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.
Apple Radeon 5770/5870 in pre-2010 Mac Pro's: What We Know So Far
- Thread starter DualShock
- Start date
- Sort by reaction score
Check your activity monitor more closely - it's using all of the cores, but not simultaneously. The OS is offloading to different cores at different times to even out the CPU load so things don't grind to a halt. OS X does this as well.
Black Ops does actually seem to be a multicore.
Most console ports these days are multicore in since the consoles are multicore.
I never said it wasn't multicore. However, you may want to check your CPU% usage in the activity monitor. I can pretty much guarantee it's not using more than three cores simultaneously, and if if even does that (2 threads for video, one for sound), it won't use them at max.
Don't assume a console port being multicore means the Mac/PC port is as well though. Coding for the processors on a console is a crapton different than the coding for multicore on a Mac/PC.
Oh, and it can't use every core as it were anyway without locking up any control input. Using all of the cores would effectively turn the game into a pretty (expensive) renderer.
For PS3 yes. For XBox 360, it's exactly the same as coding for multicore on a Mac/PC. That's why most PC/Mac ports use the XB360 version of a game as the basis and not the PS3 version.
Actually it isn't quite the same coding for mac/pc. The Xbox 360 uses a PowerPC architecture CPU, much like the G5 only more scaleable. So unless the Mac you're referring to is a 5+ year old Mac G5, it is nowhere near a straight multicore port.
Multicore programming doesn't depend hardly at all on the CPU architecture. It's always abstracted away by libraries, whether you are working on Windows, Mac, or the XBox. I know for a fact that the XBox 360 codebase is used more often for ports because the threading paradigm is the same. Even without the libraries, threading is still pretty much the same.
The PS3 is the exception of course.
(I do research into multicore and GPU programing. Also, for reference, when I moved my threading code from PowerPC to Intel Mac OS X back in the day, I didn't have to change a single line of code.)
Does anyone know if running 2 5770 cards in crossfire would get around the above problem - and hence perform significantly better than a single 5870 card in a 2006 Mac Pro?
Seems like I'm going to have to find out for myself - just pulled the trigger on a second 5770 to stick next to my current one in my 1,1 Mac pro (upgraded though with 2 quad core 2.66 processors).
Will let you know the results in a few days.
Definitely looking forward to your results. Benchmarks comparing one 5770 and two 5770s CF in Windows would be excellent.
Just bear in mind that you'll need to set the bandwidth of the two slots to 8x using the Expansion Slot Utility app. But you probably already know that.
Be aware that the second 5770 will run in x8 mode, not x16, as there are a limited number of lanes available. And if CrossFireX works only symetrically in Windows (this I'm unsure of), both cards would be running at x8 as well instead of x16+x8 like they would in OS X.
It'll be interesting indeed to see what Crossfire does on a 1,1 Mac Pro's limited PCIe bus.
Also, does anybody have MDP audio working on 1,1 Mac Pros yet? This is irritating me to no end. It works perfectly in Windows, even with Blu-Ray discs played back in Cyberlink's PowerDVD Ultra, but I cannot get it initialized in OS X because the requisite files are not present, and I have no idea which files and/or versions of said files are required.
Please note I said it works in Windows. On my Mac Pro. It is not an EFI limitation. The files needed in OS X are missing, and presumably right now only the 2010 Mac Pro install discs have them. I'm just trying to find out what files control the audio output functionality.
Windows doesn't run under EFI, which is why I suggested it.
It's possible that audio needs to EFI handshake, which could be missing on the Mac Pro 2009 and lower firmware.
As of 10.6.5, the 2010 Mac Pros and the 2009 Mac Pros use the same files. That's why you can installer from scratch on a Mac Pro using 10.6.5.
I have a 2009 Mac Pro 8 Core.
Installed an Apple 5870 last night.
I'm using a 1920X1080 display on the DVI port.
Worked last night in 64 bit kernel mode. Set it to 32 bit kernel and shut down.
This morning there was no video when i booted. Monitor just had a black screen.
Hit the power button trying to reboot, but it just went to sleep.
Hit the power button again, it woke up and had video.
I don't know if this will be a recurring issue or not.
Installed an Apple 5870 last night.
