Part 1 of the 2GB DDR Kit Roundup took a close look at 3 memories based on Infineon memory chips. Since that review in October, a number of 2GB DDR kits have appeared in the market with most based on Infineon C or B die memory chips. There are, however, a few brands that have taken a different approach to memory chips for 1GB DIMMs, and we have included those in this roundup, along with the latest Infineon-based memories.
1GB DDR DIMMs began appearing in the market over a year ago, but it wasn't until mid-2005 that 1GB DIMMs with reasonably fast timings became widely available. These faster 1GB DIMMs finally made the choice of 2GB memory kits to be a reasonable choice in a market that had been dominated by fast DDR 512MB DIMMs.
There are many reasons to choose a 2GB kit over a 1GB kit or 4 512MB DIMMs. Two 1GB DIMMs on the AMD Athlon 64 can still run at 1T Command rate, instead of the 2T required by the 4x512MB DIMMs needed for 2GB with 512MB DIMMs - a definite advantage for the 1GB DIMMs. On the other hand, until recently, the available 1GB DIMMs were generally much slower than the fast 2-2-2 DIMMs that were commonly available in 512MB DIMMs. We normally saw 3-3-3 or 3-4-4 or slower timings for 1GB DIMMs. These poorer timings for 1GB DIMMs took away most of the advantage for the 1GB 1T Command Rate.
There was an additional "gotcha" with the 1GB DIMMs that many enthusiasts quickly discovered. With a starting point of 3-3-3 or 3-4-4 at DDR400, the 1GB parts did not overclock nearly as far as the 512MB parts. For all of these reasons, we generally recommended that most users were better off with 512MB DIMMs - at least until memory timings improved on the 1GB DIMMs.
The time for faster 1GB DIMMs has finally come in the past 6 months, and they are now available from almost every memory manufacturer. In Part 1, we looked at three 2GB kits from Corsair, Gigaram, and OCZ. In this part 2, we put six additional fast 2GB kits through our test bench, with some interesting results. We also updated some parts of our memory test bench, which required retesting of the original three 2GB kits.
Our memory tests differentiate memory in two ways. First, AnandTech has always been an advocate of real world performance measurements, and we've shunned using just synthetic benchmarks in our testing of every type of component, including memory. This is not because synthetic benchmarks are not useful - they are often very revealing of component differences - but rather, it is because running just synthetic benchmarks can severely distort the picture of performance with real applications and real games. That is why we always use games and the pure number-crunching Super Pi in our memory tests. It is also the reason why we test using both Buffered (Standard) and Unbuffered synthetic benchmarks. We have found in much of our testing that the less commonly used Unbuffered benchmarks mirror more closely how games really respond to memory differences.
Second, we moved to testing different memory speeds at the same CPU clock speed in our Athlon 64 memory tests. The AMD CPU, with unlocked multipliers, allowed us to finally remove the CPU speed differences from our memory tests. This allows you to finally see the true impact of memory speed increases and memory timings on performance. As you have seen in past reviews, those performance differences are very real, although they are much smaller than what many memory manufacturers might want you to believe. On the other hand, faster memory speeds and faster memory timings do improve performance, no matter what some nay-sayers are determined to prove.
1GB DDR DIMMs began appearing in the market over a year ago, but it wasn't until mid-2005 that 1GB DIMMs with reasonably fast timings became widely available. These faster 1GB DIMMs finally made the choice of 2GB memory kits to be a reasonable choice in a market that had been dominated by fast DDR 512MB DIMMs.
There are many reasons to choose a 2GB kit over a 1GB kit or 4 512MB DIMMs. Two 1GB DIMMs on the AMD Athlon 64 can still run at 1T Command rate, instead of the 2T required by the 4x512MB DIMMs needed for 2GB with 512MB DIMMs - a definite advantage for the 1GB DIMMs. On the other hand, until recently, the available 1GB DIMMs were generally much slower than the fast 2-2-2 DIMMs that were commonly available in 512MB DIMMs. We normally saw 3-3-3 or 3-4-4 or slower timings for 1GB DIMMs. These poorer timings for 1GB DIMMs took away most of the advantage for the 1GB 1T Command Rate.
There was an additional "gotcha" with the 1GB DIMMs that many enthusiasts quickly discovered. With a starting point of 3-3-3 or 3-4-4 at DDR400, the 1GB parts did not overclock nearly as far as the 512MB parts. For all of these reasons, we generally recommended that most users were better off with 512MB DIMMs - at least until memory timings improved on the 1GB DIMMs.
The time for faster 1GB DIMMs has finally come in the past 6 months, and they are now available from almost every memory manufacturer. In Part 1, we looked at three 2GB kits from Corsair, Gigaram, and OCZ. In this part 2, we put six additional fast 2GB kits through our test bench, with some interesting results. We also updated some parts of our memory test bench, which required retesting of the original three 2GB kits.
Our memory tests differentiate memory in two ways. First, AnandTech has always been an advocate of real world performance measurements, and we've shunned using just synthetic benchmarks in our testing of every type of component, including memory. This is not because synthetic benchmarks are not useful - they are often very revealing of component differences - but rather, it is because running just synthetic benchmarks can severely distort the picture of performance with real applications and real games. That is why we always use games and the pure number-crunching Super Pi in our memory tests. It is also the reason why we test using both Buffered (Standard) and Unbuffered synthetic benchmarks. We have found in much of our testing that the less commonly used Unbuffered benchmarks mirror more closely how games really respond to memory differences.
