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

Are you kidding? This thread is amazing. There's even speak of aliens.

 

I think Half-Life 3 was confirmed somewhere too... It's the X-files of the internetz.

 

_______________________________________________

Well, what we DO know, is that following the UFO alien crash in Roswell, New Mexico in 1947, the Nvidia Research and Development department was immediately dispatched to the area BEFORE farmer Mac Brazel or the Government Army Corp got there, and went through the wreckage site, and retrieved very valuable pieces of Alien technologies and artifacts.
Nvidia has now incorporated them both into the new upcoming Pascal GPU hardware, and it will be advertised heavily during the X-Files TV commercials, according to what I was told by Fox Mulder.....

 

it glows green in the dark, is powered by zero-point energy and can even make you coffee!

 

No actually its all part of Illuminati NWO. The want us to believe that Nvidia is alien technology, but inf act the beings present at Rosswell were not aliens but green lizard people. They were immediately hired by Nvidia for their R&D. You think its just a coincidence that Nidia has chosen same color as the color of lizard nation? Wake up sheeple!

 

What if nvidia executives are the lizard peoples..... and they are seeking to control us through their GPUs.....

 

You spelt that wrong my friend.



There we go.

 

In case anyone hasn't heard, Radeon's have an AMA on Reddit tomorrow (today!).

(AMD) Radeon Technologies Group Q&A is happening here on March 3rd 10AM-5PM central time.
https://www.reddit.com/r/Amd/comments/48e8rl/radeon_technologies_group_qa_is_happening_here_on/

"edit: 10 hours from now! It was very smart of everyone to post their questions here. Doing so has given AMD personnel an early look at what info they will need readily available to answer everything quickly and thoroughly."

AMD - Radeon Technologies Group AMA March 3rd 10AM-5PM Central
http://forum.notebookreview.com/thr...-group-ama-march-3rd-10am-5pm-central.788920/

 

I think we should get back to the topic before a moderator bans us all.

 

Only one response to laptop based questions so far- and the answer was one word.
Polaris.
Really weak AMA so far. Too much red tape for them to have held this.

 

Was that the question you asked? What answer did he give?

Please post in the new thread for the AMA, updates here:

AMD - Radeon Technologies Group AMA March 3rd 10AM-5PM Central
http://forum.notebookreview.com/thr...ch-3rd-10am-5pm-central.788920/#post-10212255

 

Hey,
Is it true that nvidia pascal will be a big thing for notebooks? I heard something like "double performance per watt", because it will get manufactured in 16nm instead of 28nm. I dont haven't a clue about these things, but the last time the technology/manufacturung process was shrinked down was 2012, so I guess it is a bing thingy, right?

I wanted to buy the "new" thinkpad T560 with the "new" nvidia 940mx, but I kinda have the feeling that I would buy like the third rebrand of the same model and that it makes sense to wait (which I can). Another thing to consider is, that the actual maxwell gpus seem to not really support directx 12.

And another question: AMD has also a new architecture called "polaris" upcoming, also in 16nm. Does that mean that there will be amd gpus in notebooks again?

Thank you.

 

Big? Yes. Double performance per watt? Maybe. That's what NVIDIA claims but you have to read between the lines.

We all expect at least 50% more FPS on the top-end cards. Guarantee you they'll intentionally hold back the first generation Pascal cards to re-brand for years to come. I'm sure they plan to milk Pascal just as they milked 28nm (680M, 780M, 880M, 980M).

 

Nvidia is totting massive improvements, but most of it will be in double precision calculations which only affects GPGPU applications found in scientific research. Nothing for gamers in that. In fact when nvidia released maxwell, they ignored compute performance and its even somewhat worse then previous gen kepler.
For games, it is not sure which "kind" of Pascal will be there for laptops, but roughly expect a 30%-50% performance improvement.

Maxwell does support DX12. It only doesnt support a small feature of DX 12 which is called "Asynchronous compute". There are no GPU in market which support all features of GPU . AMD GPUs dosent support "Raster Order views". But in DX12 benchmark AMD DESKTOP card do hold their own against Nvidia. In mobile, they are far behind.

