Clevo notebooks with 800M series coming out February 2014

Do u even know what context this quote is a part of?

 

Seriously, get off your high horse Sin. Try to be a little more humble and open.

Sweclockers IS a reputable news site. I tried telling you this. Mr Najsmans agrees. The reporter from Videocardz share the same view

NVIDIA GeForce GTX 750 Ti with Maxwell GPU in February | VideoCardz.com

Sweclockers is not your typical rumor site. They post reviews of hardware. They have interviews with Nvidia and other companies. They do CES/Cebit articles where they write about what they saw at the fair.


As for me being against 28nm because I want 20nm:
The only thing we will see with 28nm is architectural gains. Meaning how much? 10%? 15%? So best case, they can add 15% (?) more cores and have the exact same heat output as current cards. Kepler itself is pretty efficient since Nvidia ditched FP64 performance and made it targeted toward gaming instead.
On 20nm its a completely different story. We get the architectural gains and Nvidia can add a lot more transistors without sacrificing thermal.
GTX 680M saw 80-90% gains over GTX 580M. That is impossible to do without a new node.

I tell you, if 28nm Maxwell see 15% improvements on 28nm, it will be a huge "MEH" across all tech forums. And I wonder how much they will manage to sell, because I think most people know 20nm is right around the corner and would rather wait. Which is why I question the rumor that Nvidia is even attempting at pushing out Maxwell on 28nm when they could just wait a few months, and rather bin 20nm from TSMC and build the Maxwell`s on 20nm instead.

That is my opinion.

 

Nope "most" people don't know about die size, they just buy .

 

Really?
Well I meant the high end cards. Those who cost $500+. I`d like to think that those people in that group are pretty informed on hardware before making the final call on what GPU they will buy.
I agree that the people buying those low end GPUs thats going inside $999 notebooks isn`t caring too much about such details. They just want a notebook that works.

So yeah I guess you are right, "most people", that means the majority, dont know about die`s and such. After all those who buy $900 GPUs are a minority.

 

Lemme clarify something for everyone:

All Sweclockers told in their article which is "new info" is card`s name(750 Ti) and its precise release date. Both of them are yet to be confirmed..

1) Fudzilla did reveal 28nm was coming in Q1 2014(February being the earliest) long time ago.

2) All Sweclockers did is mention the name of the card which is, again, not 100% confirmed.

In short: They told nothing new in their article.

How many times do I have to say it?

Sorry, but your next part may only correspond to mobile part of GPUs. Even then, 880MX on 28nm and Maxwell might surprise many. Not sure why u did not think of it, considering u have been in Mgpu world for quiet sometime.

Hint: 880M is already 12% faster. So how can a Maxwell 28nm be only 3-% faster than that? Wow, u should have been very high while writing this...

 

There you go again, on your high horse.

GTX 880M is 12% faster because its clocked higher than GTX 780M. That clock comes with a price: Higher power consumption and higher temperatures.
Architecture gains can do 12% faster but with the same power consumption.

Thats the whole point of a new architecture. Getting more efficient.

I`m not gonna respond to your "Fudzilla said this, Sweclockers said that". Its seriously an useless discussion.

 

I'm with Cloud on this one. Arguing about which websites are correct sources is pointless. What we need to do is wait for more leaks and then put all the evidence together, and Voilà! The question of 20nm or 28nm will answer itself. Sin is being overly skeptical, while Cloud is actually making educated guesses based on concrete evidence.

February Maxwell - probably 28nm
June Maxwell (Computex) - probably 20nm.

BUT, none of this is concrete yet.

 

Why don't we just close this thread, wait a couple months, and find out.

 

Not happening until we get a time machine.

 

Har de har har, HT. Speculation with proof is good, but theories with no substance (a bit of substance makes them worth reading) behind them are a waste of time.

 

Thats the whole point of a new architecture. Getting more efficient.

"Architecture gains can do 12% faster but with the same power consumption." Ok, sir. You work at Nvidia and u know it better than us. I take my evidences back then.

