Radeon HD 5870: How much faster than the 5770 in 2010 Mac Pro? [Archive]

barefeats

Aug 28, 2010, 06:27 PM

mattmower

Aug 28, 2010, 06:45 PM

Pressure

Aug 28, 2010, 06:48 PM

This is what I'm working on at the moment. I've already posted results for the Radeon HD 5870 in the 6-core Westmere. I've now got a Radeon HD 5770 the same Mac Pro running my gamut of games.

Here's a teaser while I'm working.
World of Warcraft at 2560x1600 Ultra with 4x Multisampling.
5870 = 110 fps
5770 = 61 fps
(5870 is 80.3% faster)

At 1920x1200 Ultra with 4x Multisampling.
5870 = 140 fps
5770 = 85 fps
(5870 is 64.7% faster)

Is it worth paying 80% more to go 80% faster? Duh.

Of course, it could be the difference between playable and unplayable, depending on the game, when it in this case is about gaming performance.

It isn't exactly a mystery how much faster the Radeon HD 5870 is compared to the Radeon HD 5770.

It is well-known. The only thing we do not know is how the performance differentiates between the various Mac Pro models, which is where you come in handy ;)

So keep up the good work!

sirnh

Aug 28, 2010, 06:54 PM

I'm a pretty casual gamer at the best of times so I'm not sure what to make of those figures. Given the resolutions concerned are 61 & 85 FPS still respectable numbers?

Matt.

They are respectable numbers for a game like WoW, more than playable. But WoW is not very stressful on a GPU. But I am certain RobART will be testing out some of the more intensive games soon enough.

I've been too busy playing StarCraft 2 (under Windows) to try any other games, myself.

barefeats

Aug 28, 2010, 07:10 PM

They are respectable numbers for a game like WoW, more than playable. But WoW is not very stressful on a GPU. But I am certain RobART will be testing out some of the more intensive games soon enough.

I've been too busy playing StarCraft 2 (under Windows) to try any other games, myself.

My WoW test is a GPU stresser with low CPU load. There are other scenarios in WoW that stress the CPU but they are hard to test in a consistent and repeatable manner without having all the permutations of CPU + GPU in the same location observing the same activity with the same connection speed and latency. I wish there was a timedemo for WoW. I understand there will soon be one for Starcraft 2.

Meanwhile, back in the lab...
ET:QW timenetdemo at 2560x1600 High with 4x Multisampling
5870 = 109 fps
5770 = 59 fps
(5870 is 84.7% faster)

... and...
Portal First Slice timedemo at 2560x1600 High with 4x Multisampling
5870 = 138 fps
5770 = 71 fps
(5870 is 48.6% faster)

jav6454

Aug 28, 2010, 07:16 PM

Just a very crude & raw view at the processor numbers, the HD5870 is bound to be at least 2x faster since it has 1600 cores versus 800 cores on the HD5770.

However after GPU, Shader, Memory and Core clocks, I'm sure there is much more of a difference. Also, the HD5870 has a wider memory bus at 256-bit versus 128-bit memory bus on the HD5770, a double bandwidth difference between them.

MythicFrost

Aug 28, 2010, 07:29 PM

I take it your only testing games on the Mac side? If not, I don't suppose you could test Crysis Warhead or Metro 2033, if you have them.

jav6454

Aug 28, 2010, 07:34 PM

I take it your only testing games on the Mac side? If not, I don't suppose you could test Crysis Warhead or Metro 2033, if you have them.

Head over to Anandtech and check their benches for the HD5870 and compare them to the HD5770. Results will be very similar.

MythicFrost

Aug 28, 2010, 07:37 PM

Head over to Anandtech and check their benches for the HD5870 and compare them to the HD5770. Results will be very similar.
Ok thanks :)

f1restarter

Aug 28, 2010, 07:52 PM

I was very hesitant when I ordered my 2.8GHz Quad Mac Pro with the 5770 but after playing games like Crysis demo on highest settings at 1600x1200, I am happily surprised that the 5770 can hold it's own and IMO it was running pretty smooth. Now BFBC 2 plays great on highest settings at the same resolution, so I'm very happy with this card.

I updated the 5770 to the latest firmware and my 3D vantage score is showing as 10916 3D marks. Graphics score is 9655. I'm waiting for crysis 2 and see if I would need to upgrade to the 5870 but for the current games this card for me is perfectly fine. :)

Maybe this is just the effect after dealing with the 6800 Ultra for almost 6 years. lol

barefeats

Aug 28, 2010, 07:53 PM

I take it your only testing games on the Mac side? If not, I don't suppose you could test Crysis Warhead or Metro 2033, if you have them.

