If you’re trying to find information on the NVIDIA 9400M Apple is adding to its lower-end MacBooks, here’s a hint: it doesn’t exist. (Not literally, anyway.)

Plain-English summary: NVIDIA’s new graphics card architecture used on the MacBooks combines a chip on the CPU (as found on the old MacBooks) with a dedicated GPU (as found on the MacBook Pros). The dedicated, “discrete” chip is designed for lightweight, power-efficient use, but it should be a boost from the old model. “9400M” applies to both those chips, which have … uh, different numbers. The problem is, no one knows exactly how the combination will perform until they test it; we expect to have more from live visual developers soon.

Updated: Actually, it seems “9400M” on the Mac refers to a new, single-die chip that NVIDIA considers equivalent to their two-chip (discrete + dedicated) combination with the same brand name 9400M. An anonymous reader points us here:
GeForce 9400M G

That last “G” apparently refers to this being a single-chip, motherboard graphics solution. Remove the “G”, and you have a different (though similar), two-chip motherboard + discrete solution. Both seem to have 16 stream processors total, even though the 9400M not on the Mac splits it across two other chips. (Confused yet?) As if it weren’t hard enough to follow whether that “G” is there or not, Apple calls the 9400M G the 9400M. So that means a lot of what I say here refers to the 9400M as on PCs, not the 9400M as on Macs. Ugh.

The information on the site above was added since I first posted this story. Update coming soon, though I’m going to sort my facts first, as this is not exactly Great Moments in Lucid Branding.

Exact details: Yes, as a bizarre twist of branding, the “9400M” is just a brand applied to the combination of two other GPUs. Put a GeForce 9100M G on your motherboard, and add either (bizarrely) a 9300M GS or or 9200M GS as the discrete GPU, and you get a 9400M. What’s the difference between the 9300M and the 9200M (aside from, uh, “100”)? Only the 9300M GS supports hybrid SLI (a technology for combining the two GPUs for better power/performance balance), and Blu-Ray playback.

And that brings us more questions than answers — some questions that are answered for Windows users, but even some questions for them, too (like real-world performance):

  • Will Apple use the 9200M GS or 9300M GS discrete graphics card?
  • Will the Mac support dynamically switching between integrated and discrete graphics to conserve power (as Windows Vista does), or will it do so manually? Updated: Glad I asked. Turns out you’ll have to do it manually; see Engadget video below. That means either Apple isn’t using the 9300M GS that is capable of switching, or they haven’t finished drivers yet, or both. Note that whatever the reason, this could get fixed later on.
  • Will we be able to manually turn on the GPU for live performance? If not, will dynamic switching have an adverse effect on live visual software?
  • What’s real-world performance like?
  • What does this architecture support on the OpenGL side? (We only know it supports DirectX 10.0, but that’s irrelevant on the Mac. OpenGL 2.1 is likely, given parallel models’ support, but that’s unconfirmed.)
  • How well will the Mac drivers work?
  • How well will they support Core Image?
  • With fire-sale prices on older MacBook Pros, with FireWire 400 onboard and a superior GPU (relative to the MacBook), why are you wasting your time reading this story?

Got all that? On the upside, the 9400M series does represent a theoretical gain over the previous 8400M (though they do have very different architectures), with a higher clock speed and more minty freshness and such. And HD video decoding ought to go pretty nicely. NVIDIA also touts support for its processing-on-the-GPU standard, CUDA, though I wonder what that means for Apple’s (and others’) upcoming support for OpenCL. And there’s PhysX support, which probably doesn’t mean anything to almost anyone.

There’s really a deeper question here, though, one that goes way, way beyond a new Apple laptop announcement. That question is, just how will these hybrid machines perform? AMD/ATI is working on it, NVIDIA is working on it, Intel is working on its own GPU solutions. We’ve got a lot of questions ahead. Stay tuned.

9300M GS at notebookcheck.net

notebookcheck.net benchmark list (note that the 9300M GS lies somewhere just shy of ATI X1400 land, though that’s not a real-world indicator, really, let alone one that tells us about all the potential compatibility pitfalls in live VJ software)

NVIDIA GeForce 9100M G mGPU motherboard GPU architecture, 9300M GS and 9200M GS discrete GPUs

Updated: Here’s what you have to do to switch to low-power, integrated-only graphics. Not too big a pain here, but what if you’re on a plane trying to save battery life and had a bunch of work open, etc.? It’s a pain. They’ll fix this eventually, I’d guess. (Early shipments of AMD/ATI’s competing tech had the same deficiency, before they … fixed it.)

  • Johnny DeKam

    Interesting that only one of the chips supports SLI – I highly doubt we'll get SLI out of this model. Apple will likely first have to complete SLI support at the Core Graphics Level — if they ever do this it won't at least until Snow Leopard 10.6

    Apple posted a doc today that states one can switch between the two using the energy saver control panel. So it seems out of the box support is to use one or the other chipset.

