Mobile Polaris Discussion

This XFX RX480 has great OC performance, low temps, low wattage, and isn't water cooled... if you are in a hurry - start at 7:30

This RX 480 DOESN'T SUCK!

 

Forwarded this to a friend who is kind of anti-nGreedia but was forced to buy a 1060 because RX 480s were taking forever to arrive in Canada. He definitely wants to go AMD next time and this looks like something to keep an eye on for future releases.

 

Well, someone tweeted out a glimpse of the Alienware 17 R4 with an RX 470.
Now if only we could get a review.

 

http://wccftech.com/amd-polaris-revisions-performance-per-watt/

 

I knew there was something wrong about their power hungry GPUs. Now the question is if those cheap MXM 480 from China are Rev 1 or Newer?

P.S. Old joke about "Don't buy AMD GPU until Rev 1.2" still holds solid.

 

Thats good news, hopefully we will see more RX GPUs in notebooks (that are not made by Apple). Still dreaming of the new AW 13 with RX 460 or even 470, but i doubt Dell will do me that favour. Just dont want to throw money into nGreedia's throat.

 

On the other hand we have nGREEDIA's milking, so the same applies there as well. You can safely say "Don't get nGREEDIA, unless it's a Ti (= rev. 1.2)". I don't know which one is better, or worse, depends on how you look at it. I do know that a lot more people buy the green stuff. And it's not like all of their products are fully functional on delivery, but that's another story. I haven't seen a flawless product (not only GPUs) for a while, so I can't understand people's urge to bash AMD in particular.

These are some great news. Let's see how this rumored mobile 95W RX 480 would turn-out and if it's going to see the light of the day.

 

So these revisions wold be the RX 485, RX 475, and RX 465? That was the whole point of the "##5".
Hopefully the RX 465 is a full Polaris 11 and not a cut-down variant like the RX 460 is currently.

 

IF this is out now can you please post (o those people who owns MSI GT60) if it can with this mxm videocard? Thanks

 

Hehe... THAT's what the E9550 is based on it would seem.
Oddly enough its just as I said... these cards are newer, and as such have improved efficiency rating due to better familiarization with 14nm, etc. (especially after seeing how the first ones performed in the field).
Granted, the undervolted 480 can still be more in line with 1060... but 95W vs 120W while offering nearly the same DX11 performance and BETTER DX12 and Vulkan performances show what AMD can do.

I wonder if Nvidia will follow suit (though to be honest, their lack of support for DX12 and Vulkan is something I'm not willing to spend my money on).

You know... this could tip the scale into AMD favor by having more OEM's open to integrate these new cards into their systems.
Let's just hope the cooling is also good as opposed to just sticking somethign in there for the sake of it.

 

Nvidia HW has DX12 both Pascal & Maxwell. It's not like they don't have support. DX11 runs fastest on Nvidia HW that's not AMD can reach, never.
DX12 is not like AMD rekts the Nvidia with the current Games and Drivers are unoptimized for DX12 for both, Look at DOOM Vulkan and see the Nvidia/AMD HW benches on that API, Just because of better hw for ASync AMD is not going to beat Nvidia to pulp. AMD is always second choice no matter what if you look at Mobile history they are long dead since the 7970M. the TDP of the RX480 is 150W the peak of MXM3.0b that's blasphemy for its performance !!

So far no gaming laptops except Alienware has AMD, the workstations might have the MXM option but no gaming laptops would and the RX470 in Pascal generation is a dumb move by both consumer and OEM (If anyone buys that GPU equipped machine..)

I think we are just looking for the Alternatives for MXM machines, atleast most of us since the Pascal dropped the MXM standard going into full custom cards, sparked interest with the legacy machine owners & unfortunately those are super low end options, Now that they also dropped LVDS the MXM cards would be more useless....This RX485-490 seems like a good choice But that won't beat a 1070, Still we don't have any machine with proper benches while Nvidia is prepping up for a v2 refresh.

Vega MXM would be good but there's no info on that at all..this is not AMD bashing but that's what I see from the current situation, They are just mediocre & Console market - glad to have them there else the Monopolistic trash ngreedia would shadows us with overwhelmingly stupid gimpware (GFE signup is first, next the Drivers with GFE will roll like a retarded crapshoot)..