Some Windows 7 (Boot Camp) tests are on the "to do" list. Crysis Warhead is on the list there.

barefeats

Aug 29, 2010, 01:23 PM

Before I post the results, I'm testing the Radeon HD 4870 (shipped with the 2009 Mac Pro) on the 6-core Westmere. Earlier tests on an older Mac Pro pitting it against the 5770 showed them in a dead heat.

My tests on the 6-core are showing the 4870 about 10% faster at the highest settings and resolution.

But the 5770 is a better buy (unless you already own a 4870). Plus you can put two 5770s in the Mac Pro and jumper them with Crossfire cable under Windows. Yes, the Crossfire connectors are exposed and no special motherboard is required to run Crossfire. You just enable it in the ATI "control panel."

sch8mid

Aug 29, 2010, 02:02 PM

My tests on the 6-core are showing the 4870 about 10% faster at the highest settings and resolution.

Hi Rob
Some people and " notdoomers" for sure would be interested in
OpenGl and OpenCl benchmarks
like smalluxgpu and Cinebench 11.5

it will be a far closer match between Hd 5870 and 5770

and far away from 4850 or 4870 results

http://att.macrumors.com/images/attach/png.gif http://att.macrumors.com/images/attach/png.gif

and would you mind showing as a fair comparison how far away Apples OpenGL implementation is away from Win 7 (same rig-same HD 5770)

http://att.macrumors.com/attachment.php?attachmentid=249090&stc=1&d=1283144778

Win 7 - 32 bit : 52,47 FPS

barefeats

Aug 29, 2010, 11:33 PM

smallLuxGPU is promising as an OpenCL test. Cinebench OpenGL test in a small window is not stressful enough. OpenGL Extensions Viewer has a test that lets you set high resolution full screen, multisampling, anisotropy, etc.

davelanger

Aug 29, 2010, 11:35 PM

I'm a pretty casual gamer at the best of times so I'm not sure what to make of those figures. Given the resolutions concerned are 61 & 85 FPS still respectable numbers?

Matt.

Anything over 50 is. I dont even think the human eye can tell the difference after 60, not to mention console games never get over 60

sch8mid

Aug 30, 2010, 12:16 AM

Anything over 50 is. I dont even think the human eye can tell the difference after 60, not to mention console games never get over 60

Simply not an overwhelming argument

OpenGl is not for gamers alone - pro apps like Photoshop -Motion-Aperture heavily rely on OpenGL

and with the forecoming new coded FinalCut GPus will have a hard time coming...

try to rotate a 60 MB image without good OpenGl implementation

Starfire : a new OpenCL based videoconverter uses OpenCL and you
get more than realtime conversions of HD footage (nothing exists on the dark side so far)

Rob : give it a try

davelanger

Aug 30, 2010, 12:22 AM

Simply no useful argument

OpenGl is not for gamers alone - pro apps like Photoshop -Motion-Aperture heavily rely on OpenGL

try to rotate a 60 MB image without good OpenGl implementation

Starfire : a new OpenCL based videoconverter uses OpenCL and you
get more than realtime conversions of HD footage

Since he was quoting FPS when he comes to gaming, that is what I assumed he was talking about. For gaming 60 FPS is respectable. A lot of people on older machines only get 30 FPS and that is fine for most games. 60-85 is 2 to 3 times better than that. A lot of people cap WOW at 50 FPS when playing, so 60-85 is very good.

I didnt see him quoting anything for photoshop now did you? The reply was even to someone who said he was a causal GAMER, he did not mention any other programs.

So my answer was very useful for what he was asking from a causal gamers point of view when it comes to FPS

sch8mid

Aug 30, 2010, 12:26 AM

Since he was quoting FPS when he comes to gaming, that is what I assumed he was talking about. For gaming 60 FPS is respectable. I didnt see him quoting anything for photoshop now did you?

hard core gamers on opengl based games for sure will not be very satisfied with 50 fps. :eek:

davelanger

Aug 30, 2010, 12:30 AM

hard core gamers on opengl based games for sure will not be very satisfied with 50 fps. :eek:

The poster said he is a causal game not hardcore gamer right?
So for a causal gamer its respectable. You are right though it all depends on what you want it for.

For a game like WOW its fine, but more hardcore gaming you will want more FPS.

Btw I see you edited your first post and changed the wording of your reply, thanks.

mattmower

Aug 30, 2010, 02:05 AM

The poster said he is a causal game not hardcore gamer right?
So for a causal gamer its respectable. You are right though it all depends on what you want it for.

