nVidia 2015 mobile speculation thread

Hasn't Nvidia usually released one every year though? They recently released a 980ti and I would think that's a clue of something similar coming to mobility form. I like those names though I was thinking of 970gt or 980gt since I think it would be bad if they made a 990 mobile yet there is no desktop 990. Ah well at least you bothered to respond to me lol.

 

in that sense yes, but I'm curious why they decide to drop the stock clock 200mhz below 4790k

 

Increase the clock speed in the next update of the processor.

 

Without FIVR, the voltage has probably gone up, especially if you look at the fact it's 95W TDP vs 88W for the 4790k. Also as mentioned, there will definitely be a higher clocked revision later. Either way though, its an impressive performance boost and when we get to see overclocking performance, we can do an apples to apples comparison.

 

but that will come at a price of decreasing perf/watt, and for mobile we'll have to rely on the mobo maker to, rarely, not cheap out on VRMs. my guess is that for mobile CPUs only clevo and asus and maybe msi higher end models will be able to OC due VRM limitation

 

GTX 980M MXM sells at between $800 and $900, a 30% higher price tag than GTX 970M MXM for a GPU that's about 30% faster. nVidia selling a GPU that will be 50%-100% faster (if rumor and speculation are at all accurate) at a price only 15% higher? I don't see nVidia doing that.

 

The GTX 680M = 580M SLI. Price has always been the same. If the 1080M = 980M SLI, I'm sure the price will still be the same.

Technology always gets better. Price doesn't increase with it. However, value of previous components continues to depreciate.

 

Gamers, meet your new super duper "Gaming Laptop". It's got the new super duper desktop Intel Quad Core, the hottest NVidia GPU, loads of storage, speed to die for. And fans which sound like a Boeing 747 on takeoff, But wait, there's a catch. It has around 45 minutes battery life. And it only costs $5999 and weighs 24 Pounds. Take that, Alienware. (Exit Kent T, and joking a tad). Have fun and game on!

 

Price doesn't necessarily increase with it. Sometimes it goes down as a given technology becomes a commodity. Sometimes it goes up if manufacturers believe that the market will pay even if it is overpriced. There are a lot of people paying way too much (in my opinion) or not nearly enough (in nVidia's opinion) for G-Sync. I think that nVidia will do the same thing with a Maxwell refresh because the market will bear the artificially inflated premium price.

 

yes it will go up if the market can accept, how many people do you think will accept a $1500 price tag you said? and how many people do you think will accept $1000. again this is simple macro, even if you have a monoply or obligopoly in this case, overcharging isn't gonna maximize your profit, you have to charge at what the market will accept, and I'm telling you the number of people that will accept a $1500 price tag is minimal if not no one

 

I figure that anyone willing to pay $750 for ROG Swift would be willing to pay $1500 for the fastest mobile GPU to date. I figure anyone willing to pay the approximately $500 for Oculus Rift when it ships (or for a devkit today) would be willing to pay $1500 for a mobile GPU capable of driving it at better than minimum specs.

The market is there and I expect nVidia to take as much advantage of it as they can.

 

except the market demand will be significantly higher if it's $1000 or less. would you rather have 1000 people pay at $1500 or 10000 people pay at $1000? that's the difference here. nvidia's not stupid, sure they can say some people are willing to pay $1500 for a geforce card, but they also knows the demand will be much higher if they keep price within reasonable range for both the consumer and themselves

 

I think your assumption is missing an important point. This hypothetical (so far) Maxwell refresh isn't happening in a vacuum. It's happening in order to take as much advantage as possible of Skylake's launch. nVidia don't need to worry about price-based demand. They will be riding on the demand for Skylake notebooks.

I could be wrong. It wouldn't be the first time. But I think that with nVidia's exhibited greed over G-Sync I expect them to exhibit the same greed over whatever, if anything, comes between GTX 900 series and Pascal. And with Pascal, too, for that matter.

 

I'd just leave the 980M out there to rake in some more dough. There's nothing from AMD forcing them to move on. They can then focus on the launch of mobile Pascal early next year (ahead of desktop Pascal). Fairly certain that's how it will be. Clevo is advertising their new models with the 980M, and those are due for September.

