The ECS K7S5A SDR vs. DDR SDRAM Comparison Guide v1.0
By AlphaArms
By AlphaArms
Lately, I've been noticing a lot more post about the performance difference between DDR and SDR types of RAM. Many people have been asking, "Will upgrading to DDR RAM really boost the performance of my system?"... "Is DDR RAM really worth the price to upgrade from SDR RAM?"... "Why doesn't my system seem faster after going from SDR to DDR RAM?"
Well, I've tried my best to answer these questions (and more) in this comprehensive guide of comparing DDR to SDR on the K7S5A motherboard. Hopefully this guide will help you determine if the upgrade to DDR RAM is right for you.
Table of Contents
I. Test Setup
II. Windows, Game and BIOS Settings
III. Benchmarks
IV. Conclusion
V. FAQ
I. TEST SETUP
Constants
ECS K7S5A Pro (PCB 5.0) motherboard
AMD AthlonXP 2100+ processor (Palamino) @ 1.725GHz
Quantum Fireball lct10 10.0 GB hard drive
NVIDIA GeForce2 MX 400 PCI video card
Creative SoundBlaster PCI512 sound card
3Com/U.S. Robotics 56K PCI modem
Varibles
256MB Mitsubishi PC133 SDR SDRAM
256MB Samsung PC2100 DDR SDRAM
BIOS and Software
Official ECS K7S5A Pro 030121 BIOS
Windows 2000 Professional with Service Pack 3
DirectX 9
All critical and recommended updates installed
II. WINDOWS, GAME AND BIOS SETTINGS
To keep the tests between the two types of RAM comparable, a few settings in the BIOS and Windows had to be changed.
First of all, the timings on both sticks of RAM were set as follows:
DRAM Timing Configuration: Safe
CAS Latency: 3
RAS Active Time: 6T
RAS Precharge Time: 4T
DRAM Driver Slew Rating: Normal
Secondly, the page file in Windows 2000 was changed to a minimum and maximum size of 256MB.
Finally, all the games tested were set to a resolution of 640x480 and lowest detail. This is to keep the stress more on the CPU and RAM than on the graphics card, which would have bottlenecked the system and provided useless results.
III. BENCHMARKS
Before I head into the benchmarking itself, I'd like to tell a little about each benchmark program itself.
The first set of benchmarks is the popular 3DMark series. Most gamers know what these benchmarks are and use them as a base to compare their system's performance to other computer performance levels. For this guide, I chose to use 3DMark2000 and 3DMark2001. I decided that I would omit 3DMark03 due to the fact that my video card could only run one of the four tests. Thus, the results would be pretty useless no matter what RAM I used.
The second benchmark is another program from the creators of 3DMark, which is named PCMark2002. This benchmark provides a highly more accurate result than either of the 3DMark programs would (when comparing different types of RAM) because it's function is to benchmark the entire computer instead of just the 3D portion of it.
Third, we have SiSoftware's SANDRA 2003 SP1. It includes a number of benchmarks that can test the CPU, motherboard chipset, and RAM bandwidth among other things. For this test, I used two of it's benchmarks: the RAM bandwidth test and the Cache and RAM test.
The fourth benchmark is not actually a benchmark, but rather a set of common programs that were tested to see how different types of RAM would affect their speed. The first was WinRAR, a common Windows compression program. To test it I used a 178MB .wav file and compressed it using WinRAR's best compression. I logged the amount of time it took for the file to be compressed. The second was winLAME, an MPEG-1 Layer 3 (MP3) encoder. To test it, I used the same 178MB .wav file and logged the amount of time it took to encode it to the MP3 format.
Finally, the last set of benchmarks were games. The first, Quake III Arena, is a game almost everyone is familiar with and has become a standard for benchmarking computer sytems. The second, Jedi Knight II: Jedi Outcast, is a Quake III engine-based game which is also growing in popularity to benchmark systems.
That said, let's get to the numbers!
The DDR results are on the top or left, while the SDR results are on the bottom or right
3DMark2000
Only a 48 point difference between DDR and SDR, well within the margin of error for 3DMark2000.
3DMark2001
Again, a very nill 10 point difference between the two RAM types... very much within the margin of error.
PCMark2002
PCMark tells a much different story, with a 447 point increase with DDR RAM over SDR RAM.
SANDRA 2003 SP1
RAM bandwidth test
Cache & RAM test
The RAM bandwidth test takes a sizeable performance gain going from SDR to DDR, but the Cache & RAM test seems to be unchanged.
WinRAR Compression Time and winLAME Encoding Time
DDR Compression Time for 178MB .wav file - 2:09 (129 seconds)
SDR Compression Time for 178MB .wav file - 2:20 (140 seconds)
DDR Encoding Time for 178MB .wav file - 4:16 (256 seconds)
SDR Encoding Time for 178MB .wav file - 4:17 (257 seconds)
Although the WinRAR compression clearly benefits from the upgrade to DDR RAM, the time it takes to encode MP3 files seems unaffected by the change.
Quake III Arena
Demo001
Quake III Arena gets a small performance gain of 2.9 FPS when using DDR over SDR, although it's likely that the DDR really made no difference since the results are almost identical (within margin of error).
