Brace yourself: NEW MAXWELL CARDS INCOMING!

Best case it will be 2.4X faster than 860M, and that's with very high clocks and a TDP as high as 880M's. Realistically it will be lower though.

 

Hmmmm, interesting. Much of what I've read from VideoCardz.com, regarding timelines of development or product releases, is speculation and rumors. Can you please site the source of your statement above? Do you have a link perhaps? Thanks!

 

It's all based on rumors and speculation, and rumors are based on leaks. Leaks may be real or made up. So, in the end, we still don't know for certain until NVIDIA releases an official statement.

 

Yeah, I assumed as much, but since the card designation, notebook, and timeframe were so specific, I had to wonder if I missed something. Thanks J.Dre!

 

Sorry, cant say anything about the sources.
You are right that most of what is posted is pure speculation, but the recent news are not. I can say that much.

Except the Asus G760 with 980M, that one was something I picked up on in early July where someone said 980M would launch in a "few months." But it falls pretty close to the GT72 so I have no reason to believe it is wrong.

Videocardz know quite a few important people, which is why they can make articles like that.

I think you will be pleasantly surprised.

 

Nvidia GTX 980M Maxwell GPU Arriving In October | eTeknix

 

Hehe... All these articles I've seen sources lead back to Cloudfire

 

I know it's still to soon but what are your opinion about waiting for this new GPU's to upgrade my laptop (m18x-R2) with 680M for this new ones?
I'm quite happy with my current setup but I hope new gen games come to us soon and I want to be able to play them maxed out with a good frame rate and high res. Also thinking about new monitor maybe 4K capable...

So... what do you think time to start saving money?

 

Can't hurt. I'll probably upgrade if they're decent. Just have to wait and see.

 

Dont expect 4K to run well on laptops for couple more years.

 

So what do you guys think the specs for GTX 980M will be. Dont be afraid to guess

GTX 660M GK107: 384 cores (2SMX) 118mm2
GTX 680 GK104: 1536 cores (8SMX) 295mm2 (+2.5x die size)
GTX 680M GK104: 1344 cores (7SMX) 295mm2 (+2.5x die size)

GTX 860M GM107: 640 cores (5SMM) 148mm2
GTX 880 GM204: ? cores (?SMM) (?mm2) ~430mm2 (+2.9x die size)
GTX 980M GM204: ? cores (?SMM) ~430mm2 (+2.9x die size)

 

I seriously doubt I will be surprised, 860M came out not too long ago, it is a 45W chip just being the onboard variant, the MXM is already 60W. Even with improved process the best they can do is 2.4X the performance (best case, will be lower in real benchmarks) of 860M by consuming 110-120W of power, and I am being generous here, these numbers are already breaking math.
If anyone read the comments on the WCCF article about GTX 880, they will know that most people are already disappointed about desktop 880. Remember the single benchmark Cloud found for 880, plenty of people claim it is fake or using an extreme overclock (not air cooled). Even Videocards hints that even though it may be faster than 780, most likely it won't outperform 780Ti, and the expected $500 launch price says just that.
After all this there is absolutely no way 980M will perform anywhere near a 780Ti, maybe with an extreme overclock and pulling 150W+, so these new benchmarks are fakes too. I am honestly telling you guys, don't ride the hype train, don't expect too much, keep your expectations within a reasonable range (2 - 2.4X the performance of the 860M), and in the end you won't be completely disappointed because of all the hype, just like some people who call this "the next big over hype from nvidia".

 

There aren't any real consequences if we ride the hype train , I mean the only thing that could happen is being disappointed nothing really serious , being disappointed is part of life

 

We will see how much the opinions here will change when R9 M295X hits the market. Very curious about how that one will stack up against GTX 880M.

Anyhow, anyone that wants to guess the specs on the 980M based on the above?

 

I'm willing to bet it will perform like a desktop GTX 780, perhaps underclocked a tiny bit. Not sure how many cores.

 

Wait so there will be a newer version for the current 860M? Also if so when do we expect to be able to buy them??

 

Yeah. I'd say a bit behind 780. Even a perfect power increase of 2.4 over 860m maxwell puts it between GTX770 and GTX780 in performance. A nice increase in performance.

