Apple Radeon 5770/5870 in pre-2010 Mac Pro's: What We Know So Far

[Squishy Tia]

Then barefeats prove once again they are clueless. Bus limitations affect performance by 1-2%. 5870 can barely saturate 2.5gt/s. Check benchmarks.

I think it's you who need to check benchmarks. They did comparison tests of the 5770 and 5870 in both the new Mac Pros and in the 2006 models. There's a huge difference between the two rigs uaing the same cards. I know part of it is the miproved CPU to memory controller setup and a slight boost is from clock speed bumps, but there are indeed tangible differences in framerates between the two setups.

The FPS boost from the 5770 to 5870 in the same machine (1,1 Mac Pro weren't as big. There is a difference yes, but nowhere near as big as would be seen in any of the Mac Pros from the 3,1 model on up.

That doesn't mean the 5870 is useless - just that it isn't worth the price of admisssion relative to the 5770 unless you're doing something like very large scale raids in WoW.

 

[Squishy Tia]

what you mean with YMMV,i´m playing with the same settings as i did with nvidia 1900 and runs without problems,not the best graphics in the world but running...i´m testing what settings make those things thats why i ask you about YMMV

Tomorrow i hope can test it with last Modern Warfare,original and mac native

YMMV = Your Mileage May Vary. You were running games designed for a DirectX API, and WINE enabled games essentially use a wrapper for directx functions, which is why you're seeing the oddities that were in the picture you posted.

Also, just a question, out of curiosity. Is English not your native language, or are you just...young? Aside from the broken English, you're typing using a whole lot of commas (which makes things hard to read, let alone figure out), and everything looks like one gigantic run-on sentence. So far I've been able to (mostly) pick out what you "mean", but what you've posted so far might look more like rambling to many people.

If you want to test whether or not the games are having real issues with the new video card, might I suggest you boot into Windows and run the games from there and see whether or not they exhibit the same behaviour?

 

[goMac]

YMMV = Your Mileage May Vary. You were running games designed for a DirectX API, and WINE enabled games essentially use a wrapper for directx functions, which is why you're seeing the oddities that were in the picture you posted.

Actually, DirectX and OpenGL aren't different enough to really cause issues with this. Not to mention, one card dealing with OpenGL differently than another is certainly an issue with the drivers. No matter how WINE does it's magic, if it works on one card, it should work on another.

What WINE does is actually not that scary. Portions of OpenGL itself, actually, like shaders, work in a somewhat similar manner, by dynamically recompiling your code.

That said, WINE is still pretty crappy due to the CPU overheads it introduces.

 

[Major Reeves]

what you mean with YMMV,i´m playing with the same settings as i did with nvidia 1900 and runs without problems,not the best graphics in the world but running...i´m testing what settings make those things thats why i ask you about YMMV

Tomorrow i hope can test it with last Modern Warfare,original and mac native

Nvidia 1900? What the heck is that?
Ho and lol CoD4, another fail mac port.

Cheers,
"engrish" classes for everyone^^

 

[lloyd709]

I think it's you who need to check benchmarks. They did comparison tests of the 5770 and 5870 in both the new Mac Pros and in the 2006 models. There's a huge difference between the two rigs uaing the same cards. I know part of it is the miproved CPU to memory controller setup and a slight boost is from clock speed bumps, but there are indeed tangible differences in framerates between the two setups.

The FPS boost from the 5770 to 5870 in the same machine (1,1 Mac Pro weren't as big. There is a difference yes, but nowhere near as big as would be seen in any of the Mac Pros from the 3,1 model on up.

That doesn't mean the 5870 is useless - just that it isn't worth the price of admisssion relative to the 5770 unless you're doing something like very large scale raids in WoW.

I think there might be something fishy about Barefeats results.

I say this because I don't think there is a significant difference between the chip architecture of the 2008 Mac Pros (I think 3,1 - the 2.8 2 Quad core machines) and the 1,1 2006 Mac Pros and yet the 2008 Mac Pro seems to be able to use the card to it's full potential - according to Barefeats.

