Hello people, sorry been REALLY busy the past week, so couldn't update the review and graphs.
Snaga had offered to help me on the graphs the day I asked on the forum, and soon after I sent him the raw data, I got some nice graphs back!!
I have since modified the graph and did some colour coding to make it easier to understand if someone were to look at the graph without reading the previous write-ups, and will soon publish them.
I've got to thank Snaga again for the very fast reply and wonderful graphs, and sorry I have had to put this off for so long!
Right. So the "eagerly awaited" sequel to my quick & dirty 3DMark06 test for the HD3870XT, Crysis benchmarking results, which pit the ATi 3870 single and in CrossFire configuration against the mighty nVidia 8800 Ultra.
Crysis will represent a much more realistic comparison between these different graphics card configurations against each other, as synthetic benchmarking software such as 3DMark tend to get special attention with GFX card manufactures, and are more prawn to tweaking (read Cheating) in the driver to obtain higher scores.
However, bear in mind that Crysis is almost a joint project between nVidia and Crytek, and the game is heavily optimised for GeForce cards, nVidia will have an advantage over the ATi/AMD solution and because of that reason, it is the perfect test for CrossFire’s real world performance when it is not specifically optimised.
Let's start off with a comparison between a single overclocked 8800 Ultra, single overclocked HD3870XT, and a pair of overclocked HD3870XT.
The variables are the application of Anti-Aliasing, and weather the game is ran under DirectX9 or 10. Note that all game settings were on "HIGH".
This graph shows a lot of useful information. First off, notice that single card setups take a much smaller performance hit when DirectX10 were enabled in Crysis; whereas when CrossFire was enabled, the performance droped below that of a single 8800 Ultra.
The minimum frame rate was due to the low amount of ram on this test system, which caused the system to access the Hard Drive frequently, and causing hiccups to the frame rate. The 8800 Ultra also suffered this drawback, but was not recorded due to the nature the data was collected, and the fact that it has more on board Video Memory.
Where CrossFire really shine is when it ran on DirectX9, bear in mind this setup is much cheaper than a single 8800 Ultra. Both with the AA enabled and disabled show a fair lead over the 8800, almost 10FPS faster without AA enabled. It also shows great scaling compared to a single 3870, with a 70% increase in frame rate with a second 3870 added.
A single 8800 Ultra outright beats a single HD3870XT, but that is expected as they are placed in different segment of the market, with a huge price margin between them.
The second graph is a comparison between two systems running a single 8800 Ultra, with the CPU at the same clock speed.
When running on DirectX10, a CPU with a larger integrated cache definitely helps with the frame rate, as more CPU is utilised to render the frames.
DirectX9 seems to benefit little with the different system setups, and obviously runs smoother then on DX10.
Interestingly when running on DX10, the minimum frame rate is higher then running on DX9.
The last graph shows the effect of CPU speed on single and CrossFire setups.
DirectX10 seems to take more of an advantage with faster CPU speeds on both single and CrossFire configurations then DX9.
Well, here it is. When I was testing CrossFire with Crysis using DX10, the image was flashing, as if you can see the alternate frames rendered in front of your eyes, showing poor initial compatibility of CrossFire and DX10, or may be a bug of CF. This is a great problem if you plan on running this game with DX10 using a CrossFire setup, although I suspect it will be addressed in the next release of Catalyst driver. The performance of the pair of 3870 is admirable under DX9, gaining a significant frame rate advantage over the more expensive 8800 Ultra, even when 4xAA was enabled.
I maintain my previous opinion that a HD3870XT CrossFire Setup is a better buy then a single card nVidia solution, as with a cheaper price you get better framerates and image quality (under DX9).
Setting up CrossFire in Windows was as easy as clicking a box labled “Enable CrossFire”, and you get the benefit of the fastest setup on an Intel platform.
The problem comes when you consider DirectX10. In my opinion DX10 does improve image quality, slightly, and comes at a cost of large frame rate lost. Even when running on current generation nVidia setups, it is suggested that you turn off DX10, to gain a performance boost with small image quality lost. Because although it comes at a smaller cost then with a CrossFire setup, enabling DX10 drops your frames to choppy levels on the beefiest nVidia cards when playing Crysis.
So why buy a more expensive nVidia graphics card if you are going to run on DX9 anyway? And with future driver optimisation of Crossfire setups, I don’t see how you can lose.
I would like to thank Snaga for providing me the graphs to work on this Crysis benchmarking round-up, his great work made this review a reality!
Any updates on this set-up ? Did you sort out some watercooling for those ATi cards yet?
Hello Luanswan, been back to Taiwan on some family business, but returned just tonight, so wasn't around to update this thread.
However as some of you may be aware, a lot of my own spare time and others has gone into benchmarking the setup, and graphs made so that the data can be easily interpreted by everyone. The total lack of enthusiasm, or even appreciation, of these very time consuming work has me thinking that my time was wasted, and probably better spent recommending PC specifications for new customers.
Therefore I will not post anything that takes more than 2 minutes of my time on this forum in the future, and redirect my work to more dedicated hardware forums.
But yes, I have indeed watercooled my pair of HD3780XT:
would add on a personal note that I’ve found all of your in depth reviews extremely interesting, but will admit to having no where near enough technical knowledge (yet) to be able to add any useful feedback / challenge. It's safe to say that the amount I have learned in the couple of months I've been frequenting this forum is tremendous and would love to be able to play a more active role on the hardware debating / reviewing / challenge in the future. Would be a shame if we were to loose your obvious talents and enthusiasm....
NB. Today was the first time I've seen this review, somehow managed to completely miss it during the Xmas revelries!!
Im with JP on this one.....my tech knowledge of pcs is minimal.....however since joining this forum i have learnt quite a bit from the likes of yourself and others!
Personally im still undecided about sli/crossfire...But i do like reading your reviews and looking at the great pics and even wondering where you get those acrylic cases from?
However i cannot really add depth to your reviews coz i wouldnt know what im talking about...
well 1189 views so far... lack of replies doesn't mean lack of interest. I actually missed the last bit as I have been having problems with my PC at work and sometimes IE blows up in my face and I lose all the topics I haven't yet read... and it is quite frequent but, I did read in the beginning and was awaiting for your results. I have told some people here not to go for 680i chipsets and either go with a single nVidia card or even try the crossfire solution with the 3870s on x38 chipsets.
? Did you sort out some watercooling for those ATi cards yet?
