Memory Test Configuration
This comparison of OCZ DDR3-1800 to other DDR3 memory uses a completely different test methodology than the earlier DDR3 reviews. Our reviews of DDR3 vs. DDR2, Intel P35 Memory Performance, Kingston DDR3, and Super Talent and TEAM DDR3-1600 were based on a Core 2 Duo 2.66GHz CPU with a common memory speed at 1066 and 1333 speeds. Since ratios were not available for higher speeds like 1600 and 2000 those speeds were achieved by overclocking the base system. This made comparing performance at the same processor speed, a standard procedure in our earlier DDR2 and DDR memory tests, all but impossible. The rapid ramping of DDR3 memory speeds coupled with the lack of ratios made those memory speed comparisons less useful.
New Test Methodology
Readers suggested a number of memory testing ratios and speeds which would allow a better comparison of memory performance. After considering the suggestions and the real ratios available the decision was made to settle on a new 3.0GHz test speed driven by a processor that could deliver multipliers to at least 9X and FSB speeds to 550+. This would allow DDR3-800, DDR3-1066, and DDR3-1333 to be tested at 3.0GHz at 9x333 settings. DDR3-1600 and DDR3-2000 would be tested at 3.0GHz at 6x500.
The requirement of a 9x or higher multiplier combined with the ability to run with stability at a 550+ FSB proved more difficult than anticipated. A very early Intel X6800 had unlocked multipliers but would not even boot above about 450 FSB. A very recent Q6700 met the 9X requirement but topped out in FSB at about 470 stable. Tests by other AnandTech editors confirmed that most quad core processors have a difficult time operating at 500 FSB, let alone the higher 550 requirements for overclocks above DDR3-2000. We finally located a recent X6800 that could do both the 9X multiplier and the 550 FSB requirements. This became our new processor for the memory test bed. Other components remained the same as the earlier DDR3 test bed.
It should be pointed out that the new test bed does use the same processor speed at all tested RAM speeds, but that different FSB speeds are used at low and high memory timings. Testing at the same processor speed does allow a better comparison of isolated memory performance, but it is not an exact apples-to-apples comparison. The variation in FSB speed at the same processor speed does affect the performance of the memory at the higher FSB speeds. This ranges from negligible in some tests to measurable in other benchmarks. For more information on the impact of FSB speed on performance you can refer to Intel P35 Memory Performance: A Closer Look.
A true apples-to-apples comparison of memory performance will only be possible when the additional ratios of 1600 and 2000 are available in the BIOS at a base 1333 memory speed. Still, while not perfect, the comparison of all memory speeds at 3.0GHz processor speed is much closer to our testing ideal than the previous test methodology.
All DDR3 Memory Retested
With the change in processor and base speed to 3.0GHz, all DDR3 memory was retested to provide data for comparison of all memory speeds at the same 3.0GHz processor speed. We have established high-performance DDR2 memory performance baselines and we will no longer be actively testing DDR2 memory performance for our DDR3 reviews. You can see the comparison of DDR2 and DDR3 on the next page, DDR2/DDR3 Overlap Speeds. Additional benchmarks were run to provide the most complete comparison at the common RAM speeds of 1066 and 800 MHz.
Past performance tests of DDR2 memory on the Intel P965 and P35-DDR2 platforms are included for comparison. All current memory tests use the Intel P35-DDR3 test bed (ASUS P5K3 Deluxe) with the DDR3 memory under evaluation. As detailed on the results pages, overlap speeds were tested at 2.66GHz for compatibility with previous results. The DDR3 full performance pages include results at 3.0GHz, and also include a retest of the 1066 and 800 memory speeds at 3.0GHz CPU speeds.
The CPUs listed above in our table are 1066 FSB processors, but all ran fine at 1333 FSB at default multiplier and default voltage. New 1333 FSB processors with the same ratios may be substituted for these processors as soon as they are available, providing they meet the requirements of 9X or greater multiplier and stable operation at 550FSB..
