nVidia 2015 mobile speculation thread

Maybe I'll have to invest in a Blu-ray player then, or some kind of subscription service - I'm late to the party, only got a 1080p HD TV last year! I got this one: http://www.hdtvtest.co.uk/news/sony-kdl42w653-201308113237.htm
Even calibrated the screen myself using my Spyder4Pro colorimeter, but I had to modify it to use free HCFR software as the Spyder software for TVs is pathetic!

 

It's Fox M ulder

I named my two dogs at the time Mulder and Scully, that's how much I was into the show. Did you know there's a half season of new X-files coming up next January!? With Duchovny and Anderson?

 

So you had a smoking hot redhead itch? Noice!

Yes, they are in it!
So so happy!

 

If you guys haven't already, check out The Fall on Netflix... she is still smoking ha, plus I enjoyed the show

 

may have to do this

 

I already have. Hannibal as well. So sad it was canceled

 

Ahh dam. I really enjoyed that series alot. Dam

 

Me too...

 

Yeah, that was a good show, I saw that on BBC (or was it CH4).

 

My wrist got slapped. I will say no more in this thread. You will have to wait like everyone else. I quoted early figures. That were taken out of context. Feel free to speculate all you want but I am pulling myself from this thread except as a mod.

 

Lol at the OT in this thread. Pretty much sums up we how little news to go on.
Anyway I'm saving for Pascal. If AMD can produce a faster GPU by then I'll reluctantly give them another chance because I haven't enjoyed rewarding Nvidia for their shenanigans of late. Being burnt by 7970m dying on me the wound is yet to heal. I'm closely following their newer HBM products reliability.

 

Yeah, I'm with you on that! I can definitely hold out till Pascal or perhaps an amazing HBM chip from AMD, haven't settled on desktop or laptop, but being a long time laptop owner I might flirt with a desktop as a change (building my own of course!) & keep my now Windows 10 upgraded laptop for comfortable internet & office tasks throughout the house.

 

That's the best option.

 

the rumor says this 990m will be close to 980m sli performance, but wouldnt 970m sli performance be more realistic?

 

I'm gonna just wait and see. I need to update my computer and tablet

 

No.

 

at least in the middle?

 

20-30% over 980M tops is my speculation.

 

hopefully it doesnt have heating issues like the 880m

 

Probably not as bad (hopefully), but it will definitely run toastier than 980M.

 

for me i think about cheery picked (the lowest voltage possible) full GM204 chips or little cut-down ~1800 units instead full one. No way for 980SLI performance but 40% max might be possible.

 

If anything it will be a full chip with the lower clocks and respective 20-25% increase to performance.

 

GTX 970M draws up to ~100W. Therefore GTX 970M SLI draws up to ~200W.

MXM 3.0b provides 100W on spec with a practical limit of around 125W. Therefore GTX 970M SLI performance is not possible.

 

On a side note, a friend of mine grew up with Gillian Anderson (Agent Scully), and went to grade school/High school together in Grand Rapids, Michigan in the 1970's.
She was really a wild child back then, and hung out with the freaky people, who used to be called the Stoners, druggies, hard rockers, etc.
What a contrast to whom and what she is today, in being an straight and narrow actress on the X-Files.
No one who knows her now would ever think that is whom and what she used to be way back when.....

 

There were some hints here and there in the X-Files. Supposedly Scully was also a wild child.
"Better hide your Megadeth albums, Scully"

 

Considering how overclockers on this forum have been able to draw well over 500W from the wall using two power bricks with SLI notebooks, I think 125W isn't the highest limit. No way a CPU + system draw alone caps out ~250W

 

Each card is capable of drawing over 150W when overclocked (with a modded vBIOS). I think Mr. Fox managed to get a 980M to draw 171W? (Don't quote me on this.) I'm sure the 990M can be a higher wattage card. I don't believe 980M SLI performance, though. If that's the case, the 1080M must be like 4x 980M, unless the 990M is better than the 1080M, which would be just plain stupid on NVIDIA's part.

990M rumors are screwing up all 1080M rumors. The 1080M rumors were originally 980M SLI performance.

 

That's the limit that nVidia supports at this time. But let's noodle.

GTX 980M SLI plus a fast mobile i7 is ~320W, not 250W. Plus ~20W for everything else in the system: RAM, storage, display, etc. Call it 340W.

Assume 85% efficiency from the power supply. If it draws 500W from the wall then the system can use 425W peak. That leaves about 80W overhead to hit the 425W limit. If we split that 80W between the two GPUs we get 165W per GPU. If instead the 500W figure includes the 85% efficiency calculation then it becomes around 200W per GPU.

165W is reasonable. That's GTX 980's TDP. Assuming the MXM socket can provide that much power (65% over spec) and the cooling system can handle the heat (not many can) then you could conceivably have a full GTX 980 in a notebook.

The 200W figure is where GTX 970M SLI would be, but 200W over an approximately 1" square spreader in an air-cooled notebook? Good luck with that.

 

Fixed that for ya with the observed peaks using 980M SLI

 

Hey guys

Is it worth it to invest in a g-sync 980M MXM at this point? I wonder if Pascal 1080M will have MXM, or is Maxwell the end if that.

