Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
fastest gpu 2x6 pin - nvidia 680?
- Thread starter nox-uk
- Start date
- Sort by reaction score
The MSI 680 4Gb Twin Frozr was 2x6-pin too.
You may want to have a look at the nVidia 770s because I believe those are practically identical to the 680s but I don't know if there are any 2x 6-pin ones available yet.
You may want to have a look at the nVidia 770s because I believe those are practically identical to the 680s but I don't know if there are any 2x 6-pin ones available yet.
As far as anyone can tell, the 6+8 680's use the same amount of power as the 6+6 680's. I believe some companies plopped an 8-pin on there to market to the overclocker/overvoltage crowd.
From this article, the PCIe spec allows for a card to accept either 6-pin (75W) or 8-pin (150W) power sources on an 8-pin connector. The "sense0" wire tells the card that 75W is fine. The "sense1" wire, only on 8-pin cables, tells the card to go ahead and pull up to 150W.
If everything is following the rules, the card will know to only pull up to 75W on the 8-pin connector with a 6-pin cable, and this is exactly how I run my GTX680 6+8 pin card (two 6-pin cables).
Yes, all of the 680s I have seen with 6 & 8 will boot and run just fine on 6 & 6.
I think I'll do a power draw test later today and see if the draw is asymmetrical as on 8 pin connectors on other cards.
770s require "tricking" final sense pin, the 680s do not.
I think I'll do a power draw test later today and see if the draw is asymmetrical as on 8 pin connectors on other cards.
770s require "tricking" final sense pin, the 680s do not.
I can add that even when overclocked considerably, my GTX 680 with two 6-pin connectors, power draw was always under 200W and the draw from the two connectors and PCI-e slot were always quite even. About 60-ish watts from each.
Did you run Furmark while testing? That is the only place that I have been able to make cards draw obscene power. I'm just curious.
Triple screen Eyefinity running Crysis might also show some serious draw, I wonder if there is a Windows app that could track MP power draw, probably not.
Triple screen Eyefinity running Crysis might also show some serious draw, I wonder if there is a Windows app that could track MP power draw, probably not.
I ran Furmark prior to testing the overclocking limits of my card. The results were largely the same.
I would run it now on my current clocks but I'm in Mavericks and Furmark doesn't run in Mavericks. Computer is busy doing some stuff and I can't reboot in to Mountain Lion at the moment.
I would run it now on my current clocks but I'm in Mavericks and Furmark doesn't run in Mavericks. Computer is busy doing some stuff and I can't reboot in to Mountain Lion at the moment.
Attachments
Last edited:
Attachments
Got it. I was running 0.5.0 which doesn't work with Mavericks. 0.4.0 does.
Attachments
Last edited:
Just realized that there are a bunch of OpenGl benchies in there that use OpenGl 4.0 that we can now use.
I also note that the GTX780 is 40-50% faster than that 680 you tested. GK110 is truly a generational step up from GK104, at least for running Furmark. Or perhaps GK104 knows to ease throttles to keep power draw low? I don't recall the barefeats tests showing 780 to be 50% faster than 680.
Power consumption is higher. Running test I hit 5.48A, 8.35A, and 4.02A using scale you did.
I also note that the GTX780 is 40-50% faster than that 680 you tested. GK110 is truly a generational step up from GK104, at least for running Furmark. Or perhaps GK104 knows to ease throttles to keep power draw low? I don't recall the barefeats tests showing 780 to be 50% faster than 680.
Power consumption is higher. Running test I hit 5.48A, 8.35A, and 4.02A using scale you did.
Yes, it appears the GK110 is remarkable at GPGPU tasks. Definitely a huge step forward from the GK104.
^^^^Yep, as the Doughnut Man knows, my MacVidCards flashed GTX780 is really remarkable IMHO. Now that Open CL works, in Mavericks, my experience is complete:
https://forums.macrumors.com/threads/1650842/
Lou
https://forums.macrumors.com/threads/1650842/
Lou
many thanks
I'd love to get a 770 or 780 from MVC but international shipping, flash costs on top, etc and not to mention the continual nagging in my head of putting a 6+8 pin card in a place for a 6+6... and I do need the boot screen unfortunately
At least until bootchamp starts letting me boot vmware off a usb stick.Time to track down a flashable 680 whilst they still exist
again, many thanks for the info everyone
Nox
GPUZ is usually what i use to track power draw in winslows.
