nVidia 2015 mobile speculation thread

In your scenario, I think the GM204 would be more stable, especially with overclocking. It would not only be 30%~ faster than the 980M at stock, but could probably be overclocked another 20%. The GM200 version, while it may be a good 40%-50% over the 980M at stock, would probably remain unstable when overclocked, and be quite a hot card. Technically, they'd perform similarly. But the GM204 should be more stable and run cooler. Just my guess... Can't know for sure. We've never seen a mobile GM200 card before, so we have no idea how it would handle thermals inside current systems.

As always, my vote has been for a full GM204, even though I kind of want to see GM200 (just cause).

 

You (and the copypasta places like CEG/Xtreme PC Gamer) are alone on the GM200 wagon. Everyone else is saying GM204.

 

So funny. It would actually be the other way around

 

Well, if they perform the same, doesn't matter which way. Stability matters. GM204 would be more stable.

You so funny, bubba.

 

Wouldnt you want a GPU that you can overclock higher if you have enough PSU juice and at the same time dissipate heat better?

I would 100%...

 

Again, another issue. Why would manufacturers start making larger PSU's all of a sudden after moving to smaller ones?

Doesn't make sense to pick up and turn around.

 

But you don't actually care. "RIP mobile gaming" and you moved to desktops ages ago, remember? You're just here to stir the sh*t.

 

No I dont care so much about 990M since I`m not buying it. From a tecnical standpoint I`m interested, but I said adios to notebooks a while back.

I`m actually waiting to see what the new R9 Nano`s from AMD can do. They would be extremely interesting in a notebook too btw.
But I think HBM and MXM conflicts are maybe holding them back from releasing it on a mobile card

 

You really are an odd one, arguing tooth and nail over a figment of your imagination, as if you knew something the rest of us don't, which it turns out you never actually cared for all along

 

Imagination is one of the most important things in life. Don`t forget that

Oh well, time to go. Nice chitchat

 

GM200 / GM204, I can see both working and being pretty good options, I'd like to see GM200 for the spectacle & the potential for it to be pushed further with extreme overclocking efforts. But, regardless of which option NVidia decide to go with - when can we expect to hear some official announcements do you think (or dead cert leaks)?

 

Shouldn't be too long now. September is the last month until Q4.

 

I will settle with both setup. But I don't think I will buy one until thanksgiving, consider my semester is about to start...

 

Still zero mention in any drivers or software.

Flashback to GTX 780M:
Driver leaks in March 2013. Launch in May 2013
Source: http://forum.notebookreview.com/thr...upcoming-gtx-780m.710409/page-18#post-9132631

 

This has been an entertaining thread. Speculation threads are always entertaining. I'll just bite my lip this time around and enjoy the show.

 

Quite honestly, I don't even know why everyone is so hopeful nVidia will provide something exciting. AMD is now completely irrelevant in the mobile space, and even if nVidia were to charge 20% more for the existing 900 mobile GPUs, people would still buy them, because they have no other alternative.

Yes I know historically they've always had a 12-16 month release cycle, but again AMD provided some degree of competition all the way up until 7970M, and 8970M you could at least argue was a value alternative to 780M. For 970M and 980M, AMD has absolutely nothing to counter with, and M295X in the very best case scenario still falls short of 970M while consuming 50% more power. nVidia has no competition left in the mobile space, just like Intel has no competition left in the high end desktop segment. To think that nVidia would go out of their way to provide an earth-shattering mobile GPU is just deluding yourself.

So basically, this is my tl;dr way of saying the best we can hope for is a full GM204 mobile chip, but be prepared to pay BIG for it.

 

I agree entirely. Nvidia won't bother with a GM200 chip at 2x 980m performance at $1600 even if it is physically possible. That wouldn't milk the proverbial financial cow. Offering 25-30% performance boost, basically a desktop 980 in a mobile platform, for $800 would garner more significant financial gains. I'm not sure what a soldered only solution would bring for Nvidia. This means it would have a very limited user base, even moreso than MXM based laptops.

 

And finally, it makes zero sense for their future. If Nvidia gives us 2x the 980M now, how will they build hype for Pascal? The well fed stomach does not run to the dinner bell. Unless they know right now that the mobile flagship will be 3x to 4x the 980M, why would they release such a powerful card on the eve of their biggest release in years? The market for x90M flagships is incredibly small, and the number of yearly upgraders is downright minuscule to nonexistant. How many people will buy a $1500 2x 980M this year, then turn around and buy the Pascal laptops next year?

Not even the laptop manufacturers would want such a card to drop in 2015.

 

Cool, so once we see some mentions of the 990M in drivers or software it could well be just a couple of months from launch.

 

NVidia can still make some extra money by releasing an impressive mobile GPU even if there is no competition from AMD. Yes, I agree, whatever will be released will be expensive!

I don't think the GM200 chip would be 2x980M performance (just not really practical from a thermal & current consumption perspective), I think that's a red herring, maybe 40-50% more performance maximum in my eyes. Full GM204 at 25-30% extra performance over 980M would still be pretty awesome too.

 

I think people need to remember that with Pascal we get:
- Insane increase in transistors (atleast 2x as many) vs Maxwell
- We get FinFET design. What makes it differ from normal SOI process we have now is that FinFET reduce leakage, which in turn reduce TDP. Which means Nvidia can increase clocks and cores for mobile where TDP is important. Its one of the key reasons why Intel ditched SOI and started using 3D transistors with 22nm and why Intel have such big lead against AMD in CPU designs which is without foundry and had to rely on TSMC and GloFo which only had SOI to offer for their CPUs.
- We get HBM which in turn will increase performance over GDDR5 cards. Not drastically more but enough to be noticed. On top of the extra performance this will give, we also get a power reduction vs GDDR5, which means they can allocate more power for the GPU cores which could further increase performance.

