Sidebar Sidebar
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

Confirmed AMD Radeon HD 7970 GHz Edition Running on Mac Pro

hehe, what a funky thread...

revelstudios said:
ClassObject said:
I think the purpose of a forum such as this is exactly what the OP did - give real world experience...
Thanks for the info revelstudios. Most of us are intelligent enough to appreciate your post for what it is.
Click to expand...

Thanks for the nod, Class. The worst part of this entire debacle here; I base my business and the livelihood of my family and the people that work for us around this kind of technology. I'm not flaunting any of this stuff to run benchmarks.
Click to expand...

Yeah, but keep in mind most Apple users (MacPro users too!!!) are an audience of folks who wish others to do the thinking for them. It's not a put-down it's just a stereotype. Most folks here don't even know what kind of RAM their machines accept and will prefer to purchase (follow) 2x or 3x priced "kits" from sites (leaders) who stamp "Apple approved" next to the listing. It's pretty sad IMO but OTOH I guess it enables those who can think (and research) for themselves to make a killing and live at a higher standard than those who wanna be led or can't figure out how to research something.

And while it really is "most" there are still a few reading here (usually the silent types) not fully under the Apple spell, who remain independent thinkings retaining their common sense. So for us, your posts are very appreciated. Please continue and ignore the spell-bound followers if possible. We'll get it. ;)

Thanks for your post and your efforts!
 
Last edited:
barbaresque said:
Tank you for the reply Daniel.m
I couldn't find a 7970 with two 6 pin connectors but I found many 7950 with this configuration.
If I buy a 7950, what kind of cable do I need to buy and how to plug them in the
Mac Pro's Power Supply ?

Best regards
Click to expand...

If you look in the expansion card bay, you will see two chunky 6-pin sockets directly mounted on the back board. These are the sockets that you need to power the graphics card from.

If you search for "mac pro 6-pin cable" here on the forum, you should find some threads that give links to the exact cables you need. I think they are not quite the same as Windows PC 6-pin cables so you need to make sure they are the Mac Pro ones.

Maybe someone could re-link the appropriate cables for barbaresque?
 
Thank you very much guys.
I will try to find them in Paris (i asked the two Parisian Apple Store and i they didn't want to sell them).
 
Real World Data

Hello everyone. I just read through this entire thread, then spent a couple hours doing first hard experiments on my hardware and I have information to share.

First a bit about me. I am a certified Apple technician, and video cards have been my hobby for 20 years. I also currently design, market, and sell a brand of low voltage cabling. I have been Apple trained, I understand low voltage, and it’s my hobby. For these 3 reasons I think I am in a unique position to share my findings.

My Rig:
2010 Mac Pro with Dual 30in Cinema Displays
4 x Intel SSD Array, 16GB Ram, Bluray burner
Current Video Cards: Genuine Apple 5870 and 5770

Everyone here is arguing about power consumption and nobody has any data to back it up, right? Right. So I pulled out my "Kill-A-Watt" power meter and did some base line testing between my OEM 5870 and the new OVER CLOCKED 7970 MSI card I just bought. I am not a "pro" user in the traditional sense of the word, I use this computer for gaming (amongst other things). Here are the findings:

Total Mac Pro Power Usage w/ Only Genuine Apple 5770
Idle Power Usage - 106w
In Game (Starcraft II) - 189w
Peak Wattage - 214w

Total Mac Pro Power Usage w/ Genuine Apple ATI 5870 + OEM 5770
Idle Power Usage - 137w
In Game (Starcraft II) - 245w
Peak Wattage - 281w

Total Mac Pro Power Usage w/ MSI R7970 Over Clocked to 1010 mhz + OEM 5770
Idle Power Usage - 150w
In Game (Starcraft II) - 245w
Peak Wattage - 311w

*Keep in mind, that is the power draw FOR THE ENTIRE COMPUTER with 2 video cards, 4 hard drives, etc. Game running with extreme graphics at 2560x1600

Versus a stock 5770 Mac Pro, here is what adding the video card will do to your power usage:
At Idle –
5870 +31 Watts (over single 5770)
7970 +44 Watts (over single 5770)
Delta +13 Watts (Between Apple Approved 5870 vs Overclocked 7970)

In Game
5870 +56 Watts(over single 5770)
7970 +56 Watts(over single 5770)
Delta = N/A

Peak Power In Game
5870 +67 Watts(over single 5770)
7970 +97 Watts(over single 5770)
Delta +30 Watts (Between Apple Approved 5870 vs Overclocked 7970)

As you can see, there are lots of conclusions you can draw from this real world data.
1 – A 7970 only uses an additional 97 watts of power during peak gaming. A standard 6 pin supplies 75 watts. One 6-pin plug alone is nearly enough to power the entire card. 2 x 75w is PLENTY.

