We proceed with the low-level analysis of the most important characteristics of high-performance DDR2 memory modules using our RightMark Memory Analyzer. Today we are going to review the first offer in the highest-speed unofficial category — DDR2-1066 (PC2-8500) — dual-channel kit of Corsair XMS2-8500 memory modules, a logical continuation of high-speed XMS2-series solutions. We have already reviewed XMS2-8000UL modules from this series of the DDR2-1000 class. Frankly speaking, DDR2-1066 solutions (for example, from A-DATA) have been nominally known from the middle of the last year. Nevertheless, our modules under review are the first offer on the market, which can really operate in the unofficial DDR2-1066 mode, that is with the memory bus clock of 533 MHz.
Module manufacturer: Corsair Memory
Photo of the memory module
Module Part Number
Module Part Number Expansion
The manufacturer's web site does not publish the DDR2 Part Number expansion of XMS2-series memory modules. The booklet about TWIN2X1024-8500 specifies that this product is a kit of two 512MB modules (Module Part Number - CM2X512-8500), based on eight 64M x8 chips. The manufacturer guarantees operation of these modules in the fastest (so far unofficial) DDR2-1066 mode with 5-5-5-15 timings and 2.2 V voltage. But the default mode written in the SPD chip is the maximum standard mode - DDR2-800 with 5-5-5-15 timings.
SPD chip data
Description of the general SPD standard:
Description of the specific SPD standard for DDR2:
SPD data are not quite standard, but usual for Corsair modules. The fastest mode these modules are capable of is characterized by the cycle time of 2.5 ns (400 MHz, DDR2-800). This mode corresponds to the first supported value of tCL = 5, the full timing scheme is written as 5-5-5-18, which slightly disagrees with official characteristics, published in the datasheet on these modules (DDR2-800, 5-5-5-15). Reduced CAS# latency (CL X-1 = 4) corresponds to the non-standard cycle time of 3.7 ns, 270 MHz — it probably means DDR2-533 and the cycle time of 3.75 ns. Nevertheless, the incorrect cycle time results in fractional timings, which can be written (rounded to one figure after the dot) as 4-3.4-3.4-12.2, which will most likely be rounded up by most BIOS's to 4-4-4-13. Nevertheless, it should be noted that you will hardly need such a slow mode as DDR2-533 for these high-performance modules.
Manufacturer’s JEDEC ID Code and Part Number of the module are correct. Nevertheless, as in the other Corsair modules, the SPD chip of these modules lacks information about the manufacturing date and serial number.
From this review on we upgrade to RMMA 3.65 with increased size of the allocated memory block and consequently modified default parameters in the majority of subtests. In particular, the memory block size in subtests for analyzing memory characteristics has been increased to 32 MB to reduce the effect of large L2 Cache (2 MB), which is often used in modern processors of the Pentium 4/Pentium D class. In this respect, characteristics listed below are much more precise in reflecting real memory bandwidth values and cannot be compared on the quantitative level with characteristics, published in the earlier reviews.
As usual, in the first series of tests we used the timing scheme, set in BIOS Setup by default (Memory Timings: "by SPD"). The modules were tested in three modes — DDR2-667 with 200 MHz and 266 MHz FSB (memory multipliers - 1.67 and 1.25), DDR2-800 with 200 MHz and 266 MHz FSB (memory multipliers - 2.0 and 1.5), as well as unofficial "DDR2-1066" with 266 MHz FSB and the maximum memory multiplier possible - 2.0.
BIOS on the testbed set 5-5-5-15 timings by default for DDR2-667 ("at random", as this mode is not written in the SPD chip), DDR2-800 mode — 5-5-5-18 (strictly by SPD), and finally the fastest DDR2-1066 mode — 5-6-6-18. As we know from practice, it's the slowest timing scheme the ASUS P5WD2-E motherboard (we use it for testing high-speed memory modules) can set by default.