I've formed this opinion from some benchmarking I carried out when I was considering replacing my 2006 1,1 Mac Pro when the 2008 models were released. My tests consisted of processing batches of medium sized images (jpgs) in Photoshop (which makes little use of multiple processors). I found the 2008 Mac Pro about 5% faster - bang in line with the increase in processor speed (hence my conclusion that apart from this modest increase and the additional cores there was little difference in the the 2006 and 2008 Mac Pros).

Now, I started by saying I THINK there might be something fishy about barefeats results. I say this because IF the 5870 performs to it's full potential in the 2008 Mac Pros because it needs and makes use of the additional cores this computer has then barefeats results make sense to me. However, if it's not because of this then they don't.

It gets a little bit more complicated when you start to think about how many cores and/or processors the games might be making use of. If the games only really make use of one processor then there is a good chance that the 5870 card is hindered under on a 2006 Mac Pro because it would only be working with 2 cores - which I could believe may not be enough.

I've just upgraded the processors in my 2006 Mac Pro to Quad Cores so I may take a leap of faith soon (depending on cash flow) and buy the 5870 and see if I can get to the bottom of this.

(Please don't flame me if you see error in my thinking - these are only my thoughts that I am sharing and would like to hear what people think of them)

 

[Squishy Tia]

I think there might be something fishy about Barefeats results.

I say this because I don't think there is a significant difference between the chip architecture of the 2008 Mac Pros (I think 3,1 - the 2.8 2 Quad core machines) and the 1,1 2006 Mac Pros and yet the 2008 Mac Pro seems to be able to use the card to it's full potential - according to Barefeats.

I've formed this opinion from some benchmarking I carried out when I was considering replacing my 2006 1,1 Mac Pro when the 2008 models were released. My tests consisted of processing batches of medium sized images (jpgs) in Photoshop (which makes little use of multiple processors). I found the 2008 Mac Pro about 5% faster - bang in line with the increase in processor speed (hence my conclusion that apart from this modest increase and the additional cores there was little difference in the the 2006 and 2008 Mac Pros).

Now, I started by saying I THINK there might be something fishy about barefeats results. I say this because IF the 5870 performs to it's full potential in the 2008 Mac Pros because it needs and makes use of the additional cores this computer has then barefeats results make sense to me. However, if it's not because of this then they don't.

It gets a little bit more complicated when you start to think about how many cores and/or processors the games might be making use of. If the games only really make use of one processor then there is a good chance that the 5870 card is hindered under on a 2006 Mac Pro because it would only be working with 2 cores - which I could believe may not be enough.

I've just upgraded the processors in my 2006 Mac Pro to Quad Cores so I may take a leap of faith soon (depending on cash flow) and buy the 5870 and see if I can get to the bottom of this.

Nuumber of cores is almost certainly a non-factor. 2006 Mac Pros have a total of four cores, and no game uses all four cores -not even WoW does (it can use three cores - one each per openGL thread and one for audio).

the likely difference is that on a PCIe 1.0 slot, you can execute one instruction per cycle, whereas on a 2.0 slot you have additional lanes, and even though the card can't saturate a 1.0 slot in terms of bandwidth, it can saturate the command queue, which is indeed limited when a 2.0 slot enabled card is used in a 1.0 slot.

The Mac Pro 3,1 (2008 model) also has a newer PCIe bridge controller. The one on the 2006 Mac Pro is very limited (24 lanes total, in preconfigured slot assignments). The ones that came after the 2006 model aren't limited by this.

 

[lloyd709]

Nuumber of cores is almost certainly a non-factor. 2006 Mac Pros have a total of four cores, and no game uses all four cores -not even WoW does (it can use three cores - one each per openGL thread and one for audio).

the likely difference is that on a PCIe 1.0 slot, you can execute one instruction per cycle, whereas on a 2.0 slot you have additional lanes, and even though the card can't saturate a 1.0 slot in terms of bandwidth, it can saturate the command queue, which is indeed limited when a 2.0 slot enabled card is used in a 1.0 slot.

The Mac Pro 3,1 (2008 model) also has a newer PCIe bridge controller. The one on the 2006 Mac Pro is very limited (24 lanes total, in preconfigured slot assignments). The ones that came after the 2006 model aren't limited by this.

