Mobile Polaris Discussion

I don't really have any brand loyalty. I have had the 9800 XT, X800 GTO, Voodoo Banshee, S3 Savage, 7800 GT, 7950 GT, 8800 GTS, 8700M GT, 260M GTX, 880M, 980M, 780 Ti, and Radeon HD 5830 which is the one that died last night. I've had a bunch of other cards too that I can't remember. I think I even had a Cirrus Logic card in a Linux box back in 1994 of 1995.

Failures: 7800 GT, 8800 GTS, 8700M GT, 260M GTX, 880M, 980M, 9800 XT, and now the 5830.

The 8700M GT just overheated because Toshiba didn't have a proper cooling system which caused it to fail, the 880Ms and 980Ms is likely my laptop, I fried the 9800 XT with overclocking, I fried the 260M with a bad BIOS flash, the 7800 GT had a capacitor pop off, I fried the 8800 GTS pushing it after I bought the HD 5830 to replace it by pushing the clocks too high (so it was basically intentional), and the 5830... Well it was a Gigabyte product so I am amazed it didn't give out sooner.

The Voodoo Banshee was by far the best of the bunch. Never even crashed back in the days where ATI and nVidia cards needed different drivers depending on the game you were playing. Loved that card.

The X800 GTO was a great card too. It was the fully unlockable card.

Anyway, I buy whatever meets my performance and budget needs. To be fair, AMD stopped supporting the 5830 so I was running the Crimson beta in Windows 10 which is likely why it died. It was a really old build. Q6600 2.4GHz, 6GB DDR2-800, Intel Bad Axe 2 mobo... No real loss.

 

AMD has officially thrown in the towel, on mobile.

Sad times are ahead.

 

Oh man you're taking me back. I pre ordered a monster 3D (voodoo gfx). I remember I was like a kid on Christmas so I could play the 2 games that supported it. Tomb Raider and GLQuake. I think TR was first. Man seeing 16bit 640x480 move that fast was like seeing the dawn of a new age. Before that S3 had the best 2D cards for DOS games by far. Remember that benchmark with the spinning gold cube everyone used for 2D card throughput testing?

Sorry I digress going off topic. Between DooM bring released and this it's a nice nostalgic weekend.

Edited: I almost forgot Motoracer. 3 whole games. And POD! Man I loved POD.


Sent from a 128th Legion Stormtrooper 6P

 

I was hooked on the original Doom, Duke Nukem, Quake... Ever seen Unreal Tournament on a Voodoo card? Ridiculously smooth. Made fragging on a 28k connection actually have a leg up lol. Ah the days of 300+ms ping times. People don't know how good we have it now lol!

 

Yeah, I played a ton of doom on modems. Good ole days

Sent from a 128th Legion Stormtrooper 6P

 

Except that Polaris 11 and 10 were confirmed for mobile... they will likely be available concurrently with the rebranded options... in which case, we need to look out for Polaris 11 and 10... not the rebrands.

 

They are listed on there and if the 1250MHz clocks are correct, the 10s will be 970M on the low and 980M +/- 10% on the high. 11 isn't even worth mentioning. Low TDP joke. AMD is not interested in the performance crown, they've said it and this line up shows it. Unless AMD can price the cards extremely competitively, the OEMs don't even have a reason to add them to a market nVidia dominates.

 

They've rebranded the entire low end, which matters way more than releasing mid-to-top end chips.

The 1040M is going to have free run of the market.

 

There goes my plans of replacing the 980Ms....unless the polaris mobile cards can be had for 200$

 

Honestly, I can't understand why AMD can't sell MXM GPUs to usual customers! Give 1 year warranty, make nice compatibility and sell it for 300$ and still there gonna be a LOT potential buyers! Because nowadays you can see used GPUs from 2011-2013 and they are so much overpriced!

P.S. Yes, I know that in this case their partners will not be happy so much hence it gonna lower buyers base. But then at least sell us previous Gen GPUs from warehouse!

 

They don't make their GPUs, nGreedia doesn't either.

If there were topics about nGreedia failures, you would've known that it's not your laptop the one to blame. You were very helpful to point me at "various topics about 7970m failures" (not exact quote, but it was among the lines), but I can't see yours. Did I missed it? I think that the NBR community should be aware, just like you made sure that everyone is aware of 7970m failures, dismissing the later iterations, claiming that AMD can't be trusted and that was it. I can assure you that since 680m, there's NO nGreedia flagship without failures and from the looks of it, 980m would top the chart.

Or it could be placed below the rebranded R9-M295X because of the TDP, but the actual performance be different.

