Izzard

I'm just going to mention, once again, that my GTX 970 + AKiTiO + 2011 Mac Mini + Windows works absolutely beautifully *BUT* only works at all if I switch the eGPU on at exactly the right moment. I have to power on the Mac, then the eGPU about 1 second later. If it's too early or too late then I get a black screen instead of Windows booting. Waiting for the Mac startup chime is far too late in my case.

Here's my visual guide to opening the ATiKiO case.

I feel bad for saying this because I can see you have put some effort in, but actually I think your guide could confuse people. It doesn't go into enough detail (e.g. step 1 says to drag PackageMaker into your Applications folder; that requires more explanation). Then "copy one line of kext file click desktop to reviel finder in top left corner" doesn't really make much sense at all. By all means have a go at putting together a guide for newbies, but as a downloadable PDF it's very hard for you to revise it to correct mistakes without leaving erroneous versions scattered around. How about you just start a thread instead? Please don't take my criticism too harshly.

I realise this thread is all about AMD cards but I just want to say that this sounds similar the same baffling issue I have with my NVIDIA setup. (Mac mini 2011, GTX 970 in an ATiKiO). In other words, if I don't modify IONDRVSUPPORT kext then it boots but shows only as "NVIDIA chip model" in system information. Whereas if I correctly modify IONDRVSUPPORT then Yosemite kernel pancs at the point it tries to start the login screen / desktop. (And it works great in Windows.) None of us has come up with the answer to this but the only key difference with my system to other Mac minis is that it has a Radeon dGPU.

Guide to opening up the AKiTiO 2 Other reference: http://forum.techinferno.com/enclosures-adapters/7205-us%24189-akitio-thunder2-pcie-box-16gbps-tb2-113.html#post128029 Guide Powering the AKiTiO via the barrel socket with your ATX PSU Other reference: http://forum.techinferno.com/enclosures-adapters/8317-%5Bguide%5D-making-molex-barrel-adapter.html#post113069 I hope these are a useful reference. Happy eGPU-ing!

@notrodash I wrote some comments about that issue here.

Not as far as I can find. Only different cards, different minis and different OSes. But never mind - I don't mean to threadjack.

I really meant a Wikipedia-style wiki that anyone can edit. But maybe that would be a disaster with some of the... um.. characters around here

@MVC I know it's still a mystery why my GTX970 eGPU kernel panics Yosemite booting on my Radeon dGPU 2011 mini (despite working great in Windows) - but do you think it's worth trying to flash it as a solution?

I've just read the last few days' worth of posts and I'll repeat what I've said before: a wiki would help at this point. There is information being repeated (wasting time) and misinformation being repeated (wasting money). Plus some useful hints and links which only exist in this Mac-oriented thread which our PC-only friends are likely to miss. Let's get the best advice, best photos and best links in one place?

I noticed the same discrepancy. MaximumBurrito's, symptoms, as described, don't make sense to me unless there is some sort of hardware issue. MaximumBurrito, perhaps you should double-check the two boards of your AKiTiO are connected properly. Is the short black/yellow power cable dislodged? (Although I wouldn't expect the GPU fans to turn if the GPU can't see the PCIe bus.) If I were you, I would think about getting a simple PCIe card (ethernet adaptor / USB ports or similar) to see if those work. They are cheap and if you value your time/sanity then it's a worthwhile troubleshooting step.

When you did that, did you then get the complaints in the log/ colsole about IOPCITunnelCompatible being required? Have you ever seen them? Because if not, nothing you’ve said suggests the ATiKiO thunderbolt board can talk to the PCIe board. I think your symptoms are what you’d get if it were supplying power to the GPU and nothing else.

Yeah… right back in my first post about this problem (different install of Yosemite and the previous NVIDIA driver), that same console output was right there at the end. MVC, do you still think it's wrong that it says this in the console? Does yours say something different for "loaded and registered"?

- - - Updated - - - There's no kext called NVDAGM100HAL in my Extensions folder: And looking at the plist file for NVDAGM100HalWeb.kext it has an IOClass value of "NVDAGM100HAL": …could it be that the correct kext loaded but it reports as NVDAGM100HAL? In which case, why doesn't your console do the same? This is the new driver that came out today.

Hmm, well I'll explore that angle next but this is a (relatively) fresh install of everything. A few days into this saga I wiped the SSD and rebuilt from scratch with a totally clean install of Yosemite (straight to 10.10.3) and installed nothing but the latest NVIDIA driver - and I still had the symptoms (i.e. kernel panic instead of login screen). And I've posted this part of the console output previously too because I was asking about the "[AGPM Controller] unknownPlatform" - but I'll have to read back through my posts now to see if it was the same. (It probably wasn't, considering your surprise at seeing NVDAGM100HAL).

Yes, everything looks fine and console reports: 12/05/2015 22:32:21.000 kernel[0]: NVDAStartup: Web 12/05/2015 22:32:21.000 kernel[0]: NVDAGM100HAL loaded and registered (although that is always immediately followed by "12/05/2015 22:32:22.000 kernel[0]: [AGPM Controller] unknownPlatform" and I don't know if that is something to be concerned about or not). Anyway… new NVIDIA driver for Yosemite just came out today (346.01.02f03) so I'll try that. But I won't get my hopes up.

