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

Update: First day of GDC, and AMD announced their Radeon Pro Duo, maybe Pascal will give a showing at GDC...

AMD Radeon Pro Duo (Fury X2)
http://forum.notebookreview.com/threads/amd-radeon-pro-duo-fury-x2.789406/

--------------------------------------------------------------
Exactly, with Nvidia you never know when they will drop something interesting, not just Pascal

Spoiler

All GDC Sponsored Sessions, page search for Nvidia, sessions spread over the week.
http://schedule.gdconf.com/format/sponsored-session

NVIDIA Sessions, Thursday, March 17
A full day of presentations: Thursday, March 17, 2016, West Hall, Room 3014
http://www.nvidia.com/object/gdc2016-thursday-sessions.html

GDC Nvidia BOOTH INFORMATION
Moscone Convention Center, South Hall, Booth #824
Wednesday, March 16
10:00 AM – 6:00 PM
Thursday, March 17
10:00 AM – 6:00 PM
Friday, March 18
10:00 AM – 3:00 PM

If anyone is going, please ask about Pascal and let us know what Nvidia says.

AMD's GDC info:
http://www.amd.com/en-us/who-we-are/corporate-information/events/gdc

Capsaicin brought to you by AMD Radeon Graphics
http://www.ustream.tv/channel/gQETjZzLnf9?sf22465712=1

GDC 2016 Youtube videos, new ones throughout and after conference this week:
https://www.youtube.com/results?search_query=gdc+2016

 

First day of GDC 2016, and AMD announced their Radeon Pro Duo, maybe Pascal will give a showing at GDC...

AMD shows Radeon Pro Duo (Radeon Fury X2) playing Hitman 1440P + AMD Vega + HBM "3" discussed

AMD Radeon Pro Duo (Fury X2)
http://forum.notebookreview.com/threads/amd-radeon-pro-duo-fury-x2.789406/

Capsaicin brought to you by AMD Radeon Graphics
http://www.ustream.tv/channel/gQETjZzLnf9?sf22465712=1

 

Ah yes, I saw that news on Guru3D this morning, they have a performance graph there too: http://www.guru3d.com/news-story/dual-fiji-amd-radeon-pro-duo-announced.html

I commented on Guru3D there: "That's weird, the new Dual GPU card is 20% faster than the 295x2 at two of the resolutions, but at the middle resolution it's 38% faster - Firestrike graphs. I just don't think that's a good enough chunk faster than 295x2 - I think this card is a bit pointless, especially as Polaris is supposed to be coming out quite soon too.".

Maybe Polaris is not coming out as soon as we thought!?

 

Delays acrosses the board it looks like.

 

AMD's roadmap suggests mid-late 2016 for Polaris, and HBM2 early 2017.

99% certain NVIDIA will beat them to the punch again. But that's not even the slightest bit surprising, given their history.

 

Haha... So no real performance card in 2016 from AMD. Lovely.

 

To me, from that roadmap, it looks like Polaris is slap bang in the middle of 2016, rather than the "mid-late 2016" that you pointed out. Hopefully Polaris is coming out soon, could do with some change-up, plus will get NVidia's ass into gear!

 

Mid-late = mid to late.

June / July is "mid," right around when NVIDIA is planning to launch. But AMD is always late.

 

man this sucks. Seems both polaris and pascal won't be embracing HMB for the time being.

 

How much would the lack of HBM affect the gpu?

Sent from my SGH-M919V using Tapatalk

 

Except that if you look at the graph, it would appear that Vega (with HBM2) is slated for end of 2016/early 2017 (visually wise, and at least to me, it looks more like late 2016 than early 2017).

I don't think this was unexpected.
Both Nvidia and AMD said that they will be likely using GDDR5 and/or GDDR5x (likely for entry and mid level GPU's).
Higher end GPU's are probably going to be slated for HBM2 on both ends.

 

I think the mobile space this was expected however the HBM should be in high end desktop GPUs... Hopefully the mobile ones at least have GDDR5X and if not HBM1 would be good..

 

I could do with HBM1... though honestly, it doesn't make too much sense to use HBM1 when HBM2 is ready to go and offers higher RAM capacity (not that 8GB of VRAM would matter too much if you are gaming on 1080p for instance - though, on higher resolutions such as 4k and above... yeah).
Sure its likely in limited supply, but this is probably why HBM2 will be used on top-end GPU's.

We have 0 word on HBM1/2 use on laptop top-end gpu's though.
It would be nice of AMD to surprise us like that if it can (and if it can, then it might end up surpassing Nvidia in the laptop area with it too - which would mean that AMD would have a far more attractive product at their hands)

 

no hbm2 until 2017. so gddr5/5x/hbm1 for 2016 mobile gpu yeah.

also looks like pascal just isnt ready so 1080m could be just a die shrink from current 980m maxwell.