2 –If you chose the 7970, under the worst conditions I could find it’s only a 30w difference under peak load versus the Apple card which I only experienced for brief flashes. 30 watts is nothing.

3 – Based on real world power consumption, the 8 pin connector is hardly necessary. I can only assume its for extreme over-clockers. It also explains why MSI supplies the adaptor, because they are safe in saying there is MORE than enough power on a 6-pin, but allows enthusiasts the option of 8-pin if they need it.

4 – I used an over-clocked card in my comparison, so a regular card would be even safer.

I also fired up Diablo and experienced similar power usage numbers. Of course, I have no pro-software to test but can’t imagine the results being radically different. In my opinion, it’s more than safe to put this one to bed AT LEAST for gaming.

I even got out my infra-red thermometer, and measured the temperature on the cable, 6 pin connectors, and on the logic board itself. It was 85 degrees after heavy gaming. Both plugs were the same temperature even though they had different voltage running through them. There was absolutely nothing to point to the board overheating. This really is a non-issue.

To conclude, running a 7970, even lightly over-clocked, even with a 2nd video card and lots of other accessories should nowhere approach the limits the machine can safely handle. I think this data pretty much shuts down any possible argument, but I would love to hear anyone try and poke holes in my research.
 
Last edited:
  • Like
Reactions: Zwhaler
100Daily said:
I think this data pretty much shuts down any possible argument
Click to expand...
Not so fast ;)

http://www.techpowerup.com/reviews/AMD/HD_7970_GHz_Edition/26.html

Power draw measured from the slot and PCIe power connector. Most credible IMO. Blue - regular 7970, green - OC'ed to 1050 MHz.
Your is running @1010 so it should be somewhere between regular and GHz Edition.
Reads you have on Kill A Watt could mean:
a) Kill A Watt is crap
b) Starcraft II does not tax GPU to its last juice (or it's CPU bound)
c) OS X drivers cannot squeeze full potential from the card
 
666sheep said:
b) Starcraft II does not tax GPU to its last juice (or it's CPU bound)
Click to expand...

SC2 does not stress a mid-range or high-end GPU, no. A high-end GPU is very likely to be idle for 50% or more of the time. Certainly not a good test of GPU power draw.

I agree that something like FurMark or Unigine Heaven would be a much better stress test.
 
Yeah, FurMark stresses my cards more than any other. It's probably the best to use. Small and free to DL too.

Oh, and 100Daily, thanks for your efforts so far! Very cool!

Oh, oh, and the Kill-A-Watt while not industrial strength equipment should be perfectly fine and accurate for exactly these kinds of measurements.
 
Last edited:
Hello again! I am back again from making my $4k Mac Pro into a test gerbil. Why would I do such things? Confidence.

Ok, there were some good points made and what I considered some not-so-good points. First off, Kill-A-Watts are actually fairly accurate in this voltage range. Do some reading on it before disparaging it. Second point, Starcraft is nowhere near CPU bound. It actually barely taxes my Mac Pro's CPU, but it does get the GPU fan going.

However, I do agree that Starcraft or Diablo are not the be-all, end-all tests. I mainly put that up there for fellow gamers. But since you guys wanted me to do a FurMark test I decided I would help out. I did 8x AA. I got about 42 FPS. As you guys thought, it did tax my card quite a bit more. It started off with about 360w then worked its way up until it I hit a peak of 380w and it was steady at about 370w after letting it run for 20 mins. It wasn't getting any hotter or using any more power, and the fan was going crazy so I figured 20 mins was enough for my test. This is quite a bit more than Starcraft, so you guys were smart to suggest that.

