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

It is so sad that we can only assume things.

If they would release the base benchmark score for a single one, we could assume much more precise from there what to expect.

I have been thinking about how much realistic improvements we should be really expecting.

GPU as far as I understand is based on those many many little cores, shader cores or so. Titan X has 3072 of those. If they would bump it to 4096 cores, it would mean a 30%-50% improvement in performance. Then a GTX980 would be bumped from 2048 cores to 2500 something something cores. OR 3072 cores at best. And that would be equal to the actual Titan X. GTX980 can still play most games at 4K fluent, and FHD with ULTRA settings.

In another order of thought, GTX860M, which is the upper mid end has seen the most of an increment between Kepler and Maxwell, about 80% increase in performance between GTX760M and GTX860M and a very small 8% increase between GTX860M and 960M. The main question is how better would the mass produced GPUs also be. because, if you consider that most people will either have a GTX950m equivalent, you ought to consider that they might produce a tesla or titan or top of the line harnessing 10X performance for 4X the price or something, but if it all reflects in a small improvement for most users, it is just sad.

I am just here scratching my head on the thought that they might come up with something amazing, and all of people who buy something before the new toys come out are going to lose a lot.

On another note, GTX860M is roughly equal to a GTX680M. And GTX860M can play most games at almost Ultra with ease on FHD.

At the same time a GTX 880M is roughly 10% better than a GTX780M, where a GTX980m is over 50% better than 880M.

All of these, while GTX980 notebook is even better than all of the rest by another 30% - 40%.

The way these are evolving is not one bit linear, and while technology always advances, there is never a best, or absolutes, just temporary. It is just so sad that 980 notebook still struggles with 4K gaming, because it might have been the piece needed to make it not worth to upgrade to 1080 anymore. On the other hand, I think that with medium settings on 4K, if you pull 60fps constant + Gsync, you are better off than with ultra on FHD anyways.

Decisions, decisions and first world problems which don't let me sleep

 

That's why you need two desktop 980's in your p870dm-g. Gotta get your hands on some MSI cards . . . hehehe

 

Actually the 180W cards that are found in P775 would fit as SLI GTX980 Desktop in P870DM-G. But you would need at least 2 power adapters, and the cooling might not be enough to cool them off. They are being tested at the moment to be released as SLI GTX980 setup in P870, and if it is possible, from what I know.

As for me, I am a 1 GPU setup man. Why? ?Because all of my most GPU intensive software supports only single GPU setups, and will not be able to use the second GPU. I might want a SLI GPU setup for gaming, but for now, I think that having one very strong is the best for me. madVR ( the most intensive GPU thing I know of) only supports single GPU, and so do most software for media creation / 3d modelling / software development.

 

oooh - I didn't even realize the P775 was using 180 watt cards. Do you know if there's any place to buy these standalone? I've been trying to get my hands of a pair of MSI cards to no avail.

 

Your percentages are inaccurate because you're assuming a Pascal core performs the same as a Maxwell core, which is like saying a Kepler core performs like a Maxwell core. Pascal cores will be much more efficient than Maxwell, so you'll need many more Maxwell cores to match the performance.

Pascal will offer twice the performance per watt and two times the efficiency Maxwell offers. From this, we conclude that a single Pascal core will perform near the equivalent of two Maxwell cores. This means they can put 2048 Pascal cores in the 1080 and it will still perform 70%+ over the 980.

 

Spoiler

 

We do know at least two things about Pascal.
1. It is a die shrink from 28nm to 16nm.
2. It is an architecture upgrade from maxwell with some neat features.
These two things alone should be sufficient to be confident that 1080m will out perform GTX 980. I have no doubt that Pascal will have the performance crown once it is released. The only question is when.

 

I think the 1080m will perform within 10 percent, below or above, the 980 ti. obviously nvidia would want it to out perform the 980 laptops, and plus pascal will be more efficient, which clearly is better for laptops, and they keep pushing on shrinking the laptop/desktop gap. they wouldnt want their trend of the past decade to suddenly end. the 980m and 980 were a lot closer than the 680 and 680m. they would want the 1080m and 1080 to be even closer. so if some rumors are correct, the 1080m should perform equally or slightly better than a titan x/980 ti

 

I think you can buy off ebay or ask Clevo resellers.

