Socket A SDRAM Performance Comparison

Pages: 1 2 3

A Graphical View

So lets take a closer look, this time with bar graphs and percentages, using the PC100 200MHz FSB CAS 2 (second column) as a base (100%). Hopefully in this format you can get a ‘feel’ for the actual performance differences. Don’t just look only at the difference between the lengths of the bars in the graphs because they can visually fool you in to thinking the performance differences are larger then they really are. Keep the actual percentage numbers in mind while looking at the graphs.

First let’s look at the Winstone 2001 scores, which should give a good indication of performance difference using business based applications:

Pretty much what I would expect, scores getting progressively better as the memory speeds and CAS settings get faster, same with the FSB speeds. Disregarding the 145MHz-overclocking test, notice that the scores are within 5% of the base, and even the 145MHz results don’t even get close to 10%. I think this is a good indication of what to expect using different memory (and settings) with different FSB speeds, especially in a business environment.

But look at the SYSmark 2000 scores. Even though it’s also an application-based benchmark the results are not quite the same:

The 200MHz FSB PC100 CAS 2 and both PC133 scores are real close; the 266MHz FSB score does show a gain in performance, but only about 5%.

Next let’s look at a 3D based test: 3D Winstone 2001. You pretty much can ignore the actual test score because the video card is pretty much max’ed out. But the CPU scores themselves are of value and a good indication of what to expect when using different memory and FSB settings:

Now here we start to see some significant difference, especially with a 266MHz FSB over the base. Again a nice progressively faster set of scores as the speeds and settings go up.

Quake III – always a good indication of CPU, bus and memory speeds:

Shows a nice progressive scale, very much what I would expect to see. But the gain, unless overclocking, is still under 10% (but getting close to it). Quake sure does like the overclocked 145MHz HostClk results though.

Now here are some interesting results – Video 2000. I would expect these to be very sensitive to both memory performance and FSB speeds.

But the results are not what I would have expected to see. Notice how the PC100 & PC133 200MHz FSB scores are so close to each other, but the overall performance difference is still quite minimal. I really would have expected these results to be sensitive to memory speed & CAS settings, also the 266MHz FSB.

And 3D Mark 2000…especially take note of the CPU scores:

Very interesting to see the PC133 results so close – The overall scores show a nice progressive stepping but the CPU are a different story and give some interesting results. I’m not too sure what is going on here.

Finally some Final Reality scores. This program has always been good at showing differences in CPU, FSB and memory performance.

Pretty much what I’d expected, but one of the few benchmarks to really show any significant gain in performance. It is an older benchmark, could have some bearing on the type of results it gives.

Pages: « Prev   1 2 3   Next »

Be the first to discuss this article!