5870 Vapor-X, Dirt 2, Q6600, CFX vs Non-CFX

[zc1]

You might remember my previous thread (click), in which I looked at the performance of the 5750 on my system, running NFS Shift. Out of curiosity, I went and picked up a couple of 5870 Vapor-Xs to see how they perform.

Happily for me, the 5870s came with Dirt 2, which I wasn't expecting, so I was able to run the Dirt 2 benchmark rather than just running FRAPS in parallel with a game. I was able to get my hands on a Q9650 and will include the data from testing that processor in my system as well, even though it throws in some other variables (45nm vs 65nm, 1333 MHz FSB vs 1066 MHz FSB, 12 MB L2 Cache vs 8 MB L2 Cache).

Here's the background info:

My System (Click for Gallery):

-Processor #1: Core2 Quad Q6600 @ 2.40 GHz
-Processor #2: Core2 Quad Q6600 @ 3.60 GHz
-Processor #3: Core2 Quad Q9650 @ 4.05 GHz
-Motherboard: Gigabyte EP45-UD3P (Rev. 1.6)
-Memory: 8GB Mushkin Blackline PC2-8500 @ 1066 MHz
-Graphics: 2x Sapphire Radeon HD5870 Vapor-X 1GB GDDR5
-ATI Catalyst Version: 10.3a
-Display: 3x SyncMaster 2243 (Eyefinity)
-Hard Drive: OCZ Vertex Turbo 120 GB SSD
-OS: Windows 7 Ultimate 64-bit


5870 Vapor-X Specs:

-GPU Clock: 870 MHz (vs 850 MHz for standard 5870)
-Memory Clock: 5000 MHz (vs 4800 MHz for standard 5870)


Game Settings (Dirt 2):

The only settings that were changed were resolution and AA (0x, 2x, 4x or 8x); All other settings were left at MAXIMUM/ULTRA


Effect of Processor Speed on FPS (Single Card Mode)

1. 6024 x 1080, 2xAA

2.40 GHz, Non-Crossfired --> 37 avg, 30 minimum
3.60 GHz, Non-Crossfired --> 37 avg, 31 minimum
4.05 GHz, Non-Crossfired --> 37 avg, 30 minimum

2. 5760 x 1080, 2xAA

2.40 GHz, Non-Crossfired --> 38 avg, 31 minimum
3.60 GHz, Non-Crossfired --> 38 avg, 31 minimum
4.05 GHz, Non-Crossfired --> 38 avg, 33 minimum

3. 1920 x 1080, 8xAA

2.40 GHz, Non-Crossfired --> 60 avg, 52 minimum
3.60 GHz, Non-Crossfired --> 61 avg, 54 minimum
4.05 GHz, Non-Crossfired --> 61 avg, 54 minimum

Conclusion: In my system, with a single graphics card, increasing processor speed has no effect on FPS at low or high resolution.


Effect of Processor Speed on FPS (Dual Card (Crossfire-X) Mode)

1. 6024 x 1080, 2xAA

2.40 GHz, Crossfired --> 54 avg, 46 minimum
3.60 GHz, Crossfired --> 60 avg, 51 minimum
4.05 GHz, Crossfired --> 61 avg, 51 minimum
*13% increase in FPS w/ increase in processor speed

2. 5760 x 1080, 2xAA

2.40 GHz, Crossfired --> 54 avg, 46 minimum
3.60 GHz, Crossfired --> 63 avg, 53 minimum
4.05 GHz, Crossfired --> 63 avg, 50 minimum
*17% increase in FPS w/ increase in processor speed

3. 1920 x 1080, 8xAA

2.40 GHz, Crossfired --> 57 avg, 48 minimum
3.60 GHz, Crossfired --> 72 avg, 62 minimum
4.05 GHz, Crossfired --> 90 avg, 76 minimum
*58% increase in FPS w/ increase in processor speed

Conclusion: In my system, with two, crossfire-x-linked graphics cards, increasing processor speed has a greater effect on FPS at lower resolution (58% increase in FPS achieved) than at higher resolution (13% increase in FPS achieved @ 6024x1080).