But again, how do you cool these cards?

 

This has been discussed before if you're looking for responses, answers, etc.

 

when is this being released? hopefully by june? if not sept?

 

Are you serious?

This entire thread has been about people making slightly educated but mostly wild guesses at its performance and availability. Nobody knows.

 

Wait until GTC on April 4th. That's something legit.

 

One week is not a huge amount by any means, when you think of how wild were some guesses!

 

then i guess its end of the year if not early next year? time to go for polaris man

 

___________________________________

If mobile Pascal comes out by September, I will be shocked. I have seen a lot of comments here and elsewhere about it's release, and some have stated/guessed/speculated Q3 of 2016, and some say it's late fall, but my hunch for release is with the latter, not the former.
The release of 980m & 980 desktop was relatively late last year, and the spacing of the Pascal by this summer seems too soon.
As said by many here already, Nvidia is going to milk the 970m/980m with rebrands first, and that will stretch out their timetable release for at least 6 months after that, I believe, which puts Pascal to October, just in time for the Christmas Holiday rush.
No one knows when or even IF it will be out in 2016, either. There is simply no word or any indication that it will be released by Nvidia this year, or if they will wait until Q1 of 2017.
I am very strongly anticipating Pascal with great excitement because that is going to be my next gaming laptop purchase, and I really hope it's sooner than that, but I am not going to hold my breath on it for this summer, though.
I want to be so wrong about this....

 

thats what i speculated and expected 980mx but they pulled that apparently.

 

some news about polaris and samsung.. http://techreport.com/news/29899/samsung-updates-its-m-2-ssds-for-oems-with-its-polaris-controller

 

Those are some damn impressive SSDs but isnt this off topic for this thread?

 

well, no, because using polaris chip as it's controller. if pascal isnt coming until end of this year or early next year, then polaris is the way to go this year. AMD mobile GPU lets GO!

 

I dont think this Polaris has anything to do with AMD Polaris.

 

lol

yesss, new m.2 nvme ssds with integrated highend gpus, nice! *rofl*

Sent from my Huawei Mate 8 NXT-AL10 using Tapatalk

 

Some people can be so naive and not think before saying something!

Sent from my LG-H811 using Tapatalk

 

Seemed like trolling tbh..

Sent from a 128th Legion Stormtrooper 6P

 

No way in hell it will come out in October.

Sent from Mars by Matt Damon.

 

Matt, we hate to break it to you, but nobody knows about you being alive on Mars, or when Pascal is coming out...

 

polaris!!!!!!!!!!!!!!!!!!!!!!!!!!

 

POLARIS!!!



http://www.polaris.com/en-us/slingshot

 

lol, so random

Sent from my Huawei Mate 8 NXT-AL10 using Tapatalk

 

Spoiler

 

Ban hammer incoming.

Feels like GTC is so far away.

 

we can argue if the new cards will be 100 percent better or 50 percent better, but what about at 4k? will the better memory speeds help it scale better? 120fps at 1080p doesnt always equal 60fps at 4k. will pascal not be worked as hard at 4k?

 

Can't really answer this just yet. We know it will be better. That's about it.

April 4th is when Pascal should be announced. They'll also share a launch date at the announcement.

 

Do you think I can install it in my Clevo P870DM or will it require a new motherboard?

 

You have the "desktop" 980, so you'll have to wait and see what they do with the "desktop" 1080, if there is one.

That just came out a few months ago, so I bet the 1080 version won't come out until the end of 2016.

 

If they don't go crazy with making a new MXM standard, you should, theoretically be able to install the new GPU in P870DM. But expect it to take a year before they release the card in MXM form, and another few months until the firmware and vBIOS and BIOS are fully compatible.

I am also trying to be patient with this, because I really want to know what Pascal will be able to pull, because if it is able to pull enough over 980 desktop found in P870, I might prefer to order after it is released.

