ASRock 775Dual-VSTA: Does DDR2 matter?
by Gary Key on August 8, 2006 6:35 AM EST- Posted in
- Motherboards
Application Performance
We decided to test some real world applications that typically stress the CPU, memory, and storage systems to see if the results from our synthetic memory tests carry over to the desktop. Based upon those results, our DDR2-533 memory settings should outperform both the DDR2-667 and DDR-400 configurations, with all three outperforming DDR-333.
Our tasks include three activities that are common on the desktop. Our first test was to measure the time it takes to shrink the entire Office Space DVD that was extracted with AnyDVD into a single 4.5GB DVD image utilizing Nero Recode 2. Our second test utilizes WinRAR 3.6 and measures the time it takes to compress our test folder that contains 444 files, 10 folders, and 602MB of data.
Our third test consists of utilizing Exact Audio Copy as the front end for our version 3.98a3 of LAME. We set up EAC for variable bit rate encoding, burst mode for extraction, use external program for compression, and to start the external compressor upon extraction. (EAC will read the next track while LAME is working on the previous track, thus removing a potential bottleneck with the drive). Our test CD is INXS Greatest Hits, a one time '80s glory masterpiece containing 16 tracks totaling 606MB of songs. The results of our tests are presented in minutes/seconds with lower numbers being better.
We see that our DDR2-533 memory setting places first in all applications but the largest margin of victory is a 16 second advantage in the Nero Recode 2 test over the DDR-333 setting (a 3.9% performance advantage). Our DDR-400 results are impressive as they finish slightly ahead of the DDR2-667 settings, though both come close to the DDR2-533 configuration. The effects of the other platform components have basically negated the pure performance advantage of our DDR2-533 setting in the synthetic memory tests. While this particular ratio still offers the best overall performance, it would be difficult to tell the actual performance difference between it and our other memory without a benchmark.
We decided to test some real world applications that typically stress the CPU, memory, and storage systems to see if the results from our synthetic memory tests carry over to the desktop. Based upon those results, our DDR2-533 memory settings should outperform both the DDR2-667 and DDR-400 configurations, with all three outperforming DDR-333.
Our tasks include three activities that are common on the desktop. Our first test was to measure the time it takes to shrink the entire Office Space DVD that was extracted with AnyDVD into a single 4.5GB DVD image utilizing Nero Recode 2. Our second test utilizes WinRAR 3.6 and measures the time it takes to compress our test folder that contains 444 files, 10 folders, and 602MB of data.
Our third test consists of utilizing Exact Audio Copy as the front end for our version 3.98a3 of LAME. We set up EAC for variable bit rate encoding, burst mode for extraction, use external program for compression, and to start the external compressor upon extraction. (EAC will read the next track while LAME is working on the previous track, thus removing a potential bottleneck with the drive). Our test CD is INXS Greatest Hits, a one time '80s glory masterpiece containing 16 tracks totaling 606MB of songs. The results of our tests are presented in minutes/seconds with lower numbers being better.
 |
| Click to enlarge |
We see that our DDR2-533 memory setting places first in all applications but the largest margin of victory is a 16 second advantage in the Nero Recode 2 test over the DDR-333 setting (a 3.9% performance advantage). Our DDR-400 results are impressive as they finish slightly ahead of the DDR2-667 settings, though both come close to the DDR2-533 configuration. The effects of the other platform components have basically negated the pure performance advantage of our DDR2-533 setting in the synthetic memory tests. While this particular ratio still offers the best overall performance, it would be difficult to tell the actual performance difference between it and our other memory without a benchmark.
Facebook
Twitter
RSS
55 Comments
View All Comments
Calin - Wednesday, August 9, 2006 - link
Yep, I feel stoopid :)Anyway, the idea is that a change in memory patterns (SDR to DDR, SDR to RDRAM, DDR to DDR2) is a battle between old very optimised technology, and new, unproven yet technology. The small difference in speed can be explained that "current" technology in processors is built for best performance with current memory - a new memory type often is not optimised for the memory access needed by the processor.
As an example, RDRAM was (just a tad) slower on Pentium !!! (compared to high performance SDRAM). Pentium4, which was bandwidth starved with single channel SDRAM, was much faster with RDRAM (dual channel though) - as much as a P4 2000 (Willamette) with SDRAM was equal to a P4 1600 with RDRAM. As speed increased, needed bandwidth increased too - but the move to dual channel DDR was the final nail in the coffin of RDRAM on PC.
The other example - Athlon64 is not bandwidth starved on current (dual channel DDR400) memory, so doubling memory bandwidth brought no advantage. The decrease in latency was not enough to bring extra performance.
The situation is mostly similar with Core2Duo - more memory bandwidth brings little advantage.
This might change for quad-core processors, as they could use twice the memory bandwidth we see now - or on the Athlon side with a more aggressive prefetching algorithm (which will eat bandwidth bringing data that seem to be useful in the near future).
yacoub - Tuesday, August 8, 2006 - link
"When faced with a limited budget but a desire to have the latest and greatest technology, it is usually has to cut corners"should be "one usually has to cut corners"
yacoub - Tuesday, August 8, 2006 - link
page 2:"...one of the widely used setups in use today." Maybe you like the extra words but you could drop the words "in use" and still be making the same point.
"The memory features average latencies at DDR2-667 but was able to perform at lower latencies in our testing while costing around $70 for a 1GB kit.
{transcend-ddr2.html}" <--- supposed to be a link or image?
JarredWalton - Tuesday, August 8, 2006 - link
Blame the sleepy editor. :|The Transcend table was present, but the supported RAM speeds table was not. I fixed the error, as well as the other two grammar issues you pointed out. Thanks!
yacoub - Tuesday, August 8, 2006 - link
hehe no problem. =)