 

If Nvidia has HW support for DX12, then why are its gains so much smaller vs AMD's?

Also, Why does an RX 470 in DX12 seem to be trading blows with 1060 in some games, while RX 480 is overshooting 1060 and is getting close to 1070?
One cannot blame that on 'shoddy' DX12 and Vulkan implementations. There is a clear cut difference here that benefits AMD a lot more than Nvidia.

Yes, AMD has relatively poor integration for DX11, but this is why it gains a lot of ground in DX12 and Vulkan (which are the future), and its performance will likely continue to increase over time, whereas this might not be the case for Nvidia (given their track record).

Furthermore, AMD is operating on much lower frequencies than Nvidia while offering similar (DX11) and better (DX12/Vulkan) performance - performance per watt is definitely on AMD's end.

I would personally happily go for E9550 in a laptop. It's TDP is low enough (95W), provides excellent 1080p and 2k performance, and will continue to experience better DX12 support.

Also, AMD never released anything more powerful than 480 on desktop due to focusing on recapturing relevant market share.
Would I prefer that AMD released Vega when Nvidia released its GTX 1070 and 1080?
Yes, but one has to

 

Async compute!

 

You answered your own question. It's because AMD's single-threaded, high overhead driver in pre-DX12/Vulkan APIs results in sub-optimal utilization of its hardware. Instead of fixing the problem, they're washing their hands of it and telling devs to handle it. The results are as you would expect. Most devs can't code a low-level API renderer that's as efficient as Nvidia's driver, resulting in lower performance on Nvidia hardware.

AMD's ALU-heavy GPU designs have greater chance of idle execution bubbles which means more opportunities to exploit concurrency via async compute than Nvidia, but the difference is a few percentage points at most, not a night-and-day difference. The overwhelming bulk of AMD's performance gains in DX12 and Vulkan are the result of shedding the inefficient Radeon driver layer.

 

Well, I have 7870M which is exactly half of 7970M and it runs DOOM at 45 fps +/-5 on middle low. I seriously doubt that same year and class Nvidia GPU can even that. Not just because of Vulkan but because Nvidia's old GPUs manage worse with newer games and newer drivers (#gimp).

P.S. To be completely honest 7970M plays DOOM way better on older 15.11 driver than on new 16.3.2 but 16.7 DOOM driver kinda recovers situation.

 

Yeah the Driver gimping is nasty from Ngreedia, Can't help unfortunately.

Edit : Hope Ngreedia doesn't destroy the Maxwell HW, Since Pascal is just a Node shrink OCed Maxwell.

 

Some of the differences between AMD and Nvidia have been a lot more than just a few %... in most cases, AMD is faster by default... but in Doom and Deus X... it seems AMD has a much higher advantage in not just DX12 but Vulkan too.

That's my beef with Nvidia... their 'gains' in DX12 are minor or go into the negative, while AMD always gains performance... and subsequent game patches and driver releases further increase this performance.

Plus, Nvidia stops supporting their older gpu's relatively fast (hence the gimping people likely experience)... and actually, in the case of Doom, Nvidia new drivers have in various cases resulted in performance degradation, while its been the opposite for AMD.

AMD is likely to continue support for its older gpu's 3 or 4 years down the line... Nvidia is highly unlikely to do the same, especially when Volta gets here.

 

http://wccftech.com/battlefield-1-directx-12-benchmarks-amd-nvidia/

Q.E.D. Look how poorly 780 performs, while AMD's 4,5 year old 7000 series are still packing a punch.

 

I think the more impressive thing is the RX 460 4 GB averages above 30 FPS at 1440p. For a GPU that's between the GTX 750 ti and GTX 950, it delivered good numbers.

 

So basically it is likely in a year the 7970M will match a 970M in new games haha

If this is true it will already be uncomfortably close to 780M levels!

Perhaps a well binned M290x isnt such a bad card afterall........

Sent from my SM-A500FU using Tapatalk

 

Lol more likes than I expected. This is sad Nvidia. Times are tough for enthusiasts I wonder if any revival will happen...

