Conclusion: Performance
Where to start? The promise of Hydra has been one of great expectations, but has really only garnered disappointment overall. The idea of breaking up a scene dynamically, and assigning workloads to the different cards, is one that has a lot of potential. As we saw with the scaling in 3D Mark Vantage, the potential is certainly there. Unfortunately, that level of optimization that we see for Vantage does not apply to most other games, with Far Cry 2 being the lone outlier due to its popularity as a benchmark.
The four video cards under consideration here today. Note again how much longer the GTS 450 is as compared to the more powerful GTX 460s.
We did see some interesting results with the mixed/Hybrid mode. There was often a small but measurable increase in performance over the single GTX 480 when combined with the HD 5870, but in other applications we saw the combination going slower than a single GTX 480. One interesting potential combination is to have the Eyefinity Edition HD 5870 card as the master, and the GTX 480 as the secondary. It would be interesting to see if performance scales when dealing with Eyefinity resolutions such as 5760×1200. One downside to Hydra vs. a native SLI solution is that of NVIDIA Surround. Hydra does not support that functionality, and the only way to get multi-monitor to work is to have an AMD based board as the master card, and utilize the Eyefinity functionality.
Obviously the best results were those taken from cards that are identical. While Hybrid mode works, it is not necessarily recommended by AMD, NVIDIA, or Lucid. In fact, when Hybrid mode was enabled I was unable to access the Catalyst control panel for the AMD card. This was with the GTX 480 as the master card, and it is quite likely that the NVIDIA control panel would not be available if the AMD card was in the master slot.
Hydra scaling seemed to be better for the AMD cards used in the test vs. what we saw for NVIDIA. Of the NVIDIA cards used, the GTS 450 and GTX 460 1 GB showed the best results. Overall scaling was poor across the applications used for the majority of cards though. When going back to the Crossfire vs. Hydra scaling results, we see the much more mature Crossfire able to accelerate most applications (with the exception of LOTRO). Hydra was able to scale on most applications, but usually not within the same realm as Crossfire in terms of overall performance.
Enabling Hydra is a simple click away. Considering how often it acts as a decelerator in many situations, the ease of enabling/disabling is appreciated.
The one major issue that I seem to have uncovered from this review is that Hydra does not play well with cards that have less than 1GB of memory. How Hydra works essentially uses the primary video card as the compositor. That compositing takes up memory. In older, less memory intensive applications we likely would not see much of a performance difference. In modern games which can utilize large amounts of memory, adding in the space needed to composite the frame output from the different cards can have a very detrimental effect on overall performance.
The GTS 450 card often scaled very well with Hydra, but unfortunately the overall results were not that much different from a single GTX 460 1GB board. It certainly is not a great idea to purchase two GTS 450s in a Hydra solution, only to have them be slightly slower than a single GTX 460 1GB board that costs less than the two GTS 450s. The only real decent combination that is within most people’s budgets will be two GTX 460 1GB cards. They should cost less together than a GTX 480, but will at least perform at that same level.
As in for up-to-date
As in for up-to-date applications, I am looking for a reliable reader of exe files for some of my gadgets: zune, smart phone. I don`t know for sure how these things work, but as long as we`re talking about serious evolution here, futuristic design, I believe that there is a direct proportional development of easiness to handle technology and revolutionary shapes which I happen to admire a lot.