 

FWIW, the later revision 7970Ms, 8970Ms, and R9 M290X cards do not have the reliability issues of early 7970Ms. It was a 28nm teething issue what was resolved later on

 

Ok... I think people need to chill down a bit.
The NDA is still in effect, and Polaris 11 and 10 hadn't even been released yet.
As such, we don't know anything conclusively, least of all of their maximum performances, or what kind of form factors they will come in (unless someone managed to get a hold of legit info from OEMS?).
No?

As for rebrands... yes, that is definitely not good, but do we even have any indications that Nvidia will be releasing mobile Pascal cards anytime soon?
Furthermore, Polaris 11 is likely going into most solutions and will probably be adopted by most OEM's... the rebrands might predominantly be used in Apple based products.
Again, we don't know anything conclusively.

Can we just wait and see what happens upon actual release when we manage to get a hold of proper test of both Polaris 11 and 10 before we go spinning doom and gloom?

 

polaris 10 desktop at 5.5 TFLOPS aint cool... with the 1070 giving you 6.5 TFLOPS, making it (in theory) 18% faster!

as for mobile rebrands: cmon people, thats been the name of the game forever now. introduce a new gpu gen for the top 2 models and rebrand the old stuff to fill the lower end gaps. nvidia will do the same with their 10xxM cards, just wait and see

Sent from my Huawei Mate 8 NXT-AL10

 

I guess going for TFlops solely is not agreeable for example - GTX 780Ti is at 5.04 and the ASUS Strix GTX 970 is at 4.2..
I'd love to watch Red and Greed fight in a brand new AAA DX12 and DX11 game.

 

You don't spend much time in the Sager forums apparently because literally everyone who does knows about all the issues I've had with my machine. Don't pull the fanboy card on me because I have zero brand loyalty. Yes, statistically the 7970M has been more prone to failure than nVidia cards (especially given their lower share of the market) but I've had GPUs from both camps fail and I've never been quiet about it. Since the machine is a Sager machine, my complaints ended up in the Sager forum. You also apparently didn't read the 880M threads around either because I was one of, if not the most, vocal people in those threads because I was having so many problems. I personally don't like AMD or nVidia and miss the days when we had the option to go get a Voodoo card and not have to mess with either of the clowns.

I just don't see the point in using their limited capital to produce a GPU so weak when they have plenty of chips that they could have rebranded. Maybe they'll get lucky and the "OMG I can haz Crysis 3 in a half inch thick ultrabook?!" crowd will fall for it but I'm not. They should have shelved Polaris 11, tweaked 10 to be more competitive, and put their resources they used on 11 into Vega instead.

 

No, Teraflops alone are not a good indication of GPU's overall performance, features or capabilities.
Remember the Fury line and how many more flops they were pushing?
Those did not translate directly to gaming performance for example.
They might be more indicative of compute performance... and besides, do we know if we will get a full unlocked Polaris 10 for mobile or not?
I'm guessing that if both Nvidia and AMD will aim for a specific TDP (say 125W max), AMD could easily get away with giving us a full blown unlocked polaris 10 on mobile and vastly outperform Nvidia... but this is pure speculation as of yet.

If you go by Teraflops alone... what will be Nvidia's response for mobile in that case, and would they be able to match a fully unlocked Polaris in that case that's also on 14nm?
I personally don't know... but given that Pascal is a modified Maxwell with improved compute capabilities... my guess is that at least on mobile high end, they might come out equal to each other... while on the desktop, Polaris 10 might approach 1070, or possibly rival it... depending on how good it is (but considering Polaris 10 already surpassed FuryX/980ti in Hitman Pro at Dx12 and 1440p, there's a chance that even for a mainstream desktop product, it might surprise us).

 

It's kind of funny how I was preemptively disappointed by old news a while ago, because AMD purposely never targeted high end with polaris. So I think expecting 980m level of performance, or at beast, desktop 970/r9 390 are realistic.

And sadly this will be the case from nvidia too. I doubt they will release a powerful mobile GPU because they have no competition. Their 1080m will most likely give desktop 980 performance. At best.

What an underwhelming development, only due to shortages of GDDR5X and HBM2, plust development. I am only happy about pascal improving VR, but I already expect very little gains over 980m. I hope AMD or nvidia actually try to wow me.

We are running on like... 2 benchmarks. And Polaris 10, if I remember correctly, is supposed to sell at 350dlrs price range, so it has to beat 980ti on everything to have a good value over GTX 1070.

 

Exactly. AMD are specifically targeting mainstream consumer VR with Polaris 10. Desktop GTX 970 and R9 390 are the minimum requirement for consumer VR rigs. Anything less than these, including GTX 980m, can't deliver. Thus, Polaris 10 will be at least as powerful as GTX 970 and R9 390. I speculate that actual performance will be on the order of 25-50% better than this in order to deliver more than the absolute minimum VR experience.

 

I feel that the node shrink and clock speed alone will push 1080m beyond GTX 980. How much more, that still remains to be seen.
And they have competition, its from themselves. If they dont offer enough advantages people wont simply buy their new and shiny GPUs.

