New details about Nvidia`s Maxwell

They are ADORABLE, aren't they? Probably will show up to announce another Pitcairn derivative/rebrand then get tuckered out and go back to sleep up in Sunnyvale because the little guys need they're nappy time.

 

Guys, i have a question. On the benchmarks at notebookcheck.net, why is it that there is always a massive gap between 'high' and 'ultra' setting framerates?
Sorry if it's a noob question, i'm a console gamer (considering a switch to pc)

 

If you hover over the settings, low, med, high are all usually just 1024x768 or 1366x768 and Ultra is 1920x1080 with all details at max. I've always criticized notebookcheck for that. They really need a 720p low, med, high ultra, and 1080p low, med, high, ultra. Or at least a 1080p high added.

 

Thanks a lot. Their benchmarking methodology seems quite inconsistent. I was shocked to see bioshock3 go from 120+(med) to 110(high) to 42(ultra) on the 860m review

 

they should do an Ultra 1080p without the useless 4xAA that they add everytime and that no single card will handle apart the x80M's !

 

Totally agree

 

Agreed 4xAA at native resolution is ridiculous and absolutely not required.

 

Guys 60-70 FPS @1080P with physix is really what I am looking for, my options are 860,870 or 680, can all of them pull that off?and whats with Blu-ray 3d being a new feature in 800 series?I thought a chart in the nvidia website said all GTX 280+ can pull Blu-ray 3d off?

 

It's changing a bit.

The 850M is also being called a GTX part, but the 860M is the lowest one that will support SLI officially.

 

lol :laugh:

 

Come on man, that's such a vague question depending what game you're talking about and everything. I know you're having problems surfing some of the net with your machine, but really this website is going to answer loads of your questions if you take the time to browse into the depths of it. Start off at this web page, showing game comparison frames per second with different cards:

Computer Games on Laptop Graphic Cards - NotebookCheck.net Tech

(Take what you know about the specs & 3DMark results that have been leaked for the 860 & 870, and then compare those against similar GPUs using that site, which will then enable you to draw some comparisons to game performance in the web page I linked you, 680M results are right there to be seen too).

I think you'll learn more becoming familiar with notebookcheck.net than you will from asking random questions here.

 

It's actually looking like 860m stock is lying right in between 770m and 680m. With some overclock will exceed 680m even at 1080p. Very impressive.

 

It seems the question is not so random after all, and 3Dmark 11 only measures 720p, not 1080p! and 870's 3D mark was 7280 compared to 680 5789, now because that is almost 20% of a difference which does not translate in 20% actual performance so actual differences are yet to be known...

 

To be honest your questions are getting a little boring and repetitive. That website, notebookcheck.net, will go a long way to give you the understanding that you seem to want.

 

Notebookcheck has neither FPS results from actual games nor benchmarks for 870 ,and comparing to the 680MX or 680 is not a simple task because the impact of ROPS is not easily measurable and cloudfire himself supported that. Notebookcheck is by all means a useful source but as HTWING said above the FPS results from all video cards only show 1080p ultra with so many F/MS/AA that all cards would stutter under 30 fps and therefore the difference in 1080p gaming without additions is not known between 860,870,680 which is exactly my point:S!

 

Maybe because it's too new? You need to wait...

 

So i understand that both GTX 850m and GTX 860m has same chip and overclocking GTX 850m you get 860m or no ??????????

 

Yes, but be careful. There are two completely different cards called 860M: one is Maxwell and the other is Kepler. There are also two different cards called 850M: they're both Maxwell, but one uses DDR3 memory and the other GDDR5. Assuming you get the Maxwell GDDR5 versions of both, the only difference between the 860M and 850M is clock speed. Any other combinations will be quite different from each other in performance and/or heat.

 

my guess is the gtx 850m would be a downclocked 750ti while the gtx 860m maxwell version is a binned gtx 750ti.

 

Just letting you know what I think. From your questions I can see you could learn a lot from that site, which was why I linked you to it. You can do what you like of course!

 

Yes i know that, just wanted be sure that gtx 850m overclocked is same gtx 860m. I will buy either MSI GE70 2PC or MSI GE70 2PE. And for more informetion about gtx 850m, MSI GE60/70 2PL uses GTX 850m 2GB DDR3, it's a budget gaming laptop. But in what laptops will be GTX 860m Kepler version???

 

max overclocked clocks on 850m will be lower than 860m's max. and plus you will have to ask specialists for unlocked bios. I still would choose 850m unless laptop is just 50$ lower

 

Are you sure they are using the ddr3 versions? Whats the point of buying those over the gt 750m ddr5/gtx 765m versions then?

 

He must be wrong about GDDR3

 

Just check MSI homepage
MSI Global - Computer, Laptop, Notebook, Desktop, Motherboard, Graphics and more

 

A gazillion models. Holy biiiip

 

no announcement of 790?

againn

 

wuts TL DR?

 

"Too long, don't read..." = TL;DR

Usually a quick summary of their statement (or their main point) follows that abbreviation.

 

i thought its too long didnt read

maxwell 980m LETS GO with 20nm

 

You say tomato... I suppose for the reader, it would be "didn't" because he didn't read it. But for the poster, it would be "don't" because he's telling the reader to not read it (because it's too long).

 

"Too long, dont read" would suggest to the reader that they should not read the lengthy text above which the write took time to make.
"Too long, didn`t read" means that the reader didn`t bother and prefer a short summary which the writer made.

