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

And I never said that you did. Don't apply different meaning to my words and to yours

 

Cos judging people by a cover and telling them what best they should do with their money if very open minded and virtuous right.. teach me more sir

 

A bit dramatic when you're only talking about someone who just decides they wanna buy 4 GPUs!

 

In all fairness, buying four 1080s is a bit dramatic too.

 

"Dramatic" in the sense of PC ownership, but not in terms of an individual's wider context - not when using OTT phrases like "being a wasteful being" and "[putting] more into the world than you take out" were being used earlier. Haha, that's enough from me on this topic, don't want to derail the thread, even though it's quite used to happy derailment, we know a little derailment is OK!

 

This isnt price vs performance notebookreview.com.

I am not exactly sure how is someone buying 4 x 1080s affecting anything at all? By your judgement, is LN2 overclocking bad too? PC hardware is a hobby, just like some people like cars. If someone gets enjoyment from benchmarking with 4 x 1080s, who are we do judge them?

I imagine that would be the exact same as Split Frame Rendering. Which have horrible scaling I believe.

 

I'm feeling dramatic. Was just seeing how people responded. Don't really care.

Your money, your life. Waste away.

 

Quick, define trolling in 3 short sentences.

 

Guys this topic is a little long and full of information.

For the summary. Are announced GTX 1080 Mobile and GTX 1070 Mobile their full desktop counterpards or they are crippled M series just without M in the name?

Thanks for replying

 

Crippled desktop cards

 

At this point of time there is a conflicting info.

Some rumours have them as desktop GPUs but with tweaked TDPs (kinda like notebook GTX 980 was), but the most tangible leak of notebook Pascal (in MSI GT73) had GPU specs which were unlike any existing desktop Pascal GPU.

 

Interesting idea. I actually honestly have never though of doing this. Could see it working, if I can find a bunk bed small enough. Definitely worth looking into! Cheers for the idea

 

I ended up calling up MSI once more on another subject but we discussed the GPU upgrade options and the guy talking let slip that the MXM cards this year will feature a lot of physical incompatibility with a lot of other brands on the market. (Not just MSI cards.) It seems the cards are MXM but will not be the same exactly type and will be different in size, along with another rumour i heard from a clevo reseller that only 2 clevo models will be able to accommodate the card, it seems we're in for a few changes...

If I had to take a guess I think "M" versions are done, crippled desktop cards seem much more likely.

 

@Omnomberry Social experiment!

@marios50 Are you saying you believe there's a new MXM type (e.g. type C)?

 

That's mad money. I'll stick to clevo with something cheap for half the cost with my one graphics card LOL.


Sent from my iPhone using Tapatalk

 

I don't want to read really deep into what he said but my guess is, Nvidia will be producing different MXM cards for different notebook manufacturers with slight physical adjustments. (Or something along those lines.)

 

This is exactly my idea on what SLI should do. I mean, they could design a processor, or even say, require a 4th card that distriutes the data among the other three, and we could be playing the games on three monitor setups better than if all cards were trying to draw one big scene. Even the micro stuttering would disapear in that case.

Nope and nope.

Hardware for deep learning is much more expensive / in both price and quantity of hardware required. One GTX 980 can only process so much, we are talking about true deep learning and designing an AI here.

And cannot use any datasets available online or anywhere, that is what I am building. A complete new algorithm, not just redoing all that was done over and over again. Give the software the ability to read, interpret visual cues, understand music through data processing, differentiate things, learn, and ultimately understand basic emotions. Haven't said this will be finished in 10 years from now, but it will be something interesting to see.

Deep learning provided a bit of things so far, the future is at hand

 

Uh, good luck with that lol. Deep reading is in its infancy right now, let alone general purpose AI. I also think you drastically underestimate either the amount of effort needed to generate data or the sheer number of training examples you need for deep learning, especially for the task you're trying to solve.

The hardware also isn't even that expensive. It only takes a few weeks for a few K40s to run through a multi-TB dataset, and a K40 is much weaker than a Titan X when it comes to FP32 performance. For a few thousand you can get 4 of them, which is peanuts compared to the sort of capital you need for lots of other scientific research.

