Clevo notebooks with 800M series coming out February 2014

Well the 780m had plus points over the 680m.

Nothing quite like gaming 24/7 at 1045/1750 (less than 10% behind a pair of desktop 770s).

The 880m will behave identically just with worse overclocking ram.

 

I am with Cakefish.

I have been talking about MX version coming later this year.

Here is what is gonna happen from my POV:

If 860M is Maxwell on 28nm then 880MX will mostly likely be a 28nm as well. Depends on the 860M performance over its predecessor. If the higher the % the more chance of 28nm MX 880

Desktop: I categorize the most in this area.

Yes, you are absolutely right. Desktops will receive 20nm first but "when"- whole different question.

So far the flow is taking 28nm current, for the whole Desktop line up. In which case, might as well forget about 20nm for some time.

 

No more bottlenecked by DDR. No more Optimus-like-features only for 60Hz displays. No more Optimus hassle where it doesnt reckognize the need to switch over to dGPU on certain games.
Way better drivers compared to Intel`s like you say which will better game performance. Only 1 driver needed, not 1 from Nvidia and 1 from Intel. Reduced power consumption vs IGP on Intel because the chip may only need to run the ARM part to run the display.

Yeah I can see a lot of benefits going this way.

This may also be highly related

Intel will pay $1.5 billion in Nvidia settlement

 

How do you manage to get acceptable temperatures at those clocks? The only thing that stops me from overclocking that far is the temperatures.

In my situation with a 780M I guess the sensible thing to do would be to resist the massive urge to upgrade to 28nm Maxwell and wait for a 20nm Maxwell. That's gonna take some strong self control to avoid the temptation and lure of shiny new architecture. If only I could get the thermals of this Clevo under control so I can overclock it would be a much easier proposition...

 

But if Maxwell brings 30%+ performance, will your urge take you over?!

30%+ from 880M

 

Having a Clevo, recently notorious for locking out upgrades, the decision on whether or not you'll move on to Maxwell may be taken out of your hands.

 

I don't know how they're gonna play it with the ARM CPU in terms of what it's function is going to be, but I reckon it's going to be there to support smooth operation of the GPU, somehow removing it from any CPU limitations of the main processor CPU (Intel or AMD). Kind of like how Mantle gets round some of the CPU limitations, I have a feeling that the ARM CPU on the Maxwell GPU will accomplish something similar. It's just a hunch, and I don't know much about these things, but that feels about right to me.

 

I just read that tsmc CEO stated that they will start production of 20nm SoC in Q1

I just hope that this is what he meant for the dear precious GPUs and not only arm SoC

That is according to the source Taipei times, where all the rumor from 20nm mass production began

dosed, mixed and stirred, not shaken, from taptalk

 

Very likely.

 

Meanwhile on Desktop side of news:

GeForce GTX TITAN Black Edition and GeForce GTX 790

Maxwell soon, anyone else?!...

 

Guys I see a lot of contradictions, if the 880 is a rebranded 780 with OCing and more Vram and an 880 as always would be priced the highest at a 1800+$/laptop then how would maxwell come even after the 880's release...I mean how can the highest card in the whole series (Top nvidia card) get swallowed by cheaper maxwell versions just 1 or 2 quarters away that would be bankruptcy for laptop manufacturers adopting the card because their $1800+ laptop would be weaker than a $1500 laptop 1 or 2 quarters away ? unless cheaper maxwell versions do not offer any performance or efficiency (since they are 28nms) increases which makes them pointless?

 

We need more info about this one..
if its a rebrand we will have to wait for the GTX 860M to get some maxwell ? but i saw that the GTX 750Ti = GTX 860M Underclocked most likely is weaker than a GTX 660 = gtx 770M Underclocked..
so GTX 860M = GTX 765M maxwell eq ?

 

I think if NVidia release a Maxwell card with more performance than the Kepler 880M, then they will name it to show a higher performance - like 880MX or maybe even 980M, and then laptop manufacturers will just simply charge more for that new 880MX laptop. In that case, 880M would not be King of the Hill for long. (Any card that performs faster is not going to be cheaper).

 

The first cards is the low end Maxwell`s. They will never beat the high end Keplers, GTX 780M and GTX 880M. Not even close. They start with low end, then replace high end Kepler with Maxwell later. Thats what Nvidia do with Mobile GPUs.

Yeah, GTX 860M could be based on the same chip as GTX 750Ti.

GTX 660M was a 28nm Kepler card. GTX 560M was a 40nm Fermi. GTX 660M was 30% faster. No way a new architecture thats still on the same 28nm will give similar improvements. More like 10-20% if we are lucky. That is why I have been talking about 20nm. The resent GTX 750Ti benchmarks was dissapointing as hell. GTX 750Ti actually performed worse than GTX 650Ti believe it or not. GK107 vs GM107. So there better be some power consumption improvements there or it will be one lame introduction of Maxwell.