As for M295x, I bet it will be slower, falling in line with R9 280x, so more like a GTX770 rather than 780

 

I'm sure 980M SLI will be able to at "acceptable" frame rates. Although, in 2016, when Pascal and Skylake hit the market, we should be able to finally game with 4k displays. Unless they skip it or delay it forever, I'm pretty sure 2016 is going to be an amazing year for gaming.

 

LOL! I just Google'd 'NVidia GTX-980M', and ALL of the tech sites are referencing this thread, and Cloudfire as the source. At least most of these articles indicate, 'speculation or rumors'. I am just surprised with all the frenzied response, all over speculation and rumors, and the folks at NVidia are sure to be rolling their eyes..... but I guess it all makes for good press, keeping us excited while we wait, doesn't it? I mean, geez, we all know the new tech is coming, but other than the manufacturers, no one really knows the details, and even 'leaked' information cannot be deemed reliable. All of this reminds me of the myriad of 'stock-tip' speculation. Too funny!

And to think all this began with an .INF file from laptopvideo2go.com..........

 

That's not surprising. Anyone have a time machine, yet?

 

You should have been here when Kepler speculation was the big thing, Striker.
It starts with driver leaks, move on to AIDA64 updates, to leaked benchmarks to notebooks listed with the cards, to people getting their hands on one, to discussions how good they are compared to previous generation, to heated discussions about which brand is the best and then finally silence and acceptance once the dust have settled.

Always fun to be part of the whole speculation all the way from the start

 

It gets more fun as time passes by Cloud hehehe I'll replace my year old 780m once the new card comes out. Need all the GPU power possible with my notebook right now. I'm doing some research regarding a new kind of high energy beam particle detector CMOS sensor and analysing all that sample data (~1million samples per sensor for a total of 3 sensor prototypes) sure takes it's time with the CPU so we developed a GPGPU OpenCL solution Quadro cards are out of my budget. The only option is consumer Geforce/Radeon series. Since my GT60 would cause me a bit of trouble with a R9-295mx my only option si the new toy from Nvidia. I have a good feeling

 

I'd be very happy if this thing simply surpasses the desktop GTX 770 by a small margin. A laptop with a GPU that surpasses the 770 would be drool worthy.

 

There are three desktop GPUs coming in September/October:
GTX 880
GTX 870
GTX 860

I have a very strong feeling that all those three are based on the GM204 chip. So there is very big chance that the lowest one, GTX 860 will be a 128/192bit GM204 with a lot of cores disabled.
And there are big chances that there will be a mobile version very similar, so yes, I expect a new improved mobile GPU called GTX 960M, unless it will be called GTX 970M. I think there should be something for everyone when 900M launches.

But latest rumors say that the GTX 860 wont arrive until late October, so you might have to wait until late October/November until you get the mobile equivalent

Good bet. I also think around GTX 780 sounds very plausible. Some overclocking and we have ourselves a GTX 780 Ti

Gotta stay entertained.

Have you seen the OpenCL benchmarks from GM107? A ton better than the GK107 (GT 640). They are both 60W. Maxwell is very good at GPGPU it seems.
http://www.tomshardware.com/reviews/geforce-gtx-750-ti-review,3750-16.html
GM204 will probably be beastly against GTX 780M. Hopefully they increased the L2 cache greatly as well.

Do you happen to work at CERN btw?

 

780 Ti? I think you're dreaming just a bit too hard....

Sent from my HTC One_M8 using Tapatalk

 

We will see Ethrem. I still expect close to this for 980M
http://i.imgur.com/iZFDPHn.jpg

 

We can only hope. It would be great to have that level of performance combined with the "overclockability" the GTX 680M provided. Now, that'd be something. Also, here's me hoping the 880M issues are resolved for the 980M. If they're not, wow. NVIDIA would have really dropped the ball, then.

 

Fingers crossed someone is ready with a 3K/4K display. Would be such a perfect match.
I have talked with someone who works for Alienware and they said they have been testing out higher resoultions (than 1080p) when I asked about the next Alienware 18, so there are hope they have something new there.
They didnt want to tell me anything other than that :/

 

I don't see it. 250W TDP Kepler performance will fit into 110-120W TDP Maxwell in the first generation? Makes no sense.