But lets keep a few more things in mind:
1 – The Kill-A-Watt is measuring the load from the PSU. PSU's are not 100% efficient. If you want to calculate the true load of the components within the system, you have to adjust for that. I have read from 20% to 30% depending on the efficiency of the PSU is pure loss. So that 380w is actually closer to 304w in actual system wide load, on the most efficient PSU, in the harshest conditions you guys collectively could think of.

2- I got out my infrared thermometer once again, and the board and connector were once again 85 degrees. Both connectors were the same temp, even though they have differing voltages which indicates to me there is no increase in resistance (heat).

3- Its still only 380w at most, during an intentionally harsh test. From my earlier base line, an idle 5770 system uses 106w. I got it up to 380w, an increase of 274w. Sure some of that is certainly CPU but lets assume its ALL GPU. Given an efficient PSU that puts the GPU increase at (274*.8) = 220w in the worst scenario. Again, the card is receiving 75w from the board, and 75w x 2 from the power cables. 225w is more than 220w. The card has enough power and is within spec, even at peak times, even assuming worst case no CPU usage.

Final Point – I think its especially true of mac people that they are the most overly cautious people on the planet earth. Not be a hater, but I think Jesus could appear and tell some people something was ok, but if it doesn’t have an Apple logo on it, they are scared. Sometimes that is a good thing, but then I realize there are LOTS of just scared retarded Apple owners (present company excluded). I am a former tech, after all. But, then there are rational people. Lets look at the facts. You have multiple people stress testing this. You have reliable usage in actual power macs in the real world. You have data that shows its well within tolerances even when intentionally pushed. You have thermo readings that corroborate all of the other information. I rest my case.

Which begs the ultimate question... Why WAIT to buy that 7950 Mac addition, that costs MORE money, and is SLOWER?
 
Last edited:
  • Like
Reactions: Awediohead
100Daily said:
Second point, Starcraft is nowhere near CPU bound. It actually barely taxes my Mac Pro's CPU, but it does get the GPU fan going.
Click to expand...

Define "tax" exactly? SC2 should be running at least one CPU core at 100% all the time if you're using a high-end GPU. You need to view CPU usage on a per-core basis, such as with Activity Monitor's "CPU Usage" or "CPU History" windows. If you have 4 cores with Hyperthreading, then pegging one core will only show up as 12.5% "total" CPU usage across all cores, but that doesn't change the fact that the app is CPU bound.

Edit: And it's worth pointing out that this can still show up as 50% across two cores or 25% across four cores, as the OS can and will move the application around across different cores.
 
100Daily said:
3- Its still only 380w at most, during an intentionally harsh test. From my earlier base line, an idle 5770 system uses 106w. I got it up to 380w, an increase of 274w. Sure some of that is certainly CPU but lets assume its ALL GPU. Given an efficient PSU that puts the GPU increase at (274*.8) = 220w in the worst scenario. Again, the card is receiving 75w from the board, and 75w x 2 from the power cables. 225w is more than 220w. The card has enough power and is within spec, even at peak times, even assuming worst case no CPU usage.
Click to expand...

Are we really going to have to have this discussion again? Nobody is suggesting the power supply itself can't drive the card. The 5770 has a TDP that is less than or equal to 225W and can safely run in the Mac Pro. The real question is how much power a card with one or more 8-pin connectors will draw, e.g. the 7970.

How much power are you measuring when you run FurMark on the 7970 with everything cranked up? That card is specced for more than 225W, and that's the crux of this discussion. Specifically, whether the traces and connectors on the motherboard that drive the auxiliary 6-pin power cables can handle loads above 225W.
 
Asgorath said:
Are we really going to have to have this discussion again? Nobody is suggesting the power supply itself can't drive the card. The 5770 has a TDP that is less than or equal to 225W and can safely run in the Mac Pro. The real question is how much power a card with one or more 8-pin connectors will draw, e.g. the 7970.

How much power are you measuring when you run FurMark on the 7970 with everything cranked up? That card is specced for more than 225W, and that's the crux of this discussion. Specifically, whether the traces and connectors on the motherboard that drive the auxiliary 6-pin power cables can handle loads above 225W.
Click to expand...