 

That's also how I believe it should be handled.

 

Guys "crippled" desktop cards are only crippled if they have cut ROPs/ TMUs. If they are just underclocked you can overclock them again so no problem here. I would like to have desktop gpus in next laptops but IF there will be 18" laptop. If not then im not changing my M18x

 

You can still buy Gt80 or what the new name will be This New mobile will never come with max 1080 chips now. Fully 1080 chips with +- 125w..., Ehh what a joke

 

MSI Gt80 price is ridicoulus.

TDP can even be 200W for each card, I don't give a damn. Just make a water cooling system or something.

 

GG !
MXM 3.0b is done then ? I'm broke now, what about the 77xZM /DM then if only 775DM and 870DM are supported does the 775DM host the non-MXM3.0b standard card ?

 

I doubt that the MXM standard is changing. If it was, Clevo would have announced the newer system refreshes.

 

Yeah.

It's nothing luck. It's science.

It can be done and it will be done. Not today and not in less than 5 years, as far as I can forsee.

It was misunderstood so far. when trying to make a true AI, people tried to make it work for specific tasks, it always lacked some other departments that make a human being work. For example, we can have an AI read and understand things, and process them, but if it lacks other type of information, like simulation of senses, simulation of images, simulation of space, lateral thinking, it will never be able to do that.

The problem posed can be solved by multi processing, it would take a sever the size of a larger house, made out of i7+GTX980 cards all running in parallel and a single board distributing the data to actually be able to process this. The problem posed by this architecture is it's inefficiency. An efficient similar construction would rely data between nodes and process the data as sets of specific data, with a singular node controlling the type. In this case, a single AI mind can be created from something like 100 GPUs + few CPUs, given they would have great interconnects. It would be able to think, come with new information based on the old information, and do other fun things. The problem with the design is that it cannot access internet. Ever. Random data exposure would either overload it and break it, or make it learn how to re-program itself, infect itself in all the other computers, creating a singular autonomous mind. Of course, with today's internet, it would be a failed abomination, it won't be able to neither sustain nor communicate well with itself, but it would pose a problem



Back to graphic cards:

 

I only see the Phoenix being upgradable to be honest, all I'm saying is, I've heard that new cards won't fit physically in other machines at least thats what I make of it.

 

It's ironic.

P870DM is the only laptop with that shape of MXM and 200W video card XD

 

Exactly... you know how the 200W card is bigger than the others?
There's the key

 

No. It's just that notebook manufacturers will have to design their own boards because Nvidia chose not to design a reference MXM boards anymore.

 

As long as the card has the same TDP and similar physical appearance it should be compatible with most of manufacturers vendor. Unless the shape is that kind of mxm that was used in Asus G750JH/JZ iirc. Very weird pcb shape indeed.
MSI, Clevo, alienware (if not BGA lol) should have non peculiar MXM GPU
But if it is just BIOS or software wise, prema/svl7 will always take care of it.

Weird mxm from G750JZ gtx 880M

 

Maybe ASUS signed a decade long contract on supply MXM boards with Ugliest Board Manufacturer Inc.

 

Really these corporations hare hell bent on pushing this BGA filth and lockdown bullcrap on us....If ASUS, Dell (Precision, Alienware) also stayed with reference PCBs Nvidia would't have done this and that AMD sitting just plain and happy with the console profits ultimately made the green greed raking profits everywhere....

 

This smell more and more desktop for me!! First BGA filth then different form factor. Goodbye notebooks!!

 

The cards will definitely be 1060, 1070 and 1080. I've heard this from 3 difference sources, one of them is usually 99% correct about these things.

Also they're gonna be hella expensive too, we're talking more than a 980M module, quite a bit more.

I always believed they would have ditched the "M", the 980 signaled it.

 

oh crap, this doesnt sound good at all..

- nvidia stopped giving out mxm reference designs from pascal onwards, so every OEM is doing his own thing now. good-bye mxm standard and hello mxm NON-standard! @as many different form factors as there are laptops

- only two clevo machines being supported: thats 100% P775DM and P870DM, since they can both take non-standard mxm boards, all others are EOL...