Second, we moved to testing different memory speeds at the same CPU clock speed in our Athlon 64 memory tests. The AMD CPU, with unlocked multipliers, allowed us to finally remove the CPU speed differences from our memory tests. This allows you to finally see the true impact of memory speed increases and memory timings on performance. As you have seen in past reviews, those performance differences are very real, although they are much smaller than what many memory manufacturers might want you to believe. On the other hand, faster memory speeds and faster memory timings do improve performance, no matter what some nay-sayers are determined to prove.
Facebook
Twitter
RSS
51 Comments
View All Comments
JarredWalton - Monday, January 23, 2006 - link
I don't anticipate DDR1 prices changing much, and once AM2 comes out I expect prices to start climbing as DDR1 productions halts. DDR1 prices seem to have hit bottom about 6 months ago, and they're already starting to rise - at least on the budget parts.Thor86 - Monday, January 23, 2006 - link
Did I miss this in the review? If not, it would help if you told us what memory settings used in the DFI bios for these sticks reviewed.Wesley Fink - Tuesday, January 24, 2006 - link
I used the 7/06 BIOS, which is basically the Bigtoe BIOS. I have found this BIOS to be remarkably compatible with a wide range of memory. All memory settings were left at default except the 4 reported memory timings (CAS to TRAS) and memory voltage in testing all the memory. We did not do any special tweaking, which is another reason we reran tests with the 3 2GB kits tested earlier. You can definitely achieve better bandwidth and higher scores than we did by tweaking memory in the DFI BIOS.keldog7 - Monday, January 23, 2006 - link
I was surprised to find your review stating the Corsair 3500LL could only reach 492 MHz...especially when out of the box, I clocked mine to 500 MHz. This required no special tweaking of the other timings, other than what's listed below. Any careful research on this memory, using the many online reviews of it, show that the RAM runs fine (at slightly relaxed timings), well into 500Mhz territory, and beyond.In my case, I've got it on an A8N32SLI, at 2.7V, running 2.5-3-2-7 1T with a 1:1 divider at 250Mhz. In my case, this has been verfied to be Memtest86+ stable for at least 12 hours (last run was 90 hours...then I had to stop it to actually get some work done!)
I wonder if the motherboard used in your review has biased your overclocking results?
Wesley Fink - Monday, January 23, 2006 - link
According to my review notes, the Corsair 3500LL Pro cloked to almost 260 (DDR520) on Super Pi and Sandra memory tests, but the highest we could do with stability on 3D tests was 246 (DDR492). There is always varaition in memory results.Since you criticize my results as being too low for your memory and another reader criticized my results as being too high for his memory, it is likely my results are typical of the range of results readers might achieve. Some will do better than my results and some will do worse. Overclocking results do vary among memory samples.
wildstaroct - Monday, January 23, 2006 - link
Unless Anandtech got really good batches of Infineon ICs, I don't buy the analysis. There have been numerous reports of 3D instability issues...I've experienced them myself with the Redlines. They are primex2 and memtest stable at fairly high speeds (270) but only 3D stable near spec (250ish).entrecote - Friday, January 27, 2006 - link
Neither do i buy it.I invested in a pair of the mentioned 2GB Mushkin Redline PC4000 memory. I´ve stubbornly been trying to tweak them on my DFI Lanparty + X2 Toledo processor for two months. This Redline kit of mine refuses to surpass 252MHz / DDR504. One module does 256MHz / DDR512, the other one does 252MHz / DDR504 regardless of timings. Trust me, I´ve tried all combinations of different drive strenghts, Trefs and so on. My case is not isolated, I have the same symptoms as the perhaps majority of 2x1GB Redline owners, "3D instability".
My thoughts on the matter is that Mushkin was aware of this bad batch of theirs, possibly due to quality fluctuations of infineon modules. Then they quickly began to separate a few review samples for marketing puposes. I went through the same dissapointment one year ago when buying PC3200 Ballistixs after quickly reading through among other reviews anadtechs infamous "Ballistix cherry pick" DDR roundup.
I put great trust in anandtech, please do me a favor in the future to only review retail samples of memory. Some of these mentioned brands purposefully send special samples to rewievers.
Wesley Fink - Monday, January 23, 2006 - link
We reported what we found, and 3D tests are part of our benchmarking. The OCZ for example was Super Pi and memtest stable to 290, but 3D dropped the stability to 275 (DDR550) which we reported. The Redline went well over 300 on memtest and Super Pi, but 293 (DDR586) was our highest 3D stable.What memory controller revision are you using with the Infineons? Our tests are with a Rev. E, and we have seen lower performance on the older clawhammer memory controllers with the 1GB dimms.
wildstaroct - Monday, January 23, 2006 - link
I have a rev E6 Opteron 170. Perhaps I just got an unlucky set of sticks, but the forums are loaded w/ people with similar issues, particularly around 3D.Duonger - Monday, January 23, 2006 - link
the mushkin parts have been back in stock for over a week now and is available (if not sold out already) on Newegg and on Mushkin`s Site.Wes- the ocz parts u compared. those are 2x512mb kits. why did u include it on this 2gig round up?