Nvidia launches five 40nm mobile GPUs

The GeForce 200M series just got a whole lot bigger. Nvidia has announced five new mainstream entrants to the mobile GPU lineup, all of which pack a brand-new 40nm graphics processor with DirectX 10.1 support.

Unlike the GeForce GTX 280M and 260M, which feature a tweaked 55nm G92 GPU, these latest mobile offerings are derived from the same architecture as the GT200 GPU seen in desktop GeForce GTX 200 cards. Nvidia has gone a few steps further, moving to a state-of-the-art 40nm process, adding DirectX 10.1 and GDDR5 memory support (both firsts for GeForce products), and making "adjustments in the micro-architecture to improve battery life and overall graphics performance compared to the previous generation."

If you were expecting a super-high-end 40nm behemoth, though, think again. The fastest of Nvidia’s new mobile offerings only has 96 stream processors—a far cry from the GT200’s 240 SPs. Here’s how specs break down across the five newcomers:

Model GeForce G210M GeForce GT 230M GeForce GT 240M GeForce GTS 250M GeForce GTS 260M
gigaFLOPS 72 158 174 360 396
SPs 16 48 48 96 96
Core clock 625MHz 500MHz 550MHz 500MHz 550MHz
Shader clock 1500MHz 1100MHz 1210MHz 1250MHz 1375MHz
Memory clock 800MHz 800MHz 800MHz 1600MHz 1800MHz
Memory 512MB GDDR3 1GB GDDR3 1GB GDDR3 1GB GDDR5 1GB GDDR5
Memory bus 64-bit 128-bit 128-bit 128-bit 128-bit
TDP 14W 23W 23W 28W 38W

Thanks in part to architectural improvements, relatively high clock speeds, and GDDR5 memory support, Nvidia tells us the GTS 260M is actually almost as fast as the older GTX 260M. (As we told you in March, the GTX 260M has 112 SPs, a 256-bit memory interface, and GDDR3 memory support.) In case Nvidia’s model numbering scheme doesn’t make sense to you, here’s how the new cards officially match up against previous offerings:

Looking at the bottom of the ladder, Nvidia claims the new GeForce G210M delivers up to twice the performance of the old GeForce G110M with 50% lower idle power draw.

The company goes on to say we should see the new GPUs in "over 100" notebook designs by the end of the year. Competing designs will likely carry AMD’s Mobility Radeon HD 4860 and 4830, which came out three months ago and also have 40nm DirectX 10.1 GPUs with 128-bit GDDR5 memory interfaces.

Comments closed
    • eitje
    • 10 years ago

    q[

    • ssidbroadcast
    • 10 years ago

    Alright, now taking bets for Number of Renames this GPU will undergo!

    Who’s in for two?? Three? Four??

    • ltcommander.data
    • 10 years ago

    The big question from this announcement is what is happening with DX11? nVidia vehemently denied the usefullness of DX10.1 when ATI was promoting it. Now when DX11 is just around the corner, nVidia suddenly decides to take the effort to redesign the GT200 architecture to incorporate DX10.1. More likely than not, this may be an admission that nVidia won’t have DX11 GPUs ready as soon as they hoped. Or at the very least DX11 mobile GPUs are a long ways off and the mobile line finally gets a refresh seeing that the 9xxx/GT1xx lineup draws it’s roots as shrinks from the original 80nm cores released in 2007.

    • lamparalaptopiaguita
    • 10 years ago

    i am really really sick of nvidia’s naming schemes

      • Silus
      • 10 years ago

      First it’s because it’s confusing to have GTX 200 cards with 8800s and 9800s in the mix. Now that at least the mobile segment is solely using the new naming scheme, you’re still sick of it ?

      Talk about nitpicking or simply being a fanATIc…

        • OneArmedScissor
        • 10 years ago

        They have had countless very similar cards labeled 8000, 9000, 100, and 200.

        Go look up a list of them all. It’s extremely confusing just within one of any of those “families.” I don’t really know a single model apart from the others. They constantly rearrange it so that laptop models which aren’t anything special have numbers indicating they’re as good or better than desktop parts.

        It also doesn’t even tell you if they’re 80nm, 65nm, 55nm, or 40nm. There will now be GPUs of all types out there, and nothing tells you what is what but looking up the specific model.

        Originally, they were supposed to be changing to an all 200 labeling scheme to indicate that everything was 55nm. They have now ruined that, as well.

          • SPOOFE
          • 10 years ago

          As confusing as nVidia’s naming scheme is purported to be, I’ve consistently noticed a lack of confusion over nVidia’s naming scheme. It seems like everyone already knows the deal… when whining about how confusing it is.