Interesting (and worrying) - at last some tangible sense making details on why the 5870 won't perform in 2006 Mac Pro.

 

[goMac]

I say this because I don't think there is a significant difference between the chip architecture of the 2008 Mac Pros (I think 3,1 - the 2.8 2 Quad core machines) and the 1,1 2006 Mac Pros and yet the 2008 Mac Pro seems to be able to use the card to it's full potential - according to Barefeats.

There are very significant differences, especially on the northbridge.

(Not as big as the differences between Harpertown and Nehalem, but still major differences.)

 

[Major Reeves]

Quick question.
In case you need to reinstall osX, using the original 10.6.0 install cd, will the machine launch the installer or kernel panic?

 

[goMac]

Quick question.
In case you need to reinstall osX, using the original 10.6.0 install cd, will the machine launch the installer or kernel panic?

I'm assuming that it would launch and run, just in VGA mode. That's what my machine does when I screw up my graphics drivers.

 

[Major Reeves]

I'm assuming that it would launch and run, just in VGA mode. That's what my machine does when I screw up my graphics drivers.

In some cases yes, but we have seen in the past cases when the driver wasn't present, you would get a kernel panic. Specially with nvidia cards.

 

[Redneck1089]

I can now also confirm that the 5870 works fine in a 1,1 Mac Pro 3ghz with a .B08 Boot Room. No issues with it what-so-ever! I definitely notice a speed increase in Windows games when compared to the flashed 4890 it replaced -- especially in Crysis.

 

[hoose]

Is English not your native language, or are you just...young?

I´m from Madrid and i usually don´y write in english or speak it so i forget a lot of things and comas...sorry,in spanish we use a lot of them,well,good luck,you understand my speech...i don´t have windows installed on the mac

But i now what settings make the game do those oddities...
Today i will test it with an original mac native game don´t want install windows.

Thanks.

 

[lloyd709]

I can now also confirm that the 5870 works fine in a 1,1 Mac Pro 3ghz with a .B08 Boot Room. No issues with it what-so-ever! I definitely notice a speed increase in Windows games when compared to the flashed 4890 it replaced -- especially in Crysis.

Do you have any approximate fps statistics comparisons you can quote?

Does anyone know how the 4890 compares to the 5770?

 

[yoomy]

I received my 5870 two days ago replacing my 4870 in my 08 2,8ghz Octo Core Mac Pro. Using Windows 7 I was hugely dissapointed by the pretty much non existent performance increase in COD Black Ops and Starcraft2. I measured FPS in BlackOps at a certain Position on the same map with the same settings. The 4870 gave me 39fps the 5870 40fps. My friend with a 2010 2,8ghz Mac Pro and the 5870 had 65fps at the same location with the exact same settings.

That result makes me wonder about the barefeats results where the 08 Pro was really close to the current generation with the same gcard.

Later that night my screen went dark and the new 5870 died just like that. Hopefully that card was crap to begin with. I will be receiving my replacement card today. I have also used Futuremark with my current 4870 today so I can compare results a little bit better. Will update later today.

 

[Major Reeves]

I received my 5870 two days ago replacing my 4870 in my 08 2,8ghz Octo Core Mac Pro. Using Windows 7 I was hugely dissapointed by the pretty much non existent performance increase in COD Black Ops and Starcraft2. I measured FPS in BlackOps at a certain Position on the same map with the same settings. The 4870 gave me 39fps the 5870 40fps. My friend with a 2010 2,8ghz Mac Pro and the 5870 had 65fps at the same location with the exact same settings.

That result makes me wonder about the barefeats results where the 08 Pro was really close to the current generation with the same gcard.

Later that night my screen went dark and the new 5870 died just like that. Hopefully that card was crap to begin with. I will be receiving my replacement card today. I have also used Futuremark with my current 4870 today so I can compare results a little bit better. Will update later today.

Your card probably had a bad voltage regulator.

 

[AnimeFunTv]

I can now also confirm that the 5870 works fine in a 1,1 Mac Pro 3ghz with a .B08 Boot Room. No issues with it what-so-ever! I definitely notice a speed increase in Windows games when compared to the flashed 4890 it replaced -- especially in Crysis.

