Blog Post

Graphics Problems Surface With 17″ MacBook Pro

Stay on Top of Enterprise Technology Trends

Get updates impacting your industry from our GigaOm Research Community
Join the Community!


At this point, it seems like having some graphics issues is more of a “coming of age” thing for a new Mac product than something that surprises or worries us, though I’m sure if you have one of the affected MacBook Pros, you think otherwise.

Users are reporting in an Apple Support thread that the recently shipped 17-inch model is acting up, displaying green lines and ugly screen artifacts. The problem is apparently tied pretty clearly to the NVIDIA GeForce 9600M (the dedicated card), since the problem doesn’t arise while running the integrated 9400M, only appearing when you switch to the more powerful card.

No word from Apple (s aapl) yet about the issue, although they will likely try to resolve things with a firmware fix before trying anything more serious. One customer in the support thread said he received word from Apple that he would have to have his laptop replaced, although I doubt they realized at the time that the issues were so widespread.

It’s hard to believe that a machine so expensive, after such a long delay, could have such serious and quickly apparent problems. I suppose you could excuse Apple and just chalk this up as another NVIDIA screw up, but Apple must do some serious testing with these machines before they release them to the buying public. The alternative, of course, is that they were well aware of the problem but opted to stick to internal deadlines and deal with the problem down the road with a firmware update rather than delay the ship date of the 17-inch MacBook Pro any longer. Whatever the issue, Apple seriously needs to stop shipping significantly flawed products, or their reputation for quality isn’t going to last.

13 Responses to “Graphics Problems Surface With 17″ MacBook Pro”

  1. The problem can be an reliability issue which can not be screened out by test in production of the GPU. The root cause can be a design issue, that is the GPU is not designed for the tolerance of performance shift after long term of operation , not enough margin is reserved. Generally , this type of failure will appear after the GPU has been operated for more than one year. It is a kind of soft failure , and if you increase the Vcc of the power supply, the system will be back to normal, but the battery needs to be recharged more frequently, and the chip will be operated under a higher temperature internally due to high power consumption as a result of the increase of Vcc , or you can reduce the clock speed by sacrificing the performance. So, technically speaking, there is no cure for this problem unless you relax the spec significantly. Even you replace with a new GPU on the mother board, the new chip will end up failure as well after one year of operation , but you can extend the waranty from one year to two years with two GPUs, and after two years, the chip may die anyway.So, the cure is to redesign the chip so that it can tolerate the internal shift for long term operation, say , more than 4 to 5 years.And generally, if you see such problem, majority of the dices on the same wafers will fail with such problem, and most wafers will end up with this same failure , it does not matter which wafer lots or which fab the dices were coming from, because this is a design issue, not enouth margin was reserved for internal shift , such resistance of metal lines , to tolerate the high internal resistance as a result of long term operation of the GPU.
    The cause :as we all know that during normal operation of the GPU, the metal line was stressed continuously by current flow (DC), electrons flow from the surface of metal grains to grains , seas of electrons, as the GPU continue operation day and night , eventually small metal grains will combine with each other and forms large metal grains, as electrons flow on the surface of metal grains , the metal resistance will increased as small grains combine together to form larger grains, because the density of electrons on the surface of larger grains is smaller than that of small grains . As internal metal line resistance increased, the GPU was slow down significantly due to high internal resistance, to maintain the same clock speed, you need to increase the Vcc to increase the driving of internal of the GPU, and this will result in high consumption of power and internal temperature which will also increase the internal resistance of silicons due to mobility is reduced due to scattering effect .At this point, if you keep the same Vcc and clock speed , the chip will eventually lose its driving capability and system will show all kinds of strange performance , such as lines, dots..etc . the electro migration effect . What is worse is if the chip was over heated during operation, it will increase the life acceleration factor of the chip which will shorten the operational life of the GPU, and make the operational life of the chip even shorter.
    The fix : 1) reduce the internal metal resistance to reserve more margin by increase the physical width of the critical metal lines, this will reduce the current density during operation, and reduce the stress during normal operation to extend the operational life of the GPU, this needs a re layout of the whole chip.
    2) Increase the metal thickness of the metal line to reduce the resistance of the critical metal signal lines.
    3) optimize the power distribution of the chip, and increase the power supply bonding pads.

    This can also be caused by the metals in the package but less likely, because internal the chip , we are talking nano meter of metal width, and in the package , the metal with is much larger and the resistance is much smaller.

    To verify if this is caused by the dice of GPU or the package or the motherboard:
    1) remove the dice from a failed GPU and rebond it with another package. If the system still fail, then it is the dice of GPU caused the problem, if not , then it is the package caused the problem
    2) replace the failed unit with another mother board. If the system still fail, then the problem is caused by the GPU unit, otherwise, it is caused by the mother board

  2. Feldwebel Wolfenstool

    Good thing you have nVidia to blame. Don’t they test any of their machines before release? Or do they want to sell absolutely every unit to max the $-flow?

  3. Jennypen

    It’s not related to the 9600, or if it is, I’m just unlucky. I’ve NEVER used the 9600 on mine, only ever the 9400, and I’ve had the problem since yesterday. Switching to the 9600 does nothing. It’s something that independent.

  4. mbonifa

    It looks like apple might be trying to solve the issue already. A couple of times the shipping date for the MBP 17 was moved back (after the 3-4 weeks was replaced by a 7-10 days)… but maybe it is just the naive in me :)

  5. miguelv

    I think, and lets hope i’m right, that apple will acknoledge this problem and fix it with a firmware update. anyways, GOOD LORD that poor macbook looks like its suffering from a terrible illness.

    good luck 17″ pro owners… :(

  6. I think all of the new unibody Macbooks are having graphics issues. Sometimes mine has problems, other times it doesn’t. I depend on my machine though so going without it for a few days while Apple replaces the motherboard is pretty much unacceptable.