These modules demonstrate good performance results — maximum memory bandwidth values in official modes are approximately 6.4-6.5 GB/s and 8.2-8.7 GB/s at the FSB clock of 200 MHz and 266 MHz correspondingly. That is they almost equal the maximum theoretical bandwidth of the 200 MHz and 266 MHz FSB (it's actually a tad higher due to the effect of 2 MB L2 Cache, which can "cover" 2MB/32MB * 100% = 6.25% of memory access). Note that the maximum real memory read bandwidth both with 200 MHz and 266 MHz FSB grows much when we go from DDR2-667 to the fastest DDR2-1066 mode.
These modules are also on a par with previous high-speed solutions from Corsair and other manufacturers in latencies. Together with the above mentioned increase in memory bandwidth, faster modes (going from DDR2-667 to DDR2-800 and DDR2-1066) as well as higher-clocked FSB (going from 200-MHz to 266-MHz FSB) result in noticeably lower latencies. So the minimum memory latency is demonstrated in DDR2-1066 mode (266 MHz FSB) - from 39.7ns (pseudo-random walks, hardware prefetch enabled) to 103.2ns (random walks, hardware prefetch disabled). It's a tad better than in the previous Corsair XMS2-8000UL (DDR2-1000, 250 MHz FSB, 5-6-6-18 timings) as well as in Kingston HyperX DDR2-900 (DDR2-900, 270 MHz FSB, 5-6-6-18 timings).
Timing values, except for tCL, were adjusted "on the fly" due to the built-in RMMA feature that allows to change dynamically memory settings supported by the chipset. Memory operating stability was evaluated with an auxiliary utility RightMark Memory Stability Test, included into RMMA.
Minimum timings, we managed to obtain in DDR2-667 mode with the recommended increased voltage of 2.2 V, are just 3-3-3 (as usual, the last parameter tRAS does not participate in the overclocking scheme, as it's ignored). We managed to set the 3-3-2 scheme "on the fly", but it immediately resulted in failures. What concerns this parameter, our modules under review are outperformed by the previously tested Kingston HyperX DDR2-900 (3-3-2 timings at 2.0V) and, especially, by early Corsair XMS2-5400UL (as we know from practice, it demonstrates the highest overclocking potential — 3-2-2 timings at 2.1V).
Our memory modules operate well in DDR2-800 mode with 4-4-3 timings (the voltage is also 2.2V). Note that we managed to set the 4-3-3 timing scheme here as well, but it immediately resulted in errors. And again these modules are outperformed by the older Corsair XMS2-8000UL (minimum timings possible in DDR2-800 mode are — 4-3-3 at 2.2V) and Corsair XMS2-5400UL (4-3-2 timings at 2.1V).
And finally, operating in the fastest "native" mode DDR2-1066, these modules allowed to set the minimum timing scheme - 5-5-4. Not bad actually, as it's a tad lower than the 5-5-5(-15) scheme, published by the manufacturer. At the same time, we should remember that the previous solution from Corsair — DDR2-1000 modules could operate in this mode with 5-3-3 timings.
As usual, "extreme" timings only insignificantly increase memory bandwidth, as it's still limited by FSB bandwidth. Maximum effect of overclocking timings is noticeable only in terms of latencies, only in case of random memory access and DDR2-667 mode (the best at overclocking) at that — latencies are reduced by 10% here.
Corsair XMS2-8500 proved to be high-performance high-end modules, which can operate in the official DDR2-667 and DDR2-800 modes as well as in the unofficial fastest DDR2-1066 mode (at the moderately high voltage of 2.2V). Note that these are the first modules, which are really designed for DDR2-1066 and which operation in this mode is really stable. These modules demonstrate high performance results, but relatively moderate overclocking potential of timings (compared to the early top offers from Corsair) — in official DDR2-667 and DDR2-800 modes they can operate with 3-3-3 and 4-4-3 timings (voltage - 2.2V), while the fastest DDR2-1066 mode requires increasing this scheme to 5-5-4, actually a tad lower compared to the recommended 5-5-5-15 scheme.
Dmitri Besedin (firstname.lastname@example.org)
May 18, 2006
Write a comment below. No registration needed!