 

Nvidia's new Youtube Channel "NVIDIA GAMING NETWORK", first video posted 39 minutes ago...

What's Going On at GDC 2016?

 

Hey, HTWingNut, do you think somebody stole our creation?! (Well your creation actually, I elaborated on it a bit!)
http://www.guru3d.com/news-story/nvidia-geforce-x80-and-x80-ti-pascal-specs.html

 

Pascal Specs Leaked?

http://www.guru3d.com/news-story/nvidia-geforce-x80-and-x80-ti-pascal-specs.html
http://wccftech.com/nvidia-pascal-specs/

 

__________________________________

Is this confirmed?
I would hope that it's not, and that Pascal is going to be out this summer, as earlier rumors had indicated.
We had heard reports that Pascal would be out in the 3rd quarter (after July 1st) for mobile, but if what you allude to is correct, then it seems like it won't be out until at least 4th quarter this year or maybe sometime a year from now.
I am shying away from Maxwell and the mobile Nvidia offerings for my next gaming Laptop until Pascal comes out in what seems like now late fall of this year, but I am hoping that for something much sooner than that.

 

not sure however if theres just HBM1 it'l be different design of card, means existing heatsink prob wont work. we can't upgrade anyway.. given fact that they design the same way desktop did with HBM and furyx

 

I don't think we have any confirmations, however, AMD did manage to demo Pascal at the very least twice now. There was nothing from Nvidia, apart from separate websites claiming Pascal is on track for 2016... and that maybe it could debut first Pascal gpu's in May.
Given that we know little about Pascal along with lack of any demo releases... it is possible Pascal will be delayed - by how much, dunno.
The upcoming releases could be mere demonstrations of Pascal without actual gpu's being ready.

 

______________________________

Yeah, upgrading from existing laptops is a definite no-go from my 880m that I have now, which is obviously due to architectural design differences, stemming from it's 16mm die size.
I had planned to swap out my 880m to the 980m, but I was told that there would be only a slight positive impact on gaming performance- ie, FPS rates, so I decided to wait for Pascal and a new Laptop this fall IF it comes out by then, that is.....
Based on past offerings from Nvidia in the past 6 or so years, from the 580 to the 680, 780/880 to the 980, Pascal should deliver an extra 25% performance boost from the previous mobile GPU (980m), so that would mean that we could be seeing performance numbers and FPS rates equal to the current 980 Desktop mobile card for Laptops, and that is reason enough to sit tight and wait for this October if and when it comes out.

 

i'd prob go amd this time if they manage to catch up on performance. nvidia got no compute which in some area i need

 

Polaris was confirmed to have 2.5x performance per watt vs Fiji.
Fiji was already quite competitive with Maxwell in both performance and TDP (yes, its results were slightly lower with a bit higher power draw), but, when you apply 2.5x perf. per watt increase for Polaris, and that Pascal was mentioned to have 2x perf. per watt... Polaris should be able to provide relatively better performance than even Pascal (or at least, it has that prospect).

I'm personally waiting on AMD this time around as I would prefer not going with Nvidia unless it becomes really necessary... though I don't think this will be the case.

 

all about dat async for me. since i game at just 900p and medium-high settings, i dont need to max out on anything. 100-120w card would be perfect for a bit of future proof, though 980m is already way beyond capable of doing just that, and this up coming graphics expected at least 50% increase in performance? IM GAME.

@hmscott looks like its a good thing i skipped out on GT80s, 6820hk can't really overclock at 40x thats nothing lol and 980m is simply outdated technology, IF new GPU can fit into our existing mxm slot then we all good to go, if not then it's 980m upgrade for me.

 

Yeah, Alienware is dead for Pascal most likely. Maybe someone will pick up the cause and get it working, or are you going to try to do that yourself?

The GT80(S) / GT72(S) are promised 2 generations of GPU upgrades, so that's nice, both Pascal and Volta I assume.

The next generation GT80's/GT72's could come with whatever CPU replaces the 6700k/6600k, but draw less power.

That would be fun, lower power higher performance CPU's/GPU's - maybe that laptop won't need 2 x 330w power adapters just to run.

 

2cents x10s

 

Possible Pascal mobile GPU 3Dmark 11 benchmarks:

http://www.3dmark.com/3dm11/11061015

http://www.3dmark.com/3dm11/10969217

http://www.3dmark.com/3dm11/10852916

 

It's starting

 

That first result is a desktop mainboard



http://www.gigabyte.com/products/product-page.aspx?pid=3771#ov

Sent from a 128th Legion Stormtrooper 6P

 

Stop! It's a leak I tell you!

