news 40nm desktop geforces appear on nvidia website

40nm desktop GeForces appear on Nvidia website

Yesterday, the Taiwanese rumor mill suggested that two desktop GeForce graphics cards with 40nm GPUs would launch in September. That was incorrect, apparently: information pages that correspond exactly to the rumored GPUs have already appeared on Nvidia’s website.

The company provides specifications and images for the GeForce GT 220 and GeForce G210. Both cards look like budget offerings with much more spartan specs than AMD’s Radeon HD 4770. Both also carry an "(OEM Product)" label, a likely sign they won’t show up outside of pre-built PCs from big manufacturers. Still, here’s how the specs break down:

  SPs Core clock Shader clock Memory Mem. clock Mem. interface
GeForce GT 220 48 615MHz 1335MHz 1GB GDDR3 790MHz 128-bit
GeForce G210 16 589MHz 1402MHz 512MB DDR2 500MHz 64-bit

Other features include support for DirectX 10.1, OpenGL 3.0, and CUDA. The reference GT 220 design includes VGA, HDMI, and DVI ports, while the reference G210 trades the HDMI port for a DisplayPort output. Oh, and both cards have half-height circuit boards, as you’ll see in the gallery below. (Thanks to TR reader SH SOTN for the links.)

0 responses to “40nm desktop GeForces appear on Nvidia website

  1. Haha, posting a “Charlie” article ? Maybe GrooWanderer decided to change his nick to RagingDragon.

    Charlie’s a moron and is 99% of the time incorrect, especially when it comes to NVIDIA news.

  2. Er…no. All the other chips are GT200 based. G210 is the only one that doesn’t make sense at first sight, since the SP clusters should be 24 and not 16. Some people speculate that it’s an error in the spec sheet.

  3. Not necessarily.

    A smaller, simpler, chip is cheaper and faster to shrink to a new process, then to test and validate on a new process. Also smaller process reduces unit cost, and that’s more important on a cheap product like these than for more expensive products. Furthermore, cheapo products are lower visibility if something goes wrong (i.e. high end products get more media attention), and thus a good option for experimenting with a new process.

    That said, TSMC apparently admitted they initially had low 40nm yields, but now claim to have resolved those issues.


  4. One subtle point for CUDA programmers… these are the first single slot cards that have compute model > 1.1. This means many CUDA operations are a lot more efficient (memory coalescing especially) than on older G90 cards. These are compute model 1.2 which means they don’t have native double precision support though… that’s model 1.3 and reserved for the double-slot G200 cards.

  5. ION 2 will probably be based off this GT 220. It certainly fits the rumors about it.

  6. Which is precisely why the 40 nm woes at TSMC must be true. If they weren’t, we would be seeing much more powerful cards at this point.

  7. This might be crazy
    but they should have 2 skus for low profile card (the long one)
    and the shorter card.

  8. If TSMCs 40nm yields are true that’s another good reason to do small chips on 40nm first.

  9. Actually these cards will be pretty good for Windows 7+aero. I noticed that the window-preview on the task bar doesn’t last long on systems with little graphics memory. Is Windows using the graphics card to store that info? If so, cheaper cards with more RAM would make sense.

    For a business/non-gamer machine, I want a card that is
    – cheap,
    – low power,
    – quiet (preferably fanless),
    – easily handles Aero,
    – supports dual hdmi/dvi,
    – can relieve the cpu a little from full screen hd video

    In fact, since I mainly game on console’s now, I would like this for my home computer too!

  10. I bet the fans aren’t doing much and are dead silent. It’s probably a cost cutting measure, so they don’t have to tack a fancy heatsink of any sort on there.

    I have a fanless 9600GT, very similar to the GT 220, and the heatsink on that thing is outrageously large. It can still get kind of hot.

  11. You have to remember (or maybe just go see for yourself) that lower resolutions don’t take fark all, and that’s what most people are running, even if they play games frequently. On the last Steam survey, about 75% of people were using the equivalent of 1280×1024 or less. I ran Supreme Commander maxed out on a 6600GT at 1280×1024, and that thing is a total POS even by cheapo standards today.

    It’s no wonder they focus on these first, as they’re sufficient for a huge chunk of the market, very affordable, and I’m sure they make a tidy profit, considering how tiny these chips must be.