Sent from my SM-A500FU using Tapatalk

 

I don't think so, As long as iSheeple exist, we are being overwhelmed and defeated, This trend will get worse and extends to everything....look at that 3.5mm jack fiasco, Analog for digital and 360 deg vs 2 deg of freedom.

 

8970M already matches or exceeds 780M when both cards are at latest drivers. They have made huge improvement over the years to driver performance for the 7970M/8970M.

 

That is quite a feat when you think of it! Ngreedia should be ashamed!

Iam tempted to put a 8970M or M290x into the M15x as a last upgrade now. Will probably end up just as good as the 970M in the end. Unfortunate that the fans have to be manually controlled though?

Sent from my SM-A500FU using Tapatalk

 

I just picked up a Clevo 8970M, will report on the fan behavior in the M15x when it arrives. I know the Dell R9 M290X did NOT have native fan control. svl7 reported the 2GB MSI 8970M had native fan support.

 

2gb sounds a bit meh at this stage. 4gb is really the minimum nowadays. Your Clevo 8979M has 4gb?

Also you reckon the m290x chips clock better/ better binned?

Sent from my SM-A500FU using Tapatalk

 

Too bad the R9-M295X was deemed unworthy, no? We'll see if the story is going to repeat with RX 4x0, but so far it seems like it.

 

Yes my clevo 8970M is 4GB. I know for a fact the 8970M and R9 M290X are better binned than the 7970M early batch cards. I've been able to clock 8970M/R9 M290X as high as 1000/1300 at stock 1.050V, so I figure undervolt to 900/1200 at 0.975V should be feasible in the M15x.

Well, we never really were able to test the R9 M295X because of its extremely limited availability. At 425$ a card, its not really cost effective to test.

 

Not just that but you need to order a batch of 10. RX deserves to sink pulling LVDS like ngreedia...

Sent from my SM-A500FU using Tapatalk

 

Do you say that they removed LVDS support from mobile RX 4xx? Please, don't!

 

Ixnay on the VDSLay.

Unless you have eDP port (good signs are your model was offered with 3D or 3K/4K screen options but you can check with HWInfo &/or panelook.com) sorry but you're uffedstay for both Pascal and Polaris upgrades.

 

ZM series was offered with both LVDS and eDP. both solder points are there, but only one port is actually connected. usually, they had LVDS with 1080p panels and eDP with 3K/4K

 

Even M15x had a 30 pin eDP point. The hassle to solder a connector anf fit a new screen without even knowing if the socket has proper traces or not seems a bit much!

Sent from my SM-A500FU using Tapatalk

 

I'm pretty certain that the traces are there, but it lacks a handful of resistors, capacitors and a few other stuff. It would be quite the soldering task. Also there are test points for the other 2 lanes, but making that happen would be even more hardcore, but who needs 4K on 15" anyway?

I was pointing at the OEMs, and I still point at them. They never gave us the choice, it seems like this is the case once again. The Zentrica modules are as good as non-existent for us.

Both the GeCube RX 480 and the E9550* are targeted to a different market, so you can't say with 100% certainty that IF there happens to be a proper mobile RX 480 it would lack LVDS.

* Yet we still don't know whether or not it lacks it, since no one tested it... because it's still not out there.

In all honesty I don't think that there's LVDS on the chip, but there are small-chip solutions that can convert up to 1920x1200 @60Hz, so that would be ideal, but drives the cost of the module up. There's the other way - with proper knowledge one can make a small board, but this one requires quite the dedication.

 

Gecube seller confirmed to someone in the MXM 3.0b thread it doesn't do LVDS, but yeah, E9550 is unknown ATM (as is whether the drivers/support cost/price/minimum order qty render it useless as a consumer product) but 95W is certainly a lot tastier than 150W.

Don't knock it until you've tried it. Text on 1080p 17" looks pixellated as crap sitting next to it. Source: my desk, a month ago

edit: LVDS -> eDP (20 pin) converter boards exist http://www.ebay.com/itm/LVDS-to-eDP...IY-Embedded-DisplayPort-Monitor-/161745320023 but it'd be a brave soul to test jumper combos and risking blowing up panel and/or mobo and 20 pin / 2 lane limits you to 1080p 60Hz AFAIK
and have fun soldering in an eDP port if you're not an electrical engineer like this guy: https://www.techinferno.com/index.p...clevo-p650sg-sager-np8652-4k-display-upgrade/

 

Yeah, but it was just that - confirmation. It's not clear whether or not it's a chip thing, or because they didn't wired it, because no one asked about the details.