Srsly, I had a laugh when reading this. Hearing it from someone like u is very disappointing to say the least...

Yea, Fudzilla spoiled the the light on most of the things. Sweclockers just added small bits which are yet to be determined as right or wrong.

In regards to others, funny how u say there is a "concrete information" and after believing in 20nm a few days back change your opinion now. Good u read my prove of 28nm coming first, but don't be shy to mention my name there :laugh:

 

Seriously what a clown.

Like always I have to spell things out for you. That 12% was an example of what new architecture can do based on your 12% number for GTX 880M. There have been tons of examples of what new architecture on same node can do. Try for example looking up Intel and educating yourself.

The comments about Maxwell on 20nm was based on late march launch for the first Maxwell's since nothing was revealed at CES. February launch kinda changes things how it will pan out since they started production in January at TSMC. Nobody have said anything else. I said earlier that since no "MOBILE" Maxwell was launched at CES, we wouldnt see 28nm since that meant Match-April release for them. This is a mobile forum, not a desktop forum.

And you still keep on dragging out the Fudzilla/Sweclockers discussion...

Now are you finished embarrassing yourself for the forum or do you want to continue acting like a ignorant new user?

 

Lets simplify this, when did the factories start producing 20 nm chips?and when was the first shipment of 20 nms? (Jan 2014 I assume). Now without any complications, we all know if the cards are coming in Feb. 2014 then there obviously was not enough time for 20 nm integration, late march maybe but still unlikely..., Computex very likely. Based on clevo's road map and the short time frame they had with production and integration of 20nms, I can humbly conclude 28nm is showing up

 

They will have to have time to do some validation testing too.

Beamed from my G2 Tricorder

 

Wow, hold your horses man.

I am just answering to the same questions u ask over and over. Getting hard for ya, right?

First:

When u say: "That 12% was an example of what new architecture can do based on your 12% number for GTX 880M" U shave have said it earlie, to make things clear. If I knew this, I would not have written what I did. It is your communicational fault, since words do matter when explaining something.

U are confusing, the misconceptions...

U said Maxwell is coming 20nm a few pages pack and backed yourself up with 20nm production rate. Okey.

From your latest post, I cant understand what side u are on now? As it seems U just write a bunch of things and jump from one idea to another with no proper conclusion, making your post carry 0 meaning.


I reference to desktop cause it is the first platform to receive a process. It is a straight forward connection with Mobile end(smth I repeat 10 times for u).
So, according to u I am now allowed to talk about something which is correlated to the topic?! It is absurd.

If u are done with your arguments, why start harassing other people? It is not nice.

 

Finally, smth connected with the topic from u..

 

Among the reasons why I said mobile Maxwell`s will only be made in 20nm was that it will feature ARM CPU. Which will need room on the PCB, it use power and it output heat.
We have all seen how cramped MXM cards are and how little room there is to spare for extra die. Here is Denver, that ARM which is said to be included along with the GPU
http://images.anandtech.com/doci/7621/NVCES-053_678x452.jpg.

To quote Extremetech:

http://www.extremetech.com/computing/174023-tegra-k1-64-bit-denver-core-analysis-are-nvidias-x86-efforts-hidden-within
I don`t know which variant is going on the mobile GPUs though. But it will undoubtly need power, it will output heat and it will require room.

Desktops don`t have problems with fitting a die like that one the PCB since they are huge compared to the MXM cards we use. Mobile hardware have much stricter requirements for heat and power, than desktop hardware.
Nvidia may not have much choice than building the mobile Maxwell`s on 20nm because of size, heat and power requirements. They may have to wait until late March/April before pushing out mobile Maxwell.

The only card we have been hearing about is a GPU launching on desktops. It is still a rumor, and have not been proven to come yet.

I do take risks when I try to make forecasts like this, but atleast in my head they do make sense. I do not know. I`m only trying to assume based on things I know.

 

Where did I say something that wasn't?

