Hi everyone. I decided to take the time to do an update on this guide. Apparently, a lot of people didn't like the first guide because (and I admit) it was too heavily bottlenecked by my old video card, a GeForce2 MX 400 PCI, and did not show a "real" comparison of DDR and SDR flavours of SDRAM.
However, I am now an owner of an MSI GeForce4 Ti 4200, and have some real comparitive results for everyone.
------------------------------------------------------------
Contents of this guide:
1. Introduction
2. Test Setup and Methodology
3. Benchmarks
4. Conclusion and Final Thoughts
1. Introduction
I first began this guide because when I first joined here, there were at least two to three threads a day in which users would ask the same question: "Is upgrading to DDR really worth the price?" Well, at that point in time there did not seem to be any definitive answer for this question. Many believed that the performance gain from DDR SDRAM warranted the price difference, but just as many believed that SDR SDRAM was still good enough to keep in their systems.
After much thought, I decided to put together a small guide comparing the performance differences between SDR and DDR types of SDRAM. The guide itself goes from benchmarking raw data bandwidth to some of the latest games on the market.
This is the update to that guide, with easy-to-read bar graphs, better detailed explanations of results and much more comparable results. Hopefully, this guide will become a good reference answer to the original question.
2. Test Setup and Methodology
For this guide, I used my very own home computer which is based on the K7S5A.
Hardware
Power Supply: 550W Antec True Control
CPU: AMD AthlonXP 2100+ @ 1740MHz
Motherboard: ECS K7S5A Pro v5.0 w/ HX OC 021209 BIOS
Video card: nVIDIA GeForce4 Ti 4200 64MB AGP
Sound card: Creative Soundblaster PCI 512
Modem: U.S. Robotics 56K v.90 hardware
HDD: Quantum Fireball LCT 10.0GB
RAM: 1 x 256MB Mitsubishi PC133 (133MHz) SDR SDRAM
1 x 256MB Samsung PC2100 (266MHz) DDR SDRAM
Software
OS: Windows 2000 Professional SP4 (with all critical and recommended updates installed)
DirectX: 9.0a
Video drivers: nVIDIA reference driver v44.03
AGP drivers: SiS AGP driver v1.16a
Test Methodology
All game tests were run at 1024x768 with maximum detail and no anti-aliasing or ansiotropic filtering enabled unless otherwise specified. V-Sync was disabled during benchmarking. The SDR SDRAM stick was set to CAS 3 throughout the test, while the DDR SDRAM stick was set to CAS 3 and CAS 2.5 for comparitive purposes. The other memory settings were set to Normal, 6T and 3T. A thorough system cleaning and deframentation was performed before benchmarking. Benchmarks were run three times and then averaged to make a final score. After each benchmark, the system was rebooted.
3. Benchmarks
I'll first start with memory bandwidth benchmarks to show what kind of raw bandwidth gain users can expect when migrating from SDR to DDR.
As we can see from the SANDRA memory bandwidth test, users can expect a very sizeable gain of raw memory bandwidth by upgrading to DDR. However, under normal circumstances, PC2100 DDR SDRAM usually never hits its theoretical bandwidth of 2100 MB/s... explaining why the graph shows results of about 1800 MB/s.
PCMark also shows the same results, gaining almost 1000 points more in the memory test. No surprises here. What we really want to know is how much this extra bandwidth will benefit in games.
Quake 3 Arena seems to enjoy the extra bandwidth, gaining roughly 16 FPS when using DDR SDRAM. It's interesting to note that the lower CAS latency only gives us a gain of .1 FPS which is within the margin of varience.
It's not surprising to see JK2 do almost equally as well, since both games use the Quake 3 engine. It is, however, surprising to see that JK2 actually lost some FPS when using a lower CAS latency.
Flyby gains eight FPS and Botmatch only gains about three. Still, any speed gain is a nice one... especially in a game as taxing as UT2003. There is no difference between altering CAS latencies here.