As the original responder yes I am a casual gamer. I've had Diablo 3 on pre-order since 2009 and, whenever it turns up, that's probably the next game I'll play. Oh I mooched about in Torchlight via Steam for a while a couple of months ago. Other than that the only game I play regularly on my Mac is Out of the Park baseball which isn't graphical to any great degree.

What I'm hearing about the 5770 is that it should offer acceptable performance for someone like me although, in time, not at the highest resolutions and/or multisampling multipliers.

Is that a fair assessment?

Matt

barefeats

Aug 30, 2010, 07:37 AM

Just posted this...
SERIOUS FUN: Radeon HD 5870 versus 5770 and three other GPUs (4870, GTX 285, FX 4800):

http://www.barefeats.com/wst10g4.html

(Includes Portal, WoW, X-Plane, Motion, OpenGL Extensions Viewer, and SmallLuxGPU results.)

Now I'm packing up to go test the 5770 on the 2006 Mac Pro at PowerMax.

barefeats

Aug 30, 2010, 10:52 AM

I'm back at the PowerMax lab. I'm running the Radeon HD 5770 on the 2006 Mac Pro. Working fine.

Right now I'm checking out my hypothesis that it is just as fast as the 5870 on the "first gen" Mac Pro.

Let's say the gap has narrowed to the point that the cost of the 5870 is questionable for a 2006 Mac Pro. Here's one example:

X-Plane 9.6.1 at 1920x1200:
5870 = 83 fps
5770 = 80 fps

That's much closer than the gap between the two cards on the 2010 Mac Pro.
Full report coming later today.

barefeats

Aug 31, 2010, 10:43 PM

ovrlrd

Aug 31, 2010, 10:48 PM

Here's another twist.

Radeon HD 5870 and 5770 in the 2006 Mac Pro versus four older GPUs:
http://www.barefeats.com/wst10g5.html

Next assignment? Compare the two cards to each other in the 2008 and 2009 Mac Pro.

Nice job! So it seems like if you have a 2006 the best value is a 5770. At the same time it doesn't seem like there is much difference between it and the 4870 anyway, so if you already have one of those it might not be worth the money.

Looking forward to the 2009 results!

fsron

Aug 31, 2010, 10:55 PM

[QUOTE=barefeats;10961139
Next assignment? Compare the two cards to each other in the 2008 and 2009 Mac Pro.[/QUOTE]

Thanks for your work on the 2006 Mac Pro. My interest though is the 2008. I look forward to your next assignment and to see if the 5870 makes more sense for that generation of Mac Pro's. Right now I'm leaning toward the 5770.

svale

Sep 1, 2010, 12:22 AM

Here's another twist.

Radeon HD 5870 and 5770 in the 2006 Mac Pro versus four older GPUs:
http://www.barefeats.com/wst10g5.html

Thanks barefeats.

Any plans to run some tests in Windows on the 2006 Mac Pro? I'd be surprised if the 5870 didn't outperform the 5770 significantly in Windows.

Also, did you check whether the 5870 worked with the MP11.005C.B08 Boot ROM?

Bob Kiwi

Sep 1, 2010, 08:20 AM

Indeed, thanks for this round of testing.

I have a Mac Pro 2,1 and want to get a 5870 purely for FFXIV in Windows, but be able to take advantage of it in Mac OS X as well. Good to see how much it smokes my x1900 XT.

Apple at this point is 2-3 weeks for a 4870 @349, 2-4 for a 5770 @249, and still no data on the 5870 @449. Guess I'm gonna miss the launch while waiting :mad:

maccompaq

Sep 1, 2010, 09:23 AM

X-Plane 9.6.1 at 1920x1200:
5870 = 83 fps
5770 = 80 fps

I built a Core i7 Hackintosh running Snow Leopard 10.6.4, 6 GB RAM, with a GTS 250, 1 GB graphics memory, primarily to run X-Plane 9. With all settings at Max, driving a 23 inch HD monitor at 1920 X 1080, my FPS is all over the place: well over 300 fps in sparse areas and sometimes dips as low as 30 fps for a few seconds, in the largest cities.

How did you arrive at the 83 fps when the rate changes constantly?

I used the Gigabyte LGA 1366 mobo so that I can switch out the 4 core processor to a 6 core processor when they come down in price. By then, X-Plane 10 should be out which will take full advantage of the powerful processors and video cards.

On a whim, I bought MS FSX with the intention of installing XP Pro, but the mobo would not let XP install. Since I had already purchased FSX, I bought Windows 7 64 bit. After running FSX a few times, I find that I rarely use it since I really like X-Plane.