From some posts, people are saying that Maxwell is a dying breed and whatnot.

Sent from my H60-L04 using Tapatalk

 

You know, the really old paper might say there is a 100W limit, and you kinda adjusted to 125W, just by guessing or something. Reality is a bit different though. Get at random any schematic of MXM-B machine and you'll see PWR_SRC (usually more than 19V) which exactly means - wired to the power source, no limits what-so-ever. On some schematics it says 10A max, which means 19V * 10A = 190W. This is most likely the physical limitation of the pins, not anything else, since you can only push that much juice through them. So if we are going to play the guessing game, let's play it right, with some facts behind. The only REAL limitations of MXM are mostly cooling, then space, but second wont be that much of a concern with HBM. Having the vRAM next to chip leaves all the space for power delivery stuff, so the cooling is the only hold-back.

 

I don't even know what to think anymore. From Skylake, to Batman 2.0 and of course the refresh... I just want everything revealed by the end of the week.

 

Hm. Then at least one of my assumptions is off. There's no way those cards are pulling 215W directly. That's getting into Titan X territory.

Or, if they are then I don't know how you're cooling them.

Not so much guessing as extrapolating from what nVidia have done with it. Certainly, vendors can wire things up differently but that's off spec, as much as when they wire up PCIe slots to deliver more than 66W. If you're a PCIe board maker then you cannot trust every mobo vendor to have done this or to have done it consistently if you expect your product to work everywhere. The only numbers that you can rely on are the ones in the spec. nVidia have the benefit of having written the spec .

That said, if we make the assumption that most/all MXM slots have been wired for 19V, 10A then there is just barely enough power for a single-package GTX 970M SLI equivalent.

Cooling it is still a problem. You need a lot of airflow or a water cooling rig to cool a 200W package. There aren't many (perhaps any) unmodified notebooks out there with that kind of cooling capability.

 

Since you are so anal about specs - getting power from the power source is BY spec. That's why there were 120W GPUs, followed by 125W ones, and etc, without any change in the design. Oh wait, the shocker - they work on older machines and can even overclock if the cooling is there! As mentioned previously, you wouldn't have been able to overclock, without frying something or overload cutting out, but yet the possibility is there. So where's that 100W limit again? All MXMs are wired the same - power from the power source, if you can cool it, boy are in for a fun ride. The opposite of what you say is true - if a vendor decides to limit it to certain power threshold, now that's another story, and certainly not by spec. By spec MXM pulls directly from power source unrestricted, that's how it is since at least the old MXM 2.1 standard. As for the upcoming GPU, you still miss the eventual refinement of the process as well as lowered clocks. Not to mention mobile GPUs usually get higher binned chips, with mobile professional GPUs the very best. Overclockers around here would show where those GPUs can get (because underclocked mobile, that's why).

 

My P570WM got up to 600-650W from the wall with the 780M sli benchmarks at full pelt, even with 175W to the CPU the rest is pretty much the GPUs (minus power supply losses).

 

Haha, that's insane (!!!), but obviously possible because you did it!

 

I've always wondered whether there's a concern that you might physically melt the pins if you tried to pull too much power through them.

 

No. Fifty-fifty solder, the kind used in BGA for example, melts at (checks) 183C; gold melts at 1090C.

 

Of course, it's just like the cable you plug into your laptop, it's rated for that much Amps.

The pins are gold plated, it's copper afterwards.

 

Copper melts at 1083C.

The amperage rating on power cables is how much of a load they can safely carry. Go above that and they can overheat. The copper won't melt but the insulating material will soften or melt, exposing the copper inside. It can also ignite. Plastics have ignition temperatures between around 125-200C.

Just saying that if your notebook is drawing enough power to melt gold and copper then you've got bigger problems on your hands.

 

What about the underlying material - the PCB? It would catch fire long before that and I only stated the materials used for clarification. You obviously don't know what goes into PCB production. It's funny how you stated that there is a 100W (or 125W, whatever) limitation, without any additional information and yet you are trying to convince me (us) actually how hard it is a failure to occur. There are limitations, but not the ones you are saying. I have nothing more to add.

 

Why are you guys still debating fire and wattage? We've seen 100W AMD cards burn up before. Doesn't really matter how many watts.

Let's talk about the GTX 990 and 990M.

 

Sorry for the off, I just said that it's not impossible to have it where most desktop GPUs are, given you can cool it. As for burning GPUs, let's not call names, since the other one has quite the track record as well.

 

This is what I meant when I said "melt the pins". Ok I could've been less ambiguous I guess, but it should've been somewhat obvious.

 

If there's a GTX 990 (desktop card) released Q4 2015 or Q1 2016, do you guys think it will be basically 2x 980's or 980Ti's?

If there's a 990M, must be a GTX 990 coming.

 

2x980s would be the more sensible thing to do. 2x 980Ti's would require significant reduction in clockspeed in order to fit within 375W TDP.

 

Hypothetical 990 would have to be 2x GM204. 2x GM200 would be Titan X Black or something.

 

Titan XYZ

 

Better call it YYZ.

 

It's only fitting...
The Nvidia Toronto!