Not sure what else NVIDIA can with Maxwell except overclock it. GM200 (mobile) is a long shot.

 

They could release a full GM204 for notebooks if they really wanted to. They did it with the 780M, absolutely no reason it couldn't be done with Maxwell. But like you said, lack of competition means there's 0 incentive for them to do so, plus full GM204 for mobile means they have to cherry pick the perfect GM204 dies, whereas selling cut down parts means less work involved overall.

As for mobile GM200, I'm calling it right now: never gonna happen.

 

Except that there actually is competition. nVidia recently cut the prices on GTX 970 and 980 due to AMD's Radeon 300 announcements, and just yesterday nVidia's AIBs quietly cut 900-series prices even more to further improve the Maxwell price/performance ratio with respect to Radeon 300 (I just discovered this). We here at NBR might not be taking AMD seriously right now but nVidia and their partners definitely are.

 

AMD's stock is plummeting, market share declining... NVIDIA has little to worry about. All of the new stuff AMD just launched was a response to NVIDIA Maxwell, which came out almost a year ago. AMD is way behind the curve, particularly in mobile.

The sad thing is, the entire GPU market is at such a low point right now, it's starting to affect gameplay (i.e. driver support).

 

except the demand for skylake cannot justify for the huge increase in price, think about it, broadwell is only marginally more expensive than haswell, how much sense does it make for price to be jacked all the way up for skylake launch? full GM204 will probably come at $1000 matching the titan, and there is an incentive for them to release it, that is they know many people be up for it

 

Hi all

A lot of speculation in this thread, good to see the forum alive on that front.
I`m back with a little more information about GTX 990M

Remember back in the days when 2GB was mainstream? And we got 4GB cards and thinking "wow awesome"? Then followed by 8GB cards with a lot of scepticism although we had some games break the 4GB usage if we had enough available.

Well Nvidia thinks 8GB is too little, because apparantly GTX 990M have 16GB GDDR5. Yes you are reading correctly, twice as much as GTX 980M. 4x as much as the 4GB cards

I also contacted some people and they said "more than twice as fast as GTX 980M". I don`t understand it, but these rumors shure are very interesting. And apparantly people have started recieving samples.

Take care

 

16GB of VRAM on a mobile gaming GPU? That's trying to solve a problem that doesn't exist.

 

Another feature that will allow them to double the price

 

But it's a great (fsvo "great") way to justify a $1500 price tag.

 

When can we expect a release? This was unexpected... What about BGA gtx990m? Is there room for 16gb VRAM on the motherboard? Perhaps OEMs need to remove one or two M.2 slot on the motherboard? LoL
They choose rather 8GB VRAM I think.

 

If it ends up being something absurd in price I'll just settle with the 980m on the Batman 2.0 since if it can handle GTA5 it'll be able to handle whatever else comes down the pipe for three years. still that is a massive jump when gaming barely pushses 8gb much less 16gb if true.

 

September-October launch is my guess.
The people with the cards say they have issues with the vbios as well as the bios, which makes me wonder what chip this really is.

BGA 16GB? I have no idea. I`m trying to understand how they even got 16GB VRAM in a MXM card.
GTX 980M have 16 GDDR5 SDRAM chips. The biggest GDDR5 chips available at the moment is 4Gb or 500MB I think. Atleast from SK-Hynix.
16x500MB = 8GB.

Lots of unknowns about this card for sure.
But very interesting rumors

 

There may become a problem to use old heatsink in older laptops In such a case.

 

I think the 990M may be real... After digging around on the internet, I've found there is going to be a dual-GPU GTX 990 based on the GM200 Maxwell architecture (like the 690 was Kepler). Rumors are going around that it will have 24GB's of VRAM. And if this is true, there may actually be a GTX 990M.

Hopefully these rumors are true... We've never seen anything like this before in mobile. I'm kind of excited. There was never a GTX 690M. So, pricing wise, we have no idea what it may be. But it's safe to say it'll be more than the 980M.

http://www.tweaktown.com/news/38953...-its-geforce-gtx-990-in-early-2015/index.html

Kind of makes me worried about Pascal. Is it going to be delayed again?! Will the performance jump from 990M to Pascal (1080M) not be so large?

