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

You just summarized my entire thought process, verbatim, before I pulled the trigger and opted for the 570m and the machine in my signature. I also really needed a new notebook, so that made the decision even easier. Glad I went with a 570m.

 

Did anyone but me notice that the new GTX 660m will have "up to 384 cuda cores"? While the 570 and 670 still are 336? The 580 and 675 are at 384 still too? I am suspecting the higher end models will show up in Sagers and MSIs and the gimped GTX 660m (sub 384 cuda cores) will go to Asus.

GTX 660m GDDR5 128-bit
http://www.geforce.com/hardware/notebook-gpus/geforce-gtx-660m/specifications

GTX 670m GDDR5 192-bit
http://www.geforce.com/hardware/notebook-gpus/geforce-gtx-670m/specifications

GTX 675m GDDR5 256-bit
http://www.geforce.com/hardware/notebook-gpus/geforce-gtx-675m/specifications

The only thing that has changed is the memory bus/type from model to model in the GTX lineup.

 

Poor Asus Looks like the new G55&G75 series is going to be a horrible fail.

Haha, but I can't afford a 680M probably, and if the 660M won't come out with a 192-bit (it better), then 570M I shall buy.

Time to head to the MSI forum to figure out more

 

You are right about the performens but look at the link i posted eirlier the 680 on full oc was sucking just under 400 watts wile 7970 was just belo500

 

I see, you are right, I think we are thinking pretty much the same, I thought you are saying it is not a good card, it is really good, but it actually shines in performance/power I was not expecting that it was going to crush 7970, but it is clearly the winner (mostly in benches though, as you are saying, in games they trade blows), we'll see, what I am hoping right now is a really nice GK104 based gtx 680m (maybe wishful thinking )

 

basically the new kepler cuda cores are not as fast as the old fermi cores, but since the kepler cores are significantly more efficient and smaller they can fit much more of them on a chip. my guess is that the 680m will be based on the gk104 and it will have somewhere between 768 and 1536 cuda cores.

then after nvidia gets their gk100 all straightened out, we will see a gtx685 as their new flagship.

 

The GTX 660m, is better price/performance right? I mean I only play games like Diablo, WoW, and SC2, maybe GW2 if it's good. So it would fit me right with the new Ivy Bridge stuffs.

I'm hoping to get Clevo/Sager, NP8131, or any 15" model. Which would you recommend I pre-order?

 

Can't go wrong with a Clevo.. or an Alienware.. But as for pre-orders.. i would just wait until everything is available and then decide.

Tom's Review puts it best;

 

Articles : GeForce 600M Notebooks: Powerful and Efficient - GeForce

Official article, only showcase up to 650M. So 640M is supposed to be equal to 460M or 285M.

 

Looks very impressive in my humble opinion but very little 60FPS gaming with those settings (no wonder)

 

The thing is with that particular test, you are comparing over 30% overclock of the radeon with an overvoltage applied, compared to 20%ish of the GTX680.

Thats why the power consumption of the radeon skyrocketted. On average the difference is between them on stock is around 8watts and 25watts... which is also in the article you posted.

Hell even in SLI/Crossfire the difference was 53 watts. With the Radeons having lower power consumption on idle.

 

Monsternotebooks with the P150EM:
GTX 680M "July-September 2012 Delivery"
Monster

 

This could just be because it's in Turkey. The ETA in the US should be shipped out on April 29th/30th

 

Really? Where have you heard that?

 

Well it's the same day IB releases.

Also Logical Blue One - LBO - Horize P150EM Clevo Notebook

While this is an Australian site, it still makes sense that the P150EM would be released in the US first.

 

That link has the 675M as a $399 upgrade from the 670M, that seems really high, even after conversion.

EDIT: Actually, that just sounds like the upgrade price we've had for going from 570M to 580M.

 

The 460M is much faster than the 285M as is, so they didn't run a very good comparison.

 

They are actually rather close in performance. Even though the 460M has more shaders, 285M's shaders are stronger, and bandwidth is close.