  12. I played fallout 3 max everything at 1400/1040 with a ddr3 variant of the 9500 and did so perfectly fine.. until my computer went boom…

    This would be an itty bitty bit weaker mem wise but alot better shader wise. So ya it can handle fallout 3 and im sure it can handle sacred 2.

    Not on big monitors but anyone getting this likely doesnt have past 22 inches.

  13. I was looking at an HP system that was on sale and these cards came up as a customization option under Graphics:

    §[<<]§ At the time I had no idea what the G210 or GT 120 were supposed to be, and I ALMOST fell for the '1GB of onboard memory for the AWESOME!' ploy. That is, until I figured that if its ONLY $60 more than friggin' Intel graphics than it can't be that much better.

  14. The fact that they aren’t may suggest further evidence for problems with TSMC’s 40nm process (and/or nVidia’s design).

  15. definitively this 2 dont aim to the radeon hd 4770…
    maybe at the 4670 and 4550

  16. The 220 definetly could play fallout 3 very well also sacred 2 and quite a few other titles. Heck even the 210 could play fallout 3.

  17. ACTUAL game tests. None of that synthetic testing s***! When nVidia makes drivers to give their card an “automatic” 10,000 point boost in pcmark….mmm, that is not a test. Play something. See how the card handles the game, there is your benchmark.

  18. Benchmarks. And, within families of graphics cards, shader count as a rough guide.

    These are the rubbish end of the market.

  19. It looks like they have finally had some success with smaller chips in a smaller process, hopefully they can translate this to more complex designs i.e. the upcoming GT300 series.

    Plus, these have to be better than integrated Intel solutions, right?

  20. Something like: ‘With 1GB (or 512MB) discrete GeForce graphics for Xtreme gaming!!!!234’.
    These cards are proof that there’s no shame in their game when it comes to slapping some useless RAM on a card and then marketing it as competent for gaming. Pathetic.

    Edit: I see this point has been made already. Ah well, call this nodding in agreement.

  21. With 1GB of video RAM as well! A waste on a card that’s slower than even a 9600GT. *Depending upon the architecture these use. Nonetheless the 1GB of video RAM is funny and obviously just a marketing ploy.

  22. Why would OEMs want to SLI these cards, when buying something twice as powerful would be only a few dollars more in that market segment?

  23. It really implies that this refreshed GT200 architecture uses 2 SMs per TPC for 16 stream processors. This probably makes sense for a mainstream architecture.

    I’m thinking we may see a true mid-range GT200 based graphics card to finally replace the often renamed 9800GTX+/GTS250. Perhaps 10 TPCs for 160 shaders with a 256-bit memory interface and GDDR5. With high enough clock speeds from the 40nm process, it could match the GTX260, be cheaper to produce, and be a good complement for upcoming DX11 GPUs which will no doubt initially launch as high-end only just like the 8800GTX and GTX260/GTX280 did.

  24. There’s something very important missing to these new additions, namely the SLI connector…

  25. There’s something very important missing, namely the lack of an SLI connector…

  26. Yes, I know, I’m not saying AMD is any better.

    Claimed design wins aren’t equal to final products either, on both sides.

  27. What still puzzles me is G210. Based off GT200 and with 16 processors doesn’t seem possible. 24 SPs would be more likely.

    GT 220 seems to be what ION 2 will be based upon.

  28. Er…AMD “introduced” their HD 4690 mobile parts sooner (about 3+ months ago) and no products with it exist either.

    Mobile parts are about design wins and according to reports, NVIDIA has 100+ of those for the new mobile chips. That means 3-4 months, before real products appear. I don’t remember any reports of design wins for AMD…

  29. Well from their point of view, it’s almost bad to tell someone the truth. Then they will think that the 50$ graphics card is a PoS and go over to walmart where someone else will give them the same graphics card and tell them that /[

  30. At least they’re better than what usually ship with OEM’s.

    The problem the arises with these is the Best Buy sales guy will say “And this computer has a discrete nvidia graphics, perfect for gaming”. Then people will go home and wonder why games look like slide shows.

  31. I would ignore when Nvidia says they’re launched and wait for the first products to appear. Nvidia’s 40nm mobile GPUs turned up in real products yet? No.

  32. They look incredibly weak. Here’s to hoping that big manufacturers don’t market them as ‘awesome discrete graphics’ (hint: they will anyway)