US$189 AKiTiO Thunder2 PCIe Box (16Gbps-TB2)

[Oscar J]

Morv: have you been doing any further testing in OS X? Are the stability issues there too?

[Morv]

I did not use the R9 280 any further as I needed to sell it again as soon as possible due to it being useless to me.

@ithildin: Will you provide the pictures? You may have found a general solution to the stability problems which is important.

[Sp0t]

Sure AMD works as well, but I think there will be no "internal display only":

http://forum.techinferno.com/implementation-guides/8585-%5Bguide%5D-2014-15-macbook-pro-2x-r9_280x%4016gbps-tb2-netstor-na211tb-win8-1-osx.html

AKiTiO is unstable with the AMD cards on Win8.1:

http://forum.techinferno.com/diy-e-gpu-projects/8844-targeting-akitio-thunder2-pcie-boxs-stability-issues.html

External monitor is not a requirement, internal screen full screen apps will be accelerated by eGPU, if you install Virtu MVP 2.0:

http://forum.techinferno.com/diy-e-gpu-projects/2967-lucidlogix-virtu-internal-lcd-mode-amd-egpus.html#post120413

Thank you for letting me know

As you've probably heard, AMD has lowered 290/x prices due to the recently revealed 970 issue, so I was thinking about getting a red one.

So if I buy an AMD card and use it with my rmbp 750m, then I will still be able to run applications on my internal screen (at a certain penalty, of course), did I understand you correctly?

Do you think that the issue with Akitio being unstable with AMD cards will be fixed any time soon?

I am in the process of earning money now, so I have around 2 month to weigh everything.

[Morv]

@goalque: Please don't use one example of an AMD card not working as proof for every AMD card being unstable on Windows 8.1. I only had one AMD card and up to now 3 Nvidia cards. 2 Nvidia cards were/are stable(GTX 480, GTX 960) and one wasn't(GTX 780). One simply has to try which cards work and which don't. Do not generalize at this point. I'm still waiting for ithildin's answer regarding the mod he mentioned in my thread.

@Sp0t: You simply have to try which work. There is no proof that no AMD cards work.

[Sp0t]

@goalque: Please don't use one example of an AMD card not working as proof for every AMD card being unstable on Windows 8.1. I only had one AMD card and up to now 3 Nvidia cards. 2 Nvidia cards were/are stable(GTX 480, GTX 960) and one wasn't(GTX 780). One simply has to try which cards work and which don't. Do not generalize at this point. I'm still waiting for ithildin's answer regarding the mod he mentioned in my thread.

@Sp0t: You simply have to try which work. There is no proof that no AMD cards work.

Good, thank you.

[euqlaog]

@goalque: Please don't use one example of an AMD card not working as proof for every AMD card being unstable on Windows 8.1. I only had one AMD card and up to now 3 Nvidia cards. 2 Nvidia cards were/are stable(GTX 480, GTX 960) and one wasn't(GTX 780). One simply has to try which cards work and which don't. Do not generalize at this point. I'm still waiting for ithildin's answer regarding the mod he mentioned in my thread.

Sorry, I didn’t want to generalize at all. I have three AMDs: R9 270X, R9 280X, HD 7970 and AKiTiO is unstable with those AMD cards for sure, the same AMD cards work perfectly with the same MBP, same Win8.1 drivers and NA211TB, so the problem is with the AKiTiO itself or powering it. From the beginning of the release of the AKiTiO, I have seen only problems with AMDs on the main thread. I will try soon report my progress with AKiTiO + AMDs. I did make some progress.

[juniordiscart]

@Morv

I assume you are referring to me in one of the spoilers. I'm the person who switched out his GTX770 for a GTX760. I must admit I was a bit fast in crying victory, as I mentioned in a few post further down in that thread. The GTX760 does reduce the problem significantly compared to the GTX770 (it happens a lot less often and also much later compared to my GTX770).

However, I do have a GTX460 (1GB, 2 x 6 PCI-e power pins) as a spare graphics card and have been using it for over a week now without a single crash. It's obviously not as powerful as a 700-series card, but it runs flawlessly. I may get my hands on a GTX560 of a friend for a while and can use it to test this card as well.

Btw, I checked my event logs when using the GTX760 and GTX770, and both of them give the entries you describe in the OP. Using the GTX460 only gives one MEIx64 entry in the log after the WinLogon event.

[Morv]

@Morv

I assume you are referring to me in one of the spoilers. I'm the person who switched out his GTX770 for a GTX760. I must admit I was a bit fast in crying victory, as I mentioned in a few post further down in that thread. The GTX760 does reduce the problem significantly compared to the GTX770 (it happens a lot less often and also much later compared to my GTX770).