Easily the most popular benchmark around, I had to include 3DMark2001 in the guide... just for completeness
. Here we see a pretty nice boost in overall score, although hardly mindblowing. 617 points means A LOT to some people though, so it's a welcome boost.
Everyone's favorite benchmark of controversy also seems to like the upgrade to DDR SDRAM. Here we see a gain of 91 points. Considering the fact that 3DMark03 is very GPU limited, 91 points is nothing to snuff at since the hardware changed did not involve a better video card.
I'll finish with WinRAR compression time. Here we see DDR SDRAM shaving off about 10 seconds worth of time when compressing a large file.
4. Conclusion and Final Thoughts
Make no mistake: DDR SDRAM will definately give you a performance boost when upgrading from SDR SDRAM. Period.
However, the question to ask yourself is "Just how much of a boost am I expecting?"
Well, if you think the upgrade to from SDR to DDR is going to give you a 100 FPS boost in every game, I'm afraid you're sadly mistaken. It will, however, give you a nice performance boost in many older games such as Quake III Arena, and a slight boost in many newer games such as UT2003. In the end, it's up to the user to decide if that boost warrants the price tag of DDR.
Considering that prices on RAM (especially DDR type) are hitting an all-time low, when building a new system, DDR is really not optional, but a must-have component. The price difference between SDR and DDR is almost completely diminished and there is really no reason why a user should settle for SDR. The performance is obviously better and the older SDR-type is beginning to phase out.
On the other hand, some users may feel that they do not need the extra performance (which is perfectly fine). Thus, the upgrade to DDR SDRAM is just wasted money. They feel that SDR SDRAM performs fine for what they do on their computer, and there's no need for extra speed. Furthermore, when you really pull the figures out of context, who really needs an extra 16 FPS when both types of RAM are performing well above 100 FPS? Is 10 seconds really that important when compressing files? In many users minds, it is not.
My personal recommendation is this: Anyone who is building a completely new system (read: not using parts from an old computer) should use DDR SDRAM. There's no real point in saving $5 by going with SDR SDRAM. However, if you are comfortable with the performance you have right now with SDR SDRAM, or you are building a new computer from old computer parts, the upgrade to DDR may not be in your best interests. Ultimately, it's your call if you want to upgrade.
However, I am now an owner of an MSI GeForce4 Ti 4200, and have some real comparitive results for everyone.
------------------------------------------------------------
Contents of this guide:
1. Introduction
2. Test Setup and Methodology
3. Benchmarks
4. Conclusion and Final Thoughts
1. Introduction
I first began this guide because when I first joined here, there were at least two to three threads a day in which users would ask the same question: "Is upgrading to DDR really worth the price?" Well, at that point in time there did not seem to be any definitive answer for this question. Many believed that the performance gain from DDR SDRAM warranted the price difference, but just as many believed that SDR SDRAM was still good enough to keep in their systems.
After much thought, I decided to put together a small guide comparing the performance differences between SDR and DDR types of SDRAM. The guide itself goes from benchmarking raw data bandwidth to some of the latest games on the market.
This is the update to that guide, with easy-to-read bar graphs, better detailed explanations of results and much more comparable results. Hopefully, this guide will become a good reference answer to the original question.
2. Test Setup and Methodology
For this guide, I used my very own home computer which is based on the K7S5A.
Hardware
Power Supply: 550W Antec True Control
CPU: AMD AthlonXP 2100+ @ 1740MHz
Motherboard: ECS K7S5A Pro v5.0 w/ HX OC 021209 BIOS
Video card: nVIDIA GeForce4 Ti 4200 64MB AGP
Sound card: Creative Soundblaster PCI 512
Modem: U.S. Robotics 56K v.90 hardware
HDD: Quantum Fireball LCT 10.0GB
RAM: 1 x 256MB Mitsubishi PC133 (133MHz) SDR SDRAM
1 x 256MB Samsung PC2100 (266MHz) DDR SDRAM
Software
OS: Windows 2000 Professional SP4 (with all critical and recommended updates installed)
DirectX: 9.0a
Video drivers: nVIDIA reference driver v44.03
AGP drivers: SiS AGP driver v1.16a
Test Methodology
All game tests were run at 1024x768 with maximum detail and no anti-aliasing or ansiotropic filtering enabled unless otherwise specified. V-Sync was disabled during benchmarking. The SDR SDRAM stick was set to CAS 3 throughout the test, while the DDR SDRAM stick was set to CAS 3 and CAS 2.5 for comparitive purposes. The other memory settings were set to Normal, 6T and 3T. A thorough system cleaning and deframentation was performed before benchmarking. Benchmarks were run three times and then averaged to make a final score. After each benchmark, the system was rebooted.
