that at 110W, a mobile Polaris 10 will have equal performance of Fury X... that's without taking into account any other improvements AMD might have done to it in the interim (we also need to keep in mind that Polaris 10 was demoed running Hitman at DX12, locked at 60 FPS, which was faster than Fury X if I'm not mistaken... and given that the FPS was locked, it is possible that Polaris 10 could have even greater performance that AMD didn't want immediately revealed).

Plus, a 14nm node which AMD is using seems to be more mature than 16nm one used by Nvidia.

At it's worst, I think Polaris could easily be comparable to Pascal, and at best, faster (especially in DX12).
Also keep in mind that Fury X (with it's Crimson drivers and subsequent driver releases) and 980ti are basically equal in terms of raw performance, with not that much edge going to Nvidia in terms of TDP at average thermal loads either (at least during tested gaming loops).
Maximum power loads (torture stress tests) are very different however and 980ti consumes about 296W vs Fury X's 453W - however, I think I also read something about Fury's power spikes being rather unique... and the overall average TDP consumption is greater on Fury X cards (but that can easily be attributed to the number of COMPUTE units on board - which do nothing special for gaming... but are excessively more powerful compared to anything on Nvidia's end in professional software, which is probably where all the extra TDP is going).

Now, I wouldn't expect the differences to be drastic... but I think they might translate to better power efficiency and raw power for the AMD team (of course, we won't know anything for sure until we see final products from both AMD and Nvidia to compare... right now, we are running on raw numbers we were given officially)... or AMD might have slightly lower efficiency... but if Nvidia doesn't really address it's compute performance (and some preliminary indications points towards that they won't), then overall, AMD would still be the better choice (because its hardware also seems to be better equipped for DX12 vs Pascal, if those same preliminary reports are accurate).

 

Think of it like Intel. You have 3 tiers with core i3, i5 and i7 and then you have individual models in between. That is the reasoning behind R#-### naming scheme.

 

Polaris 10 confirmed to be the high end mobile offering.

http://videocardz.com/59248/amd-officially-confirms-polaris-10-and-polaris-11-market-positioning

 

Just waiting how it fares against the 980M, even though I cannot upgrade anytime soon. I guess waiting one more year for the real GP204 core and maybe Vega mobile would be better than just jumping off right now, hope things are fine without that full UEFI mess like the nvidia on Alienwares.

 

Ashtrix -

This will probably sound as if I'm repeating myself (and I probably am):
110W Polaris 10 = Fury X from mere 2.5 perf. per watt benefits without any further optimizations.
So, with optimizations in mind, I think that we could see 90 - 100W Polaris 10 that's still faster than Fury X, or has at least comparable performance - though I think we can see OC 980ti performance (not radically OC-ed, but max OC 980ti performance could be there if the thermal headroom is available... and 14nm should allow for higher clocks too).

Might be that Vega mobile with HBM2 could be reserved for next top end in mobile space (but we have nothing to base this on as of yet since Vega's release is scheduled for early 2017).

 

Well if that's the case the new chips - they'd rip apart the current gen lineup then, poor people who have upgraded like me they'll be in deep turbulence then since the driver mess at nGreedia is like a mercury now, It's totally chaotic. Should brace ourselves they are close now ! & endure until the Vega and GP204 come, Also the kicker for MXM+HBM might need a redesign just not a simple turd BGA mess..

 

That is my worry too with HBM2.. However if AMD release a high end MXM card, I will jump and get it.. Had enough of NVIDIA..

Sent from my LG-H811 using Tapatalk

 

I might also jump on mobile Polaris 10 if it delivers at least FuryX and possibly beyond performance levels.
It just doesn't make sense that AMD would intentionally nerf this card for mobile at those TDP levels (which are more than acceptable for laptops even without binning that usually takes place), and besides, AMD needs a strong next generation mobile GPU to reclaim laptop market-share.
The desktop version however will likely have a higher TDP than 110W though, and of course greater performance. But if Vega already was already rated 180W, then I'm curious where Polaris 10 desktop ends in terms of TDP (150W?)

Also, Link4's link says that Polaris architecture will bring 2x perf. per watt. This is incorrect, since multiple other sources claimed (along with AMD if I'm not mistaken) 2.5 perf. per watt.

 

Time to sell everything and make funds available for Polaris 10 mobile!!!!

I know there will be others willing to join me after the mess nGreedia has left us with

 

As for me, I have been with ATi/AMD since HD4000 series and I cannot remember any significant challenges so far. In fact, my current 7970M still serves me really good with almost any game at High at 1080p, though I remember the mess with drivers during the first launch, but I bought my current rig almost a year later I could not believe at that time that it's Nvidia counterpart (680M) was like 300USD more expensive! Hell, even significantly less powerful 675M was (a bit) more expensive than 7970M at that time, it was so as ridiculous as Ngreedish! I'm really hopeful that the year when I change my rig will mark the so-much-anticipated comeback of AMD to mobile!