 

Those results are underwhelming unless they are the midrange or something. They are 980ti or less. Like way less?

 

If it is MXM. My M15x will go pascal...cause supported. Question is will the newer craptops support a new architecture on mxm with all the system bios blocks, graphics switching etc. I DOUBT IT


Sent from my SM-A500FU using Tapatalk

 

The article claimed these were likely Pascal mobile cards, but hfm indicated at least one is a desktop GPU.

 

doubt it, no UEFI, unless you're AW15 not m15x or m15x r2. m18xr1 is dead on 780m can't go further apparently and thats at least 1-2 generation after m15x r2.

we just now need polaris benchmark for a show down, hopes on AMD.

 

Clearly you no little about the Dellienware M15x everything from gt 240M to 980M works in the M15x plug and play.

There are no issues except for in some cases heat. Yes every single mxm 3.0 card released in 7 years works. Basically the system bios really does not discrimate against any particular card nor care whether it is UEFI based.

My 980M works perfectly.

Sent from my SM-A500FU using Tapatalk

 

note to clevo: pls derive ALL future clevo system bios versions from the one used in the m15x

 

True that... You listening @Prema? Pass the message on to Clevo

 

clearly i dont, just from speculation as i dont own m15x. if m18x r1 doesnt take it and it's released later, then likely no support for it. definitely won't be plug and play though, as u'd need vbios mod and bunch of other stuff done to bios EPROM etc.

 

If only! I think it's just pure greed that is the main motivator for not creating system BIOS like the M15x - these companies just don't want us keeping our laptops for 10 years! Some of the motivators for these more locked down system BIOS could be security reasons though from what I've heard, but it's still in their financial interest to prevent upgradeability of laptop platforms - the swines!

 

In a nutshell mate

Untiyole. Why speculate when you have hard facts. Sadly M15x is more likely to work with Pascal than new maxwell mxm based laptops.....

Sent from my SM-A500FU using Tapatalk

 

That's because Dell made no use of the spare OEM customizeable MXM PINs in the M15x model (NVIDIA altered their use with every new generation of cards and broke many systems in return)...so turns out that less is more.

 

That is interesting to read Prema! Thank you for. sharing

So if the reason is so simple and success so obvious it is clearly a sign of greed on OEMs behalf.

Sent from my SM-A500FU using Tapatalk

 

True, about the OEM, but there's always the BIOS to mess the support for AMDs in my case... No OEM use here, just something nasty in the BIOS (maybe part of the M7820 vBIOS) that screws compatibility for any other AMD but the HP M7820. Almost all nGreedias work fine with a few exceptions that don't work at all (660m and 765m).

 

I am a bit curious when will Pascal be coming in Desktop form to laptops. Like how GTX980 is to a laptop. Because I highly doubt that a mobile pascal with m in it's name would beat 980 desktop in notebooks.

EDIT::: Going through this thread, I saw this.

http://cdn.wccftech.com/wp-content/uploads/2016/03/NVIDIA-Pascal-Zauba-Spotting.jpg

Out of curiosity, the price per unit for one of those things is 1200USD. I think that Pascal will be faster, but much more expensive if this is the case.

EDIT2::: This thread started at the end of 2014.

 

Mobile Pascal will definitely beat desktop GTX 980. 980M is 9700 points in Firestrike, 980 desktop (in a notebook) is 12400 points in the same test. The GTX 980 is only 28% faster than the 980M, whereas Pascal will ultimately double the performance of the 980M, so you can see that mobile Pascal will thrash the 980 in a notebook.

 

I have been thinking about this thing with doubling the performance.

Ultimately, the core GPU is the true bottleneck, even though making vRAM faster should improve performance.

Now, the 16nm are supposed to bring 2X the performance per watt, but look at it this way: Intel processors for laptops 4710hq and 6700hq. 4710hq is build using 22nm and has a score in R15 of 135 per single core and 646 per all cores. 6700hq is built using 14 nm FinFET transistors and has a score of 144 and 676. The improvement in performance is not great, considering that they have the same TDP.

This makes me question if Pascal would really be able to bring double the performance, or it is a Nvidia marketing thing.

Furry X which uses stacked memory is pretty fast, and faster for certain things, but it certainly was not the one fastest card on the market, even at it's best point.

 

u cant compare cpu gens like that, since intel is so focused on improving igpu performance as opposed to actual cpu performance. comparing sandy bridge with skylale u have a ~30% improvement in IPC / performance per clock, whereas igpu performance improved by ~850%

Sent from my Huawei Mate 8 NXT-AL10 using Tapatalk

 

We don't know this yet, and even if one Pascal-chip does it's not certain a mobile Pascal will.

 

I missed a lot while on my trip! Wow, leaks and everything.