- nvidia dropping the M in their mobile lineup has just one single reason: make regular mobile gpus and sell them for a hefty price premium due to average joe thinking: ZOMG! i can haz desktop gpu in crapple laptop :-O "founders edition", anyone?!?!

man, this is SO gonna suck for mobile enthusiasts... *sigh* and yeah no, not talking about the BGAlienware gang here...

Sent from my Huawei Mate 8 NXT-AL10

 

stop buying alienware then your telling them its ok.... sheesh

 

hunh?

 

Thing is, performance will be very close to the actual desktop card at least in the 980 it was and I suspect such will be the case with the rest. As for a price tag, expect over 1100+ Euros if i had to make an educated guess. At least for the 1080.

 

New idea for nvidia: rename the 1060 and 1070 as "1080" next.

 

thats the thing though: mobile 1080 will be a desktop 1070 performance wise, so why not just call it that? price! and yes, id wager wed have to pay the same exact price for a "non M" mobile desktop gpu as we do now with the 980 mobile. big difference there: the mobile 980 DOES actually perform like a true desktop 980 and is not disguised as a desktop 970 in 980 flavour...

so basically: throw a 980M successor onto the market, call it 1080 instead of 1080M, DONT change performance compared to desktop, ask for 50% price premium and be happy that everyone's still buying your crap like crazy.... *ugh*

Sent from my Huawei Mate 8 NXT-AL10

 

So when actually will the real early bench of mobile 1080 and 1070 are coming?at least early legit benchmark with full specification, it gives me hopes.
I'm getting tired of waiting.
Almost pulled a trigger on a 999$ USD alienware 18 i7-4930mx+SLI gtx 780M. Is it a good deal or is kepler getting real cheap lol.
From my point of view, at least mobile 1080 must beat 980 mobile by 50% at least (26K ish graphic score 3DMark11 or 19K graphics score 3Dmark firestrike). All stock. Then i am sold

 

I don't see actual cards being sold before late September/early October tbh.

As for performance, 40-50% sounds about right for the 1080.

 

If you're right, and mobile versions are roughly the same as one-step down from their desktop counterparts, and knowing that a desktop 1060 is roughly on par with a desktop 980, then...

mobile 1080 = desktop 1070
mobile 1070 = desktop 1060 = desktop 980
mobile 1060 = desktop 1050? = 980M?

Time will tell...

 

The funny thing is, for a mobile version of gtx 980, they are charging at least 2700$ and 4500$ for its SLI(only god knows why) at launch.
If they do really stepping their greed, i can't imagine anything less than 3000$ for a single mobile gtx 1080+i7 6820HK. Supported by the fact tjat Asus G752VS with gtx 1070 mobile will start around 2000 Euro. Might as well buy a gtx 980M SLI really and get the performance of mobile gtx 1080 (if it really perform like desktop 1070). Ahh.. 3 months to save before burning my pocket

 

If really nothing fits the batman, I'll just get some dirtcheap 980m
Or we can wait if there will really be a prema-edition laptop

 

that sounds about right...

Sent from my Huawei Mate 8 NXT-AL10

 

Ill stick with my 980 if that's the case, and work on some solution to get the desktop 1080/1080ti working externally.

 

The P870DM should work with whatever they have.. The P775DM should be fine too I guess.. It us P771ZM/DM owners that are screwed... I especially wouldn't be very happy with Clevo if I were a P771DM owner but hopefully we are not left in the cold...

 

Problem is the gimping. If the 1080 is not up to spec to compared to the desktop iteration, i really wont want it. Specially for the price premium than that these things will retail for. If they do release the 1:1 version of the 1080, then i will jump on it. For that 15-20% gain (980 vs the 1070/1080m?) forking out $1100+ isn't my cup of tea.

 

P870DM is a 100% P775DM is the second model if i had to make one guess.

I don't see it being 1:1 with the desktop card. I see it being about maybe 30-40% better than the 980 desktop perhaps.