However, I do have a GTX460 (1GB, 2 x 6 PCI-e power pins) as a spare graphics card and have been using it for over a week now without a single crash. It's obviously not as powerful as a 700-series card, but it runs flawlessly. I may get my hands on a GTX560 of a friend for a while and can use it to test this card as well.

Btw, I checked my event logs when using the GTX760 and GTX770, and both of them give the entries you describe in the OP. Using the GTX460 only gives one MEIx64 entry in the log after the WinLogon event.

[ATTACH=CONFIG]13892[/ATTACH]

Yes, it was you I was referring to but I didn't want to search for your post as I didn't know anymore in which thread I had seen your post and I didn't know your name either :S.

Still, thanks for your information. Did you have Optimus enabled with both the cards, e.g. access to the internal display? Like the others stated they only get it when Optimus is enabled and I can agree to this with my new GTX 960. If I boot and it boots with iGPU enabled I get these entries and a load on the System task of about 15-20%.

Also, I remember the output on the external display felt somewhat slow or laggy with the GTX 780. Running the system with the GTX 960 feels way smoother(cursor movement, window movements).

If you like to, check about this, too.

If you want a stable fast card, take a Gigabyte GTX 960 ITX. That's the one I've got and i did not have a crash until now, it's about a week nearly since I've got it. 3DMark11 ranges from ~8700 to ~9400(+200MHz on each core and memory, stable, without increasing voltage).

It's really sad that ithildin doesn't answer. There may be a solution to use bigger cards but I don't know that much about electrics, so I still don't know what exactly he meant. @goalque: Are you any smarter about this already? You wrote that you came to the same thought as he did but I guess you don't want to blindly harm your Akitio board just like me

[juniordiscart]

Ah, sorry, I hardly paid attention to disregard the information in the spoiler.

Maybe a dumb question, but how do I disable optimus then? I do have 2 GPUs listed: the 750m and (for this moment) the GTX460. Can I just disable the 750m from the device manager? (The MacBook doesn't recognize the Intel iGPU in Bootcamp, so I can't disable that). Or should I do something in the nvidia control panel?

Thanks for your reply and effort in investigating this.

[Morv]

Ah, sorry, I hardly paid attention to disregard the information in the spoiler.

Maybe a dumb question, but how do I disable optimus then? I do have 2 GPUs listed: the 750m and (for this moment) the GTX460. Can I just disable the 750m from the device manager? (The MacBook doesn't recognize the Intel iGPU in Bootcamp, so I can't disable that). Or should I do something in the nvidia control panel?

Thanks for your reply and effort in investigating this.

This is very interesting information. Optimus can only be enabled with the iGPU present. Your Macbook having the GT 750M but giving the same messages hints that it isn't Optimus or iGPU related.

As I previously wrote in the first post, something is going on at boot which leads to those messages. I'm starting my Macbook by pressing Alt until the boot menu appears, then I plug in the already powered eGPU, wait a second or two and then I boot Windows. Same procedure all along but sometimes it loads with the iGPU enabled and I do not know why, it also happens if I don't load OS X in the meantime. I then have the messages like you but I only have the iGPU in my Macbook which is why we thought it's Optimus related.

I don't know, maybe this is some weird allocation failure happening at boot? I'd still like to know if this the reason for the crashes or if it's really something electric related to the Akitio board.

[tamer1009]

No offence, but that seems to be a very old PSU?! Try to get another one, please.

I have tried it today with a new Corsair PSU. Everything works fine. But randomly when I start my macbook I get a blackscreen (and no startup sound).

So conclusion; sometimes it works and sometimes not. I also tried to start using alt and then plug in the TB cable, wait for a few minutes and select Windows, then the boot selection menu freezes.

Anything I missed so far?

[Dschijn]

Sorry no clue… some people have that problems (sometimes it works vs. sometimes it doesn't work).

But when you boot with it, it works and you can play games? Have you tried to improve stability by testing using the powered riser?

I never had that problems. My eGPU etup boots flawless everytime with full performance!

[ithildin]

It's really sad that ithildin doesn't answer. There may be a solution to use bigger cards but I don't know that much about electrics, so I still don't know what exactly he meant. @goalque: Are you any smarter about this already? You wrote that you came to the same thought as he did but I guess you don't want to blindly harm your Akitio board just like me

I'm not sure what you mean by me having found a solution. My Macbook Air + AKiTiO + EVGA GTX 670 2GB is definitely not stable (no Optimus either). Like for @juniordiscart , my GTX 460 is more stable but I've noticed that Windows 8.1 is occasionally unresponsive.