Sent from my HTC One_M8 using Tapatalk

 

GTX 650 Ti Boost: 134W
GTX 750 Ti: 60W
(GTX 860M: 45W)
Almost the same performance between 650 Ti and 750 Ti (-8%)

Apply that to:
GTX 780 Ti: 250W TDP
GTX 880: ?? TDP
(GTX 980M: ?? TDP)
It makes perfect sense...

Whatever, believe what you want. Just wait

 

I guess we will find out. I'm going to remain pessimistic and say that the 980M will be around 10-15% faster than the 880M would be if it worked right.

Sent from my HTC One_M8 using Tapatalk

 

The 780M is approximately 20% better than the 680M (stock), and that was a re-brand. I'm pretty sure an entirely new architecture will provide more than 10% - 15% performance over the crippled 880M, considering the 780M can outperform the 880M, lol. I wouldn't even count the 880M because it performs worse than its predecessor.

The performance increase from the GTX 580M to GTX 680M is what we should see from the 780M to the 980M, which is approximately the performance of a GTX 780. When it's overclocked, it may be within the reach of the GTX 780Ti...who knows.

EDIT: Originally, I had thought the 980M would perform at about the GTX 680M SLI level. The two cards after the 680M have just been re-branded.

We'll have to wait and see.

 

I said 880M if it worked right meaning a 993MHz 780M with 8GB of RAM.

I don't think we have the thermal capacity to have the power of GK110 in a laptop and that's regardless of whether GM204 is even able to achieve that kind of performance in the first place. GK110 was and probably still is the most powerful graphics core ever created, I just don't see Maxwell shrinking that power into something that is usable in a laptop.

There's a great thing about being a pessimist though - if things go better than we expect, we're pleasantly surprised rather than sorely disappointed.

 

The thing about 580M to 680M though is that it was an architecture change AND a die shrink.

I know we mentioned bigger dies here but I wonder just how big you can push the die on mobile MXM cards before they blow up in your face. Certainly to reach desktop 780 performance you will need a 50% bigger die (Cloud gave some pretty good calculations a few pages back).

 

Here is what I have pieced together from multiple leaks that have been repeatedly posted from different sources on different sites at different times.
Most of this are not just rumors but almost certain at this point.

GTX 880: GM204, 20SMM, 2560 cores, ??? core clock (base), 256bit bus, 4GB memory, 7GHz+ memory clock, ~= 430mm^2 die area (approximation), $500 launch price (rumor), September launch
GTX 870: GM204, <= 18 SMM (estimate, most likely 16), <= 2304 cores (estimate, most likely 2048), similar or lower core clock, 256bit bus, 4GB(?) memory, <= GTX 880 memory clock, ~= $400
GTX 860: GM206, 12SMM, 1536 cores, 1??? core clock (base), 196bit bus, 3GB memory, 6GHz+ memory clock, ~= 250mm^2 die area (approximation), October launch

Now some guesstimates
GTX 980M: GM206 (if not a severely cut down GM204, which is less likely, they can just release it later as 980MX when they use a smaller manufacturing process or manage to get the power consumption down some other way), 12SMM, 1536 cores, 1??? core clock (boost), 196bit bus, 6GB memory (most likely option), > 5GHz memory clock (most likely 6GHz), ~= 250mm^2 die area, ridiculously high price, october launch.

GTX 970M: even higher chance of being based on GM206, probably some SMM units cut down, same 196bit memory bus but lower clocks, 3GB(?) memory (most likely), much lower price (although still overpriced), same launch time as 980M

GTX 960M could be anything really, including a 860M rebrand with slightly higher clocks.

 

Well, it could be that, too. We still don't know anything for sure. At this point, it can neither be confirmed nor disproved.

 

Not gona happen on 28nm, maybe 30-40% at best but nowhere near 580M to 680M jump.

 

Not gonna happen, huh? Did you have dinner with the CEO of NVIDIA last night? Oh, wait. No, you didn't.

Just giving you a hard time. I know this is unlikely, but I said, "should" and not "will" or "would." Just like the 880M should have been better than the 780M. Unfortunately, we don't live in a perfect world.