 

I'm not playing any card, but the informative one. Someone has to be up front, beating the drum, seeing that a lot of others are hiding in the shadows. There IS a problem, people have to realize it. It might not be in your case, but there IS one. We are approaching the warranty invalid times (2 years mark), if people don't uproar then, they are helpless and deserve whatever the "mighty" nGreedia throws at them. And trust me, there would be problems, these GPUs wont magically convert to reliable over the night. As far as I know Alienware dropped support altogether, for the only remaining MXM, or rather inventory cleaning, AW18.

If you haven't looked at the sales, these crap books actually sell. I'm not happy about that either, but they have to make money somehow. Still think that Polaris 10 would deliver. Might not be earth shattering, but it would be enough to not give a damn that there's another option. Actually is there another option? Having a GPU that would fail 2-3-5 times during warranty period is not exactly having and if it happens to dodge the bullet, the Big Green would make sure to wipe them clear with a driver or two.

 

It's possible the 1080m could only be a level of desktop 980. The 880m vs 980m stock was about 35% difference. That would stack up if the post was same. But I highly doubt that to be the case. 980m was 80% of the 980 at stock when it was scoring around 11000. It's not that I doubt it can't be true because nvidia said they are closing the gap Between mobile and desktop. That means the 1080m would far exceed your expectations if it did keep that gap between each other. If the 1080m got a fire strike score of 13000 (980)vs the 21828(1080). Then that's only 59% of the power of the desktop. That would be going backwards on closing the gap and on what they said. If they maintain 80% (I think about 70% is reasonable if any drop in the 1000 generation) would score 17462. That might be hard to comprehend as that beats a 980 TI. but that's actually in line of what other member of this forum are hoping or predicting the mobile version to match In the pascal forum.
To further fuel my logic. The 980m can match the 970 with some overlock. That keeps them at about the same power.
It is said the 1070 is at the titan x performance which scores 17396 which is in line with a 980 TI which means those scores at 80% of a 1080 makes sense because the 1080m would be scoring similar to the 1070 like it did in the 900 series.

 

Here we go folks! Polaris 10 (r9 480 and 480X) benched:
http://videocardz.com/60253/amd-radeon-r9-480-3dmark11-benchmarks


If I remember correctly, 7970M was also based on the second tier architecture Pitcairn and it's performance was somewhere between 7850 and 7870 (closer to the latter). So if we follow the same logic, then the mobile Polaris 10 should be closer to R9 480X (67DF:C7 in the graph) in terms of performance, which would make it a bit faster than desktop 980. If mobile Pascal is delayed, AMD would have a golden opportunity to claim the performance crown at least in mobile space (though really sucking in desktop).

 

The poster mentions in the comments that it's missing the supposed full chip (2560 cores), if there is such thing and if it ever makes it outside iMacs (and if it does, let's hope it wont be couple of years late, like the W7170m). If the C4 is the mobile version, it would mean 980 desktop mobile (boy, is this a dumb name) performance (and I mean the proper one - P870DM) for about half the wattage. There's still a lot of info missing (A LOT of ifs), but yeah, looking good.

 

Looks good so far. Hoping Clevo adopts polaris mobile so I can purchase some cards..

Any info? @woodzstack

 

lets just hope AMD actually DECIDES to bring the actual, full polaris chips to mobile...thats not a given, really!

Sent from my Huawei Mate 8 NXT-AL10

 

Oh yeah one major thing I am forgetting..I hope we will get XDMA support instead of those stupid crossfire cables. That would be fantastic. Technically this has been available since Tonga...but since nobody seems to have Tonga MXM cards we never really knew if mobile Tonga had it or not.

 

^ This exactly. I always wondered if one was to get 2x Zentrica R9-M295X would they run in CF without cable? They are full-blown Tongas after all, something that even desktops haven't seen (not under 2xx series at least).

@jaybee83 - It's up to the vendors as well. Not that AMD has zero part in it, but the vendors have to say " We want to use your GPU design" as well. Let's just hope that things would pan-out nicely and no bribing is involved. (Sorry but I can't think of anything good coming from the green team for now)

 

funny thing actually: the full tonga mxm boards actually went predominantly into base systems, such as casino slot machines *lol*

Sent from my Huawei Mate 8 NXT-AL10

 

It's still a game, I give it that

 

I haven't bothered to ask those that would know. Sorry, means I know nothing ATM.

 

Good news, if any... I will happily take GTX980+ level performance in my laptop if Polaris MXM delivers .. Big IF though..

