Is Maxwell for soon ? Yes and no - Hardware.fr

[PedantOne]

dont forget, it is primary notebook chip!

[tajoh111]

dont forget, it is primary notebook chip!

I was thinking about that. The gt650 and gt 750 in the macbooks are a bit long in the tooth.

Most notebook OEM's, particularly apple wouldn't want to be using keplar again for the third time. Particularly with a 7790 or 260x derivative mobile likely possible, Nvidia probably wanted a new chip available whenever broad-well hits this year.

If I was nvidia, I would be fighting tooth and nail to maintain some form of partnership with Apple. Apple support of OpenCL is a big wedge so I have a feeling if they were to begin to lose Apple, they are going to have a very hard time regaining that partnership. Right now they have the imac and macbook chips which are likely lucrative for Nvidia. I have a feeling losing one is going to result in losing the other.

None the less mobile reviews are boring. Usually they turn into laptop reviews quickly in all but the most highend chip meant for clevo's.

[Olivon]

http://wccftech.com/nvidia-maxwell-g...-arrives-2014/

[w0mbat]

Maxwell looks really good IMHO. If this even remotely true, having close to 2x the Perf/W is a huge leap. Imagine twice the power of a 780Ti within the same TDP or the same power in half. Valve's "Steam Box" just got a lot more powerful!

[zalbard]

Maxwell looks really good IMHO. If this even remotely true, having close to 2x the Perf/W is a huge leap. Imagine twice the power of a 780Ti within the same TDP or the same power in half. Valve's "Steam Box" just got a lot more powerful!

Keep in mind that's only DP GFLOPs. DP GFLOPs of Kepler arch is not too impressive...

[AliG]

Keep in mind that's only DP GFLOPs. DP GFLOPs of Kepler arch is not too impressive...

Still a big accomplishment. This pretty much makes the Titan obsolete

[Darakian]

I doubt that the 2x Perf/W is going to be for double precision. Nvidia gimps their geforce line so it can sell teslas to researchers at a huge mark up.

[Hornet331]

well we gona see in 2 days what i really can do.

My personal guess is, that it pretty much performs just as you expect from the price class, not much better then the previous cards or the competition, but what we will get is better power consumption.

[Dimitriman]

If the top dog Maxwell gets around 50% more performance than 780ti then it will be a great achievement. Anything less is business as usual. Better thermals and power are always welcome though.

[snoid_zero]

[Olivon]

My guess is that Big Maxwell (GM200 or GM210) will double performance over GK110, like they did with GF110 -> GK110.

[Darakian]

My guess is that Big Maxwell (GM200 or GM210) will double performance over GK110, like they did with GF110 -> GK110.

That would be very nice. I certainly hope nvidia decides to make a ghetto tesla again (titan)

[GoldenTiger]

My guess is that Big Maxwell (GM200 or GM210) will double performance over GK110, like they did with GF110 -> GK110.

That's my guess too.

[Dimitriman]

That's my guess too.

Just because it has much better perf/watt does not mean it will double the overall perf. At some point you run into size restrictions. Remember GK110 is already 500+ mm2. Even at 20nm, you won't be able to make anything much larger than that from the Maxwell uarch.

[boxleitnerb]

Perf/mm2 also increases with new architectures, not just due to node shrinks. And remember, GK110 already has 3840 ALUs - it's just that 960 of them are DP ALUs. If Nvidia could somehow use them for SP, too, that might save a lot of space too.

[Sh1tyMcGee]

Dominating in France maybe. I know here stateside I can throw a rock and hit a 780, but 290s are as elusive as bigfoot.

Thats the most flawed statement i have ever heard, thats because for every 100 gtx 700's there is 5 r9 290s lol.....

[Hornet331]

So yeah first benchmarks are out:

http://www.computerbase.de/artikel/g...xwell-im-test/
http://ht4u.net/reviews/2014/nvidia_...50_ti_im_test/
http://www.anandtech.com/show/7764/t...review-maxwell
http://www.tomshardware.com/reviews/...view,3750.html

So far a mixed bag imho. The serverly cut down texture units cost massive performance in some game, best bang for the buck is the 265 right now, but if you consider the power consumption the 750Ti is the better choice.

To bad they only released the real lowend cards now, wonder when they gona bring the better ones and how godo they actually scale and if they can also bring down power consumption with the higher end cards.

Whats kinda interesting is that the 750Ti has nearly the same performance in litecoin mining as a 265 but with more then 50% less power consumption. Wouldn't be surprised if the first miniers will jump on it quite fast, since 4 cards will outperform a 290x and consume less then it (and are also cheaper then the 290x, in the US at least)

[zalbard]

Whats kinda interesting is that the 750Ti has nearly the same performance in litecoin mining as a 265 but with more then 50% less power consumption. Wouldn't be surprised if the first miniers will jump on it quite fast, since 4 cards will outperform a 290x and consume less then it (and are also cheaper then the 290x, in the US at least)

Not a good idea since you'll need more CPUs, motherboards and such... Taking a slot by itself has a certain cost associated with it.

[Hornet331]

Not a good idea since you'll need more CPUs, motherboards and such... Taking a slot by itself has a certain cost associated with it.

Why should you need more cpus? you can easily put 4 cards on one board, doesn't even need to be a 16x slot.
The only thing you probably would need is cards with a power connector since all the cards combined would draw to much from the board alone or get pci-e raisers with power cords.

In the end it still would be cheaper then a solution with a 290X.

Then again even with this you hardly will make any money, but that hasn't stopped the mining craze till now.

[Dimitriman]

Time for sun powered bit coin farms...

[stuffme]

Not a good idea since you'll need more CPUs, motherboards and such... Taking a slot by itself has a certain cost associated with it.

Not a problem!

http://www.asrock.com/mb/intel/h81%20pro%20btc/

[paulbagz]

$/watt/kh isn't there for that 750.

Maybe the higher end Maxwell stuff will be nice but can't even see the top end stuff surpassing 7950/70s greatly just yet.

-PB

[Hornet331]

$/watt/kh isn't there for that 750.

Maybe the higher end Maxwell stuff will be nice but can't even see the top end stuff surpassing 7950/70s greatly just yet.

-PB

Well going by available numbers 7950 have 0,38W/kh and 0,43$/kh (but thats with EU prices not the inflated US prices) (going with 650kh for the 7950 with 250W)
The 750ti comes to 0,23W/kh and 0,57$/kh (MSRP) (going with 260kh)

Even going with the best 280X results available (~750kh with 250w -> 0,33W/kh) you cant beat the W/kh of the lower maxwells. So yeah for mining the 750ti will amortize much more quickly then any amd card out there currently.
Nvidia really delivered a damn power efficient piece of hardware there.

[NEOAethyr]

If they can get the watts down by 8x more times...
That's what I'de like to see (15w apu with the gpu of the 750ti give or take...).
I wonder, how many years that will take?

If I remember right they haven't even done a die shrink yet on the new chips.
This is just the new arch that lowered the wattage.
Maybe we will see some 45w midrange gpu's next year...

[snoid_zero]

Not a good idea since you'll need more CPUs, motherboards and such... Taking a slot by itself has a certain cost associated with it.

nope...it's just as simple as this.