    • UberGerbil
    • 10 years ago

    Looking at the top two models: ten watt jump from a relatively modest bump in clockspeed (36% power rise from a 10 -13% clock increase) — but that’s what happens when push a design past where it wants to be.

      • ew
      • 10 years ago

      Wait a second. I thought TDP != power draw. ๐Ÿ˜›

        • Silus
        • 10 years ago

        And you’re right. TDP is the the worst case scenario power draw, which doesn’t happen in the majority of situations obviously.

      • UberGerbil
      • 10 years ago

      It doesn’t. Notice that the 230 and the 240 have the same TDP despite having different performance (and presumably different power draw).

      However, if nVidia could fit the 260 into the TDP of the 250, they would have. That doesn’t mean that it’s /[

        • Lans
        • 10 years ago

        Seems very strange though…

        GT 230M -> GT 240M: 0 change in TDP for 10% bump to core and shader clock.

        GT 240M -> GTS 250M: 5W TDP delta for 2x or +48 SPs and 2x memory clock (GDDR3 -> GDDR5), and small bump in shader clock but lowered core clock.

        GTS 250M -> GTS 260M: 10W TDP delta for 10% bump to core and core and shader clock and 12.5% bump to memory clock…

        From that, it looks to me that:
        1.) 500 or 550 mhz core clock is very little diff to TDP.
        2.) 1100 or 1210 mhz shader clock is also very little diff to TDP… should be under 1.0W diff. So 2x (+48 SPs) adds 1 to 2W diff give or take?
        3.) So GDDR3 800mhz to GDDR5 1600mhz looks the most power hungry.

        Based on that, I could have gotten 550mhz core, 1250mhz shader, 1600mhz GDDR5, and 112 to 144 (96 + 16 to 96 + 48) SPs at around 30W to 33W instead of 38W?!!!!

        I guess they are memory bound or 10% to shader clock helps more than 50% more SPs???? Or, more cynically, don’t want to overshadow the GTX 260/280M???

        Will have to wait for benchmarks and over and underclocking results! ๐Ÿ™‚

          • UberGerbil
          • 10 years ago

          The TDPs aren’t necessarily indicative of actual power usage, especially in the middle of the lineup where multiple models share a common TDP. And power usage is not linear with clockspeed outside of a certain range (at the high end it’s profoundly non-linear, in fact).

      • HighTech4US
      • 10 years ago

      Intel’s TPD also seems very strange …

      Intel’s Pentium Dual-Core E2220 (2.4 GHz) and the E2140 (1.6 GHz) both have the same 65 watt TDP.

      WOW A 33% clock decrease yet the same TPD. Something must be wrong with Intel’s process. Lets repeat this over and over again until you believe.

      ————————————————————-

      Now do you see what an idiot fanboi charlie is. Nothing is wrong with either Intel’s or nVidia’s processes. TPD is a design point and what the manufactures can guarantee their part won’t go above.

      For all we know 260m could have a TPD of 29 watts but the limit of 28 watts puts it into the next rung at 38 watts.

      Also expect future parts (280m) to be placed into the same 38 watt TPD.

        • Silus
        • 10 years ago

        Heh, you only said TDP correctly…once ๐Ÿ™‚

    • kilkennycat
    • 10 years ago

    Now, all we need is a 40nm version of the GF9300/9400M (Ion and Macbook core-logic) to complete Intel’s current core-logic misery in the lap-top/net-book market. Souping up a 40nm 9300/9400M with a 32 stream-processor version of the IGP instead of the current 16 stream-processor would be icing on the (nVidia) cake. Pairing it up with a dual-core Via Nano would give a netbook or long battery-life low-cost laptop “made in heaven”.

    • vikramsbox
    • 10 years ago

    For a moment, I thought that I was in heaven! 45W TDP for a GPU. Then I saw the “M”. It would be nice if they could design powerful desktop GPUs on 40nm with TDPs of around 60-70W. That would be progress.

    • pogsnet
    • 10 years ago
      • Silus
      • 10 years ago

      And you know that how exactly ?

        • asdsa
        • 10 years ago

        Reading the news? And very nice naming scheme again nvidia (Not).

          • Silus
          • 10 years ago

          News ? There are no “news” about this, just rumors…

          But it seems you are considering the BS out of the idiot Charlie’s mouth to be “news”. Obviously the other rumor about NVIDIA being first, doesn’t matter to you ๐Ÿ™‚

          And what exactly is wrong with the naming scheme ? Following their new naming scheme is wrong now ?

