GTX 690 in MacPro4,1 WITH external PSU + Add2PSU

Discussion in 'Mac Pro' started by 5050, May 2, 2013.

  1. 5050 macrumors regular

    Joined:
    May 28, 2009
    #1
    OK, so I'm renaming this thread because I've transitioned away from the idea of running the GTX 690 on the MacPro4,1's internal PSU and the plan is to now use an external PSU (Seasonic Platinum 860W) to power the card. More power than what's needed to run one GTX 690, and really future proofing for a potentially 2nd GTX 690 using the same power supply, but I digress . . .

    I'm currently seeking a solution that will allow my external PSU to power up/down and sleep/wake with my workstation. The basic idea is if the external PSU is in sync with the internal PSU, the GTX 690 will also be in sync (and thus acting like a normal PCIe powered graphics card).

    The best solution I've found is the Add2Psu. Here's the link for the Add2Psu adapter:

    http://www.add2psu.com

    My "original" plan with the Add2Psu was:

    SATA 15-pin connection (from 2nd optical drive bay) > SATA 15-pin to Molex 4-pin adapter > Molex extension cable > Add2Psu > 24-pin power supply cable > external PSU

    The Molex cable would run out through the back of an open PSI slot cover into the Add2Psu and eventually into the external PSU.

    However, getting the Molex cable out through the optical bay has proved beyond challenging, and I'd prefer not to splice the Molex cable simply to feed the wires through the very small openings out of the optical bay. And feeding the Molex extension cable through the optical drive tray cover is also not an option.

    The 2nd more viable option I'd like to :

    PCIe 6-pin connection on motherboard > PCIe 6-pin to Molex 4-pin adapter > Molex extension cable > Add2Psu > 24-pin power supply cable > external PSU

    The path out from the PCIe 6-pin connection through the PCI slot cover is MUCH easier to navigate than the optical bay. My question is, will the PCIe 6-pin provide the correct voltage necessary to power on/off, sleep/wake the external PSU via the Add2Psu? Will it provide too much wattage than what's actually necessary?

    Also, if anyone has a more practical idea, it would be very much welcomed!
     
  2. bsbeamer macrumors 6502

    Joined:
    Sep 19, 2012
    #2
    have you looked into the GPU expander chassis options? doubt it's as cheap as this solution, but it is another idea - if you're looking for one.
     
  3. 5050 thread starter macrumors regular

    Joined:
    May 28, 2009
    #3
    Yes, both Dynapower and Cubix. Will probably grab one of these further down the road, but as you noted, the Add2Psu is the "cheapest" (and it's a relative term, right?) option now.
     
  4. GP-SE macrumors 6502

    Joined:
    Feb 27, 2013
    #4
    I think the Optical bay Molex is 5v and PCI-E is 12v
    what about a SATA to molex adapter, then use a SATA port to trigger the external PSU.
     
  5. 5050 thread starter macrumors regular

    Joined:
    May 28, 2009
    #5
    Just a quick question before diving into my reply, but what voltage should be going into the 24-pin connection on the external PSU to trigger the powering on/off, sleep/wake of the external PSU? 5V or 12V?

    So the original plan was to draw the power from the spare SATA 15-pin connector in the 2nd optical bay (as you're suggesting) using a SATA 15-pin to Molex 4-pin adapter, but I can't seem to find a way to get the Molex cable out of the optical bay (see original thread details above). If the SATA 15-pin to Molex 4-pin is the way to go (instead of a PCIe 6-pin to Molex 4-pin), how about an adapter chain that converts the PCIe 6-pin into SATA 15-pin connections before going into the Molex 4-pin? Would that work?

    Here's how that setup would work:

    1) PCIe Mini 6-pin (motherboard female connector)
    2) PCIe Mini 6-pin (male) to PCIe 6-pin (male)
    3) 6-pin PCIe (female) to SATA 15-pin (male)
    4) SATA 15-Pin (female) to Molex 4-pin (male)
    5) Molex extension cable (female to female) > Molex 4-pin (male) on Add2Psu

    And here are the cables . . .