Beamed from my G2 Tricorder

 

They more than likely recieved 20nm samples long before TSMC reached mass production. They have probably already made many 20nm ES of a model.
Mass productions means that the process is ready and the tools in the factory are working like they should.

Here is another reason why 20nm will be here faster than anyone expects.

I follow these things. I make forecasts out of them. I dont base them on nothing
For those who don`t know: 2012 was the year 28nm Kepler came out.

http://www.xbitlabs.com/news/other/display/20131022230815_TSMC_Shares_More_Details_Regarding_16nm_FinFET_and_20nm_Progress.html

 

Not every Maxwell will feature them. Only the high end, which wont be here for a while. What VC wrote in their latest article seems very plausible:"The new feature of the Maxwell architecture is a coherent memory, which is basically a special buffer for both CPU and GPU. This feature is not exactly new, it is available through CUDA interface, but this time NVIDIA will take this to the hardware level."

Desktops too, have their standards to follow. Plus, Nvidia did not clarify what Maxwell will feature ARM processor.

All I write is not based out of nowhere but has its prove of existence. Sometimes when things are not clear, I have to respond how it reads which is definitely not my fault, since I always describe everything as deep as possible.

20nm wont do any good for Nvidia, since it is quiet costly to make.

 

"Can we close this thread until 800 come out"

It is a few pages back.

 

Found 880m SLI FireStrike Graphics Score 11708

NVIDIA GeForce GTX 880M SLI - Notebookcheck.com Technik/FAQ

 

And there we have it

Article from yesterday:

TSMC Begins Volume Production of Chips Using 20nm Process Technology - X-bit labs



Combine that with this article back in October last year:

TSMC Shares More Details Regarding 16nm FinFET and 20nm Progress - X-bit labs


Hello 20nm :thumbsup:

 

Yai.

So we gonna have in late 2014, good news!

 

No, you should try reading the articles again.

 

Do u want to make a bet between us?? Officially.

If not, then don't contradict me more.

 

Here is some food for thought, Mr tough guy:

28nm reached mass production, October 24th 2011
20nm reached mass prodution, January 16th 2014

Nvidia`s 28nm Kepler saw its first light in late March 2012. 5 months after 28nm mass production started.
Yields were horrible on 28nm for Nvidia.

20nm production will be 30% faster than 28nm production according to TSMC.

Come on. You have a brain in there somewhere.

 

I do, but u must have it too, somewhere.

U want to make a bet between us? I say one thing u say the other. In the end will see who is right. Victor takes all reputation and the looser backtracks all this statements and publically acknowledges that he was wrong.

Deal or not?

 

Why should I bet with someone who doesn`t have any arguments to back up his claims?
You haven`t posted anything with even the tiniest useful content in this thread yet

 

I think Cloud has a solid source. "At present TSMC is producing chips using 20nm fabrication process at select modules of fabs 12 and 14. Starting from May, the company will initiate 20nm production at the fab 15 modules 3 and 4.". Cloud can you explain what the difference between fab 12,14 and the 15 one and does that mean since they will start in May, the 20nm will reach the market by June 2014 (Computex)? or since the 12, 14 fabs are in progress we can get them by May 2014?

 

Same with you. U talk about general 20nm production, which not only Nvidia is involved in. Plus, there were and still are the charts what show how much more expensive 20nm is to make, compared to 28nm.

Secondly, I have marketing arguments to back me up as well.

Oh, well. Looks like u are scared. But it all right, that is to be expected

As in the end, should not u be brave enough to bet with a person who in your opinion (does not have any back ups)..

 

And there he went to my ignore list. Congrats, you are the first one to reach that list.
I`m not wasting my time with this guy anymore.

 

Fab 12 and 14 is the ones producing most wafers of all TSMC`s fabs. It is really a good thing that both of them are producing 20nm right now
http://www.tsmc.com/download/3Q13_installed_capacity.pdf

I`m not sure how much each module produce though.