    BUT in theory it may be possible to use the integrated chip for the LED display, and the dedicated GPU for the external DVI. This would be akin to using a 16x PCIe for video output and a lesser speed card for your software interface. A marginal, but useful performance increase – especially if you're running a Matrox2go solution.

    Johnny DeKam
    Founder VIDVOX

  • Hey Johnny, great to hear from you!

    Readers, meet Johnny, Vidvox founder and one of our favorite VJs on the planet!

    What you're seeing in the video above is just that, switching between the chips. As I said, there's an ATI design that takes the same approach. The problem is, I don't think these are engineered to work simultaneously, at least not on the NVIDIA. They may not be designed to even power both at once. I think ATI actually does allow the two to work together internally, and in that case you don't even have to dedicate one to one screen or the other.

    It's all speculation, though, because even on the PC side I've seen pretty limited real-world benchmarks.

    My complaint is that, using the 9300M GS, you can switch dynamically — no logout/login required — on Windows Vista. Now, maybe this machine actually wound up using the nearly-identical 9200M GS that doesn't support that feature, maybe there's something different about Mac OS that doesn't let you do that, maybe Apple will do this in a driver update later. No idea.

    Of course, the bottom line is this sort of technology is better than Intel's integrated graphics, but not yet ready to compete with dedicated GPUs. It's sort of not that exciting yet. But I do imagine it's almost certain VDMX will be happier on the new MacBook than the old one. We'll know when it ships.

  • Just to be clear, you're stating the older Mac Book Pros have better GPUs than the current, that is false. the new MBPs have 2 GPUS, one integrated (although you *could* call the 9400 discrete), the other the 9600M GT. The older 8400Ms are not a faster GPU than the 9600 GT, and I see no mention of them on your page :

    if you want to know about Hybrid SLI, you should read this:


  • Vade, sorry, I was unclear. The new MacBooks have *only* the 9400M, with its hybrid integrated/discrete approach. It's much better than the old MacBooks, but not better than the old MacBook Pro.

    The old MacBook Pros had the 8600M GT, which is most definitely faster than a 9400M. I don't recall Apple ever shipping an 8400M.

    So, anyway, my point: you've got $1300-1500 burning a hole in your pocket. You can choose between:

    * an old MacBook — you get FireWire and slightly faster CPUs, but an inferior screen, and lousy integrated Intel graphics, so I wouldn't bother
    * a shiny, new MacBook — no FireWire, and the good but middling 9400M
    * a fire-sale discount MacBook Pro: FireWire (800 + 400, albeit on the same bus), the beloved old video dongles, and the not-too-shabby 8600M GT

    I'd choose the MacBook Pro. If you really don't want to jump ship for PC because you can't afford the new Pro, there's really no reason to.

    Hybrid SLI — awesome link. But yeah, still leaves questions about support on Mac OS and these machines.

  • Johnny DeKam

    I don't agree that the discrete GPU is slower than old MBP Pros. According to Barefeats, on spec its 43% faster than the 8600M


  • I'm going to have to start referring to the non-Pro MacBook as the "MacBook Amateur" so this is clear.

    What I'm saying is — and I don't know how to say this any more explicitly — *IF* you don't have $2000 and are considering the MacBook Amateur ๐Ÿ˜‰ the 9FOUR00M (not 9SIX00M GS) is slower than the 8600.

    So, in descending order of speed:

    9400M (which is really 9100+9200 or 9100+9300)
    Intel integrated graphics

    Now, we don't have very solid numbers on the 9400M product yet, but it's pretty safe to say this relative comparison should be accurate. There aren't a whole lot of processors even on the 9400's discrete chip. And while the Pro lets you switch between 9600 and 9400 (which actually means it has no fewer than *three* GPUs, two discrete and one integrated), the MacBook Amateur does not.

    So I'll stick by what I'm saying. If you don't want the added cost of the MacBook Pro, you get a GPU on the *new* MacBook non-Pro that's still less powerful than the GPU on the *old* MacBook Pro.

  • Anon

    Sorry Peter, you're absolutely wrong on this. The MCP79/Geforce 9400M mGPU chipset is a brand new single chip design, there's no other chips involved.


    Take a good look at the teardown pics. The MCP79 powers the regular Macbook, the Air and the Pro. It's a single chip.

    And information about it does exist, you're just not looking in the right place.


  • Anon



    Have a good look at the MCP79 vs an Intel 2-chip chipset. It's not a rebranding or whatever nonsense you're suggesting, it has 16SP(stream processors) GPU and the entire chipset core logic in 1 chip. The 9100 only has 8SP GPU in it, so again, you're spreading wrong info.

    Apple wants to save space and a highly integrated design like the MCP79 is a huge advantage over the 2 chip designs of Intel chipsets.