    PCIe Mini 6-pin to PCIe 6-pin
    http://www.amazon.com/gp/product/B002UR1654/ref=oh_details_o00_s00_i01?ie=UTF8&psc=1

    8-inch SATA 15pin to 6pin PCI Express Card Power Cable
    http://www.monoprice.com/products/p...=10226&cs_id=1022607&p_id=8494&seq=1&format=2

    6-inch SATA 15pin Female to Molex 4pin Male Power Adapter
    http://www.monoprice.com/products/p...=10226&cs_id=1022604&p_id=8797&seq=1&format=2

    Molex(5.25 Female) / Molex(5.25 Female)
    http://www.monoprice.com/products/p...=10245&cs_id=1024501&p_id=1317&seq=1&format=2

    Thanks!!
     
  6. GP-SE macrumors 6502

    Joined:
    Feb 27, 2013
    #6
    instead of using the optical drive SATA can you use a free SATA connector on the mobo? or do you have all 4 drive bays filled?
     
  7. Tesselator macrumors 601

    Tesselator

    Joined:
    Jan 9, 2008
    Location:
    Japan
    #7
    AFAIK the molex 12v rails are 40W each (max). So you need one from two different rails to do it that way in perfect safety if using only MP4,1 PSU. All the native PSU feeds run near (thru) that ODD bay so you might be able to nab a line going to SATA (backplane) connector that you're not using. The most intelligent I've seen to run the 12V lines down out of the ODD bay (to the GPU) to thru the wiring harness opening in the corner of the ODD bay. You might have to remove the molex connector to lace it thru but that's easy - it's just a tiny flat driver to depress the spring-loaded catch on each line and the leads slip out.

    I wouldn't recommend this tho.

    I would much rather go with your original intention of using an auxiliary PSU. In no case would I ever use a SATA power connector to molex or to 6-pin DIN - that just seems nuts for several reasons.
     
  8. 5050 thread starter macrumors regular

    Joined:
    May 28, 2009
    #8
    Thanks Tesselator for your thoughts. I already have the external PSU, the issue is getting it to power on/off, sleep/wake with the rest of the workstation. What is your suggested solution to this?
     
  9. Tesselator macrumors 601

    Tesselator

    Joined:
    Jan 9, 2008
    Location:
    Japan
    #9
    Me? I already answered that in your other thread. I would look at a diagram of the MP you're using and splice directly into that mechanism. Alternatively I would look into 1600 to 1800W PSUs that I could fit into the same area as the MP's native PSU and just change it out.
     
  10. 5050 thread starter macrumors regular

    Joined:
    May 28, 2009
    #10
    I actually don't feel comfortable modifying the Mac Pro's factory PSU or any type of splicing but thanks for the suggestion (maybe one day I'll get there!). In the meantime, I'm looking for the "simplest" way to make this work without altering the original components of the Mac Pro.

    I tested the SATA 15-pin to Molex 4-pin into the Add2Psu and the external PSU/GTX 690 powered up/down, slept/woke with the rest of the system, so that's a go. The only problem is that I have the cables feeding out of my optical drive bay cover and I'd prefer to feed my cables out through the rear of my Mac Pro.

    Again, the PCIe 6-pin connection on the motherboard is the easiest way to get the cables out through rear (using the open PCIe cover). Is there any danger in using the PCIe 6-pin connection (the same way I used the SATA 15-pin which worked above) to sync up my external PSU with my internal PSU?

    Forgive my ignorance of basic voltage, but what does the difference between 5V and 12V mean when triggering the external PSU?

    Thanks again for all the feedback I've received here.
     
  11. Tesselator, May 4, 2013
    Last edited: May 4, 2013

    Tesselator macrumors 601

    Tesselator

    Joined:
    Jan 9, 2008
    Location:
    Japan
    #11

    I think there are just going to be problems inherent to running the AUX PSU externally. How you address them will depend on a whole bunch of factors you personally determine as being acceptable and not. IMHO, the only sane way to go about it is to have the AUX PSU internal and in the ODD bay. If that meant I had to buy a cable harness for the SATAII cables and run the drives in a cradle or enclosure then so be it.

