|
With the aforementioned issues relating to bandwidth, I was interested to see exactly how far the card would go in terms of overclocking to maybe aid the card with a little extra leeway in terms of MHz.
I spend 2 days overclocking the card, testing stability over a various number of scenarios. I tested on the all the benchmarks in this review, and also several other games I play regularly to ensure maximum stability at the overclock I managed to attain.
For overclocking, I chose to use Radeonator version 2.00. There are several different programs that one could use to overclock the card, but ever since it's release, for me, Radeonator has proved the most reliable and most user-friendly.
I should note that I did not in anyway modify the card, add or remove cooling, and the cooling for the card was simple that which was supplied.
At defaults, the card ships with a core clockspeed of 275MHz, and memory speeds 550MHz DDR.
I was able to overclock to card to an impressive 322MHz Core, a 47MHz jump from retail speed. I was also able to increase the Memory speed to a very respectable 602MHz DDR.
I could get significantly higher than this, but after an hour or two or gaming, their would be visual anomalies. Therefore, the above speeds on my system were 100% stable with zero visual anomalies. I can't help but feel with the addition of some extra cooling, I would have been able to go significantly further, particularly with the memory.
So, what kind of performance increases does this overclocking warrant you may be asking. We used some of the tests integrated into 3DMark 2001SE to determine the performance increase this overclock yields. All tests were run at 1024x768x32.
The first test is a theoretical fillrate test, both single and multi-textured, and is a valid way of testing the increase in bandwidth across the card.
Fillrate performance when overclocking
As mentioned in the section describing changes to the core, one of the disadvantages to the changes made to the RV250 was that it would severely affect performance of multi-texturing. We see considerable jump in multi-texturing performance with the overclocked card, which after investigating was mainly down to the increase in core speed.
The next test is a simple run of the default benchmark.
3DMark 2001SE performance when overclocking
A very impressive leap score all the way up to 8648 3Dmarks, comparible with some solutions outside of the budget segment.
A third and final test, as demonstrated earlier, the Nature Test. This test is probably the most stressful for the hardware that has been used in this review, so if anything were to translates the theoretical increase in fillrate into actual performance, it would be this.
Nature test when overclocked
As you can see, the overclock yields an increase in performance under the most intense of scenarios of around 10%, not bad considering there was no modification to the card cooling wise, and it was entirely flawless in operation at these settings.
I do believe with an improved cooling method, the card could yield even larger overclocking potential. As a whole, the card proved to be a solid overclocker out-of-box, and although many users will not choose to overclock their card, based on the stability of this actual overclock, I would recommend at least considering overclocking, as the increase in bandwidth on an a bandwidth starved solution is highly beneficial.
|