I'm using a 1920X1080 display on the DVI port.
Worked last night in 64 bit kernel mode. Set it to 32 bit kernel and shut down.
This morning there was no video when i booted. Monitor just had a black screen.
Hit the power button trying to reboot, but it just went to sleep.
Hit the power button again, it woke up and had video.
I don't know if this will be a recurring issue or not.
I just got my 5870 installed in my 2008 Mac Pro earlier today. Here's what I have so far:
The card is installed along with a HD 2600 XT which is in slot 2. I drive four displays.
1) 30" Cinema on DVI of 5870
2) 15" Dell on MDP > VGA of 5870
3 & 4) 15" Dell on DVI > VGA of 2600
I am regularly having to unplug the MDP display and back in again after a reboot or cold boot, as it isn't detected otherwise. If I go to Detect Displays it lights up though.
Oddly, when it's blanked, my cursor still "goes to it" between the 30" and the next display to its right.
This happens in Windows too. Here's my big issue though, which I also had in my 2.4 GHz 17" MacBook Pro (which I no longer own).
The 30" has flashing intermittent blue lines on it only on Windows, and only when the MDP display is plugged in.
On my MBP the blue lines (which look exactly the same) were on the 30" all the time.
30" (DVI) + MDP = Blue Lines
30" by itself = No blue lines.
Two 15" displays, one on DVI > VGA and the other on MDP > VGA, no blue lines.
So it's only when the dual link DVI is being used it seems, and only when the MDP is being used along with it. And it's only on the Windows side, which is nice. On my MBP it was in Mac and Windows.
The card is installed along with a HD 2600 XT which is in slot 2. I drive four displays.
1) 30" Cinema on DVI of 5870
2) 15" Dell on MDP > VGA of 5870
3 & 4) 15" Dell on DVI > VGA of 2600
I am regularly having to unplug the MDP display and back in again after a reboot or cold boot, as it isn't detected otherwise. If I go to Detect Displays it lights up though.
Oddly, when it's blanked, my cursor still "goes to it" between the 30" and the next display to its right.
This happens in Windows too. Here's my big issue though, which I also had in my 2.4 GHz 17" MacBook Pro (which I no longer own).
The 30" has flashing intermittent blue lines on it only on Windows, and only when the MDP display is plugged in.
On my MBP the blue lines (which look exactly the same) were on the 30" all the time.
30" (DVI) + MDP = Blue Lines
30" by itself = No blue lines.
Two 15" displays, one on DVI > VGA and the other on MDP > VGA, no blue lines.
So it's only when the dual link DVI is being used it seems, and only when the MDP is being used along with it. And it's only on the Windows side, which is nice. On my MBP it was in Mac and Windows.
I'm assuming you're using a Y-splitter for one of the PCIe power connectors? Unless I'm mistaken, those two cards require a total of three connectors between them.
Hrm. I wonder how they managed to not need a power connector for that particular card when both the 1900XT and 3870 require one. Was the 2600 really that pathetic a card? ^_^
Initial results of 5770 v 5770 crossfire v 5870 in a 1,1 mac pro!!!
To further the knowledge of mankind (in particular 1,1 owners!) I took the plunge and bought a 5870 in addition to a second 5770 (hoping to send one or two of them back)!!!
Anyway, I started out by setting up a crossfire with the two 5770s. Setting it up wasn't a problem using ATI CCC (ATI driver software) which gave me confidence in seeing some good results - it said it was all working I would see great gains!!
However, I didn't see any improvement what so ever with the games I tested it on: CoD Modern Warfare 2, GTA, F1 2010 and Splinter Cell Conviction. It might be that these don't support crossfire but research on the web tells me at least CoD and GTA do. I'm guessing but I think the no performance increase is due to having to run both cards at 8 rather than 16 - but I'm open to suggestions. I was only using one crossfire bridge - there seems to be space for two but I understand a second bridge isn't necessary.
Now for the good news. The 5870 flys!!! On average, on all the above games it gave me between 50% and 100% improvement. It was hard to tell exactly because the fps moves around a lot on most of them but it's definitely a big jump. Now, I'm still not sure if Barefeats only tested their 5870 on OSx (which I haven't done - all my results are in Windows 7) but this goes against their findings and conclusion that 1,1 owners should run with the 5770 - I believe this is roughly what they said.