 

I would take those benches with a grain of salt. The poster also mentioned that those were not likely top clocks for Polaris. Plus we don't know how Polaris will perform in actual games and pro software. I guess we have to wait and see, but I'll gladly take furyx performance in a laptop

 

For me those benches look realistic with the exception of that underwhelming crossfire set up. Keep in mind, VC also provided benches for 1080 about two weeks before the announcement and they proved to be true. As for performance, we can expect Fury (non X) level of performance straightaway, so some OC should make it to fury X levels

 

Not too much longer to wait now, AMD's Live webcast from Computex starts on May 31st, 10 PM EST (June 1st in Taiwan).

 

Uhm... as even the poster himself stated... benchmarks like these are done with samples running at relatively lower clocks...
Furthermore, we know benchmarks aren't indicative of real life performance.
All in all, we don't have a complete picture as of yet.

 

Clocks don't seem that far-fetched to me... 1250-1300 was the rumor for P10 (1250 specifically) and P11 is rumored between 800 and 900 (850 specifically). Not too long to find out.

 

However, rumors are one thing, verified evidence is something else (people have been putting up all kinds of unsubstantiated articles based on rumors and ignoring actual info coming from the companies themselves.
Who is releasing the rumors in the first place? Do we have any reason to think they are genuine?
14nm is a more efficient process, so there might be a possibility we will see higher clocks in play.
Either way, we won't know more until likely next week, or possibly later.

 

Not too far away now, lets wait instead of speculating.. Hope AMD can blow us away.. But lets see.. They haven't properly released a MXM GPU since 2014..

Sent from my LG-H850 using Tapatalk

 

Wait is too long!! Really want a lower power part (80W max) with gtx 980 performance out of the box! Fingers crossed!!

Sent from my SM-A500FU using Tapatalk

 

I am hoping Polaris will come with at least fury x level performance for top end laptops, but that would be a personal wish (not necessarily something that will materialize... still, fury (non x) is also really good, but it wouldbe nice to see a full core Polaris on mobile).

Out of curiosity, is there any information available about mobile Vega next year, or are those strictly desktop grade parts?

 

You can see how tight lipped they are about Polaris, so Vega is even bigger question mark, especially for mobile.

Part of the blame should be redirected to MSi and Clevo.

 

Also AMD.. Their failure to produce anything sellable doesn't help.. Also high failure rates with 7970M etc..

Sent from my LG-H850 using Tapatalk

 

That's why I word it the way I did - part of the blame. I for one would've bought R9-M295X if available. As for failure rates it seems that for one company it's a problem, while for another "it is not".

 

I think the ODMs figured out that the Tonga core was so hot and power hungry it really had no place in a laptop, so they didn't bother to offer it. I'm sure engineering samples exist out there. I've seen a Dell R9 M295X engineering sample from a seller in China before.

 

That's interesting to know, but yeah, I'd guess so. With 880m one would think that they had the means to cool such chip, or as we can see, the other way - don't ever put 125W chip on MXM-B ever, ever again. It's not like 100W MXMs last all that much to push things even further, damn lead-free solder (or rather the cheaper versions of it). Since we are talking about high wattage MXMs, does anyone knows if the 150W M5500M is happening? I would love to see how this would pan-out.

 

What's even more confusing about the R9 M295X is that it ended up in the Alienware 15 of all places...soldered. Who the hell knows what they were thinking, but Dell has been making stupid decisions as of late. Looks like Clevo and MSI are not so stupid.

 

But see what they did in the case of M295X... the OEM's decided to put in inadequate cooling and power supply... so the card was effectively gimped from the get go.
However, when it comes to Nvidia, most OEM's will go out of their way to cater to their 'high end', keep it nicely supplied with power and also adequate cooling.

I agree there's a problem with AMD not releasing anything 'new' over the past few years, but still... having those solutions on offer would have been nice at the very least.

Perhaps OEM's should have advertised AMD cards a bit differently... something along the lines of: 'adequate for gaming, and an excellent performer in professional software (where Nvidia lacks tremendously - which makes me think that most of the Stream Processors and TDP on AMD side went into that - Nvidia intentionally gimped their compute capabilties by comparison, which is probably why they managed to improve on power efficiency and gaming).

 

Aside from the iMac's Tonga GPUs, do you have any evidence that the Tonga GPUs were improperly cooled in the 15 R1, 15 R2, or Dell Precision M7710? Because from the reviews I read and the accounts from those who had them (of those of Alienwares, which is rare, sadly), there weren't any overheating problems nor thermal throttling (and with the 15 R2/M7710, the power supply shouldn't be a problem). Could be a case of limited evidence of not showing the whole picture, but if it performed slightly under a GTX 970m without problems, it performs slightly under a GTX 970m.

 

sure its happening, otherwise nvidia wouldnt have officially presented the card still waiting on actually available MSI workstations with that board though, should pop up sometime in june... would be interesting to see if they would be available as stand alone mxm boards anytime soon...and at an affordable price