 

i was not saying that the 7970 are not great card
And if you read through the overclock part you will se thatthe 680 are at its top staible oc as are the 7970
And to be honest the 680 is a bit of a disapointment in my eyes hope the gk 100/110 comes soon i think thats the one that will do the job proper

 

Cinebench OpenGL test GTX680 a mediocre card:

http://www.pcwelt.de/produkte/3D-Leistung-Nvidia-Geforce-GTX-680-5110282.html

So it seems that GTX680 is a competitive game card only -Forget CUDA calculations and 3D applications and OpenGL/WebGL games.

 

Lol...
If this is the case, it's a major fail.
But it's also possible that Nvidia is actively working on separating the gaming and 'professional' cards intentionally (even though the hardware between the two is exactly the same) so people would be forced to get either one or the other (A LOT more expensive for 'pro cards') depending on their needs.

 

guys, 680m is the best gaming product for centuries, and I think it is enough to be that way, you cannot do everything.. yeah I think they will separate professional cards further (they already started with Tesla, who cares about quadro)

 

Yep the GTX 680 sucks in that respect. Even the Radeon 7870 fares better iirc. But I personally don't give a damn and I'm willing to bet that 98% of its potential buyers don't either. GPGPU focus results in big, complex, expensive, inefficient chips (Fermi) and can be detrimental to gaming performance (Cayman).

With Fermi Nvidia basically said screw that, let's build the fastest gaming & compute chip, even if it has to be >500mm² and inefficient. It proved quite troublesome at first but in the end they achieved their goal. AMD stayed reasonable for Tahiti but the fact that Nvidia improved efficiency with Kepler and went for a ~300mm² chip specifically designed for gamers looking for the best bang for the buck is a bummer for them, as they used to rely on that as their competitive advantage, at least since the 4000 series.

If you factor in the fact that the green team is scheduled to later release their own big GPGPU oriented chip, I think AMD have taken the risk of looking plain silly in the upper-mid end/high-end segments. Since they no longer have the efficiency crown, and they're not in a position to set the trend (since their chips are bigger and more complex), and Nvidia will have 2 better performing chips that will cover both gaming and GPGPU markets.

 

except that the pro and regular cards are identical.
They mostly separate them either via VBIOS or software... or through a different measure.
Besides, CUDA was a fairly recent product introduced... why didn't they simply remove that from gaming gpu's in the first place?

It's idiotic and unfair.

 

dude, not for nvidia, quadro / tesla / geforce are totally different (well, spec-wise, manufacturing point of view I have no idea)

 

GT650M DDR5 benches:

http://forum.notebookreview.com/sager-clevo/647054-clevo-w110er-ultraportable-16.html#post8399634

 

Thanks man. It kinda looks like the GT 650M is equal to 560M since it scores about the same in 3DMark06 but Kepler does it better with DX11 games like have been pointed out earlier and thats why it scores better in 3DMark11

 

This is totally untrue. They are the same cards.

 

so you are thinking number of CUDA cores equals to same cards... I said I don't know the manufacturing process, but all specs except CUDA cores is different, show me quadro 6000's equivalent in gtx family?

 

graz`zt is right, Quadro and Tesla cards have more vram and are gpgpu orientated, whereas the GTX series is gaming orientated

 

Our Chinese friend 慕容蛛蛛 just said that the Clevo GTX680M 4GB is now expected end of June...

 

4GB will not matter for most gamers. Even most notebooks it is destined to be in support max 2 displays at any one time. What I am interested to see is the 2GB variant.

 

That sounds like overkill about as much overkill as having 32 GB of RAM.

 

not to mention the 64gb ram capabilities of those new x79 boards, but back on topic, why would nvidia put 4gb on a mobile chip and only have 2gb on their desktop version?

 

probably they are considering surround display for mobile platform too (otherwise no idea)

 

4TB option for GTX 680 in June!

 

well actually the new p150em has dvi,hdmi, and display port and nvidia says the gk104 can output up to 4 displays so it makes perfect sense to have that much vram if it can support that many displays, and anyways they always stick more vram in on notebooks anyways.