 

About that, my theory on it is that the 2.5X performance/watt improvement was over Tonga, and the 2X improvement is over Fiji (Fury X). I think I have said in this thread already a while back that the 2.5X improvement should be over Tonga (it makes sense if you look at the Capsaicin roadmap) so nothing really has changed. A Fury X is about 70% faster than a Sapphire Nitro R9 380X which uses 190W (Tom's Hardware review) And Sapphire Nitro R9 Fury uses about 260W. If you do some calculations the Fury X is about 25% more efficient which is exactly 2.5 / 2 = 1.25.

Fury X vs 380X performance http://www.eteknix.com/amd-r9-380x-4gb-graphics-card-crossfire-review/3/

 

With recent nGreedia drivers I wont be surprised if it has worse performance than 7970m with recent Crimsons.

 

I can guarantee this... I overclocked my 780Ms by 80mhz on core and clock and barely broke 13000....7970M CF can easily, easily beat that almost 2 years ago.

 

Actually, previous reports, including various slides have stated that Polaris 'shines' 2.5 times brighter and that it will have 2.5 times performance per watt of previously most powerful GCN offerings (which would likely mean the Fury line).

Here's the link:
http://www.extremetech.com/gaming/2...s-2-5x-performance-per-watt-may-utilize-gddr5

However, we have no real evidence to support the idea that 2.5 perf. per watt relates to Tonga.
One has to take into account that AMD did demo Hitman 12 at maxed out settings with Polaris 10, and the gpu was able to achieve 60 FPS (locked at those FPS levels - so we don't know its full potential). If I'm not mistaken, this surpasses Fury X... so we're looking at about overclocked 980ti level of performance for mobile Polaris 10 (I hope) at baseline. And with 14 nm, it should be possible for people to raise clocks up.

Btw.. if we went with the above suggestion that Polaris offers 2.5 times more perf. per wattt vs Tonga... wouldn't it make more sense to apply this idea to Vega instead?
Vega uses HBM2 (which is more power efficient than GDDR5/X, and if AMD tweaks it right by limiting its power consumption, like it did with Nano, then they could easily come up with a very efficient and powerful GPU).

 

If they can compete with Nvidia's second tier card in terms of performance in the mobile market and have their offering be a bit cheaper, I'll be happy.

 

Hopefully they go one step further and give Nvidia a complete ass kicking. Nothing would make me happier.

 

According to the following sources, Polaris 10 performs close to 980ti:
http://www.tweaktown.com/news/51814...close-to-nvidia-geforce-gtx-980-ti/index.html

http://wccftech.com/amd-polaris-10-desktop-polaris-11-notebook-gpu/

It has a maximum tdp of 175w (though said it will consume much lower amount)... Which bodes well. We might just get 980ti/FuryX level performance in a top end laptop, or very close to it.

Plus, if what those sources say is accurate, then Polaris 10 is confirmed for high end notebooks

 

Don't get me wrong, I would love that. I've been preaching it here for like a year now. But it's just really unlikely, at least on the mobile side.
EDIT: Especially given the fact that what AMD has been pushing with Polaris has been perf/watt and not necessarily raw performance.

 

We'll find out soon enough I guess. Hopefully we get something to get excited about at the very least.

 

AMD will not give full Polaris 10 to mobile, not in 12 months.
Also the one who release first looses because next one can overclock their card a little to overpower on few % in benchmarks and few games - that's all that's needed. And AMD is in bad situation: Nvidia goes first and everybody runs buying it. AMD goes first and everybody goes... waiting how that corresponds to Nvidia's offer. That is a lower ground due to 25% market share.

 

Well, since Polaris 10 is confirmed for mobile and considering the full chip is as fast as 980 Ti, then the severely chopped down one might be like 980 performance, which is like 2.5 better than M390X, which will still be a HOLY COW-like gen to gen jump in performance. I'd be damn happy with that. Hell, even 980M performance would be satisfactory as long as AMD gets some share in mobile segment.

 

I don't think the mobile Polaris 10 will necessarily be 'severely chopped down'.
It might... though, we have 0 reason to think so as of yet.
It might just be low enough to be close to 980ti performance wise, but with overclocking headroom to reach it.

Or, if they will be holding back on the performance, they might want to release a new version of Polaris 10 that's better tweaked, or possibly Vega with HBM2 for mobile early next year.

 

http://arstechnica.com/gadgets/2016/04/amd-polaris-will-be-a-mainstream-gpu/
If the best we'll see is about the performance of the GTX 980 ti, I expect (well, hope for) a very competitive or appealing price.

 

*groan*

You blew it AMD. You really, really blew it. Radeon division needs to be sold off to Intel, who can actually fund proper R&D into graphics performance. I'm tired of AMD's false promises and nGreedia's unethical tactics.

 

How exactly did amd blew it? I think they did pretty good thus far from peeling news (plus, we don't even have solid benchmarks yet).
If the numbers pan out, then what amd said about Polaris power efficiency and performance will be accurate. Companionship are targeting lower power consumption. But I think that fury x level performance in a laptop (or close to it) is really good. Nvidia still has to release solid evidence of their Pascal cards mot to mention numbers

 

Raja came out saying " we will take the performance crown" and now AMD is saying "we're targeting mainstream"

No. They have to gun for the crown, but they don't have the resources to do it. Time to sell it off and let somebody actually take proper control and slap nVidia around.