    [​IMG]

    If you're using a large full-sized ATX type PSU that might also mean you will have to run your ODD externally as well. Cases are cheap these days if you don't become stupid and opt for something like the Drobo. Around $30 to $70 should do it: http://www.bhphotovideo.com/c/produ...350SUAB2_Hi_Speed_eSATA_FireWire_USB_2_0.html and http://www.bhphotovideo.com/c/product/551632-REG/Macally_G_S350SU_G_S350SU_eSATA_and_USB.html

    The ODD can run off a simple USB2.0 connection fine so in total all you really would need is that cable bracket above and an enclosure that can accept one 5.25" and two 3.5" dives if you don't wish to stack them. If stacking is acceptable ODD enclosures can be had for around $25 to $50.

    IMO it's either that or be confronted not only by the struggle of trying to connect a PSU externally but also face the messy look it's going to create sitting nearby somewhere on the outside. UG!


    If you're using a cable adapter (xxxx --> Molex) or the ODD bay located existing molex, then there's no need to worry about this. If you're wiring it yourself have a look at that little black box:

    [​IMG]

    And see if needs 12V or 5V. It's just an electromagnetically opened or closed circuit (switch) and uses the voltage to energize a coil which "turns on" the "magnet" making the two ends of the (spring-loaded) metal leads "stick" to each other. :p That might be over-laymanized but you get the idea. You should use whatever voltage that black box (relay) is rated for or change it out to suit the voltage you wish to use.
     
  12. IceMacMac macrumors 6502

    Joined:
    Jun 6, 2010
    #12
    I'm finding this thread very depressing. I too would be interested in a straightforward solution, even it it set me back a bit of money. I have no interest or time to mess w/this kind of thing.
     
  13. Tesselator macrumors 601

    Tesselator

    Joined:
    Jan 9, 2008
    Location:
    Japan
    #13
    Straightforward as in just shoving something in place and have it all "just work"?

    By the very nature of this mod that's not going to happen! You will have to use a screwdriver, you will have to string wires, and you may even need to use a soldering iron.

    But I guarantee even a ten-year-old of average intelligence can do all of these things. So it's only your own reluctance to go outside your comfort zone that's prohibiting you. Mac users have a VERY narrow comfort zone - it's almost disgusting how narrow it is. :p
     
  14. 5050 thread starter macrumors regular

    Joined:
    May 28, 2009
    #14
    Thanks, your explanation of the relay mechanism is very insightful.

    So you're saying, as long as the adapter I plan to use feeds into a Molex 4-pin, I should be fine with the voltage (i.e. it shouldn't matter if it originates from a SATA 15-pin, PCIe 6-pin, etc as long as the other end is a Molex 4-pin)?

    Also, does it make a difference if the ends of these adapters are "single" Molex connectors or "dual" Molex connectors (most of the adapters I've found that lead into Molex 4-pin connections are generally "dual" Molex connections instead of "single")?

    Thanks again!
     
  15. Tesselator, May 5, 2013
    Last edited: May 5, 2013

    Tesselator macrumors 601

    Tesselator

    Joined:
    Jan 9, 2008
    Location:
    Japan
    #15
    Yep, that's right. Of course assuming the manufacturer of the adapter cable didn't blow chunks or something. But that's a pretty safe assumption... ;)


    You mean where it's a Y configuration... one six-pin connector and two molex connectors? No, that wouldn't matter - either will work (straight or Y). You can remove the extra one if you like or just leave it on. If you leave it on it might be a little better to cap it off. Either with a plastic cap made for that or just wrap it in some electrical tape. The idea behind this is that those molex connecters have their leads not-so-deeply recessed and can actually touch some protruding (usually off the main board) shunt posts and so forth.
     
  16. MyMac1976 macrumors 6502

    MyMac1976

    Joined:
    Apr 14, 2013
    #16
    I sorta half followed the the original 690 thread..