I've been busy at work and I haven't gotten around to testing the setup again to look for MEIx64 entries in the logs. I should have some time for this tomorrow.

EDIT: I think you got me and @ha1o2surfer mixed up:

I would just cut your 12volt cable going into the thunderbolt card and wire them up to the 12volt going into your GPU. That is what I had to do to make it stable.

[ithildin]

@goalque: Please don't use one example of an AMD card not working as proof for every AMD card being unstable on Windows 8.1. I only had one AMD card and up to now 3 Nvidia cards. 2 Nvidia cards were/are stable(GTX 480, GTX 960) and one wasn't(GTX 780). One simply has to try which cards work and which don't. Do not generalize at this point. I'm still waiting for ithildin's answer regarding the mod he mentioned in my thread.

By the way, it was @ha1o2surfer that modded his Akitio daughterboard mini power cable, not me. Like you, I would love to find out exactly what he did and how it stabilized his Akitio eGPU.

[Morv]

I'm not sure what you mean by me having found a solution. My Macbook Air + AKiTiO + EVGA GTX 670 2GB is definitely not stable (no Optimus either). Like for @juniordiscart , my GTX 460 is more stable but I've noticed that Windows 8.1 is occasionally unresponsive.

I've been busy at work and I haven't gotten around to testing the setup again to look for MEIx64 entries in the logs. I should have some time for this tomorrow.

EDIT: I think you got me and @ha1o2surfer mixed up:

Oh yes, I messed that up, I'm really sorry!

@ha1o2surfer: Can you provide pictures and/or some details on how to do the cable mod? As you can see in this thread there are people struggling with this stability issue. As you wrote that you could fix it with your mod it may be the general solution. People who don't know for sure what you meant will most probably not try to mod it and harm the board but a guide with pictures should be easy to follow. I really think people would be thankful to finally have a stable setup with their desired graphics card.

[Morv]

Yes, just clarifying this at this point: I got things messed up. ithildin struggles himself with stability issues and needs help, too. Sorry about this.

[tamer1009]

Sorry no clue… some people have that problems (sometimes it works vs. sometimes it doesn't work).

But when you boot with it, it works and you can play games? Have you tried to improve stability by testing using the powered riser?

I never had that problems. My eGPU etup boots flawless everytime with full performance!

I got it working now, I went from 100 to 1222 with Unigine benchmarks.

Still wondering about nvidia settings, what should I do with CUDA? I can select both the 750M and 970. Also Physx? Any idea?

Thanks!

Update: After soldering two cables on the jackplug rearside I connected it wrong on the 12V molex, I confused the + and -. Look what happened:

That's the end of the story lol...

[ithildin]

Update: After soldering two cables on the jackplug rearside I connected it wrong on the 12V molex, I confused the + and -. Look what happened:

[ATTACH=CONFIG]13897[/ATTACH]

That's the end of the story lol...

Ouch!

Looks like one of the condensers on the TB card is fried but there's probably more damage elsewhere. I hope your GPU is still OK.

[Morv]

That's the reason I wouldn't touch this thing in such a way without a detailed super dummie guide

Sorry for your loss, though =/

[ithildin]

I've been puzzling a bit on the mod that @ha1o2surfer mentioned. I can only surmise that he hooked up the TB controller card to his PSU in addition to the powered molex. Even though I haven't modded the small TB controller power cable, I've tried connecting my PSU to the DC jack + PCIe riser + GPU simultaneously and the GTX 670 still crashed as per usual. Looking carefully at the PCB traces surrounding the DC jack and the mini power cable connector, it doesn't look like there's much there besides some condensers to provide some current filtering:

Red arrows on the picture represent multimeter's red probes and black arrows are black probes. Red ellipses are +12V and blacks are ground except the C2 where multimeter shows -12V. Yellow rectangle is the 2nd golden finger of PCIe x16 riser. Any thoughts about this? Thanks!

[ATTACH=CONFIG]12307[/ATTACH]

It looks like the mini cable is just feeding 12V to the TB controller card, which means hooking up power to the cable or to the DC jack should give similar results.

Bottomline, I think my modded molex-to-DC barrel adapter should be equivalent to ha1o2surfer's modded TB card power cable. It didn't improve matters for my setup but maybe it will help people with Macbook Pros.

If you want to test it out and need your own molex-to-DC barrel adapter, @jacobsson has a nice mod guide here. Be extra careful with polarity when wiring it up! The center contact on the DC barrel plug must be connected to positive (yellow cable)!

[tamer1009]

Ouch!