This comparison of OCZ DDR3-1800 to other DDR3 memory uses a completely different test methodology than the earlier DDR3 reviews. Our reviews of DDR3 vs. DDR2, Intel P35 Memory Performance, Kingston DDR3, and Super Talent and TEAM DDR3-1600 were based on a Core 2 Duo 2.66GHz CPU with a common memory speed at 1066 and 1333 speeds. Since ratios were not available for higher speeds like 1600 and 2000 those speeds were achieved by overclocking the base system. This made comparing performance at the same processor speed, a standard procedure in our earlier DDR2 and DDR memory tests, all but impossible. The rapid ramping of DDR3 memory speeds coupled with the lack of ratios made those memory speed comparisons less useful.
New Test Methodology
Readers suggested a number of memory testing ratios and speeds which would allow a better comparison of memory performance. After considering the suggestions and the real ratios available the decision was made to settle on a new 3.0GHz test speed driven by a processor that could deliver multipliers to at least 9X and FSB speeds to 550+. This would allow DDR3-800, DDR3-1066, and DDR3-1333 to be tested at 3.0GHz at 9x333 settings. DDR3-1600 and DDR3-2000 would be tested at 3.0GHz at 6x500.
The requirement of a 9x or higher multiplier combined with the ability to run with stability at a 550+ FSB proved more difficult than anticipated. A very early Intel X6800 had unlocked multipliers but would not even boot above about 450 FSB. A very recent Q6700 met the 9X requirement but topped out in FSB at about 470 stable. Tests by other AnandTech editors confirmed that most quad core processors have a difficult time operating at 500 FSB, let alone the higher 550 requirements for overclocks above DDR3-2000. We finally located a recent X6800 that could do both the 9X multiplier and the 550 FSB requirements. This became our new processor for the memory test bed. Other components remained the same as the earlier DDR3 test bed.
It should be pointed out that the new test bed does use the same processor speed at all tested RAM speeds, but that different FSB speeds are used at low and high memory timings. Testing at the same processor speed does allow a better comparison of isolated memory performance, but it is not an exact apples-to-apples comparison. The variation in FSB speed at the same processor speed does affect the performance of the memory at the higher FSB speeds. This ranges from negligible in some tests to measurable in other benchmarks. For more information on the impact of FSB speed on performance you can refer to Intel P35 Memory Performance: A Closer Look.
A true apples-to-apples comparison of memory performance will only be possible when the additional ratios of 1600 and 2000 are available in the BIOS at a base 1333 memory speed. Still, while not perfect, the comparison of all memory speeds at 3.0GHz processor speed is much closer to our testing ideal than the previous test methodology.
All DDR3 Memory Retested
With the change in processor and base speed to 3.0GHz, all DDR3 memory was retested to provide data for comparison of all memory speeds at the same 3.0GHz processor speed. We have established high-performance DDR2 memory performance baselines and we will no longer be actively testing DDR2 memory performance for our DDR3 reviews. You can see the comparison of DDR2 and DDR3 on the next page, DDR2/DDR3 Overlap Speeds. Additional benchmarks were run to provide the most complete comparison at the common RAM speeds of 1066 and 800 MHz.
| Memory Performance Test Configuration | |
| Processor | Intel Core 2 Duo X6800 (x2, 2.93GHz unlocked, 4MB Unified Cache) 9x333 - 3.0 GHz 8x500 - 3.0 GHz |
| RAM | OCZ PC3-14400 Platinum (2GB kit - 2x1GB, DDR3-1800 8-8-8) Super Talent W1600UX2G7 (2GB kit - 2x1GB, DDR3-1600 7-7-7) Team TXD31924M1600HC9 (2GB kit - 2x1GB, DDR3-1600 9-9-9) Kingston KHX11000D3LLK2 (2GB kit - 2x1GB, DDR3-1333 7-7-7) Corsair CM3X1024-1066C7 (2GB Kit - 2x1GB- DDR3-1066 7-7-7) Corsair Dominator CM2X1024-8888C4 (2GB Kit - 2x1GB - DDR2-1250 5-5-5) |
| Hard Drive | Samsung 250GB SATA2 enabled (8MB Buffer) |
| System Platform Drivers | Intel - 8.3.0.1013 |
| Video Card | Leadtek WinFast 7950GT - 256MB |
| Video Drivers | NVIDIA 93.71 |
| CPU Cooling | Intel Retail HSF |
| Power Supply | Corsair HX620W |
| Motherboards | ASUS P5K3 Deluxe (Intel P35 DDR3) ASUS P5K Deluxe (Intel P35 DDR2) ASUS P5B Deluxe (Intel P965 DDR2) |
| BIOS Revision | 0604 (6/26/2007) |
| Operating System | Windows XP Professional SP2 |
Past performance tests of DDR2 memory on the Intel P965 and P35-DDR2 platforms are included for comparison. All current memory tests use the Intel P35-DDR3 test bed (ASUS P5K3 Deluxe) with the DDR3 memory under evaluation. As detailed on the results pages, overlap speeds were tested at 2.66GHz for compatibility with previous results. The DDR3 full performance pages include results at 3.0GHz, and also include a retest of the 1066 and 800 memory speeds at 3.0GHz CPU speeds.