  • @Anon: That information was added since yesterday's announcement.

    But "absolutely wrong"? Jeez, who knows. Reviewing NVIDIA's *own information*:

    If you look at the 9100M G mGPU series:

    The combination of the 9100M G motherboard and 9300M GS discrete chips results in the "GeForce 9400M" brand.

    The "GeForce 9400M G" to which you refer "goes by the name GeForce 9400M" on Apple products, according to NVIDIA's site.

    If you're right, that means NVIDIA is using the same brand name (9400M) to refer to completely different products, which is really, truly a nightmare.

    So, in fact, that would mean that you're only logging out / logging in to switch on the MacBook Pro, not the MacBook Pro, because THIS 9400M is one chip instead of the other, two-chip 9400M, and Hybrid SLI is on the 9400M + 9600M GS MacBook Pro ONLY.

    I may be wrong, but that means I failed to recognize the difference betwen NVIDIA's GeForce 9400M, which includes the 9100M GS and 9200M GS or 9300M GS and NVIDIA's GeForce 9400M, which includes the 9400M GS.

    I've seen some terrible branding. That would be the worst branding I've ever seen. But it does seem to be the case.

  • Okay, fine, Anon. It's my fault. I was referring to:

    "NVIDIA GeForce 9400M"

    whereas *NVIDIA* was referring to:

    "NVIDIA GeForce 9400M"

    It's complete nonsense that I would suggest that branding is confusing.

    Give me a break. (for what it's worth, though, it does seem you're right — but this is ridiclous)

  • So, just to try to make sense out of this very odd choice by NVIDIA …

    Presumably by going to the two chips on the 9100, they get a total of 8 + 8 stream processors (8 on the 9100, 8 more on the discrete 9200/9300), which would then add up to the 16 the 9400M (which is really the single-die 9400M G)?

    But on the 9100, it's on two dies? In other words, NVIDIA is just assuming the "9400" brand refers to some combination of one or two chips with the same basic architecture for 16 stream processors?

    Want to update the story, but this is puzzling, and it's virtually impossible to get press info out of Apple hardware or NVIDIA.

  • Anon

    It's very simple really. 9100M is what you would call Geforce 8200/8300/MCP78 on the AMD side. It has the same 8SP GPU architechture and it's basically old news, the AMD version of this MCP78 chipset released back in March 2008 or something.

    MCP79/Geforce 9300/9400 is a brand new, single chip 16SP GPU plus chipset core logic design by Nvidia for Intel CPUs. It's been delayed many times, I'm sure Nvidia wanted to release this before Intel's G45 integrated graphics chipset was released but there was some problems with it I guess. Believe me, there's no confusion, this is a brand new chipset with 16SP.

    So far, the MCP79 has been extremely impressive, even able to outperform the AMD/ATI 790GX(higher clocked than the 780G) with 128MB dedicated sideport memory while the MCP79 GPU itself has NO dedicated memory, only using the much slower system memory most of the time.


  • Okay, then that side makes sense. Thanks for that.

    What still confuses me is, why was the 9100 also called the 9400M, when combined with the 9200/9300 discrete chips? Why not just call it a 9100? ๐Ÿ™‚

    Or is that to get parity between AMD and Intel solutions, ie —
    9100 + 9200/9300 = 9400M on AMD, MCP78
    9400M G = Intel, MCP79 = 9400M on Mac?

    We may need to loop back and do an overview for dummies on this. I know some system builders who have been working on this, so they've likely better keeping better tabs on this than I have.

  • Anon

    As far as I know, no OEMs would release that mess that is the 9100+9200/9300 when the MCP79/Geforce 9300/9400 single chipset has same or better performance and fabbed on smaller 65nm process, MCP78 was fabbed on 80nm process. There shouldn't be any confusion as far as I know.

    And the product pages always refers to the chipset as 9100M, so again, I can't seem to see any confusion. I'm not even gonna touch the 9100+9200/9300 mess, simply because MCP79 has made that obsolete as least on the Intel side, end of story.

  • Okay, thanks, Anon. I'll retract my earlier statement ("this branding is insane") and replace it with "the branding of the 9100+9200/9300 *as a 9400* is insane." ๐Ÿ˜‰

    Will post an updated story later tonight.

  • <a href="http://gizmodo.com/5067433/confirmed-apple-can-enable-dual-gpu-and-on+the+fly-switching-in-macbook-pro&quot; title="Apple Can Enable Dual GPU and On-the-Fly Switching in MacBook Pro" rel="nofollow">Confirmed: Apple Can Enable Dual GPU and On-the-Fly Switching in MacBook Pro

  • Stefinnee

    Can someone tell me which would be better:

    A 9400M using shared system memory or;
    A 9300GS with 512 MB of dedicated memory?