Well, no joy. Removed CUDA framework, modified the kexts again from scratch. Still shows as NVIDIA Chip Model if I hot-plug the eGPU and still kernel panics instead of showing the login screen if I boot with the eGPU connected. Funnily enough, Windows really *is* plug and play as long as the eGPU is powered at the right instant.

Yeah, actually I knew that but I also knew that wasn't causing the problem (and I have since tried with MVC's kexts - made no difference). Hmm, now this is an interesting thing to investigate since more than a few of the guides I read said to install the CUDA driver. They made it sound essential. And - unlike your Mac mini - mine *does* have an AMD GPU in it. That page says it should be fine installing CUDA if you have both an AMD *and* an NVIDIA GPU; they're advising not installing CUDA if you *only* have an AMD GPU. ...however, since my AMD GPU (soldered to the logic board, unfortunately) is also what was impeding Physx from working in Windows, perhaps it is interfering here too. Worth a shot! I will remove the CUDA driver and report back.

Thanks, goalque. I sent the AMD card back - sticking with the 970.

I've come to the same conclusion. And apart from this issue of the eGPU not quite working in OS X (I had similar symptoms with the AMD R9 280X showing as "7XXX" and not outputting/detecting a display) I don't have any other issues with this machine... plus it's working great in Windows. Since I haven't seen anyone here with my exact combination of computer / enclosure / GPU, my working theory is that it's my particular combination that doesn't work, rather than my specific mini. Sooo... anyone out there got a 2011 Mac mini, an AKiTiO and a GTX 970? goalque had that combination here... ..but only talked about Windows.

Thanks again! (Also a big thanks - again - to Tech Inferno Fan for helping me get PhysX running on the GTX 970 instead of the CPU, which happens because the NVIDIA driver always saw my Radeon dGPU and disabled Physx. The solution was to disable the PCIe Root Port hosting the Radeon in Device Manager. I will include that in my implementation guide when I get around to it.) Regarding the "nvram -p" command: yes, it's in amongst all the other accumulated data: [FONT=Menlo]$ nvram -p [/FONT][FONT=Menlo]SystemAudioVolume )[/FONT] [FONT=Menlo]boot-args kext-dev-mode=1 nvda_drv=1[/FONT] [FONT=Menlo]fmm-mobileme-token-FMM bplist00%d[etc…][/FONT]

I totally agree. Despite the best intentions of the moderators, these forums are a labyrinth with a plethora of wrong turns leading to miserable dead-ends. As I've said before, I think what's needed is a Getting Started wiki that we can all contribute to, helping newbies dig their way out of the common mistakes and back onto the path of glorious 3-figure fps gaming. The first time my eGPU finally showed up (and immediately accepted the driver and started working) in Windows - after many evenings of beating myself up for wasting money on unusable kit - I was over the moon and got that adrenaline rush you get when something finally comes together and works

As MVC has suggested, you could just be falling at the last hurdle here (at least with regards to Windows; you definitely need to sort your driver issue out for OS X). I get these *exact same symptoms* with Window 8.1 / 10 Preview on my 2011 Mac mini unless I get the timing right powered up. You could be so close and not even know it. Experiment. For me, the right timing for powering up the eGPU's PSU is between 0.5 seconds and 1.5 seconds after pressing the power button on the computer. A fraction too early = Windows boot hangs. A fraction too late = Windows boot hangs. You need to get all Goldilocks on this.

OK, I've now replaced the Asus STRIX GTX 970 with an MSI GTX 970 GAMING 4G. Again: works really great in Windows, but still kernel-panicks Yosemite and still shows up as "NVIDIA Chip Model" if I hot-plug. I give up!

Hi Tech Inferno Fan Thanks for sending Setup 1.3. This was so I could use iPort to disable the Radeon in my Mac mini, which is blocking Physx from running on my Nvidia eGPU: Hardware: Mac mini (Mid 2011) Processor Name:Intel Core i7 Processor Speed:2.7 GHz Number of Processors:1 Total Number of Cores:2 L2 Cache (per Core):256 KB L3 Cache:4 MB Memory:16 GB Boot ROM Version:MM51.0077.B10 SMC Version (system):1.75f0 [*]AKiTiO Thunder2 PCIe box [*]Corsair Builder Series CX 430 Watt ATX/EPS [*]iGPU: Intel HD 3000 [*]dGPU: AMD Radeon 6630M [*]eGPU: Asus Nvidia GeForce GTX 970 Strix Graphics Card (4GB, GDDR5, PCI Express 3.0) Unfortunately I've run into a problem. Following your instructions I installed Setup 1.3 just fine and confirmed that it could chainload Windows 10 (after changing from MBR2 to MBR). If I power on my eGPU exactly one second after the mini (just as with Windows) then I get past the grey screen and I can boot into Setup 1.3. (Just as with Windows, if I don't time this right, it goes straight to a black screen instead.) Everything seems to be detected just fine when I do this: However, the command to disable the dGPU always hangs, whether I select it from the menu or trigger it correctly from the batch file. Basically, "setpci" seems to cause this hang, as far as I can tell. Note that it doesn't go blank or stop updating the screen. Tellingly, the cursor is flashing on the next line. Other than that, it has stalled. Any ideas what I can try now? Thanks!