The only caveat is I'm running these tests on an upgraded 1,1 with 2 quad core 2.66 processors instead of the duel cores in the standard model. I'm not sure but I don't think this is going to make that much difference - especially not on all the games tested.
I'm not planning to set up the cross fire again unless anyone can give me a reason why - did I do anything wrong? However, it would be good to hear suggestions on other tests I could run between the 5870 and a single 5770.
Anyone want to buy a 5770 - I'm sticking with the 5870!!!!! I've now got the upgraded 1,1 from hell!!!!!!
EDIT
I've now completed some more controlled testing and with the exception of GTA and Splinter Cell Conviction the gains of the 5870 over the 5770 are more modest than I first thought I'm afraid. In Windows With GTA it really does improve the performance by a factor close to 2 - with everything set to high running down a particular street I get about 22 fps with the 5770 but about 40 with the 5870. Similarly with Splinter Cell the fps goes from 20 to about 38. With all the other games the performance gains are closer to 10 to 20 percent. It's interesting though because these are the only two games that I feel push the graphics (all the other games work perfectly on both cards with modest gains in fps given by the 5870).
Also, the 5870 gives little performance increase with X Plane in OSx (in line with Barefeats findings).
I'm now not so sure whether to keep the 5870 or not - it does at least make GTA playable!
To further the knowledge of mankind (in particular 1,1 owners!) I took the plunge and bought a 5870 in addition to a second 5770 (hoping to send one or two of them back)!!!
Anyway, I started out by setting up a crossfire with the two 5770s. Setting it up wasn't a problem using ATI CCC (ATI driver software) which gave me confidence in seeing some good results - it said it was all working I would see great gains!!
However, I didn't see any improvement what so ever with the games I tested it on: CoD Modern Warfare 2, GTA, F1 2010 and Splinter Cell Conviction. It might be that these don't support crossfire but research on the web tells me at least CoD and GTA do. I'm guessing but I think the no performance increase is due to having to run both cards at 8 rather than 16 - but I'm open to suggestions. I was only using one crossfire bridge - there seems to be space for two but I understand a second bridge isn't necessary.
Now for the good news. The 5870 flys!!! On average, on all the above games it gave me between 50% and 100% improvement. It was hard to tell exactly because the fps moves around a lot on most of them but it's definitely a big jump. Now, I'm still not sure if Barefeats only tested their 5870 on OSx (which I haven't done - all my results are in Windows 7) but this goes against their findings and conclusion that 1,1 owners should run with the 5770 - I believe this is roughly what they said.
The only caveat is I'm running these tests on an upgraded 1,1 with 2 quad core 2.66 processors instead of the duel cores in the standard model. I'm not sure but I don't think this is going to make that much difference - especially not on all the games tested.
I'm not planning to set up the cross fire again unless anyone can give me a reason why - did I do anything wrong? However, it would be good to hear suggestions on other tests I could run between the 5870 and a single 5770.
Anyone want to buy a 5770 - I'm sticking with the 5870!!!!! I've now got the upgraded 1,1 from hell!!!!!!
EDIT
I've now completed some more controlled testing and with the exception of GTA and Splinter Cell Conviction the gains of the 5870 over the 5770 are more modest than I first thought I'm afraid. In Windows With GTA it really does improve the performance by a factor close to 2 - with everything set to high running down a particular street I get about 22 fps with the 5770 but about 40 with the 5870. Similarly with Splinter Cell the fps goes from 20 to about 38. With all the other games the performance gains are closer to 10 to 20 percent. It's interesting though because these are the only two games that I feel push the graphics (all the other games work perfectly on both cards with modest gains in fps given by the 5870).
Also, the 5870 gives little performance increase with X Plane in OSx (in line with Barefeats findings).
I'm now not so sure whether to keep the 5870 or not - it does at least make GTA playable!
Last edited:
Barefeats's results were under OS X, not Windows 7. So your findings (while nice) don't contradict him yet.
Look at this: 5770/5870 in 2006 Mac Pro.
I'll do some more controlled tests and give some stats either later today or over the weekend - both in OSX and Windows.