Jedi Knight II: Jedi Outcast
JK2FFA Demo
Even though Jedi Knight II and Quake III use the same engine, Jedi Knight II depends more heavily on the system speed than Quake III... probably due to more advanced shading features, higher polygon counts and ect. We see a much larger performance gain here with a difference of 8.0 FPS.
IV. CONCLUSION
The results of these tests should give clear answers to those seeking to know whether or not the jump to DDR is worth it.
Today's hardware is more than up to the task of providing fast access to common programs such as Word, Excel, PowerPoint, and other applications used for getting work done. That said, increasing RAM bandwidth by using DDR RAM will seem to make almost no difference whatsoever. When it comes to modern games, it seems that more work than ever is being forced onto the video card and off the CPU and RAM, reducing the amount of performance to be had from improving RAM bandwidth and the like (although that is not always the case).
However, there are a few reasons to make the upgrade to DDR. First, there's improved system stability (speaking of terms of using DDR on the K7S5A, that is), granted that one buys a decent stick of DDR RAM. The 2 x PC133 RAM sticks will no longer be an issue, as 2 x PC2100 RAM sticks have no problems whatsoever with this board. Secondly, when you decide to upgrade, you can leave out the cost of RAM since you already have exactly what you need. And finally, there's always the slight performance gain when using DDR RAM over SDR, for those of us who are very performance saavy.
To sum up, if you have SDR SDRAM and you are quite happy with the performance of your machine, there's no need to buy DDR SDRAM just because it scores better in a few benchmarks. Don't waste your money. On the other hand, if you like to have all the latest tech, want the greatest amount of performance possible out of your system, and like to "future-proof" yourself (so to speak), DDR SDRAM may be the answer for you.
Ultimately, it all depends on what you want. I hope this guide may have helped you decide what that is.
V. FAQ
Q: What's the difference between SDR and DDR SDRAM?
A: Among many smaller differences--both physical and performance wise--the big difference between DDR and SDRAM memory is that DDR reads data on both the rising and falling edges of the clock signal. SDR SDRAM only carries information on the rising edge of a signal. This feature allows the DDR module to transfer data twice as fast as SDR SDRAM. So your system is reading 266MHz instead of a data rate of 133MHz, the standard speed of SDR SDRAM.
Q: What are all these memory timings and settings I keep hearing about? What do they do?
A: Each of these settings in your BIOS determine how well (or fast) your RAM can complete it's tasks issued by other parts of your computer. A brief explination of each setting is provided below.
DRAM Timing Configuration offers little in the sense of performance gain, but may help in curing memory problems. With some memory modules, the timings of the modules are not correctly identified using the SPD (Serial Presence Detection). In some cases, this setting can be used to tell the BIOS approximately how fast the module is and what the timings for the module are. However, most quality memory modules that are produced today work just fine with SPD.
CAS Latency controls the time delay (in clock cycles) that passes before the SDRAM module starts to carry out a read command after receiving it. It also determines the number of clock cycles required for the completion of the first part of a burst transfer. In other words, the lower the latency, the faster memory reads can occur. However, some memory modules cannot handle such low latencies and may lose data, resulting in a system crash/failure. Most memory modules also have something called SPD, or Serial Presence Detection, which automatically determines the specified CAS latency of the memory module.
RAS Active Time is the amount of time a RAS can be kept open for multiple accesses. High figures will improve performance.
RAS Precharge Time sets the number of cycles required for the RAS (Row Address Strobe) to accumulate its charge before the SDRAM refreshes. Reducing the precharge time to 2 improves SDRAM performance. However, the precharge time of 2 may be insufficient for some SDRAM modules. In such cases, the module may not be refreshed properly and it may fail to retain data.
DRAM Driver Slew Rating affects the performance of DDR SDRAM. Set this setting to "Fast" for a slight performance gain.
Memory Termination was introduced in the the 020626 BIOS by ECS, and is assumed to be a fix for the ghost memory problem the motherboard thinking it had more RAM installed than it did (shows up in the POST screen as module 0,1 for example).
Q: Can I use higher speed memory (e.g. PC2700; PC3500) on the K7S5A even though the manual says it can support up to PC2100?
A: Yes, you can use higher speed memory on the K7S5A. However, the maximum speed the motherboard will use the memory at is 266MHz (PC2100). To take advantage of the higher speed memory, you will either need to get a board that supports higher speed memory, or you could try overclocking your K7S5A.
Q: What is this "2 x PC133 RAM" problem I've been hearing about?
Q: When I use two sticks of PC133 SDRAM, my computer randomly locks up, but when I use just one stick it doesn't freeze. What's the deal?
A: The K7S5A seems to randomly freeze when two sticks of PC133 SDR SDRAM are used. For example, a person with two sticks of 128MB PC133 RAM (it could be 64MB, 256MB, or 512MB memory modules... it seems the size doesn't matter) may have random lockups occur for no real apparent reason. ECS has done little to correct this problem, or even find it's source. Some people offer the advice that it is because 2 x PC133 RAM caused an adnormal amount of voltages to be dropped from the power supply. Thus, the only work-arounds for this problem are
1. Use only one stick of PC133 SDR RAM.
2. Use DDR RAM.
------------------------------------------------------------
Comments, corrections, questions and thanks (
) are welcome. E-mail ndow86@hotmail.com or post a reply.------------------------------------------------------------
~Fin
