GTX Titan in Mac Pro Late 2009 works fine!

[Asgorath]

Latest web driver for 10.8.3 appears to be based on the same 313 branch that was used for the Titan launch under Windows. Might be worth updating to this driver to see if the card works better under OS X. Direct download link since the downloader page gives a .dmg file incorrectly:

http://us.download.nvidia.com/Mac/Quadro_Certified/313.01.01f03/WebDriver-313.01.01f03.pkg

 

[skyline r34]

Don't know if this is true or not but i guess the titan might be the stock GPU for the 2013 Mac Pro but then again this is just a rumor


http://t.co/EMiWAby9TC

 

[MacVidCards]

VERY old news.

Titan was first reported working in 10.8.4 in April. Those Architosh guys are reposting some old reposts.

I recently even posted that it works in 10.8.3 using new web drivers.

I will also say this, the EFI on the GTX680 contains lines for the GK110 family.

May mean something, may not.

In the FWIW department, the EFI for Quadro 4000 conatined references for GTX480, this was why I was able to write the EFI for it that is all over the place now. So they were either planning a GTX480 for Mac Pro or wanted someone else to finish the job.

 

[5050]

Titan was first reported working in 10.8.4 in April. Those Architosh guys are reposting some old reposts.

I recently even posted that it works in 10.8.3 using new web drivers.

I will also say this, the EFI on the GTX680 contains lines for the GK110 family.

May mean something, may not.

In the FWIW department, the EFI for Quadro 4000 conatined references for GTX480, this was why I was able to write the EFI for it that is all over the place now. So they were either planning a GTX480 for Mac Pro or wanted someone else to finish the job.

MacVidCards, can you point me in the direction of your post regarding Titan compatibility in 10.8.3?

I was unaware of any compatibility below 10.8.4.

Thanks.

 

[MacVidCards]

titan works using 10.8.3 Web Driver

I posted it at Netkas under my other nom-de-plume in a one line post in thread about the 10.8.3 web drivers.

Guess I should make a big deal of it.

I will do so.

 

[ambachnz]

Mac Pro 4,1 GTX Titan Question

Hi all,

Have just successfully installed an FSP Booster X5 and EVGA GTX Titan Superclocked in to my Mac Pro 4,1 8 Core 2.26 GHz running 10.8.4 for rendering video files in DaVinci Resolve.

Slot Config is as follows :

Slot 1 - GTX Titan
Slot 2 - Quadro 4000 For Mac
Slot 3 - Blackmagic-Design Decklink 4K Extreme
Slot 4 - USB 3.0 X 2 Card

I seem to have lost the startup screen (grey with Apple logo) when either of the GPUs are connected to my current monitor (ASUS PB278Q). I was hoping to avoid this by continuing to use the Q4000 as my GUI card (Resolve talk). Any thoughts on why startup screen has disappeared?

Also can anyone recommend an application that will allow me to monitor the temps of the Titan? I noticed the Mac Pro was almost breathing fire when rendering out a 4K video clip earlier today. The poor Titan does not have a lot of room to breathe with the Q4000 sitting directly on top of it. I really do not want to melt the very expensive Titan or the Q4000 sitting on top of it...

Lastly, under System Report, under the PCI Cards header, it says "There was an error while gathering PCI card information". Any thoughts?

Thanks in advance, Alain

 

[pastrychef]

Lastly, under System Report, under the PCI Cards header, it says "There was an error while gathering PCI card information". Any thoughts?

This is normal behavior when you have a video card in the Mac Pro that doesn't contain EFI firmware.

 

[666sheep]

I seem to have lost the startup screen (grey with Apple logo) when either of the GPUs are connected to my current monitor (ASUS PB278Q). I was hoping to avoid this by continuing to use the Q4000 as my GUI card (Resolve talk). Any thoughts on why startup screen has disappeared?

Also can anyone recommend an application that will allow me to monitor the temps of the Titan?

Disconnect the Titan and you'll get bootscreen back. My guess is that Quadro is connected via DVI, Titan via DP, correct? OS X somehow does prefer DP/mDP over DVI.
Computing-only card for Resolve doesn't have to be connected to display at all AFAIK.

Unfortunately OS X does not offer any monitoring of GPU temp.

 

[MacVidCards]

You might try putting Quadro in slot 1 connected to display and Titan in slot 2, if you can afford the slot space.

 

[Stimpyzz]

Has anyone managed to get two TITANs working in a Mac Pro?

We have a 2010 model that has two 6pin onboard (one is powering the 6pin and 8pin-using adapter)

To add the second card would it be do-able using the sata ports on the motherboard ? 15pin sata to 6pin + 15pin too 6pin to 8pin?

 

[MacVidCards]

Has anyone managed to get two TITANs working in a Mac Pro?

We have a 2010 model that has two 6pin onboard (one is powering the 6pin and 8pin-using adapter)

To add the second card would it be do-able using the sata ports on the motherboard ? 15pin sata to 6pin + 15pin too 6pin to 8pin?

Oh, yes. Got a couple Titans and tossed in a GTX780Ti for good measure.

