Pascal: What do we know? Discussion, Latest News & Updates: 1000M Series GPU's

Yeah the 1080 Mobile version will definitely be that much better. But if there is a chance that they will release a full fledged on next year, or late this year, might as well fork out for that.
The prices on these will be so high

 

Yup I have confirmation that it will be at least 200 more than current 980M mxm module perhaps even the desktop one. Even the BGA laptops will be costing over around 3k+ for a 1080.

 

I think Nvidia finally makes price 5% less than that giant speculated amount you discuss now and you will say it's so darn cheap because cheaper than expected.

Except sales won't be so much successful so prices for 1060M (I still call it until officially declared deleted M) will be at only slightly higher than usually because nvidia always drops middle end hardware after top.

 

These cards will cost a ton, you can be sure about that

 

If this is all true, I'll be sticking with a 980M for quite a while.

 

Yeah, i might as well start messing around with a Razer Core, most of the other laptops that had issues, had Optimus and and Intel iGPU enabled. Im really curious to see if it will work for us.
Other than that i even have plans on trying our the mPCIe eGPU adapters. Lets see how this plays out.

 

Oh, the irony!

When people first saw P870, everyone thought that the 200W bigger MXM module is going to be hard to sell, and that the laptop having this big module is something different.

How ironic if it will end up being the only module that actually can get upgraded due to it's size which permits it.

I actually wish that everyone will be getting upgrades, from 771 to 870 owners, even people with lower tier that have MXM should be able to upgrade!

 

I would love to see a clevo working with the Core.

Especially since the modifications required are at BIOS level as far as I know and it would be easier for the Clevo community to tweak around that. Huge help to have an eGPU solution for a laptop with a desktop cpu.

 

I can't comment about other models as I've heard nada but an educated guess would be that only the 17" latest gen inchers will be good to go. To be honest, I don't expect many people to shell out immediately 1k+$ for the 1070 or so. Plus these modules are still at least 2 months away, I expect the BGA laptops to be out in September and modules in October..

 

I prefer MSI's eGPU solution with its full x16 PCIe connection. Now, if only they'd ever actually start selling somewhere this compatible laptop: https://www.msi.com/Laptop/GS40-6QD-Phantom.html

 

That BGA tho

Although I expect there will be at least 1-2 new eGPU solutions by Q4, mostly thunderbolt tho.

Nice to see MSI rocking that x16 tho

 

Yeah, some one got it working without issues with their Dell laptop, the problem was Intel Driver / iGPU related. Since my system doesn't even show the iGPU or have support for it, thats one issue taken care off. I have asked my friend to bring his Core over later this month. Gonna give it a shot then.

Oh and dont forget about the MXM to PCIe adapter that's being worked on. That one is the most promising for full 16x PCIe eGPU use.

 

On the desktop side of things, there's more info about the GTX 1060.
http://videocardz.com/62122/nvidia-geforce-gtx-1060-official-performance-leaked
http://videocardz.com/62127/nvidia-geforce-gtx-1060-available-for-preorder
http://videocardz.com/62182/nvidia-geforce-gtx-1060-3gb-to-feature-1152-cuda-cores

 

Oh great, let us know how you got on!

I think we're gonna have good news by the end of the year at the pace that these eGPU solutions are moving at.

 

Rip rx480.

 

http://videocardz.com/62138/nvidia-geforce-gtx-1060-reviewers-guide-leaked
According to this, Nvidia used an older AMD driver for the RX 480. So, as usual, let's wait for the reviewers' benchmarks to roll in.

 

Good luck doesn't mean be lucky.

You're right about more than 5 years. Much more.

Multi-task learning has been shown to help when you have tasks with potentially overlapping latent representations such as image captioning and translation. It doesn't mean you can slap together hundreds of different objective functions and datasets and say that it magically works better.

...? Deep learning isn't some magical voodoo which lets computers learn everything by throwing more and more data and compute power at the problem. All that it is is an optimization problem using deep ANNs as function approximators. The class of functions you can approximate well with any given network architecture is extremely limited. An ANN isn't going to suddenly go Skynet unless a) you give it the ability to do so and b) you have training data and an objective function which lend themselves to the task.

 

Guys everything is so problematic, No word from Prema also....looks very gloomy. I don't care if those two Clevos only get the socketed cards. And the asking price is absurd for these GPUs if the rumors are right, I'd go for Desktop instead of shelling cash on this BGA filthware and ridiculously overpriced gimped junk, 980M will be the last MXM that will be sold more than this Green Greedy junk...Can't imagine this is happening, last time when Maxwell was announced Alienware went full BGA retarded and now Pascal coming with no MXMs, next is what ?