 

TLDW = Too Long, Don't Wait - that's what Nvidia tells gamers about 20nm high-end Maxwell GPUs release... so they would buy notebooks with Kepler based graphics and gave bigger profits to Nvidia.

 

Thank you.

 

Desktop sites are full with :don't bother waiting for 20nm Maxwell" news. Meanwhile it is just a few months from us!!

 

I just don't see Nvidia releasing mobile 20nm Maxwell GPUs as soon as people here seem to think. It doesn't make any sense for them to.
They are just getting around to releasing the 28nm 800 series for Laptops next month and then 3 months later release 20nm Mobile chips?
Not buying it, pardon the pun.
I think what Nvidia is going to do, is to release the 20nm later in the Late Fall 2014, or even a year from now. That makes more financial sense for them.
________________________________________________

 

Not sure if this applies to mobile cards but still interesting nonetheless...

Waiting on 20nm graphics cards from Nvidia and AMD? Don't. | TechSoda

 

If I'm not mistaken, wasn't the original target for Maxwell release to be in 20nm? The 28nm Maxwell seems like a testing bed or compromise while waiting to launch the 20nm chips. Not sure how holding a 20nm release would be financial beneficial to nVidia, as I'm sure they would like to get these chips to market as soon as possible to please the shareholders. They can always rebrand a surplus of 28nm maxwell chips.

 

Fact is, they can afford to take their sweet, sweet time with AMD being so lethargic in the mobile market at the moment. No need to rush when you own 75%+ of mobile market, with little signs of any real imminent counter-threat from AMD. As always I think we'll see 20nm desktop parts released before mobile. Summer is possible, but I question whether they would be so quick to replace their much-hyped 880M. The 580M came about 6 whole months after the 485M - that was the shortest gap in recent times between top-tier mobile GTX flagships. 3 months would be a very short lifespan for the 880M. It would also buck the trend of high-end GPUs based on new architectures being introduced on desktop a few months before mobile (480: 03/10, 480M: 05/12 & 680: 03/12, 680M: 06/12), unless they are planning to release the desktop 800 series any day now!

 

I tend to agree with your line of reasoning in timelines. I just don't see nV realeasing their high-end mobile Maxwell cards until Fall this year since a) it would be unprecedented for both mobile and desktop cards to release a refresh of their "new"/rebranded top card, less than 6 mo after its release b) there is absolutely no pressure from competition, and c) g104 production is still going strong and selling well.

 

Please remind me how long did it take to release 680mX, 675mx and 670mx since 600m series?

EDIT: just checked - 7 month. So yeah, it's possible that Nvidia announce delay after Dx12 announce to "prepare maxwell for new technology."

 

That's very interesting. Given NVidia's huge potential performance increase of Maxwell on 28nm, and given the increased costs of 20nm combined with lack of competition from AMD, I think it's conceivable that NVidia will just stay on 28nm, and just skip the 20nm node. I've not been reading loads of info around this subject, so maybe there are some counter arguments, but so far this makes sense to me. The only downside is if they could afford to wait for the next shrink in process until Volta (which would be 2016 I think).

EDIT: Although if they're going to want to include an ARM cpu on the same chip, then I could see them wanting/needing a 20nm GPU to allow for enough space for both the CPU & GPU on the same chip.

 

Yeah... remember Ivy Bridge? If you check the reviews the TDP was 77w for the i7 but it exceeded 70*C on stock clocks with most aftermarket coolers. Granted, that was because of Intel skimping on the cooling and using thermal past between the Heatspreader and the CPU itself instead of solder, but whatever... it ran waaay hotter than Sandy and overclocked worse. xD

 

U cant compare Desktop chips with Mobile chips.

2 different things. It can take ages for Mobile chips to be released after Desktop chip makes its appearance.

 

but NV said, that Maxwell was the first chip where they did "mobile first, desktop later".

and to switch to 20nm would be a smart move anyway:
1) you get more chips from each wafer -> more profit (if you can handle the process and get similar numbers of "ok"-chips)
2) chips run cooler -> higher clock speeds possible or less energy consumption -> customers more pleased and you can react to competitors easily (just relase a GTX865M with even higher clocks ...)

 

Just found that on the MSI Page for the new GS 70 Stealth Pro with 870M.

As i read the following text ...

"The NVIDIA GeForce GTX 870M leverages NVIDIA's revolutionary new architecture to deliver extreme performance than the previous generation. Compare to GeForce GTX 770M, users could get over 30% faster and more performance at same level new graphics card with lower temperature and watts..." ( source)

... for me that sounds like the GTX 870M even could be a Maxwell Card, which would be (for me) the most plausible answer to the question why they now can run such a powerful GPU in such a slim chassis, whereas the predecessor "only" had a 765M, not a 770M.

Maybe there are/are coming two 870M GPUs? One as Maxwell, one as Kepler, similiar to that we see with the 860m at the moment...

 

I don't think they can run Maxwell on 28nm for the 192-bit and 256-bit chips. They would be way too large. That's why they're only running the 860m and under chips with Maxwell at 28nm now. They *HAVE* to use 20nm tech (or smaller), especially for the desktop chips.

 

Because of having to package the ARM CPU on the same die too? Maxwell transistors are more dense than Kepler transistors apparently, so they've got some space savings by using Maxwell over Kepler, even on 28nm.

 

750 ti review did show that it is a cut-down version of a more powerful card, which is still 28nm and is a part of Maxwell family...