Nvidia might want to release the fastest they can with GTX 990M because Pascal will then be close by, and Nvidia might want to change the public perception that they have been slowballing mobile Maxwell. In all honesty, GTX 980M is exactly that, slowballing.
I think the high end mobile Pascal will beat GTX 990M by such great extent anyway, they might as well give it everything she got

 

I sure hope you are right Cloudfire, but call me a skeptic. But why does Nvidia care if there is a perception that they are slow balling mobile GPU's? It's not like people won't buy the new tech anyhow.

 

Ya'll be crazy folk.

Der ain't gon be no GM200 990M 'cause dem fellas at NVIDIA ain't no good at making progress without dem AMD folks givin' incentive to do so.

 

Regarding variable TDPs and naming schemes, the 860m came as both Kepler and Maxwell. Quite confusing at times.

 

Let's have variable VRAM, too. It will automatically enable (and disable) modules when needed to help with heat.

 

Crucially though they didn't vary in performance by more than around ~10%.

It's not the actual TDP being so variable that I find utterly redonkulous but rather the performance. A named GPU shouldn't vary in stock performance by more than a few percentage points, in my opinion.

 

But do you think Pascal will 3 to 4 times faster than the GTX 980M?

 

We saw 980M ID's in GPU-Z and drivers two months before release. None of that for the 990M?

 

Nvidia be like: ZOMG, 990M haz TOO MUCH POWAH! how in hell are we supposed to write drivers for dat? idea: sell extra driver package at release for 800 bucks, total package = 1600!

 

Exactly. The TDP difference b/w Kepler and Maxwell 860M (75W vs. 45W) was much greater than the perf difference because the latter was much more efficient. But since 990M is Maxwell only, a highly variable TDP like before will result in an unacceptably large perf variation.

 

Intel has been intentionally dragging their feet for 4 years now, yet people still buy their CPUs year after year, and a not insignificant amount of enthusiasts still "upgrade" every generation, even though that 5% improvement barely means anything outside of benchmarks.

So I agree, nVidia doesn't give a rat's ass what people's perception of them are (remember Bumpgate, and 970's 3.5GB vramgate? people either didn't notice or didn't give a damn), because people will keep buying them anyway. Plus when you have 80% marketshare, you automatically have an army of rabid, frothing at the mouth zealots doing PR spin for you in every which way possible, so they don't even have to worry about perception either.

 

I kinda just want the 990M to come out, so we can close this thread. Tired of seeing it at the top of the gaming section.

More 990M threads forthcoming!

 

CEG said week 40.
This guy from Clevo subforum said

Drivers and codenames are always added way before launch. OEMs need to test and verify that the card is working in their machines and need time for that before launch. So Nvidia add support and unreleased cards pop up in OEM specific drivers like Asus or Clevo.

So I find this situation very odd. Everything about this card actually

Maybe you are right HTWingNut. GTX 980M is a damn shame really. So much potential left and the GPU was how much faster than 780M? Not that GTX 980 was a disaster performance wise against 780Ti anyway. But they atleast got GTX 980Ti and GM200. Would be cool to see a 600mm beast in mobile too you know

I have no idea really. Doubt it will be that fast. I think like the desktop Fiji cards, the true strength of Pascal will be in bandwidth intensive resolutions. Will go nicely with the new 4K display for 17" as well as 3K displays for 15" and with DSR.

 

Yep, week 40 is only 6 weeks away, so as you say we should probably have seen 990M pop up in drivers already, praps it will be later than Week 40 (Week 40 = Sept 28th).

 

http://forums.laptopvideo2go.com/to...k-driver-34804/?do=findComment&comment=149477



And it was just April...

 

Don't think that's it.

 

It's very much the g-sync edition of 'it'...

 

So, you're saying the 990M "g-sync" version ID has been available since April? 100% certain, without a doubt?

Where's the regular?

 

ure such a tease @Prema

Sent from my Nexus 5 using Tapatalk

 

April fools.

 

It appeared in a Lenovo subdriver.

Are you sure its not related to the new P50/P70 Lenovo workstations with Maxwell Quadro cards?

The codename doesnt exactly scream Geforce...

 

We saw much more stuff regarding the 970M and 980M by this time last year, including 3DMark benchmarks from Chinese forums.

Let's hope some show up soon.

 

when the super full pascal titan comes out, i'll be getting 2 of them and putting them in sli, maybe 3 if nvidia keeps true to their better scaling on multi gpus. then i'll hook them up to a 5k monitor. prices should be down then too.

 

The one in April was a Gigabyte driver...
I don't know if production units will use the same ID but that's what has been used until now:

13DA: regular version
161A: g-sync version

The new Quadros are:

13B1 = Quadro M1000M
13FA = Quadro M3000M
13F8 = Quadro M5000M

 

...next to 1TB DDR4-5000 RAM, a 20-core extreme desktop cpu @5.0 Ghz and 4 PCIe 4.0 x32 M.2 NVME drives with 10TB each in RAID 0

 

I'm stepping out but if Prema is getting involved... That says it all unless he's just bored and trolling everyone lol.

 

Might as well start mining for silicon and creating my own cards XD

 

maybe skylake-e will have 10-12 cores because of the shrink. lol. or would broadwell-e have that?

 

Can we get back to reality please? Every core adds latency... This is Intel, not AMD.