 

Erm, that was the point.

I want to have AI to do two things. First, understand music, learn from studying music, and apply a set of very complex DSPs over music to make it better, more details on www.dobresculaboratories.blogspot.com
Second. This is a new project idea. I want to experiment something that is taboo. I want to actually program in a program that learns about the outside world, and can predict data based on other data, make decisions, create new data . Very very experimental, won't go full on with details.

Imagine something like a simple doll, that is able to analyse the face of a kid through cameras, understand expression, predict answers, raise mood of kids. Or go more in depth, add layers of sensors, the machine should be able to understand the time of day, what is happening inside a place, understand animals and what they communicate through a series of data on it, and many sensors. (this example is not what we're going for in the project, but the idea is good)

 

If they insist to bring the doom of laptops, I guess all of us will be using external GPU solutions, on weaker laptops. IT will be more expensive, less reliable, but in the end, we, enthusiasts, always find a way to overdo things.

 

I see nothing here about ML. Some DSP concepts and ideas, but nothing of substance as of yet. What would you be trying to train a network to do? Figure out some order and parametrization of plugins to apply to try to make music sound better? How are you quantifying your objective or reward function? Manually annotating hundreds of thousands or millions of examples in a consistent fashion to predict a processing pipeline is pretty much impossible. Crowdsourcing even simple question-answer datasets requires a ton of time and money and still produces pretty mediocre results.

What you're describing here is training a generative model.

This is to some extent plausible right now. Image understanding and captioning is a currently active area of research, and recognizing faces and emotions is actually relatively easy compared to many other tasks. Dataset generation would also be fairly straightforward.

Please don't take this as me just taking a dump all over your ideas. That's not the intention. But this is my career, and it pains me every time I see popular media coverage heralding every bit of machine learning news as a sign that the singularity is near. Anyways, I don't want to derail this thread any further, but if you'd like to discuss it further in PMs I'm open.

 

Indeed. Now, if only we could figure out how to mass produce e CPUs, we'd really be in business

 

I'd choose the 480x over 1060, knowing that AMD will continue to support the architecture for years to come while nvidia stops at year 1.

 

say what now? im all ears / eyes!

Sent from my Huawei Mate 8 NXT-AL10

 

Sorry about that

Cannot offer the actual method of what we're doing, everything after a bound is under NDA. It's not going to be free nor outsourced, it's a commercial project.

Will be clearer after we finish and patent the idea, so we can cover in depth what it's doing.

Seems that everything will become external. Might as well make a foldable display and we can take things away in... Wait a minute... That was the whole point of laptops...

Well, I can never understand what they're trying to do. I find it amusing that nowdays laptops aren't powerful enough to keep 25+ tabs of chrome open, but they keep telling us that it's better.

It's not. CPUs are getting more and more low power in laptops, GPUs are mostly BGA, they are an inch from dropping laptops all in all and make only some sort of monstrous tablet.

 

Anybody has a feeling that Nvidia throw in some Async Compute capability in 1060 as was done with H265 HW decoder in 960 in the past?

 

I wouldn't expect any big improvements there at least till Volta.

Proper full async would likely need architectural changes, while H265 decoder was probably added just "as-is" to existing architecture (add a new block to HW design vs re-design everything).

 

Don't buy Kepler, it's too old for your future proofing, unless you only wanna keep it for a year (maybe two) or if you're only gonna play old or currently existing games. I think a single 980M peforms better than 780M sli for gaming, especially as no sli issues.

 

https://www.techinferno.com/index.p...sion-project-complete-some-kinks-to-iron-out/
http://forum.notebookreview.com/thr...m-connector-an-external-pci-e-x16-box.407071/

 

How about if i get a deal of 2 pcs of 980M for 900$. Is it better than that whole system alienwAre 18 i7-4930mx+gtx 780M sli? I've got a clevo p375sm that has single 780M tho. .
Anyway my whole point is whether these cheaper last gen technologies is worth over the super expensive upcoming mobile pascal

 

980M sli is great - both for now & the future! $900 to upgrade your p375sm sounds good, don't know much if anything about p375sm so I can't comment on it's suitability as a base for 980M sli.

 

Some info about GTX 1060 (mobile) from PurePC (Polish portal, they already tested card/laptop).
Same as desktop version, but with lower clocks and TDP 65W.


Gaming benchmark results coming next month (after official GPU premiere).
Link:
http://www.purepc.pl/notebooki/geforce_gtx_1060_mobile_pelny_uklad_gp106_trafi_do_laptopow

 

Curious that for mobile we're seeing actual 1060 benches before any 1070/80.