 

How did AMD blow it? they seem like they have a really efficient uarch this time around .
980 ti performance for mainstream is nothing to scoff at, if it is all around more efficient and cheaper than Nvidia offerings, that's a big win, plus we know that Vega will be their high end, and the actual card that will go up against Nvidia's high end offerings; 1080ti and Titan, considering Vega will have HBM2 and Nvidia will be using GDDR5x there is still big chances of them taking the performance crown, but it will be end of 2016 more likely first half of 2017 .

Came across these video's pretty interesting watch (part 2 is really interesting) .





If any of this pans out I think the ATI/Radeon division is better in AMD's hands, than in Intel's, the industry needs competition or we as consumers get screwed .

 

That's a big if, and I hope it does turn out well in the end. If it doesn't, it needs to be sold off to somebody that is ready to take charge, whether it be Intel or Samsung.

 

While Intel has improved their integrated gpus, I don't know if their offers will be as capable as AMDs (cpu bit aside). Despite their massive budget, AMD has been consistently providing new innovations that are more future proof as far as their hardware goes.

P.S. Touch keyboard issues, bah... corrected the post on my laptop when I got home

 

Unfortunately, especially with the recent leaks from NVIDIA, it looks to me like prospects might be pretty disappointing across the board for both companies.

I don't know if AMD will be able to take the performance crown - especially if their mindset is to focus on the mainstream, but they should definitely be able to win back some market-share which should help future R&D.

 

Where did you read that 1080 ti and Titan will have gddr5x? I highly doubt it. But I don't get where AMD blew up too. AMD just calls it what it is. While BSeedia called 980 a High-end card while it was emm... not. Because in 6 months we got 980ti. AMD does the same just without BS, they just can't afford it now.

 

My bad, I confused GP 104, with GP 100, so HBM2 isn't out of the question, GP 100/1080ti may very well come with HBM2 memory .

 

AMD did the exact same ****. If AMD's 7970 werent so ****ty at stock, we might have big die chips at 650.

 

http://videocardz.com/59465/amd-radeon-r9-m480-based-on-polaris-11-gpu
Well, at least we know something on the mobile end of things. Though if the table is to be true, why is there a reduction in memory bandwidth?

 

Just a wild-ass guess but it probably goes with the reduced memory clock speed.

 

Or it might be that it simply doesn't need as high bandwidth?
It's an entry level GPU... not a middle-ground or high-performing one.
Or , it might be a technical error.. or, AMD may have several Polaris 11 derivatives that have different bandwidth configurations?

 

I doubt AMD will cut down mobile Polaris 10, they will just reduce the clocks. Also this time around Polaris 10 is so efficient (130-150W at most, rumors put the full chip at 135W for desktop) that they can clock a 100W M490X only 15-20% lower than the top of the line desktop card. At that point it should still be faster than a 390X.

And btw the rumors stating that Polaris 10 is close to 980Ti and Fury X performance are based on nothing but Firestrike Ultra scores. The problem with that is Ultra is a 4K test and polaris only had 256bit GDDR5 which definitely affects it's scores. Other things to consider is that the chip was most likely an Engineering Sample (lower clocks possibly) with early drivers so it's performance can still improve a lot. There is also the fact that nVidia cards have higher performance in Firestrike vs AMD cards than in real world games, and the fact that Fury X's insane memory bandwidth helps it a lot in Ultra (Fury X also doesn't suffer as much from bottlenecks because cards are well optimized for synthetics) and it ends up quite a bit faster than a 390X than it really is. So until release ready Polaris 10 with final drivers are tested we won't know how Polaris will perform. My guess is it might end up a bit faster than Fury X at 1440p but not 4K, heck I doubt anything will even challenge a Fury X at 4K until Vega 10 is released.

 

OpenGL benchmarks of Polaris 11 and Polaris 10 (apparently) :
http://techfrag.com/2016/04/30/amd-polaris-10-polaris-11-opengl-benchmarks-leaked/
http://videocardz.com/59468/amd-polaris-10-and-11-opengl-benchmarks-spotted

According to that benchmark, the Polaris 10 scores essentially slightly better than GTX 950.
Something is off and doesn't sit right. At over double the TDP of Polaris 11, I find it highly unlikely that 10 would score the same as Polaris 11 (unless this is just one version of Polaris 10?).

Polaris 11 was stated to be a direct competitor to 950 at lower TDP... not Polaris 10 (which according to this benchmark scores the same as 950 at basically just lower TDP?).

Taking this with a bucket of salt until we see a full fledged release - but I guess it wouldn't be the first time something was over-hyped (in which case, we were given false info on performance capabilities of Polaris).

 

Expect about 13-14K Graphic score from the R9 M490X

 

Will it be the top tier mobile amd GPU this year or M495X? 13-14k GPU score firestrike? Tnx

Sent from my Nexus 4 using Tapatalk