I think if we see Maxwell on 28nm, it will be only on a few cards and will serve as testing the water. Introducing Maxwell what the architecture can do in terms of power consumption savings. Plus get the general Maxwell features out to the public, UMA and perhaps the ARM core.
Then push out 20nm Maxwell cards with more cores as soon as they are ready. Those will be the performance cards.

I think GTX 880MX based on 20nm will give more improvements than 30%. GTX 580M > GTX 680M saw like 80-90% improvements. I hope we atleast see +50%. Then I will be happy.

 

580M was 40nm then 680M 28nm thats 12nm less, but this time we got only a 8nm gain..

edit : then gtx 765M maxwell eq will be actually SLOWER OMG !

 

That's still a 30% reduction in both cases.

 

40 > 28 = 30% reduction
28 > 20 = 28.5% reduction

Does it work like that? I'm not a computer physician. I'm not sure.

 

Yes, more or less because you're fitting that many more transistors on a wafer.

 

By this logic, for 20nm, the calculations go:

28nm = 1536 Shaders.

(Using Cakefish's math) 28.5/100*1536 = 437.76

Rounding that to 437,

20nm should have 1536+437 shaders = 1973. Rounding it to the nearest most plausible number, 20nm = 1920 shaders for a chip the same size as the 780M.

 

Did someone just drop his jockstrap..? Let's not forget computer engineering innovations are moving into three dimensions, not two..

 

LOL sponge. Your point is taken.

 

GTX 780 Ti is 30%+ faster than 680. While being on the same arc and tech as well.

If Nvidia is pulling 880M and the exact temps are yet to be confirmed than my prediction can come true.

Edit: Is`t it 22nm and not 20nm? 580-680 30% reduction but this number has no influence on cudas.. since 680 tripled 580`s amount.

 

I think it works like this as way of determining the factor for how many more transistors you can fit on the same sized GPU die when going from 28nm down to 20nm:

( 28^2)/( 20^2) = (784)/(400) = 1.96

So, I think you could expect to fit 1.96 times more transistors per unit area on 20nm in comparison to 28nm.

So, the 780M with 1536 shaders might turn into a GPU with 1536*1.96=3010 shaders if it happened to be shrunk to 20nm for the same die size. That's how performance is doubled with each die shrink, and I guess that's how they determine the nm size to aim for the next generation - doubling the transistors per unit area.

 

Are nVidia and AMD both using tri-gate like Intel though? I thought nVidia weren't doing that until Volta?

 

Looks like there are more people than just me that feel extremely dissappointed about 28nm Maxwell cards.
Like I have said earlier in this thread. 28nm Maxwell will give very little benefits. Doesn`t make any sense to rush the architecture out to me, perhaps other than finally introducing Maxwell after having to post pone it in 2013.

ASUS GeForce GTX 750 Ti listed, coming mid-February? | VideoCardz.com

Lets hope they won`t last long until 20nm is here and Nvidia can stuff more cores in the chips.

 

Yep looks about right. And Nvidia HAVE to go down to 20nm to be able to keep up with Moore`s law and stay competitive in the industry. For those who don`t know what Moore`s law is, it means doubling the transistors every 2nd year.

I more than welcome GTX 880MX with 3000 CUDA cores

Correct, transistor stacking is apparantly coming with 16nm (Volta). Should give some nice benefits.
https://www.semiwiki.com/forum/content/2762-tsmc-s-16finfet-3d-ic-reference-flows.html

 

Posting this for future use

 

You're right. Thanks for correcting me.

 

No problem, you're welcome! It's quite amazing how such a small decrease in architecture design size can result in such a large increase in number of transistors within the same unit area, just down to the nature of how Area is calculated. Also a great reason to wait for the next die shrink if it's close to launching!

 

By any chance is that to illustrate that Fermi has been around through 400, 500, 600, 700, and 800 series!?

 

They still have spare Fermi chips going round, huh? Maybe 900M series they will finally retire Fermi.

My challenge is to repaste my 780M until I can overclock to 880M stock clocks without setting the house on fire. It can't be too hard, +150MHz isn't a lot to ask, surely...