Thank You! And in the nick of time, my flashed 4890 dosent seem to work well at ALL with the new 10.6.5 update even with the QE patch, this has been an ussue with the card ever since 10.6.3. Is there anyway we could update the front page about this?

 

[yoomy]

So the replacement card arrived. It beats the 4870 with 15000 to 9000 in futuremark but in COD BO my friend with 2010 2,8 quad core and same card still beats me by 20 fps

 

[peskaa]

So the replacement card arrived. It beats the 4870 with 15000 to 9000 in futuremark but in COD BO my friend with 2010 2,8 quad core and same card still beats me by 20 fps

CODBLOPS is rather CPU dependent. The architecture difference between 2008 and 2009/10 machines is quite pronounced - particularly with TurboBoost.

 

[lloyd709]

So the replacement card arrived. It beats the 4870 with 15000 to 9000 in futuremark but in COD BO my friend with 2010 2,8 quad core and same card still beats me by 20 fps

If you are using Windows Home Premium then only one of your processors will be being used - you have to have Windows Professional before it will use more than one processor (not to confuse this with cores). I'm not sure whether COD would use more than one processor anyway but I would like to find out. If you are using Windows Professional (or higher) you can check whether the second processor is being utilised using Resource Monitor. It would be interesting to find out!

 

[lloyd709]

Nuumber of cores is almost certainly a non-factor. 2006 Mac Pros have a total of four cores, and no game uses all four cores -not even WoW does (it can use three cores - one each per openGL thread and one for audio).

the likely difference is that on a PCIe 1.0 slot, you can execute one instruction per cycle, whereas on a 2.0 slot you have additional lanes, and even though the card can't saturate a 1.0 slot in terms of bandwidth, it can saturate the command queue, which is indeed limited when a 2.0 slot enabled card is used in a 1.0 slot.

The Mac Pro 3,1 (2008 model) also has a newer PCIe bridge controller. The one on the 2006 Mac Pro is very limited (24 lanes total, in preconfigured slot assignments). The ones that came after the 2006 model aren't limited by this.

Would running two 5770 in crossfire get around this bottleneck?

 

[AnimeFunTv]

Now that we know the 5770 and 5890 work on 1,1 MP's I'm trying to look at some hard data on which one would be better, I would move that the 5890 is better, but looking at some of the posts, its only 'slightly' better on gaming. For me, if the fames per sec. on the 5770 & 5890 are a diffrence of one or two fps then thats not a real deal. My bigger question is how much of a diffrence when it comes down to Apps like Premiere and Final Cut video rendering.

 

[Mac Hammer Fan]

It's a pity that the drivers of the 5870 are so weak. Cinebench under Windows gives much better results than under MacOS X. I think that the 5870 is only a better buy if you are a gamer.

 

[rouxeny]

5870 arrived early.

Just installed, took about 20 minutes, by somebody who is only hasn't installed a card in a computer for a long while. The hardest part for me was figuring out how to take out the old card.

I have an Early 2008 2x2.8 Quad-Core, running 16 GB of memory, and previously had a NVidia 8800GT. I'm driving a 30" Apple monitor.

So far, Aperture 3 seems a bit snappier, photos are definitely faster to load. When on Fullscreen, the time it takes for the next photo to "snap" into focus is a bit shorter.

Tried Starcraft 3, and it's a lot better. I now have "Ultra" graphics settings and am hitting fps of 20-39 where prior I was on "Medium" at best with noticeable delays.

Finally, the computer seems marginally louder.

 

[goMac]

Just got my 5870 today.... Early 2008 octo 2.8, replacing an 8800 GT.

It seems much quieter than the 8800. I haven't heard any ramp ups of the fan like I used to, and the initial ramp up on power on is much much quieter...

The performance delta isn't what I thought it would be. On SC2 on ultra settings with high textures on my 8800 it typically ran at about 45 fps. On the 5870, on ultra settings across the board, I'm usually around 30. I know I upped the settings and that would cause a slow down, but I expected the speed of the card to make up for it.

Could it be that ATI needs driver fixes?

Edit: Huh, after a relaunch, I'm getting 62 fps on Zerg, a huge improvement over the 32 I used to get.