Like always, I`m only giving my opinions and my subjective forecasts, but why shouldn`t we have 20nm here when CeBit launches in mid June?

 

Alright guys. Your creditable source just went ahead and excluded himself from the bet, which if he was very certain at , would definitely have accepted it.

In the End, just wait for Q3, Q4 the earliest for 20nm. As all the sources plus the marketing competition state suggest.

And take this guy with a bigger grain of salt then before.

Ignored him, what a shame to have this pesky dude with this much rep.

What we end up with is 28nm till Q3/Q4. Which is one of the unique cases in gpu world.

 

Which is basically a 770m but only 2GB?

 

Cloud, forgive me for being simple minded but in your opinion based on that info with only fab 12,14 can we get 20 nm at march or will we have to wait till Computex in June. The GTX 770m is on heavy discount, I can wait to march but not june...

 

You are not simple minded.

I have no idea. It all depends on how Nvidia plays it I think. I guess with a soft launch, meaning 1 chip that maybe is used on 3-4 GPUs, a late March release would be possible. Nvidia will now have 2 months to produce 20nm wafers at TSMC until we are mid March. With Kepler it took 5 months to produce enough 28nm chips. But they were based on 3 different Kepler chips (GK107, GK106 and GK104), and was used on 7 different GPUs.

Atleast wait until March to see what happens. Nobody knows for sure.

I think May-June is most likely though. We should atleast have products ready to buy then. Among them, GTX 880MX I hope. Hello 880MX SLI.

 

Oh my word this thread has become a pissing contest, why can't Cloudfire have a different opinion than your own? it is pretty ridiculous that it came to this, what happens if I say 20nm will come February 2015? will you want to bet with me too? grow up, this is a speculation thread, no one and everyone is right (for now), and no one will care who was right when the time comes .

 

I am not saying he cant. He always got in the way when I was talking about Q3 2014 Maxwell release.

Now he says the same.

I do have my opinion and as long as I can freely express it, it is all fine. U had to read the convo carefully, before making such conclusion.

 

Good thing we have a quote function:

 

Good thing u have an edit button. As the post was submitted prior to your edit.

Anyway. Everyone will stay with his/her opinion.

Lets have a deal: Everyone continues doing what they are doing and don't disturb each other and it brings too much hassle, as show by the example.

 

I originally wrote "end of Q2 for sure" (which isn`t late 2014...) but thought I should try to pin it down to months instead.

 

This is the article I responded with : "Yai So We Gonna have in late 2014, good news!"

And not the one u posted above. With Edit function in place I can do some magic with everyone`s posts here as well

 

Bragging rights here. Good u have smth left in defence.

 

Somewhere I read (out of context) - 20nm is going to be more expensive!!

Someone care to explain how and why? I might be looking from the wrong POV.

 

Search it online, there will be a ton of info. But I can spam this thread with context from them if u really want and not being able to do it yourself.

This is info, general 20nm which Nvidia is only a part of..

 

So you suddenly changed your mind from late 2014 to Q3 after that? Its going the right direction...

Anyhow, this is seriously wasting time on bullcrap. You believe what you want, I`ll do the same.

This messed up discussion ends with that funny post.

Cloudfire out. Cheers

 

Next Gen Graphics and Process Migration: 20 nm and Beyond | PC Perspective

Quote from the author of the article: Make no mistake, there is a lot of research in a LOT of different areas to overcome the issues that the industry is running into. The challenges have always been there (breaking the 1 micron barrier was seemingly huge), but now the challenges are just bigger, more complex, and more expensive

In response to:thanks Josh. Lets hope there is that one guy who says "how about trying this?", and he changes everything.

 

I did not mean links to prove it. I am just asking with scientific curiosity. If anyone can explain I will be thankful.

To what I know, 20nm fab should lead to cheaper manufacturing than 28nm. Selling costs may be more because its a new product. Now "20nm is going to be expensive" - is it w.r.t to selling cost or is there something else to it?