Haven't tried it in any power situations on the mp, but it's sensors logging should be useful.
so some cards that require 1 x 6 pin and 1 x 8 pin connectors can run of 2 x 6 pins? what about something like a nvidia 780 running in a 2006 mac pro?
It can work, but you will be drawing more power from the Mac Pro's power connectors than they were intended for. 6-pin connectors are rated to draw up to 75W. 8-pin connectors are rated to draw up to 150W. The PCI-e slot is rated at 75W.
As you can see from my post above, my GTX 680 card with two 6-pin connectors are pulling 68.52W and 67.8W from the 6-pin connectors and 63.12W from the PCI-e slot. All are within spec since each of those connections are rated to provide up to 75W.
You can also see that the GTX 780 is pulling 65.76W, 100.2W, and 48.24W. Obviously, 100.2W is above what the Mac Pro's logic board was rated for.
thanks for the reply. since my mac pro will not run 10.8 or higher, can i use one of the newer cards in Mac OS X?
The following was taken from the FAQ on this forum.
"Note that the Fermi generation cards work with Lion, but in almost all cases the Kepler cards require Mountain Lion. You may find a GK107-based card that boots in 10.7.5, but it is highly recommended that you use Mountain Lion since there appears to be much better driver support in that OS in general.
To be clear, the stock Apple drivers now contain basic support for all currently-shipping cards. This means that you no longer have to swap out the card when doing a Software Update to the next OS version. In all cases, it is recommended that you update to the corresponding NVIDIA web driver, as those drivers tend to have more bug fixes and/or performance improvements than the stock Apple drivers."
That being said, I've read postings from other Mac Pro 1,1 and 2,1 users who have successfully installed 10.8.x and 10.9 who say the process is quite simple using Chameleon.
"Note that the Fermi generation cards work with Lion, but in almost all cases the Kepler cards require Mountain Lion. You may find a GK107-based card that boots in 10.7.5, but it is highly recommended that you use Mountain Lion since there appears to be much better driver support in that OS in general.
To be clear, the stock Apple drivers now contain basic support for all currently-shipping cards. This means that you no longer have to swap out the card when doing a Software Update to the next OS version. In all cases, it is recommended that you update to the corresponding NVIDIA web driver, as those drivers tend to have more bug fixes and/or performance improvements than the stock Apple drivers."
That being said, I've read postings from other Mac Pro 1,1 and 2,1 users who have successfully installed 10.8.x and 10.9 who say the process is quite simple using Chameleon.
So I'm curious, does Apple purposefully make drivers for off-the-shelf cards (like a kind of "soft support" situation) or is it just pure luck and coincidence that they're designed for official Nvidia chips in Macs and happen to work with unofficial cards?
Also wondering the same thing about Nvidia since technically these drivers are for Quadro cards? Just hoping some higher up from either company doesn't step in one day and rain on our parade, or that this driver development doesn't fall by the wayside when the MacPro6,1 is released.
Also wondering the same thing about Nvidia since technically these drivers are for Quadro cards? Just hoping some higher up from either company doesn't step in one day and rain on our parade, or that this driver development doesn't fall by the wayside when the MacPro6,1 is released.
The drivers for almost all NVIDIA cards have been present in the OS for a long time. When I realized that Hackintosh folk were using a wide variety of cards by just using an init script I got excited. For some time you could get most cards working by using ATY_init from Netkas. The meaning of this got me off my butt and digging through hex code to figure out how to sew EFIs together with PC bios. In effect, we had drivers just WAITING for cards.
All that has really changed was that both AMD & Nvidia added the same functionality into their drivers. The cards self initialize without Netkas' kext or an EFI.
We write EFIs for the cards and get even better support for them.
It is doubtful that they will pull these drivers, and even if they do it may still be possible to hack some. The 6,1 is worrisome in that there will be less reason to include drivers for cards that aren't in lineup but something tells me we will have support for awhile yet.
All that has really changed was that both AMD & Nvidia added the same functionality into their drivers. The cards self initialize without Netkas' kext or an EFI.
We write EFIs for the cards and get even better support for them.
It is doubtful that they will pull these drivers, and even if they do it may still be possible to hack some. The 6,1 is worrisome in that there will be less reason to include drivers for cards that aren't in lineup but something tells me we will have support for awhile yet.