HURRAY: Nvidia 600 series not just Fermi!! (Kepler)

why is What's New - GeForce in chinese!

 

because you picked the language?

 

nope click on usa and it goes to chinese

 

Hmm mine is US and displays English.

Thought I will share this for some gags.


M17X R4 with GTX 660M and GTX675M GTX970M plus HD 7970M
A few more details out on Alienware M17x R4 gaming notebook by VR-Zone.com

 

On the bright side, most of us have been way too distracted with Mass Effect 3 to worry about the sad state of GPU updates.. Guess thats a bonus! (Great game BTW..)

 

News are slow as usual. Here are two small news:

HWinfo made some changes to their software:

http://www.hwinfo.com/forum/Thread-HWiNFO32-and-HWiNFO64-v3-93-1540-Beta-released

The winds have whispered about two versions of GTX 660M, one with 128bit GDDR5 and one with a 192bit bus and GDDR5. Take this with some salt, but it sounds plausable since GTX 460M also came with 128/192bit.
Good

 

What scares me is that 680m isn't on that :|

well here's the latest info about "GTX 680", not really "new" per say

http://www.fudzilla.com/home/item/26308-nvidia-gtx-680-pixellized-in-more-detail

190 watt TDP and ~300mm die size is interesting

 

I am really curious how will the gtx 660m (if kepler) fare against 570m/580m, if it matches/surpasses 570m, then we will be having fun with 680m but looks like we still have a long way to go until 680m is released...

 

It will surpass the 570M for shure. Especially with a 192 bit bus. Then it might match the 580M too, or even surpass it. GT 640M saw a 93% GPU score increase compared to the 540M. If GTX 660M see the same 3DMark11 score boost it will get 3555 points which will put it 400 points above 580M.

 

Isn't the GTX 670M supposed to be a rebadged GTX 570M, and likewise for GTX 675M and GTX 580M?

If that's true then it doesn't make sense that the GTX 660M would exceed the performance of the GTX 570M, since if it did so then it would also exceed the performance of the GTX 670M.

 

Yeah, if anything they named them in order of performance, there's no way the 660m is going to outperform the 670m.

 

GTX660M will probably be a GT640/650M but with GDDR5 memory instead of GDDR3. Would be good for a 40% boost with the different memory alone

 

From Semi Accurate:

 

Yup, the 670M = clocked higher 570M and 675M= 580M.
If the 660M have a 192 bit bus like mentioned, it will surpass the 570M. With a 128bit bus I think it will beat the 570M in lower res but struggle with 1080p gaming.

Higher number in the same series equals better performance yes. But I don`t know, I think the mixture of Fermi and Kepler will lead to lower number beating higher number this round. Let the confusing begin

Of course I don`t have anything other than the GT 640M score to base it on. 40% boost over the GT 640M like Devenox thinks means 100 points above 570M with 3DMark11. But I am pretty shure the GTX 660M will have more than just faster memory since GT 650M is there in the middle too with faster performance than GT 640M. So either more shaders or higher clocked GPU core/shaders for the 660M compared to 640M which will lead to even higher performance than just faster memory.

Here is what Saltius think how the GK107 will turn out to be which I think won`t be too far away .
More shaders and faster memory with GTX 660M

 

My ideal 660M has 448 shaders and 192-bit GDDR5.

Realistically, 384/128-bit GDDR5 is probably what we'll see.

 

I think hey might have overshot the claims with 2-4x perf per watt improvement, which is just fantasy. Still, I'm not settling for anything lower than 580m performance this gen.....I'm expecting at least 50% improvement, even that might be too much .

-----------BTW you can't just compare models from different gens, e.g. 640m to 540m and say "ohh Kepler is something percent better", it's unfair due to the weird naming systems they have..if we had at least a TDP for the 640m it would be the best to compare the performance between gens, otherwise all these comparisons are 100% pointless.---------

The 660m will also not beat the 670m by any margin, or else they'd have switched names.

 

I think all of the Kepler News/Non News has given me ADHD.

 

for sure 660m won't beat 670m, that's out of question, but it can beat 570m (670m will be at least 10% better than 570m, like we had 580m was 10% better than 485m), then it is interesting, as 560m is about 30% slower than 570m, I think this alone is a crazy boost in performance..