3. Benchmarks
I'll first start with memory bandwidth benchmarks to show what kind of raw bandwidth gain users can expect when migrating from SDR to DDR.
As we can see from the SANDRA memory bandwidth test, users can expect a very sizeable gain of raw memory bandwidth by upgrading to DDR. However, under normal circumstances, PC2100 DDR SDRAM usually never hits its theoretical bandwidth of 2100 MB/s... explaining why the graph shows results of about 1800 MB/s.
PCMark also shows the same results, gaining almost 1000 points more in the memory test. No surprises here. What we really want to know is how much this extra bandwidth will benefit in games.
Quake 3 Arena seems to enjoy the extra bandwidth, gaining roughly 16 FPS when using DDR SDRAM. It's interesting to note that the lower CAS latency only gives us a gain of .1 FPS which is within the margin of varience.
It's not surprising to see JK2 do almost equally as well, since both games use the Quake 3 engine. It is, however, surprising to see that JK2 actually lost some FPS when using a lower CAS latency.
Flyby gains eight FPS and Botmatch only gains about three. Still, any speed gain is a nice one... especially in a game as taxing as UT2003. There is no difference between altering CAS latencies here.
Easily the most popular benchmark around, I had to include 3DMark2001 in the guide... just for completeness
. Here we see a pretty nice boost in overall score, although hardly mindblowing. 617 points means A LOT to some people though, so it's a welcome boost.Everyone's favorite benchmark of controversy also seems to like the upgrade to DDR SDRAM. Here we see a gain of 91 points. Considering the fact that 3DMark03 is very GPU limited, 91 points is nothing to snuff at since the hardware changed did not involve a better video card.
I'll finish with WinRAR compression time. Here we see DDR SDRAM shaving off about 10 seconds worth of time when compressing a large file.
4. Conclusion and Final Thoughts
Make no mistake: DDR SDRAM will definately give you a performance boost when upgrading from SDR SDRAM. Period.
However, the question to ask yourself is "Just how much of a boost am I expecting?"
Well, if you think the upgrade to from SDR to DDR is going to give you a 100 FPS boost in every game, I'm afraid you're sadly mistaken. It will, however, give you a nice performance boost in many older games such as Quake III Arena, and a slight boost in many newer games such as UT2003. In the end, it's up to the user to decide if that boost warrants the price tag of DDR.
Considering that prices on RAM (especially DDR type) are hitting an all-time low, when building a new system, DDR is really not optional, but a must-have component. The price difference between SDR and DDR is almost completely diminished and there is really no reason why a user should settle for SDR. The performance is obviously better and the older SDR-type is beginning to phase out.
On the other hand, some users may feel that they do not need the extra performance (which is perfectly fine). Thus, the upgrade to DDR SDRAM is just wasted money. They feel that SDR SDRAM performs fine for what they do on their computer, and there's no need for extra speed. Furthermore, when you really pull the figures out of context, who really needs an extra 16 FPS when both types of RAM are performing well above 100 FPS? Is 10 seconds really that important when compressing files? In many users minds, it is not.
My personal recommendation is this: Anyone who is building a completely new system (read: not using parts from an old computer) should use DDR SDRAM. There's no real point in saving $5 by going with SDR SDRAM. However, if you are comfortable with the performance you have right now with SDR SDRAM, or you are building a new computer from old computer parts, the upgrade to DDR may not be in your best interests. Ultimately, it's your call if you want to upgrade.
