Mobile Polaris Discussion

Yea, Ryzen probable won't come to laptop's till after raven ridge Q3 2017. I agree with desktops having more options for enthusiasts but its ALOT easier to pack up and take even a desktop replacement laptop with you if your not around home much. I have a nice desktop but my laptop see's 5x the use just because I use it at work and when I go on vacation.

 

I don't travel a lot, but when I do, its a lot easier for me to use my laptop on the go, or heck, even around the house.

I don't like to be chained to a desk as far as computers go... I like using my laptop on the bed. I just got really used to that.

Now, it would be a lot more preferable to find an affordable laptop solution.

I can manage with an Intel CPU, but I'd definitely prefer an AMD gpu simply because of better driver support and DX12 capabilites.
Ryzen CPU on top of that would be excellent, but as you say, it is unrealistic to expect it on mobile before the year is out (which is really a shame).

 

I get weird questions about a laptop worth double what a desktop would cost for same power. Then I just point out I can lug it outside. Or downstairs in front of the TV. Or on the dining room table for space. And >5kg plus two powerbricks that weigh more than most laptops I can still just about carry it all in one go

Intel have also said that mobile Cannonlake will be out by year end, claiming another 15% on top of kaby lake... but then they claimed KL was 15% over skylake so whatever to that. IMO theyre just trying to pee on AMD's parade coming up.


Does anyone have any input on whether RX480 is at all possible on Clevo. As I understand the two problems are:

- thermal sensor config (is this hardware and/or firmware issue?) Theoretically bypassable by mod or not?

- EDID issue (internal panels only) - again, potentially moddable into vBIOS if I understand? Would a vBIOS dump and edit be enough or is some reverse engineering or decrypting required first?

 

Has anyone tried to buy the TaoBao RX 480 MXM? How was it?

Otherwise, we may not see an upgrade until Vega from AMD. The good news is that Vega seems to have a variety of new features in it, making it a big upgrade.

 

10 Years of Flagship AMD Video Cards BENCHMARKED!

 

Hmmm I wonder if the E9550 vBios would work on the Gecube RX480 MXM

 

Very good news:

AMD Vega in notebooks
The most interesting part was covered by Scott Herkelman, a co-founder of BFG-Tech and former general GeForce manager at NVIDIA, currently a vice-president of AMD. His segment was entirely devoted to Vega.

So what did we learn? First off we were told us that Vega will go into notebooks. By decreasing the footprint of the memory by stacking VRAM dies, less space is required and therefore it’s more likely to be adopted by notebook makers. This never happened with Fiji, mostly because of high power requirements, but Vega is not just one graphics card, but a new architecture that will be adopted in various ways, one of them will be mobile Vega. Sadly, we don’t know if this means dedicated graphics cards or just integrated graphics that will go into APUs, but one way or the other, Vega will be present in notebooks. Scott explained that this will allow OEM partners to create “thinner and lighter notebooks, that still pack that punch you need to drive virtual reality or the latest and greatest AAA games” (for me that sounds like a dedicated GPU).

https://videocardz.com/67642/amd-radeon-rx-vega-is-just-around-the-corner

 

We can only hope it's MXM and the proper 3.0B type... That would be great!

 

The way it's worded makes me think MXMs are not a priority

 

Here's more information on Vega coming to notebooks:
http://hothardware.com/news/amd-radeon-rx-vega-shipping-with-up-to-8gb-of-hbm2-mobile-parts-incoming

It could be a respin of the above article, but 8GB of HBM2 might mean a dedicated gpu... then again, since HBM technology provides memory on the same die area, I don't think it would matter whether its an integrated part or not - though, I might have been under (admittedly incorrect) impression that it could be more expensive to produce (for now) higher capacity HBM RAM gpu's than lower capacity ones?
So, a gpu with say 4GB of HBM2 would (at least in theory) be cheaper than 8GB HBM2 gpu.

If that is the case, 4GB HBM2 would be likely as well as impressive for an IGP with massive bandwidth... and if Ryzen architecture can benefit from HBM2 at the same time for all workloads (without needing specialized HSA/HUMA supported software), that would be pretty good too.

Question is... would an IGP be able to even approach using 4GB of VRAM, or even 8GB for that matter?
And if the answer is 'yes'... just what kind of performance levels could we even expect?
460 to 470?