Get another power supply. A few threads on here how to do it.

Whatever you do, don't post your 3D Mark score in the "D700s ...3D Mark" thread. Will get you a nice coat of tar and feathers.

 

[OrangeSVTguy]

Has anyone managed to get two TITANs working in a Mac Pro?

We have a 2010 model that has two 6pin onboard (one is powering the 6pin and 8pin-using adapter)

To add the second card would it be do-able using the sata ports on the motherboard ? 15pin sata to 6pin + 15pin too 6pin to 8pin?

Running strong with 2 Titans https://forums.macrumors.com/threads/1472518/

Removing your last 3 hard drive bays, I seen someone get a 3rd titan in there.

Oh, yes. Got a couple Titans and tossed in a GTX780Ti for good measure.

Get another power supply. A few threads on here how to do it.

Whatever you do, don't post your 3D Mark score in the "D700s ...3D Mark" thread. Will get you a nice coat of tar and feathers.

I haven't checked in a while but the 780 ti is working in OSX now?

 

[echoout]

Has anyone managed to get two TITANs working in a Mac Pro?

We have a 2010 model that has two 6pin onboard (one is powering the 6pin and 8pin-using adapter)

To add the second card would it be do-able using the sata ports on the motherboard ? 15pin sata to 6pin + 15pin too 6pin to 8pin?

I've been running 2 Titans in my 2009 for a couple months to get me through a big Octane project, but then it saved my bacon on a last-second AE Ray-tracing project last week.

My z820 has 2 780Tis in it that I'm going to swap into the MP whenever I get some time between projects. These two computers using Octane with C4D Team Render is a great setup for my business.

----------

I haven't checked in a while but the 780 ti is working in OSX now?

Yeah, with the beta drivers. But waiting for a CUDA update.

 

[NukeUser]

Hey echoout- thanks for posting all the GPU tips, they are great! I was talking on a thread about the GTX780ti with MacVidCards and he said his 3,1 was much slower than his 4,1 with the exact same HD and same cards...... have you heard about this happening?

 

[echoout]

Hey echoout- thanks for posting all the GPU tips, they are great! I was talking on a thread about the GTX780ti with MacVidCards and he said his 3,1 was much slower than his 4,1 with the exact same HD and same cards...... have you heard about this happening?

The 3,1 has slower buss speeds, etc. Does it have anything to do with the GPU itself or is it just the computer? My 3,1 was decked out and SO much slower than my barebones 4,1.

MVC knows his stuff.

 

[Tutor]

It's how I hack my Macs. MacPro2,1s make good GPU expanders/ + U get 8 cores to help

... .
Removing your last 3 hard drive bays, I seen someone get a 3rd titan in there. ... .

I tried that last weekend - it works. I have Maxupgrades' MaxConnect for Mac Pro [2006-2008] Optical Drive Bay Disk Mounting Assembly ( http://www.maxupgrades.com/istore/index.cfm?fuseaction=product.display&product_ID=300&ParentCat=315 ) for each of my 2007 MacPro2,1s. That allows the freedom to remove the Apple factory hard drive brackets 2-4 and still have some room for some extra storage. I installed two additional 3.5" HDs/SSDs (you can install more if you use 2.5" ones - remember: you've still got the motherboard Sata data & power connectors that the three removed drive brackets had access to) in the lower level of the Optical Drive Bay of MaxConnect. My FSP BoosterX5 goes on the top level where the Optical drive would go in the MaxConnect. I used my Dremel to cut a small section out of that multis screwed (20 screws) inner assembly on my case cover (of course I removed the inner assembly before the cutting began) [see pic below] so that I can just drop the GPU power cables from the FSP straight to the GPUs and the one from the FSP that would otherwise block a PCIe slot to reach external power, now runs to a squarish hole that I cut in the grill on the back of the systems (near the door lever). Yes, this solution leaves only 19 screws now holding that assembly, but I don't feel screwed - I can live with that one screw sacrifice. With my 2007s, I have to allocate the PCIe performance to 8/8/8 for slots 1, 2 and 4. On early Mac Pros if you're running Octane, you're going to suffer a little hit with scene voxelizing speed initially in any event, but once that done (and that takes at most only a few seconds on the front end), the rendering proceeds quickly - dependent on the GPUs that you've chosen.

P.S. - CUDA 6 and OS X 10.9.2 will soon drop.

 

[Tutor]

The 3,1 has slower buss speeds, etc. Does it have anything to do with the GPU itself or is it just the computer? My 3,1 was decked out and SO much slower than my barebones 4,1.

MVC knows his stuff.

As to Octane/CUDA rendering, the GPU rendering performance on my MacPro2,1s in the x8 slots {1,2 &4} using GTX cards at stock settings leaves little to be desired vs. their performance on my systems with PCIe 3.0 x16 slots using the same GTX cards at stock settings. Of course, the time it takes to voxelize the scene to get the data into the GPU is more noticeable. But, once that's done they humbly hum along to my complete satisfaction.

 

[echoout]

The 3,1 has slower buss speeds, etc. Does it have anything to do with the GPU itself or is it just the computer? My 3,1 was decked out and SO much slower than my barebones 4,1.