I did run Furmark and posted the full results when cranked up... Smh... And I am not making any suggestion about the ability of the PSU to power the card either. Smh again.... You totally misunderstood what is going on here. And I just told you, in your quote, how much power the card was running with Furmark cracked up on the 7970. 220w. Smh a third time. Please re-read.

But as was pointed out earlier, the Mac Pro PSU is actually 85% efficient which means the 7970 peak usage actually came in at 232 watts (274w * .85). But again, some of that wattage increase was due to CPU power increases (that I could not isolate/subtract out) so that 232 watt PEAK/HARD CORE usage is certainly a bit lower in reality.
 
Last edited:
100Daily said:
Again, the card is receiving 75w from the board, and 75w x 2 from the power cables.
Click to expand...

I believe this setup is probably okay, but you are making a big assumption here that the card draws evenly from all three sources. Someone indicated in the longer thread where this all was discussed before that this is a fallacy. Board designs show that there are different regions of the board powered from the different connectors, therefore they can draw different amounts of power.

It makes sense, too. If one connector was 6-pin 75W and one was 8-pin 150W, why would the card draw evenly from them? That would be disaster for the 6-pin connector unless they never drew more than 75W evenly, in which case, why have an 8-pin at all?

To provide an example that exaggerates the problem, it's entirely possible that the PCIe is only taking 50W, and the 6-pin 50W, and then the 8-pin might be drawing 120W (after all, that port is rated to 150W on the card). In that scenario, the traces supporting the second PCIe power connector are at risk.

Again, I believe your setup is probably okay, but nevertheless I think your assumption on that one point (even power draw) is wrong.
 
ActionableMango said:
I believe this setup is probably okay, but you are making a big assumption here that the card draws evenly from all three sources. Someone indicated in the longer thread where this all was discussed before that this is a fallacy. Board designs show that there are different regions of the board powered from the different connectors, therefore they can draw different amounts of power.

It makes sense, too. If one connector was 6-pin 75W and one was 8-pin 150W, why would the card draw evenly from them? That would be disaster for the 6-pin connector unless they never drew more than 75W evenly, in which case, why have an 8-pin at all?

To provide an example that exaggerates the problem, it's entirely possible that the PCIe is only taking 50W, and the 6-pin 50W, and then the 8-pin might be drawing 120W (after all, that port is rated to 150W on the card). In that scenario, the traces supporting the second PCIe power connector are at risk.

Again, I believe your setup is probably okay, but nevertheless I think your assumption on that one point (even power draw) is wrong.
Click to expand...

Which is exactly why I measured the temperature of the traces on the board. To corroborate ALL of the other data, which it did. They were both 85 degrees. They were the exact same temp whether I was pushing the card, or at idle. Plus each port/trace was the same temp under load, but as you point out were probably carrying differing currents. To me that means that neither path was under under any stress.

BTW, to answer your question (it was posted earlier)the 8 pin has different grounds which the card detects to adjust its power consumption. The card won't just start pulling 150w... But good question for those that didn't read through all the supporting links in the thread.

I totally agree with the theory that different ports carry different currents, but regardless it wasn't creating ANY ADDITIONAL HEAT. Not even 1 additional degree to the board in the stress test. Let alone the kind of heat required to burn up the pathway on a board.
 
Last edited:
Not final answer

Something isn't quite right here.

If I put the 7970 I have in my 3,1 and boot it into 10.8.3 it works fine with 6 to 8 pin adapter.

But if I try to run "Furmark" on fullscreen on 27" Apple LED it will SHUT THE COMPUTER DOWN EVERY TIME.

Loud click, SMC has to be reset to boot again. To be fair, a full speed GTX580 will do the same thing. But in 2nd slot it runs for a few seconds, in slot 1 it will literally shut down as soon as the test starts running.

The 3,1 has been upgraded to 3.2 Ghz Dual CPUs, but has only 2 drives and no other PCIE cards.

On my 4,1/5,1 so far it hasn't choked the machine yet, but if I can smother the 3,1 with this 7970, then others can as well.

Were you running Furmark with just it's tiny little default window or running it in full screen mode?
 
MacVidCards said:
Something isn't quite right here.

If I put the 7970 I have in my 3,1 and boot it into 10.8.3 it works fine with 6 to 8 pin adapter.