Polaris is 14nm, Pascal is 16nm. But 2nm is not a big difference.
Polaris is expected to be released around mid-2016. Pascal later (maybe by end of year)

 

Alright thanks.

I just compared two notebooks, one with 560m and the other with 660m (560m = 40nm , 660m = 28nm) -> last time where nvidia shrinked the manufacturing process of notebook gpus (by 42%, 28nm -> 16nm = 42% aswell).

XMG P501 Pro (i7 2720qm, gtx 560m) http://www.notebookcheck.net/Review-Schenker-XMG-P501-PRO-Notebook.56654.0.html
- 3,1 kg
- 48 x 376 x 256 (height x width x depth)
- two fans, two heatpipes for GPU
- GPU at max 85°C
- 46.1 db max

XMG A522 (i7 3610QM, gtx 660m) http://www.notebookcheck.net/Review-Schenker-XMG-A522-Clevo-W350ET-barebones-Notebook.80018.0.html
- 2,8kg (10% lighter)
- 44 x 374 x 250 (12% less volume)
- one single fan, one heatpipe for GPU
- GPU at max 75°C (13% cooler)
- 49.2 db max

-> pretty much same performance between 560m and 660m as far as I know

In my eyes it made quit a big difference between 560m and 660m. Maybe there will be similiar difference in maxwell -> pascal?

 

Here's something to chew on from Sweclockers

"VBulletin has previously been able to uncover that Nvidia is planning to announce Pascal for laptops at Computex 2016 , which takes place May 31 The sharp release, however, is thought to later and reportedly to editors Nvidia be published on the 16 June to its partner manufacturers." (Google translated)

http://m.sweclockers.com/nyhet/2182...l-for-barbara-datorer-slapps-i-mitten-av-juni

 

About that mobile GPU...
http://www.overclock.net/t/1592444/...ted-gp104-256bit-and-will-come-with-4gb-gddr5

 

4GB VRAM is low for a 2016 flagship mobile card though. Good news that they could be basing it on an MXM board similar to layout as 980M - good heatsink compatibility for a start.

 

Woohoo, Pascal laptops launching June 16? I think I can put up with my hp a few more months.

 

lol, thats what uve been saying since december 2014

 

LMFAO.


Unfortunately I am in the same boat.

 

Still all speculation.

Sent from my VS980 4G using Tapatalk

 

Haha, I remember, MovieMarketing with his faithful 2009 Envy still!

 

Same for me with my G73JH. I think we have reached a state of Zen, where pain is gone.
I was desperate for a new laptop in beginning of last year. I was getting frustrating with playing games on lower settings and the last straw was when my screen broke. But I decided to wait and got the screen fixed. I started to play older game. Now I think if push comes to shove I can last another year with my trusty G73JH and older games.

 

Jeez, thanks for making me spit coffee all over the keyboard. Now I really need to upgrade

 

Sadly yes.... always seems to be something better just around the corner. I need to do some kind of benchmark showdown whenever I finally upgrade. Is it going to be like a 10x faster level of difference with 5830M vs whatever is the Pascal version of 980M?

 

Well a 780M runs 3DMark Vantage 6 times faster than the 5830M according to notebookcheck.net. 980M is about 80% faster than 780M, therefore 1.8x6 = 11, therefore 980M is 11 times faster than 5830M. And finally, the Pascal GPU mobile high end might be 80% faster than 980M - so 1.8x11 = 20, therefore Pascal GPU could be 20 times faster than your 5830M. That's how I see it - I'd say that's a worthwhile upgrade - that's going from 6fps to 120fps!

 

It sounds unbelievable, but likely true...

 

The likelyhood of Pascal being 80% faster than 980M is a bit out there though.
I wouldn't expect more than 50% increase since that seems more in line with historical progression.

 

Historical Progression:

285M Tesla --> 485M Fermi: Vantage P GPU: 5631 to 12191 = +116%
485M Fermi --> 680M Kepler: 3DM11 P GPU: 2709 to 5897 = +118%
680M Kepler --> 980M Maxwell: 3DM11 P GPU: 5897 to 12531 = +112% / Fire Strike Graphics: 4049 to 9682 = +139%

980M Maxwell --> 1080M Pascal ==> ??? at least +100%!?