 

I think it's possible to mod your heatsink to increase & even out pressure on the GPU die, but it can crack the GPU chip & therefore break it permanently. That's one mod that I saw someone do on the Tech Inferno forums. I think it was a user called j95 who did that mod. Part of the mod involved removing the C-clips so you could somehow get more pressure applied, but can't remember the rest. The quality of the repasting job you do on our GPU is quite influential as well (not so much the type of paste used, but the quality of the job done). I've used the spread method outlined on the Arctic Silver website to good effect with 2 different GPU's in my laptop, and it's worked really well with low temperatures (not over 67 degC fully overvolted & overclocked in my sig) - I used Arctic Silver 5 (not the best, but pretty good). Here's the link to the spread method:
http://www.arcticsilver.com/pdf/appmeth/int/ss/intel_app_method_surface_spread_v1.1.pdf

EDIT: The mod I was talking about above, it's been named the "M17x Retention Mod", but I guess it would work for others. Here's two links, first one is j95 who's done the mod, and his results, and then the second one is a whole thread on it:

http://forum.techinferno.com/alienware-m17x-aw-17/3702-m17x-r3-gpu-upgrade-gtx-780m-16.html
http://forum.techinferno.com/alienware-m17x-aw-17/24-m17x-retention-mod.html

EDIT 2: And found this on NBR forums, but don't know much about this one, or even if it's the same mod:
http://forum.notebookreview.com/alienware-17-m17x/523915-pics-cooling-mods-done-tx-4-clip-mod.html

 

Oh, but guys, the new 800m series is based on ddr3 memory. Which is totally slower than ddr5, which is made of unicorns and rainbows, you guys! It suucks if it doesn't have ddr5, or at least ddr6!

 

Yups, that Fermi have been with us for 5 years now (lol) and to store as a record if someone asks what GT 820M is.

I heard they come with an onboard Bluetooth support. Who doesn`t love Bluetooth?!?1!

 

Looks fake. GF117 should be 28nm.

 

Even so, Fermi?

 

yeah the fermi reused were on 28nm

 

They must have received ultra cheap fab time on machines for that.

 

Sigh, guys its not fake

Here is another one from Samsung NP530 that use the same chip

 

at 40nm and Fermi they must have found someone with the old fab equipment down in someone's basement. Sheesh 820m = 720m = 620m = 540m = Intel IGP.

 

Yeah, so old tech even phones surpass it now

 

I don't really know, however for you to shrink, you need to redesign some parts, it was more of a lets milk that stuff kind of decision probably.

however I disagree with a lot of things the crazy korean does

old fab is cheap, new fab is expensive.

old fab is specially cheap since you already had profit from that, and unless used, it will go to the garbage... or you can send to some developing countries and sell old cars like they are new, or thats what the US car makers did here

anyway, I still find hard to believe that they are going with fermi again, specially at 40nm, during the 600m series it was still quite possible to see those 40nm specially on just rebadge models, however now...

and specially after they developed the 64bit versions of the lower end kepler, they must have some very hard economic reasons to still use those.

anyway i think i only saw the 720m on some very low priced notebooks, one that i remember right now is the latitude series 5000, interesting to note that the 3000 has a MUCH more potent GPU, the 8850m

 

They are doing it here :
80% of notebooks got HD3000/4000 + ULV cpu's or normal CPU's with Geforce 710M (48CC barely beat the optimus's hd4000 w/ ridiculous 2gb vram)
10% :you got the asus's ulv i5/i7 dual core + GT 740M @ 64bit and the y510p GT 750M (single of course and no sli available and 768p screen)
NO full HD screens in all models except 2 asus 3rd gen and extremely expensive..
OF course old gen cpu's 6 months after they come out in the europe/US..

or you get the avg wage of an IT engineer multiplied per 2 (4000d means 2500$; all the salary not what he got left at the end of the month) and you get the ASUS g750jx 3d blu ray and full of expensive ram (all useless options : we got no 3d content inhere, no blu ray's sale and i found the 16 gigs useless)
the 780M model would cost 3000$ and remember that life isnt that cheap here)

off topic : ended
my apologies about the monstrous off topic post but my country is driving me crasy and the resellers just laugh at your face when you ask them why the laptop are so not well chosen..

 

F**king hell. nVidia, you stupid (and crafty) old b**tards.

 

I feel you

http://forum.notebookreview.com/off-topic/567605-what-bothering-you-right-now-318.html#post9537605

 

You have a point there. If they'd concentrate on the content that really matters, like only offering realistic vRAM (not 4GB for a 720m for example), only 8GB system RAM, and omit the optical drive entirely, you can get prices reasonable and don't waste money on anything extra.

 

you know the ODD is actually a great cost in the machine? sometimes the cost is higher than the screen or ram

EDIT:

I was just checking notebook check for some cpu benches and guess what? the 820m is there with a 28nm fab

 

Bug with gpu-z then.

I'm surprised Nvidia will be using Fermi again, but it's not as bad as it seems.

 

im not even surprised, Im thinking of straight rebadge, aka 720m is going to be 820m, 680mx/780m/880m, 760m, 860m and so forth

 

Yes, it seems it will be a full Kepler rebadge for 800m series this time around *if* they do release in the next month or so, with 20nm Maxwell come Q3. I would think if that was the case that they would release them as 900m series because the performance difference at 20nm should be significantly improved. Who knows. This whole rebadge thing is getting old.