 

A Damn Headache Disease? Because that's what it's giving me

I'm just trying to figure out what to get in my next laptop, but the manufacturers are making things way too difficult!

 

Levovo Y480 with 630m, german source

Maybe Yields are really that bad? The lenovo y480 can already be preordered with an ivy bridge processor but only with 630m and NOT 650m like originally announced

 

$1k is a bit pricey for what you are getting there IMO. $799 Maybe.. $699 would be even better.. But.. im sure it will come down, i mean, the Y470p went for for only $750 at launch.

 

$1000 is OBSURD!!!

 

whats even worse is battery life: 3,5 hours the article says

 

It's possible, the 470M was better than the 480M. Since it looks like the 570M and the 670M are the same, with a new architecture the 660M may be able to top the 670M.

 

I will be utterly disappointed if the GTX 660M won`t surpass the 570M. I don`t know how the 128bit will do on 1080p but with a 192bit fully GK107, I`m expecting it to beat the GTX 570M solid

Yikes. Its not impossible that Nvidia is struggling with the 28nm. Semiaccurate just recently posted an article saying Apple drops Kepler GPUs from large number of upcoming laptops. It makes you wonder if AMD is using all their 28nm die on desktop GPUs instead of notebook GPUs because they have the same problem too and thats why we haven`t heard a single word about their GCN mobile GPUs

http://semiaccurate.com/2012/03/13/apple-drops-nvidia-kepler-from-large-numbers-of-laptops/
http://www.tgdaily.com/hardware-features/62058-report-apple-drops-nvidias-kepler-from-low-range-macbooks

 

then they would have named gtx 660m as 670m and 670m as 660m, no need to confuse people, why should they even stick to a weird naming considering architecture is changing...

also gtx 480m was a trial, and they corrected it with 485m, please do not confuse 480m as the predecessor of 580m, 485m was the pred of 580m, and as soon as 485m was out there 480m was out of table.

 

graz: Embrace the chaos, don`t fight it

 

They can get away with it if they put out 670M long before 660M.

 

What both scenarios have in common is an architecture change later within the same naming series; assuming that the 660M and the 670M are Kepler and Fermi, respectively.

An improved architecture within the GTX 4XXM series is what led to the 470M outperforming the 480M. So an improved architecture within the GTX 6XXM series is what could lead to the 660M outperforming the 670M.

 

"but that the included GeForce GT640M is based on the long-awaited 28nm Kepler architecture." source

GF GTX 680 Kepler will be released on 23rd of march source(in german)

GT 640 Release date 01.05.2012 source


Edit: Just found a review of the M3 at anandtech incl. GT 640

From the review: "At 1600x900 the GT 640M is undoubtedly feeling the pinch from the 128-bit DDR3 memory bus, but it still holds its own. This is about the highest resolution you could conceivably want to run this GPU at, at least without GDDR5 (which the 640M does support)." So I Guess an 640M with GDDR5 would be possible for an ultrabook...

 

It depends on whether NVIDIA included support for GDDR5 for the GT 640M.

 

The way I understood what was written here :"At 1600x900 the GT 640M is undoubtedly feeling the pinch from the 128-bit DDR3 memory bus, but it still holds its own. This is about the highest resolution you could conceivably want to run this GPU at, at least without GDDR5 (which the 640M does support)."

One can come to the conclusion that the 640 does indeed support GDDR5(who if not NV has made it possible?!).

 

That is not to say we can determine whether NVIDIA enabled support for the GD5 VRAM in the VBIOS. However, one thing we can be sure is that most 1080p notebooks will not feature the GT 640M.

 

So when does the Embargo/NDA stuff expire? Mar. 23rd? Makes me want to hibernate and come back then since I'm only really interested in the 575m/580m(?).

 

Latest rumors:
NDA of Kepler is ending March 23. That is when we will see the official launch of Kepler.

GOOD NEWS EVERYONE

The performance of the GT 650M

 

This is my line of thinking as well.

If 660M is a Kepler chip, it's almost impossible to make it slower than the 570M. I mean, the 640M has 384 shaders and is close to the 460M, so the 660M is only going to be uphill from there.