 

There was allegedly a bench of an 1080 mobile which got removed, its swimming on some other sites who picked it up and claimed it their own though ...

Me comparing those values to my 970m

Spoiler

 

So, it's as many guessed, we will be getting desktop cards (gimped) and the 1080 MXM won't be there for sure then, following 980N, the 1060N (absurdly stupid gimp) replacing the 970M and 1070 replacing 980M.

This is the end of era, the MXM3.0b standard. To look on the positive side I hope that new one Clevo is putting will be standard and supported for upcoming cards / refreshes. One doubt is Clevo will they abandon all the 77xZM/DM owners ?? Would be really a godsend if they design the 1070 as a 1070 MXM3.0b..

Side note, If Nvidia is done with the MXM (which they designed) what about the AMD ? are there any chances that we might see those RX480s on MXM 3.0b spec ?

 

Yeah looking at the bench mark, the 1060 isnt even close to the stock 980 (notebook). With moderate 24/7 OC the 980 takes an even larger lead.

This would make sense though, because a lot of companies just released laptops using the desktop 980, just last month. Those will become literally obsolete if the 1060 would out perform the 980 based systems.

 

That is a nasty score for a mobile 1060
It's roughly 19% faster than stock 980M.
Tbh, the 980M itself was never a flagship to begin with. It was always 980 (with a huge margin).
so it is fair to assume in Laptop GPU
Mobile 1080 > titan X > mobile 1070 > 980Ti > 1080M > 980 > mobile 1060 > 1070M > 980M > 1060M > mobile 1050ti > 970M >/= 1050M

Cmiiw guys

Edit : i'm lost with nvidia naming their gpu

 

Great find, finally some more solid info. Thanks. Pity that 3dmark screenshot is so tiny (though it's the same size even on the original site).

Seems like 3dmark graphics score is 11,955, which would put notebook GTX 1060 at:

~90% of notebook GTX 980 / desktop GTX 1060
~116% of desktop GTX 970
~123% of GTX 980M
~157% of GTX 970M (which it effectively replaces)

 

65W & 123% of 980m! Thats not to bad.

Sent from my ONEPLUS A3003 using Tapatalk

 

Wow, that's some good performance for the 1060 in notebooks and not much lower clocks than desktop! 11,935/9,675*100 = 23% faster than 980M in Firestrike!

 

Not really.
This is a stock 980 : http://www.3dmark.com/fs/9224955

 

That guy has a nice username.

 

NotebookCheck has median firestrike graphics score (over 7 different notebooks) with GTX 980 at 13,047:

http://www.notebookcheck.net/NVIDIA-GeForce-GTX-980-Notebook.150599.0.html

So actually it's ~92% (keep in mind, GTX 980 for notebooks has been deployed in different chassis and different MXM formats, not all have the same TDP and the same performance; 1,228 MHz base clock would be on the high end of the range).

But even with 13,904 it would be ~86% so not really that much different .

 

@Robbo99999
Yeah just read that. You can translate whole articale on - http://www.purepc.pl/notebooki/geforce_gtx_1060_mobile_pelny_uklad_gp106_trafi_do_laptopow

And that's a great news!

 

True, but im just kinda possessive

 

Dats me bebe.

 

1. Great score on xx60. Like the idea of it being that strong

2. If It's gonna be BGA, it's gonna have optimus, it's gonna be a worthless junk, optimus means only problems for power users, and no average user buys xx60 laptops, most of laptops after xx60 don't offer lots of battery anyway, for that 10% saved by adding optimus, nVidia is shooting themselves in the leg.

3. Nvidia, We know you are listening on this channel. As well as all other channels. Please, don't drop MXM, regardless of risk and exposure. It's a needed product, there is a segment large enough to buy it. I know that eventually, there will be thicker laptops that will just have the full circuit part of desktop cards put in place by two screws and manual cooling, but until then, let's satisfy this segment, eh? And please, drop optimus in gaming laptops. Please. It's so bad that I considered buying something else, but since there is no other GPU this strong on a laptop, our only option is to hope you see that laptops with xx60 inside don't need optimus, as it locks down apps that developers need to run natively on GPU, even though they are not registered to do so.

 

They will just offer BGA soldered GPUs and you won't be able to upgrade them so you will be force to buy new laptop = more cash for manufacturer.

 

It makes less cash in the enthusiast segment, because people will avoid buying, will keep the same for a long period of time, and it will dry out. They need to think this through.

 

Put together all we in the forum with 980 and the median Firestrike score will raise to heaven And I haven't count with Overclock!!