More questions than answers... Will it be another 880M dud?

 

This is an April's Fool Article.

 

It is? LOL

What about the other one?

 

While I have no doubt that a new NVIDIA flagship is coming I cannot believe that it will be twice the performance of the 980M. That's more power than a desktop 980. There is no die shrink, there is no new architecture.

Notebooks have a strict limit in TDP of around 125W for GPUs.

It's just not possible by the laws of physics for this mythical GPU to exist with that supposed level of performance in a form that is compatible with notebooks.

I'm gonna have a prediction of my own; GM204, fully unlocked, clocked about the same as 980M, 256bit memory bus with 6GHz GDDR5 with max ~30% improvement vs 980M. I can believe the VRAM part though seeing as NVIDIA have already released the 12GB Titan X.

 

Another card doesn't necessarily mean "twice as powerful" as the 980M. Could be 30% more powerful. All rumors.

I also doubt they'd release another GPU this year that's twice as powerful, lol.

 

That a response to Cloudfire's source's claim:

 

Oh, I see... Well, I once had a technical support agent tell me a 240W PSU would make my system explode.

Can't rely on anyone other than NVIDIA's people.

 

This is my personal prediction:

GM204 fully unlocked
Clocked about the same as 980M
256bit memory bus with 6GHz GDDR5
Maximum of ~30% improvement vs 980M

I can believe the VRAM part though seeing as NVIDIA have already released the 12GB Titan X. Anything more than 8GB is wasteful though and just a marketing gimmick.

 

I agree with you. More cores is most likely. 20%-30% improvement over the 980M is realistic.

 

Just an educated guess based upon the 780M vs 680M

 

16GB would require double density modules over the titan-x, i'm not sure they exist.

 

I am doubtful, but I do think it's far more plausible than the 'twice the 980M's performance' part of the rumour!

I'm not trying to attack Cloudfire here by the way, he's just the messenger, just doubting whoever was his source

 

The first mention of a 990M that is twice as powerful as a 980M came from a China CEO of Hasee during a presentation. Commented a late 4th quarter release.

Seems unlikely but who knows.

 

Pascal must be freaking amazing then, 'cause rumors suggest one Pascal core = 2x the performance of a Maxwell core.

A single 1080M = Titan X SLI (2x Titan X or 4x 980M) performance? HOLY CRAP! Wait a minute...

These rumors are getting out of hand.

 

If this is true; Why is Nvidia so determined/keen to "destroy" overclocking on existing graphics card on laptops? This nonsense has Nvidia no reason to do if these Rumors are correct.

 

How can the rumour be correct?

Twice the 980M would put the performance above the desktop 980 and thus, TDP through the roof.

I can't see how it could ever happen. Not until Pascal.

 

My sockets are ready

 

They do, but Samsung just started volume production in January this year.

As far as 990M being "twice the performance of 980M", not a snowball's chance in hell. 980M is roughly 10-15% behind a desktop 970, and a Titan X is roughly equivalent to 970 SLI. Factor in SLI scaling overhead and 2x 980M is basically Titan X level performance.

Good luck getting that kind of performance to fit within a 125W TDP envelope on 28nm. Just not going to happen.

 

I would have believed it if it were soldered with some massive 150W heatsink requirement, but it's MXM. Ehhh.

 

150W is not enough.

 

If the 990M is real, I'd avoid it just to avoid issue. I'm sure it'll be another wild card (pun intended).

Wait for Pascal - efficient, cool, and powerful.

 

Maybe it has dual GPU's on the same PCB. But that would be near impossible on the size of an MXM card. I wouldn't doubt if a 990m would exist but no way possible to be twice the performance.

 

Dual GPU's and 16gb VRAM. Oh yes... Do not believe it until I see it . This does not happen this fall/winter in any case.

 

A Pascal based "990M" kind of fits that bill perfectly -- 2x 980M performance within the same TDP, and 16GB of HBM2.