 

About this thing, I think that as long as it is stronger, it will be able to play 4K better. Consider that 4K graphics is in general full precision, not double precision nor half precision. This means, that it depends directly on how stronger it is compared to Maxwell, not on many tricks they include, like strong half precision calculation.

nVidia seems to be pushing low power computation, and other stuff, while they keep pushing away the fact that we need stronger GPU processing for VR and 4K gaming, and (next year or so) 8K gaming.

 

Wait a minute!. I always thought that all P870DM could be installed with a single GTX_980_MXM or one standard MXM3.0b or SLI standard MXM3.0b. Because GTX_980_MXM is just a longer version of standard MXM3.0b.
Is this not correct? Is there a difference in chassis?

 

Edited sorry wrong quote

Considering 4K has 4X the pixels to render per frame than 1080p its more like 120fps @ 1080p = 30fps @ 4K

 

|There is no difference in chasis.

But there might be differences on nvidia's part (like not releasing Pascal in MXM at all?).

Anyone in here has an older Clevo (like one that came with 5XX or 6XX or 7XX GPUs) and was able to fit 9XX MXM cards in it?

I am just curious if this is a thing that happens, or it would be the first time when this will be possible (an older chasis supporting newer GPUs)?

 

Besides that, consider that there is a lot of need for 120fps at 8K for VR headsets, so no matter how good new gpus will be, we still need more. and we need ASAP. I can't wait to fully immerse in the universe of a game. Think about seeing Hyperdimension Neptunia in total immersion .

 

Each 4k frame is over 8 million pixels, roughly four times the number of pixels compared to 1080p (not double). If you hope to achieve 60fps @ 4k, I suppose it's more comparable to 240fps @ 1080p.

 

its 200 percent of 1080p. i get about 50-60 fps in bf4 with my 880m. when i play it at 4k, it drops to about 20-25. sometime 30fps depending on the scene.

 

Raising that scale still doesn't represent accurately what 4K is.

 

i connect it to a 4k monitor. its native

 

Amazing! End of March and still absolutely nothing from Ngreedia.

I hope they really screw up with Pascal after all the rubbish drivers and this super long wait! Not to mention the overclock block malarky!

Sent from my SM-A500FU using Tapatalk

 

Ah, I thought you were just raising the resolution to 200% on a 1080p monitor and saying that's 4k.

My bad.

 

I really would like to buy the better card between Polaris/Vega and Pascal but I feel like this has to be taken into account...


Sad state of NVIDIA

lying about direct x 12 / Vulkan hardware support was great ?! 80%of the dgpu market will soon be obsolete because of this.

No hardware Async and still not fixed in software apparently

Kepler is now pathetic against old GCN, Will soon happen to Maxwell

Pay more than 100$ for G-Sync instead of Freesync.

Gameworks is designed to perform bad on AMD (for example it use x64 tessellation) and AMD has to optimize drivers after release but then the benchmarks already are skewed

Pathetic lawsuits against qualcomm and Samsung.

The GTX 970 3.5gb lie

Forced to use the bloated geforce experience for driver updates that break machines

Nvidia do no innovation compared with AMD, which has far less money (example HBM)

Refusing open standards like Freesync and open/mantle instead get ''Proprietary" gimpworst and G-sync

The throttling 880m that got swept under the rug

Trying to stop overclocking on mobile cards through drivers only to repealed there mistake after people threatening false advertising lawsuits

Not saying AMD is perfect but none of this nonsense lately on its graphics side and there drivers released today are another boost is seems on my 390 and m290x. I will wait until both are released before deciding what my next card will be desktop wise the next 980ti or titan or the next fury with hbm2. But this stuff is something to think about before supporting Nvidia....I know a lot of people have Nvidia cards on this form as there hardware junkies and I will probable get called a fanboy but I think it should be said.

 

Actually 400% (roughly 8m pixels vs ~2m pixels)

 

then why is cranking the resolution scale in bf4 and star wars battlefront to 200 percent at 1080p the same thing as 4k DSR?