To be honest, it doesn't really matter that much to me. I've decided it would be best for me to wait until Pascal and Skylake hit the market before upgrading, considering they'll require a new motherboard. I hate being limited or restricted to certain things.

 

Lol when we heard that 880m was a gk104 , most people knew it wasn't going to be better

 

I know. But given that it is a new card, it should have been better.

 

Not really , unless there is a technical reason for it to be better doesn't matter if they call it a new card. Unless it is a architecture change , die shrink, or adding more cores , slight bumps in clock doesn't mean it should be significantly better.

 

In my previous post, I was trying to say, "in a perfect world it would have been better." Why are you arguing that? It's a moot point.

I'm not an idiot. I know new doesn't necessarily mean better. The 880M is a prime example of that.

 

Same here. Maxwell is supposedly the last of the MXM 3.0b cards, so kind of pointless for me to shell out $1800+ for a final gen product even if performance is stellar. Much better to just hold off until 2016 (or later) when Skylake and Pascal arrive.

 

I know your not a idiot, didn't mean any insult. I just find it that people focus too much on the name of the card rather the actual chip .

Also I think that there's a chance for a 30%-50% increase in performance but also a good chance of significant increase in tdp due to die size but we won't know till we know the finalized die size and how much they night cut down the gm204

Where did you hear that maxwell is the last of mxm 3.0b? Just curious

 

No it should be better than it is.

NVidia screwed up badly with it. They're hotter than 780M cards are when they're overclocked to the same 993MHz speeds - hotter by around 8C. It's a fact that these cards should have been much better than they are. I don't know of anyone who is able to run the modded vbios in any stock machine that hasn't had crazy hard mods to improve cooling efficiency without 90s for temps in a matter of an hour or two.

As for the new cards, I don't have a choice but to upgrade now (if I even can) because I won't have the money to buy a new machine after this money is gone.


Sent from my HTC One_M8 using Tapatalk

 

On this very forum, and maybe even in this very thread.

Seriously though I swear I saw someone mentioning this.

 

Why should it be better than it is ? It makes no sense , its the same chip , same amount of cores , its just an over clocked 780M , u will only see a 10%-15% max gain due to the increase clocks. Nothing really that great and easily achievable to get a 780M to. Only benefit is the 8gb vram. The main purpose of the 880M is to milk the gk104 and push its limits in a laptop.

Also when the stock 880M pushes the limits of the laptop cooling capacity why would you try to use a modded vbios without any cooling mods or power updates.

If you are expecting gains like 20%-30% on the same chip , that's literally impossible unless you whether 2 230 Watt PSU and liquid cooling installed in your laptop.

This will change my upgrade plans if true , so mxm 4.0 coming out , completely forgot about that the mxm could change. Hopefully happens with pascal.

 

Did you read anything or just decide to throw a bunch of numbers out there? This has been tested and what I am saying is a FACT:

The 780M overclocked to the SAME 993MHz speeds that the 880M is supposed to boost to is around 8C or more COOLER than the 880M is!

What sense does it make that the same core speed has that large of a temperature difference? It makes zero sense. NVidia screwed up these cards and now they're trying to bury that fact. The 780M actually needs a voltage bump over the 1v the 880M uses in order to maintain the 993MHz clock speeds but the temperatures are actually manageable while they aren't on the 880s.

Sent from my HTC One_M8 using Tapatalk

 

Actually the 780M needs a voltage bump even at 880M base clock (954MHz) if it wants to hold steady. I can run Unigine Valley for exactly 2m30s before my 780M downclocks to "regular" boost clock (849MHz).

That being said I may have just gotten unlucky with my 780M.

 

As above. Something has gone horribly wrong with Nvidia's release of the 880m, I can run my 780m's cards jacked up all day at 1.05 - 1.075 volts with higher clocks and they'd still run far cooler than these blast furnaces. If you peruse beyond this speculation thread its pretty apparent Nvidia need to address the 880m issue.

As for moving away from MXM 3.0. I'm guessing the general assumption is when (?) Pascal comes out in 2016* the inclusion of NV Link will require it... I'm sure it will happen one day but right now it's just more if's and but's