 

Thanks man. Nice of you to have us updated. I wonder if I can wait all the way to July. But its tempting

Yup, agree with this. For people using a single monitor 4GB is a total overkill though.

It makes you wonder if you have to pay a premium for the card, since they stuffed so much VRAM inside to cater for everyone.

---------------------

7970M is due to release anytime now btw. The new Alienwares will offer 675M, 660M and 7970M, plus Clevo/Sager have now discontinued 6970M and 6990M. Something is coming pretty soon

 

when can we expect to see one in a dell xps 17?i am think of buying one in may.i'll wait another month or 2 if it releases.

 

Unless you are working across several monitors and using the full power of your gpu, more VRAM will hardly matter.
As for being more gpgpu oriented... that's only because the manufacturer of the cards force them to use OpenGL (via VBIOS or another type of modification) and gaming gpu's use direct3d (both though have same hardware support... primary difference is in the software).

Hardware-wise they ARE identical.
You also have gaming gpu's that have different VRAM sizes/types - that doesn't mean it does them any good though in gaming.

 

man seriously? He wasn't meaning any performance, VRAM being different just plain simple makes it different, get it? it is not a BIOS modification or driver difference which makes the ram differ... *facepalm*

 

The difference between most consumer and professional GPU are double and single precision performance... Nothing more.

 

Google Translate

This shows the gt650m should play games like dirt 3 without aa I would say at over 30fps on high. Looks like a gtx660m is needed for proper gaming.

 

Oh nice a review. Thanks

They are testing the DDR3 version of GT 650M. It is the Samsung Q470 which also have a weak dual core without turbo. Which is why you see the DDR3 version crush 555M in benchmarks and shockingly is not that far away from GTX 560M lol.
I see that the notebook does very good in Battlefield with high settings and in Resident evil with high settings too. Throw in a better CPU and it would do a lot better in Dirt 3. Or use the GDDR5 version

I tried rep you but have given away too much rep today. You have to wait

 

I saw the gpuz and it is ddr3. Maybe gt650m gddr5 should be as powerful as a gtx560m then. My acer 5930g though is good enough except cpu is poor so I may wait till dell outlet sell for half the retail price l7..x and l521x.

 

The only exception i can see is for people using SLI's 680Ms on a high-rez (4k) external. That would be pretty handy. But those folks are few and far between.

a 4G Desktop card would be much more useful.

Hardware Wise, yes. But the drivers make the biggest difference. The Gamer card drivers are tweaked often for maximize game performance and patch bugs. Thats not the case with the professional / CAD video card drivers.

 

That's my point.
The only main difference between pro and gaming cards are in the drivers (software) which specializes the gpu's to a singular direction.
Hardware-wise, they are effectively the same - therefore a gaming gpu if successful in being moded to be recognized as it's pro equivalent, would behave in 3d programs like a pro card.

And Nvidia/AMD both overcharge for the difference at hand - which is a plain and simple robbery.

 

How can you say it is robbery. NV/AMD are getting payed to support their products for a market where money isn't as important as stability and good support for the programs needed.

It's simply another market. If the products are the same who cares. They serve a different purpose thus the different price.

 

for nvidia this is just wrong, please look at the specs of quadro 6000 (which indeed was a technological BREAKTHROUGH), gtx 560Ti, gtx 570, gtx 580, ALL are different... (that is why they are charging 2k for a Tesla GPU and world at large is buying it, otherwise trust me, companies are NOT stupid...) number of CUDA cores is NOT equal to the card being the same...

for amd, it is right as m8900 is equal to 6970m spec-wise (in mobile world, I didn't check desktop specs yet).

 

My big concern at this point is what comes next haha.

I mean, 600M has largely been a disappointment imo, due to the long string of rebadges. If the 680M isn't absolutely phenomenal, I wouldn't be surprised if nVidia releases something ridiculous in say....6 months after the release of 680M. Like what the GTX 485M was to the GTX 480M, if you're looking for a comparison.