    So we're trying to quad SLI two 690's in a mac pro to run Windows correct?
    Is this not what KVM switches were meant for? Two computers can share one monitor and key board?
     
  17. 5050 thread starter macrumors regular

    Joined:
    May 28, 2009
    #17
    Haha, sounds great, thanks. OK, one last question . . . I've noticed a lot of the Molex ends only have 3 contacts instead of 4 (like the link below). Is this an issue going into the Add2Psu?

    http://www.amazon.com/6-Pin-PCI-Exp...qid=1367746109&sr=1-16&keywords=PCIe+to+molex

    ----------

    Nope, just to power one 690 in a Mac Pro running OS X 10.8.3. I'd potentially add another 690 in the future to have a total 2 690s running off the Mac Pro PCIe slots, both powered by the external 860W PSU. This thread is focused on getting the external PSU to power up/down, sleep/wake in sync with the Mac Pro internal PSU (basically so the 690 "behaves" with the rest of the workstation).
     
  18. Tesselator, May 5, 2013
    Last edited: May 5, 2013

    Tesselator macrumors 601

    Tesselator

    Joined:
    Jan 9, 2008
    Location:
    Japan
    #18
    I dunno what those are. I've never seen nor heard of a 3-pin molex. But of course it's only common sense that if your device or gadget has a 4-pin plug then you need to use a 4-pin connecter. :) If it's neutered 3-pin (4 wide with one just missing) that probably means that it only supplies either 5V or 12V but not both. Look at the black box or read the instructions to determine which you need, then look of the specs of the cable you're buying to make sure it's supplying what's needed. There won't be any damage if it doesn't but your gadget there won't work - as it won't actually be connected.

    [​IMG]

    Make sure you get the genders right too. Or use a gender-changer to adapt.

    [​IMG]
    Male




    [​IMG]
    Female.

    .
     
  19. 666sheep macrumors 68040

    666sheep

    Joined:
    Dec 7, 2009
    Location:
    Poland
    #19
  20. MyMac1976 macrumors 6502

    MyMac1976

    Joined:
    Apr 14, 2013
    #20
    My point was you are going through all of this to put a dial GPU card in a Mac to run Windows. Why not just build a PC and use a KVM switch? Better yet for the same money buy a Titan and use it both OSX and Windows. No dual GPU issue and arguably better performance and more future proof
     
  21. Tesselator, May 5, 2013
    Last edited: May 5, 2013

    Tesselator macrumors 601

    Tesselator

    Joined:
    Jan 9, 2008
    Location:
    Japan
    #21
    Convenience and about a thousand dollars would be my guess. ;)



    BTW, there will still be a GPU issue if a big enough PSU isn't incorporated into the build.
     
  22. MyMac1976 macrumors 6502

    MyMac1976

    Joined:
    Apr 14, 2013
    #22
    we're dealing with a close to $1000 gpu that will only work in Windows. IMHO at some point you have stip and question whether you're putting lipstick on a pig
     
  23. Tesselator macrumors 601

    Tesselator

    Joined:
    Jan 9, 2008
    Location:
    Japan
    #23
    it only works under Windows? What makes you say that?

    My guess is that the 690 will bench in OS X better than the 680 and the 680 will bench better than the 670 and all those will bench better than the 570. So an improvement is an improvement relative to itself. Thus not lipstick on a pig.
     
  24. MyMac1976 macrumors 6502

    MyMac1976

    Joined:
    Apr 14, 2013
    #24
    The 690 is a dual GPU card and there is no SLI support in OSX. Therefore OSX might use one of the two GPU's. Titan is about the same money and can be used in OSX it makes no sense to me to use anything other than that if you need over the top graphics.
     
  25. 666sheep macrumors 68040

    666sheep

    Joined:
    Dec 7, 2009
    Location:
    Poland
    #25
    You should read OP's earlier threads to understand. OS X itself will see both GPUs but it will use one. CUDA and OCL accelerated software will see and use both GPUs. That's the point.
     

Share This Page