            • pogsnet
            • 10 years ago
    • Silus
    • 10 years ago

    I don’t see how G210M is part of the GT200 family. You can’t have 16 Stream Processors in an architecture that has Stream Processor clusters with 24 of them. The others make sense (with 48 and 96), but this one just doesn’t add up.

    As for AMD’s mobile Radeons, I have yet to see any of them on a actual product, so the three month period that they were reported to exist, seems to have been just the usual paper launch. Let’s see if these are different.

      • gtoulouzas
      • 10 years ago

      Even ati’s mobile 4850 is not exactly widespread these days ; in fact I’ve only found on msi’s impressive line of “gaming laptops”. Laptop makers seem to be missing the point that they should diversify if they’re going to entice buyers away from low end netbooks. Enough with the integrated graphics and 3540/9600 shit..

        • Silus
        • 10 years ago

        Well, I’ve seen dozens of laptops with NVIDIA’s high-end mobile graphics cards, which is why I’m saying the HD 4860 was definitely nothing more than a paper launch. I can easily find a laptop with a GTX 280M (which was released not too long ago), while I can’t find a single laptop with a HD 4860.

      • Goty
      • 10 years ago

      None of these cards are really part of the GT200 family, they’re all built off of G92 and its lower derivatives, AFAIK.

        • Silus
        • 10 years ago

        Maybe you missed this part of the report:

        “Unlike the GeForce GTX 280M and 260M, which feature a tweaked 55nm G92 GPU, these latest mobile offerings are derived from the same architecture as the GT200 GPU seen in desktop GeForce GTX 200 cards.”

          • Goty
          • 10 years ago

          Yes, yes I did.

          *stands in a corner in shame*

            • Silus
            • 10 years ago

            No cake for you ๐Ÿ™‚

    • khands
    • 10 years ago

    Well, I guess this goes to show they /[

      • UberGerbil
      • 10 years ago

      Mobile parts from red and green don’t get compared apples-to-apples the way discrete desktop GPUs do, so there’s less opportunity for a price war.

      • OneArmedScissor
      • 10 years ago

      Yes they were, and are. Go look at Radeon 4770 stock. Nada. It’s been that way since it came out. There were supposed to be “non-reference” models very shortly after the original launch, but not a single one has been able to be released. And how many 4860s and 4830s have you seen in laptops, which, as this article notes, have been “available” for 3 months?

      Just because they can pull off midrange GPUs doesn’t mean they’re making very many, or that they can do them properly. Remember, these aren’t even as good as the 8800GT, which was just a 65nm modified 8800GTX/GTS, so that’s pretty old stuff at this point. They can’t outdo that at 40nm, which is quite telling.

        • khands
        • 10 years ago

        I was talking specifically about the green team’s woes and rumor that they wouldn’t have /[

          • OneArmedScissor
          • 10 years ago

          I see, sorry for the confusion. I never read about them having issues with design, though.

          The fact of the matter is, they aren’t able to do anything terribly useful with it. If they had something a bit more high end, or an improved 9400M, that would be something but this is just more flooding the market with derivatives of cards they’ve been doing the last several years.

          It’s almost more of a failed attempt to hide the problems, if you ask me. This is a paper launch of nothing interesting. The only given time frame for release is hinted at in this one very loose statement: “The company goes on to say we should see the new GPUs in “over 100″ notebook designs by the end of the year.”

            • khands
            • 10 years ago

            The fact that, while they were able to make a mobile architecture that was based on the 200 series desktop chips, but were unable to make any of them better than their already released mobile chips does speak volumes.

            • UberGerbil
            • 10 years ago

            But does it speak to the limits of nVidia’s designs, or TSMC’s process?

            • OneArmedScissor
            • 10 years ago

            That was kind of my point, but it’s still a good question.

            khands said they had problems with the designs. I don’t know anything about that.

            TSMC is obviously having problems with 40nm, though. Still no more Radeons…wow. It makes me wonder how long they spent just building up one giant batch, and how long it’s going to take them to get things right.

            I read a rumor about AMD releasing a new GPU at the end of this month, though, and they were showing them off at Computex, so I suppose it’s POSSIBLE that they dropped the production of cheaper GPUs in favor of entirely new ones.

            • khands
            • 10 years ago

            It was a rumor going around that Nvidia was having problems with their designs, not fact, so this proves otherwise, that’s pretty much all I was saying about that.

            As for TSMC, yes, their 40nm is pretty messed up right now. And may mean moves (at least partially) to Global Foundries for either the next generation or the generation after that. It would be nice to see the spin-off be profitable.

Pin It on Pinterest

Share This