I know how far I can push my eyes, but hey, whatever floats your/or whoever's boat*. I'll repeat it again though, it would be quite the challenge to get the two extra lanes wired (you have to make a custom board for them or "hang" them in the air... yeah, no) and the link you've provided only proves my point, what it would take the current two lanes to work (well in the link they are all 4, but I hope you get what I mean).

* BTW 17.3" 1080p looks grainier than my 1200p 17", so another point for 16:10 And while I'm at it, the perfect res (for me) for 17" is 2560x1600 (15.4" - 1920x1200), no scaling, pure space.

You need the other way around - eDP to LVDS, since that's what's missing. Anyway, you can see how small the chip is. It wont be that hard to be put on an MXM board, or do a separate board, given someone wants to. And since one is going to convert eDP to LVDS, his/her panel IS 1080p 60Hz*, so nothing to worry about there.

* Otherwise it would've been eDP and all would've been right in the world.

 

funny thing about resolutions: on phones, they dont use higher resolutions for extra space but make the icons bigger so that u basically end up with wasted space on a large screen.

on computers, u get icons and fonts that are increasingly tiny and impossible to read with increasing resolution. so instead of making both look sharper, they just make them smaller.

why not just switch this around and be happy in BOTH worlds?

 

Well, on Android we do have the DPI adjustment, Icon packs which helps us with the real estate but on Winblows X it's a disastah !!

 

Hey Raidriar did you ever test out a M295x? I remember reading a few months back you were strongly considering to do so?

I am now in the same boat although I know the M15x cooling will not handle it properly it being in the same "hot" league as the 980M..

It looks like if this is the end of mxm for lvds at least amd should be strongly considered due to robust driver optimistations to GCN.

M295x might just do very well in the right machine!

Sent from my SM-A500FU using Tapatalk

 

yes indeed, DPI adjustment FTW, otherwise id just die with those HUGE icons and wasted spaaaaaaaaace.... especially on my 6 inch display!

 

I considered it until I asked the manufacturer if it supported LVDS. they said they had no idea so I didn't pursue it. Would be a very expensive gamble at 450$

 

Pretty sure it does. The card goes back to 2014 a time before forced edp only support.

Sent from my SM-A500FU using Tapatalk

 

Just checked datasheet and pin out descriptions the card does output LVDS for certain.

Sent from my SM-A500FU using Tapatalk

 

So does the RX 480 Mobile...doesn't mean the LVDS support is in the vBIOS

 

I see. That is a potential problem!

Sent from my SM-A500FU using Tapatalk

 

The main problem in both cases is whether or not LVDS is cut out of the chip, or it's a matter of vBIOS. The MXM pin-out is understandably the same, I think they just copied the MXM spec and that's it. What and how is wired, and how it is implemented and is it in the vBIOS at all, these are the questions that would give a better understanding on the matter.

Because it doesn't work with Windows PCs. That's why I always said and would continue to do so - unless Windows makes the scaling work like on Mac, it's wasted pixels for me. That's why I prefer a resolution that I can make use of and everything would look good (no crappy scaling involved).

 

I guess there is no way other than to test it. The temptation is great. If only a test card could be loaned!

Sent from my SM-A500FU using Tapatalk

 

So I got my Clevo 8970M card, tested it in the M15x and...no native fan support

However, I put it in my M18x R1, and got fairly amazing results for such an old card. At 79% ASIC, I managed to pull a +150/150 on stock 1.025V for a clockspeed of 1000/1400, scored over 8K 3dmark11
http://www.3dmark.com/3dm11/11671573

Not bad GCN, not bad at all. Just took you some time to optimize.

 

Firestrike? 1.025V at 1000mhz is good! My old early batch 7970M managed only 970mhz at 1.05V

Sent from my SM-A500FU using Tapatalk

 

I don't have firestrike or any of the new 3dmarks, since they don't accurately stress the system enough as much as vantage or 3dmark11 do.