MVC knows his stuff.

I'm not positive what we're talking about anymore, reading this thread backwards.

 

[Tutor]

I'm not positive what we're talking about anymore, reading this thread backwards.

My point is that the speed at which the GPU performs its internal computations is the same; its just fed a little slower. How were you powering it?

 

[riggles]

I'm wondering how you power 3 Titans internally, even with the FSB.

 

[Tutor]

I'm wondering how you power 3 Titans internally, even with the FSB.

Titan 1 - (a) 1 - 6 pin from MB to 6 pin on Titan + (b) 8 pin on Titan feed by 8 pin Y cable feed by 6 pin from MB and 6 pin from 4 pin molex in optical drive bay + (c) PCIe

Titans 2 & 3 - (a) FSP BoosterX5 + (b) PCIe

 

[riggles]

Titan 1 - (a) 1 - 6 pin from MB to 6 pin on Titan + (b) 8 pin on Titan feed by 8 pin Y cable feed by 6 pin from MB and 6 pin from 4 pin molex in optical drive bay + (c) PCIe

Titans 2 & 3 - (a) FSP BoosterX5 + (b) PCIe

Eek. I would be nervous about asking a SATA connector to deliver potentially up to 75W :-0

 

[NukeUser]

Tutor- show us some pics of 3 cards in there!

 

[Tutor]

Eek. I would be nervous about asking a SATA connector to deliver potentially up to 75W :-0

I misspoke earlier by calling it a SATA connector. My 2007s did not come with SATA power connectors for the Optical Drive bay. They came with 4-pin Molex connectors on a ribbon cable (I have used SATA drives via a conversion cable in those bays). I recall from having researched the Molex power issue awhile back that the molex cable can supply about 60W safely. The Graphics Card Power (W) for the Titan is 250W, but in a MacPro the Titan isn't always demanding that much power or even doing so much of the time. Each onboard 6-pin graphics power connector delivers 75W each, plus 75W from PCIe slot. So the total from those three sources = (3x75) 225W. Many folks use just that, via various hacks - such as by using a cable some video card manufacturers supply with their GPUs built with an 8-pin connector, that is 6- to 8-pin direct cable, to supply the needed power of a Titan in a MacPro or Windows system. To lessen your concern in a three Titan MacPro environment, you can lessen the power requirements somewhat for whichever Titan is assisted by molex power by simply not using that Titan for video display, i.e., use it for compute functions only. That's how I'll do it from here on out if I need to put a three Titans in one of my 2007s for regular use. Luckily, however, just because I tried it last weekend doesn't mean that I'll be doing it often. I have so many 3, 4 and 8 double wide PCIe x16 slotted systems that I doubt that it to be a regular occurrence. I did it last weekend to just to see whether and how it could work to respond to a question from someone else.

P.S. I found the article that I referenced above - http://www.playtool.com/pages/psuconnectors/connectors.html#peripheral .

 

[Tutor]

Seems as if Nvidia is giving each of its high end GPUs a day in the sun.

Tutor - show us some pics of 3 cards in there!

See my earlier post. The regular configuration (and current state) of two of my three MacPro2,1s is 3 cards; but just not the three cards that I think that you may have in mind. Those two MacPros each regularly house 2xGTX590s (2x8-pin) plus a GT640 (only PCIe power needed) that I use for 3d scene setup and tweaking. The other 2007's regular configuration (and current state) is 2xGTX590 plus an Intensity Pro video card. The latter is also the current state of my MacPro4,1 (subsequently hacked into a MacPro5,1 of sorts). The GTX590s render faster than the reference design (RD) Titan tested by Barefeats ( http://www.barefeats.com/gputitan.html ) by 1.13x each. I prefer to design my 3d scenes on the MacPro. Moreover, the 590s (used) cost a heck of a lot less - under $400. And "Yes - the GTX590s have a heck of a lot less vram (3G, but 1.5G per GPU processor - the GTX590 has 2 GPUs processors) than the Titans; but I want the discipline in scene design that imposes."

BTW - The GTX590 has 1,024 Fermi CUDA cores. The GTX690 has 4G vram (but 2G vram per GPU processor - it has 2 of them) and 3,072 Kepler CUDA cores (it's 1.2x times faster at rendering than an oRD Titan). One way of looking at Fermi vs. Kepler is that one Fermi core is just shy of being equal to 3 Kepler cores when it comes to CUDA power. The GTX790 (which should soon drop) is said to have at least 6G vram (or at least 3G vram per GPU processor - it has 2 of them) and 4,992 Kepler CUDA cores versus 2,880 Kepler CUDA cores for the soon to drop Titan Black Edition (a GTX780 Ti [1.32x faster at rendering than an oRD Titan] with more vram and somewhat faster clocks than the 780 Ti ). In other words, the GTX 790 should be the fastest GTX card at CUDA functions by a long shot, but the Titan Black Edition will have more vram. Thus, the 780 Ti will be the lower priced of the three top dogs, and almost just as fast as, but have less vram than, the nTitan.