 

3DMark will not recognise the new cards at all. The field should be blank. But I want that performance from GT 650M. Then comes the question- Is GTX 660M a Kepler? What kind of performance will it have?

From several sites that talks about Kepler, the said dates were March 12, 23, or 26. Since March 12 has passed without much information released, take these dates with a pinch of salt.

 

That is the real benchmark of the GT650M. If 3DMark doesn`t detect the 650M the bencher must have edited some settings so that the name pops up.

It is the Samsung Q470 that have been tested, here is the picture of the laptop in HIS POSESSION as proof. So please do not derail this thread with the same crappy "fake" posts as this thread saw from a few people when I opened it ( it is very discouraging). They were real and the score above is also real

Instead find some comparisons of the 3DMark11 and post them



This thread should be closed. We have the Kepler thread I made and it is functioning good. No need for two paralell threads with the same discussions.

What is written here have already been talked about in the other Kepler thread

 

The new Samsung Q470 was launched in China almost three weeks ago.

First spotted on 03 March on sale in China
http://forum.notebookreview.com/samsung/649158-upcoming-samsung-q470-notebook-nvidia-gt-650m.html

Today I found a chinese thread with 3DMark 11 score provided by a Chinese owner of the Samsung Q470 notebook.
¡¾Q470 ʹÓÃ3Ì죬¼òµ¥ËµËµ¡¿-ÈýÐÇÂÛ̳-ZOLÖйشåÔÚÏß

 

Seriously. This is just RUDE and useless. I already posted the same in the "HURRAY 600 series not just Fermi" plus some pictures more. You derailed the other thread with saying that the 3DMark11 couldn`t have detected the GT 650M but suddenly it was good enough to post it here. Not even a "Ok, I was wrong" in the other thread when I posted a picture of the laptop in his posession.

Why did you even make this thread? It have the exact same content as my thread. You just had to make your own thread didn`t you?

/rant

 

Why are you going off topic lately. I read about the chinese thread in the afternoon. I held back because of source authenticity issues. Since you said it was real, then if I were to post the link in your thread, you could blast me for spoiling your thread. The main aim when I post is to give information when possible.

And I don't even know your source in the first place.

EDIT: Saw your edited post. Ok, I was wrong. But please supply the source if possible in future for the avoidance of doubt and so that the fellows can judge for themselves its authenticity.

 

No I would rather prefer we stay the Kepler discussions in one thread instead of this thread that was created long after mine. I wouldn`t have "blasted" you.

This thread is useless, and people have to cross check different threads to find something new about Kepler

 

Why not get the mods to merge the three threads together. So, all the information will be consolidated in one thread as you wish. Then have the thread renamed to a new appropriate title that best describes its nature: Speculation.

So far three threads I found, including this very thread.

http://forum.notebookreview.com/gam...-hurray-nvidia-600-series-not-just-fermi.html

http://forum.notebookreview.com/gaming-software-graphics-cards/638044-gtx-660m.html

Yours was created one month after the GTX660M thread like this was after yours.

Would a mod please be so kind to help us so that Cloudfire will feel happy? Thanks to our mod.

 

GUYYYYYS, INCOMING BIG NEWS
Kepler GTX 660M with GDDR5 screenshot:

Kepler GTX 650M with GDDR5 screenshot:


Windows experience index of GTX 660M (I think):


The notebook with these GPUs are the Lenovo Y480/Y580


Y580 comes with a 3610QM Ivy Bridge processor


And a 1080p display


Source: http://ideapad.it168.com/thread-1958679-1-1.html

 

@Cloudfire: Could you please acknowledge the source as the pictures do not belong to you.

GTX 660M same core clocks as the GT 640M and GT 650M, but with GDDR5.

 

Huh... who would have guessed it?



p.s. - I noticed that Anandtech said the 640M's core clock was at 625MHz while gaming.

 

That is the "hot clocks" or turbo clocks actually.

 

This bench for the 640M is uber impressive IMO;

 

I am merging this thread since yknyong1 has PM'ed me and acknowledged that it is a duplicate. Sorry for the mess!

Cloudfire, next time please don't be so combative, eh?

 

Yes, I'm just just pointing it out for anyone who was wondering.

Hopefully the 660M goes much higher.....