 

That's my hope and was my original prediction. Pascal should be 100%+ faster. However, I'm anticipating them to hold back Pascal, intentionally, at least for the first release. There will probably be at least two (if not more) iterations of Pascal, 2017 offering another 50%+ over this year's release.

Pascal's shelf-life will not end this year. That we can all agree on. Re-branding is game within the game.

 

Yep. Just that Maxwell never officially got a refresh of the top end GPU's like other architectures did. Just more of the same, really. 860M and lower got refreshes to the 9xxM level, but 970m and 980m still trucking along after 18 months without even a refresh in sight.

 

The jump between Kepler and Maxwell should have been even bigger. Remember originally Maxwell was supposed to be die shrink. But they had to skip 20nm due to issues with TSMC and finfet.
And I think 780m was worthwhile re-brand of 680m. I would have been happy to see a similar re-brand of 980m.
Edit: Remember this?

 

That is not a pascal...

 

Right. Just an architecture change improved performance dramatically on the same node size.

 

Shouldnt we start to see 1080M leaks soon if its available in products in mid June? Only 3 months away

 

Yeah, should be seeing leaks around GTC in April.

If NVIDIA doesn't announce Pascal and present a working Pascal GPU at GTC, then they probably won't launch until late summer.

 

Do you imagine Nvidia will repeat the (desktop) GTX 980 for mobile experiment with Pascal's GP204 or whatever it will be called? If so, is that likely to be another year or two away?

Otherwise, the full GTX 980 with 2048 CUDA cores in P870DM for example is still going to be faster than the 1080M for the foreseeable future, no?

 

The 980M is about 30% less powerful than the 980. The 1080M will definitely be more powerful than a single 980.

From what I recall when they made the 980M, NVIDIA wanted to close the gap between desktop and mobile graphics performance, so seeing another 1080 "mobile version" is very likely, IMO.

 

Hi,
I got same questions:
- Do you think there is a significant performance improve in "casual" gpus like 1040m / 1050m?
- I read a lot, that pascal could be hot, but why? If they improve the performance per watt, why should pascal be hotter than maxwell? For example, when I compare the tests of notebooks (notebookcheck.net), kepler GPUs run at max like 10°C cooler than fermi GPUs (while having more or less the same cooling system).

 

Who knows what the performance of GP104/1080M will be.
I would guess a bit faster than GTX 980 but I do not expect a lot faster. For one the GTX 980M is a massive chip, it is 400mm2 and I think Nvidia would want to go down to 350mm2 again which these chips usually have been.
Secondly, I think it will be a 100W GPU which means it get a disadvantage against the 165W GTX 980 we now have.

If I should guess, 40-60% faster than GTX 980M and 0-20% faster than notebook GTX 980.
If its another 165W GPU, that of course changes things. But I think Nvidia will stick with 100W with disabled cores first, then followed up by a full GP104 chip later on in 2017 which is either a 105W GPU if they are successful at reducing voltage with the Pascal silicon so they can add more cores without going up in wattage. Or it is another 165W GPU.

GTX 680M was about 40-50% faster in average than GTX 580M when it launched. It was bigger, and was based on a new architecture and a new node. I thought it was almost twice as fast, but these links says otherwise
http://www.notebookcheck.net/Review-GeForce-GTX-680M-vs-Radeon-HD-7970M.77110.0.html
http://www.anandtech.com/show/6336/...book-review-nvidias-geforce-gtx-680m-in-sli/3

 

980M is really a 120W TDP MXM GPU, and if they make equivalent Pascal TDP, I would expect nearly double performance from the die shrink and architecture changes. If they drop to a 100W TDP well, then it's not really an apples to apples comparison. Would be a bit disappointing to say the least. Depends on what vRAM is used as well, etc.

 

LOL. Isnt it just a m5000m card?

 

One thing's for sure: don't expect 100% GP104 potential out the gate. We were stuck on 28nm for 4 years (and counting), and we'll be on 16/14nm for even longer, so it's gonna be a loooooooooooooooooong milking process.

 

"Double performance per watt over Maxwell if you double the watts."

 

And triple the price ofcourse......