But if I try to run "Furmark" on fullscreen on 27" Apple LED it will SHUT THE COMPUTER DOWN EVERY TIME.

Loud click, SMC has to be reset to boot again. To be fair, a full speed GTX580 will do the same thing. But in 2nd slot it runs for a few seconds, in slot 1 it will literally shut down as soon as the test starts running.

The 3,1 has been upgraded to 3.2 Ghz Dual CPUs, but has only 2 drives and no other PCIE cards.

On my 4,1/5,1 so far it hasn't choked the machine yet, but if I can smother the 3,1 with this 7970, then others can as well.

Were you running Furmark with just it's tiny little default window or running it in full screen mode?
Click to expand...

I ran FurMark 2 different ways. I ran it in OSX in the windowed screen. I tried to get it full screen but the software wouldn't cooperate.

Then I switched over to Windows via Boot Camp and ran it in like 20 different ways in Windows. I cranked it up to 2560x1600 at 8x AA and let her rip. Same exact 380w. I expected more current draw at full res, but it was actually the exact same as the Mac OS.

Mac 3,1 isn't on the approved list. Good to know.
 
100Daily said:
I ran FurMark 2 different ways. I ran it in OSX in the windowed screen. I tried to get it full screen but the software wouldn't cooperate.

Then I switched over to Windows via Boot Camp and ran it in like 20 different ways in Windows. I cranked it up to 2560x1600 at 8x AA and let her rip. Same exact 380w. I expected more current draw at full res, but it was actually the exact same as the Mac OS.

Mac 3,1 isn't on the approved list. Good to know.
Click to expand...

I'm confused, too, but I may have misread your post. Why would my machine w/7970 freeze in less than a minute after booting vs your being able to run stress tests? I did have the card connected to 3 27" Apple displays but simply booted into Windows and let it sit. This was a fresh W7 install so nothing crazy was being loaded during startup.

I had what may be the same card MSI 7970 connected via 1 6p and 1 6p-8p connector in my 2010 MP. I had no other video or PCI cards in the machine.
http://www.amazon.com/gp/product/B007NG3WQI/ref=oh_details_o04_s00_i00?ie=UTF8&psc=1
 
Last edited:
As an Amazon Associate, MacRumors earns a commission from qualifying purchases made through links in this post.
jenzjen said:
I'm confused, too, but I may have misread your post. Why would my machine w/7970 freeze in less than a minute after booting vs your being able to run stress tests? I did have the card connected to 3 27" Apple displays but simply booted into Windows and let it sit. This was a fresh W7 install so nothing crazy was being loaded during startup.

I had what may be the same card MSI 7970 connected via 1 6p and 1 6p-8p connector in my 2010 MP. I had no other video or PCI cards in the machine.
http://www.amazon.com/gp/product/B007NG3WQI/ref=oh_details_o04_s00_i00?ie=UTF8&psc=1
Click to expand...

I don't know what to tell you. I actually just installed boot camp for the first time as well. So it was a fresh Windows 7 install. For what its worth I did some Start Craft benchmarking as well...

30" Apple Cinema 2560x1600
All Settings on "Extreme"
5770 (OSX) - 10-14fps
5870 (OSX) - 16-26fps
7970 (OSX) - 16-26fps

Both cards were literally frame by frame identical. But when I flipped over to Windows it was a whole different story. I only tested the 7970. Same settings

7970 (Windows 7) - 65-160fps

I havent had a PC for gaming since probably 2003, so its been 10 years since I really did a comparison like this if I am honest. But I was pretty shocked how terrible the OSX drivers are. I mean seriously. I knew they would be bad, but I had no idea they would be THIS bad.
 
As an Amazon Associate, MacRumors earns a commission from qualifying purchases made through links in this post.
100Daily said:
I just ordered the Mac Edition from New Egg. Next Day Air. :D

http://www.youtube.com/watch?v=5ep08vaCUnE
Click to expand...

I am really curious what the OSX driver CD is about. I always thought as the OSX containing the driver, but apparently there is a 2nd set of drivers. I will certainly do a before and after benchmark on the software.

Updated: Wow, just like that New Egg SOLD OUT! Ka kaa!
 
Last edited:
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back