Looks like one of the condensers on the TB card is fried but there's probably more damage elsewhere. I hope your GPU is still OK.

Yes, do you have any idea what type of chip/condenser this was? I can't read it anymore, I might try to solder a new component on it.

That's the reason I wouldn't touch this thing in such a way without a detailed super dummie guide

Sorry for your loss, though =/

Thanks!

[ithildin]

Yes, do you have any idea what type of chip/condenser this was? I can't read it anymore, I might try to solder a new component on it.

It looks like an 107C SMD Tantalum capacitor. You can get a replacement part from here. It's a polarized capacitor so make sure it's soldered in the same orientation as the original part (see pictures here). Like I mentioned before, it's likely that this is not the only damaged component on your AKiTiO boards. It's worth taking a closer look at both boards and checking for burned tracks or other damaged components. Some you can check with a multimeter, others (such as the ICs) won't be easy to check without more sophisticated equipment.

[euqlaog]

I re-evaluated the purpose of a small 4pin power cable of the AKiTiO PCIe board WITHOUT CUTTING WIRES. I just used a modified floppy drive connector.

OWC Mercury Helios has the same PCIe board (exactly the same part number) and maybe some of you have noticed that it’s equipped with three 4pin sockets (1 for TB card + 2 for PCIe board), but let’s focus on the AKiTiO which has two of them.

Please do not take this how to properly power the AKiTiO PCIe box. Modifying the cables by yourself is dangerous. I don’t take any responsibility if you fry your board. My tests are just a proof of concept. If you consider doing something similar, please use a multimeter and always check the voltage and polarity before plugging in any wires!

All the combinations of previous methods resulted in unstable AMD eGPU and a blank white screen during the physics part of the 3DMark11 test. I was interested in how much the J4 socket (2) consumed power, but as a side result my configuration passed 3DMark11, 3DMark13 and Valley benchmarks completely. I used two PSUs, and two risers (RSC-R2UT-E16R + the riser of Milo ML07). I can repeat this after each reboot, but second run of 3DMark11 physics part crashed. I was able achieve same results sometimes with one PSU as well.

AMD Radeon R9 280X video card benchmark result - Intel Core i7 4702HQ,Apple Inc. Mac-3CBD00234E554E41

AMD Radeon R9 280X video card benchmark result - Intel Core i7 4702HQ,Apple Inc. Mac-3CBD00234E554E41

AKiTiO has 3 power sources ready:

- 2.5mm/5.5mm DC jack, (1) in the picture

- 4pin power socket, (2) in the picture

- 4pin power socket, (3) in the picture

Additionally J6, which is connected to J4 directly - both of them have a +12V line to J11 (the biggest socket place on the other side of the board). OWC utilises J6 instead of J4.

Observations during the 3DMark 11 performance test:

- Either one of power sources (1) and (2) is compulsory. If you don’t plug anything into these, GPU will never show up in the device manager.

- Power consumption of (1) and (2) are almost equal. Constant ~9W-10W and it doesn’t change a lot during the performance test but there is a slight difference when you use another GPU. GTX780 gave a very constant 8.9W number during all the performance period, but with the R9 280X the consumption was approximately 1W greater, still very constant as well.

- Power consumption of (3) depends on the GPU. We have now a proof that x4 slot is able to output much more than 25W through the TB card + PCIe board. Note that we are not sure how safe it is for the TB card and PCIe board, because AKiTiO is designed for 25W cards and the inner lines may be too thin for 75W. The safest option is to use a normal powered riser.

Unfortunately it doesn’t seem to have an effect on stability whether you feed power directly to (3) or via (1) and (2) with the original 4pin cable. I doubt to get 100% stable AMD eGPU after every reboot, but at least I found some keys to pass 3DMark11 physics part, new powering methods and interesting power consumption numbers.

[tamer1009]

It looks like an 107C SMD Tantalum capacitor. You can get a replacement part from here. It's a polarized capacitor so make sure it's soldered in the same orientation as the original part (see pictures here). Like I mentioned before, it's likely that this is not the only damaged component on your AKiTiO boards. It's worth taking a closer look at both boards and checking for burned tracks or other damaged components. Some you can check with a multimeter, others (such as the ICs) won't be easy to check without more sophisticated equipment.

Thanks, I already ordered a new 107C to fix this. Only this component was damaged, everything else is OK.

[jacobsson]

Thanks, I already ordered a new 107C to fix this. Only this component was damaged, everything else is OK.

This is of course not very useful to hear after a disaster like this (been there myself), but always double/triple check with a multimeter for polarity and use a marking pen, especially when there are only yellow cables. I really hope you manage to revive your board with the 107C.