Effect of Crossfire on Average FPS

1. 6024 x 1080, 2xAA

2.40 GHz, Non-Crossfired --> 37 avg, 30 minimum
2.40 GHz, Crossfired --> 54 avg, 46 minimum

3.60 GHz, Non-Crossfired --> 37 avg, 31 minimum
3.60 GHz, Crossfired --> 60 avg, 51 minimum

4.05 GHz, Non-Crossfired --> 37 avg, 30 minimum
4.05 GHz, Crossfired --> 61 avg, 51 minimum


2. 5760 x 1080, 2xAA

2.40 GHz, Non-Crossfired --> 38 avg, 31 minimum
2.40 GHz, Crossfired --> 54 avg, 46 minimum

3.60 GHz, Non-Crossfired --> 38 avg, 31 minimum
3.60 GHz, Crossfired --> 63 avg, 53 minimum

4.05 GHz, Non-Crossfired --> 38 avg, 33 minimum
4.05 GHz, Crossfired --> 63 avg, 50 minimum


3. 1920 x 1080, 8xAA

2.40 GHz, Non-Crossfired --> 60 avg, 52 minimum
2.40 GHz, Crossfired --> 57 avg, 48 minimum

3.60 GHz, Non-Crossfired --> 61 avg, 54 minimum
3.60 GHz, Crossfired --> 72 avg, 62 minimum

4.05 GHz, Non-Crossfired --> 61 avg, 54 minimum
4.05 GHz, Crossfired --> 90 avg, 76 minimum


Conclusions for lower resolution testing(1920x1080):
The greater the processor speed, the greater the effect of crossfire on FPS. At the lowest CPU speed tested, CF-X actually reduced average FPS, but as CPU speed increased, so did the positive effect of CF-X on average FPS. CF-X @ CPU speeds of 2.4, 3.6 and 4.05 GHz changed average FPS by -5%, +18% and +48%, respectively @ 1920x1080.


Conclusions for higher resolution testing (5760x1080 or 6024x1080):

The greater the processor speed, the greater the effect of crossfire on FPS, and the effect was more pronounced that what was seen in the low-resolution setting):
CF-X @ CPU speeds of 2.4, 3.6 and 4.05 GHz changed average FPS by +46%, +62% and +65%, respectively @ 6024x1080.



What Does All of This Mean From a Practical Standpoint?

What does this mean when it comes to helping to make the decision on what upgrade path to take when it comes to your Q6600 system? I can only speak to Dirt 2, as that's the only game for which I have good benchmark data.

A) If you game at 1920x1080 or less:

If you game at 1920x1080 or less then your best performance will be from adding a second card and upgrading your processor to the fastest that you can get (either by overclocking or moving up to a faster processor +/- overclocking).

-Overclocking or switching to a faster processor alone --> 0% FPS improvement

-Adding second graphics card alone --> 5% less FPS

-Adding second graphics card + overclocking @ 3.6 GHz --> 20% increase in FPS (60 to 72 in my case) vs stock speed and a single card

-Adding second graphics card + swap to Q9650 @ 4.05 GHz --> 50% increase in FPS (60 to 90 in my case) vs stock speed and a single card


B) If you game at 5760x1080:

At high resolution, processor speed was not limiting, so your best performance bet is to get a second graphics card.

-Overclocking or switching to a faster processor alone --> 0% FPS improvement

-Adding second graphics card alone --> 46% increase in FPS (37 to 54 in my case)

-Adding second graphics card + overclocking @ 3.6 GHz --> 62% increase in FPS (37 to 60 in my case) vs stock speed and a single card

-Adding second graphics card + swap to Q9650 @ 4.05 GHz --> 65% increase in FPS (37 to 61 in my case) vs stock speed and a single card

[Ryoohki]

Nice result.. i get about the same as you at 1920x1080p with 8x at Ultra with a 5850 but it's heavely OC (940/1140). I ordered a second 5850 PCS+

[zc1]

Thanks Ryoohki! My 5750s overclocked nicely, but I haven't tried these ones at all.