The CPUs listed above in our table are 1066 FSB processors, but all ran fine at 1333 FSB at default multiplier and default voltage. New 1333 FSB processors with the same ratios may be substituted for these processors as soon as they are available, providing they meet the requirements of 9X or greater multiplier and stable operation at 550FSB..
Facebook
Twitter
RSS
25 Comments
View All Comments
Lonyo - Tuesday, July 31, 2007 - link
Any chances of a power consumption comparison between DDR2 and DDR3?DDR3 is supposed to run at a lower voltage, so in theory it might use a little less power. Would be interesting to see if there is any difference (DDR2/3-800 would probably be best, since that's a standard speed for both).
yyrkoon - Tuesday, July 31, 2007 - link
Does anyone even sell a DDR3 capable motherboard yet ? If so, is anyone even using DDR3 ? Personally, I think latencies need to come down, Prices need to come down,etc. Memory companies are *claiming* they are taking a beating in the market for DDR2 (claiming all time low, and losing money . . .).Personally, I think you reap what you soe, and they got what they deserved for their early market prices.Anyhow, short and skinny, I think *we* all need to take things slowly this time around, OEMS, buyers, and reviewers . . .
yyrkoon - Tuesday, July 31, 2007 - link
By the way, when I asked if anyone is even making a DDR3 motherboard yet, I was pretty much joking. Obviously if you're testing it, there has to be some form of a platform availible.You know, I cannot help but think that DDR2 was not quite 'finished' yet, and I do not understand the *need* for DDR3(unless OEMs are looking to rape our wallets again . . .). Of course, if 'Joe blow enthusiast' HAS to HAVE DDR3 memory because it gives him/her an extra 4-13 FPS in an outdated game at 2-3x the cost of DDR2 . . . well... lets just say that I expect that OCZ, Geil, and the rest would be more than happy to keep you poorer ;)
Some of us actually like to upgrade smart, using as many parts from older machines as possible to save money for other things. This sort of marketing strategy makes it hard on us who would like to do so while keeping our system upgraded once a year or so. I just got over having to buy memory, CPU, and a motherboard the last 'technolgy' advance, and I really do not wish to repeat the process.
asliarun - Tuesday, July 31, 2007 - link
Man, I never understand viewpoints such as yours. This is a technology article on the latest DDR3 advancement, and is not marketing propoganda urging you to go out and buy it NOW. Intel's latest CPU chipsets (P35/P38, IIRC) all support DDR3 (along with DDR2), so it's not like DDR3 is exactly vaporware. Only AMD is not supporting DDR3 right now because firstly, they will need to upgrade their integrated uncore memory controller, and secondly, they tend support upcoming technologies much later than Intel. Furthermore, DDR3 is definitely the future as it has much more headroom than DDR2, and is designed to work at lower voltages.In any case, my point is that we're discussing a new memory standard technology which is already in the market and is slowly being adopted. Initially, it WILL be highly priced like any other technology until volume manufacturing kicks in. However, if you are a price sensitive customer instead of a "Joe blow enthusiast" (frankly, like most of us), no one is forcing you to replace your RAM with DDR3 TODAY, least of all this AT article. Your logic of not adopting DDR3 simply because it is initially expensive and because it only gives "4-13fps increase" is however, absurd. By the same token, there is no need for ANY technology improvement, especially those that *only* result in an incremental improvement.
As a footnote, you should be grateful for all the "Joe blow enthusiasts" in this world instead of heaping scorn on them. All said and done, you and I ARE freeloading on him, as he's the one who's financing our cut rate technology purchases.
GlassHouse69 - Tuesday, July 31, 2007 - link
Oh you think so?hm.... i wonder how much Anandtech/daily got for reviewing this... hm.....