HBM allows the GPU to be scaled down by 50%.
So, I would imagine 460 or 470 level of performance for an IGP with HBM2 might at least be possible?

Mind you, with all Pascal gpu's being fitted into laptops, I could easily see the entire Vega lineup with HBM going into laptops as well at stock desktop levels (except of course for the top end Vega maybe if it has too high power consumption - but then again, HBM2 reduces power draw as well, so as of yet we still don't know what power consumption will be for all Vega gpu's).

At the very least the architecture is more advanced in comparison to Polaris as well as being more efficient.
Adoption into notebooks is another matter, but with Ryzen's viability in the desktop area, OEM's might (finally) provide us with best AMD options at a much lower price (vs Intel) without skimming on quality.

 

We'll see soon how much Vega improves. Polaris was a 'tock' - a die shrink which basically bought what was the 275W 390X down to 150W (or perhaps less with the 95W E9550 if that made its way to a comparable consumer part) and added a few optimisations here and there.

Vega is the 'tick' which hopefully drastically improves the internal efficiency. The RX480 has same ballpark peak 5-6 TFLOPS performance to a 1070 yet in games is way behind it. Hopefully its close enough that everyone bothers investing in drivers and optimisations to make it a genuinely competitive option!

AMD has a long road ahead to get there though, and their existing relationships in the mobile space are with brands I don't give the slightest fraction of a toss about.

 

Polaris as far as I understand it was an entirely new architecture in addition to being on a new manuf. process.
So, definitely not a simple die-shrink (you might be thinking of Pascal though which was touted to be a die-shrink of Maxwell essentially with even losing some IPC).

Also the 480 had some voltage issues early on resulting in higher TDP.
After undervolting, 480 seems to be able to consume as much energy as 1060, and AMD was able to further improve upon that for mobile (resulting in E9550 - making it a more efficient mobile gpu than even 1060 with comparable performance).

Vega is yet another different architecture.

TFLOPS do not translate to gaming directly... they do however indicate potential performance in professional software.
So, on a pro software note, 480 is 1070 equivalent more or less.

As for OEM's... we'll see what happens in that department, but now that AMD's stocks have gone up significantly, and Ryzen is basically a much better bang for buck than Intel, there is a chance OEM's might decide to uptake Ryen, and of course Vega - though of course, I could be wrong.

We've seen though OEM's do like to give Nvidia preferential treatment and install their power hungry desktop GPU's into laptops while completely leaving AMD out of the picture, or if they make AMD systems, they frequently use limiting RAM (even though APU's depend heavily on dual and quad channels and of course faster memory for better performance), not to mention low quality screens, too high chassis, etc.

 

FAR from a new architecture. Even including the Fury/HBM experiment, Vega is by far the biggest change in GCN since it first debuted.
e.g. http://www.anandtech.com/show/10446/the-amd-radeon-rx-480-preview/3 - Polaris uarch changes were basically improvements aimed at feeding the cores better to boost efficiency. This was the relevance of my mention of TFLOPs, comparative efficiency (not as a measure of absolute performance). The Polaris cores themselves not changed from fiji/tonga.

Polaris was yet another 'value' oriented part from AMD, and they are set to live on as RX560 RX570 RX580 again with optimisations. To get AMD making money again they need 'performance' oriented parts to earn the same kind of margins a Titan or an MXM part earns Nbloodyvidia. Yet by all rumours even big Vega is going to fall somewhere between 1080 and 1080Ti which is yet again competitive, but not outcompeting.

 

How do AMD Enduro capable MXMs work in Optimus supporting laptops? I ask because laptops that support Optimus use the iGPU as a middleman so all video technically goes through it so now I waonder did anyone try the 480 MXM in an optimus laptop?

 

http://forum.notebookreview.com/threads/possible-mxm-rx480-card-on-clevo-laptops.799952/
P150SM - fail due to separate Clevo thermal sensor reason but GPUz detected card ok (same failure as in P750ZM which has no iGPU)
I read somewhere it was tried in an MSI of some flavour but didn't work on internal display (lacked EDIDs in vBIOS) - probably, but not sure if optimus

 

MSI units tend to have optimus?

 

Take a look here (i'm not at knowledge level to explain it)

http://www.zentrica.com/wp-content/uploads/2014/04/HD7970M-MXM-3-0-SPEC.pdf