[nextpage title=”Introduction”]

With the introduction of AM2 socket by AMD all new high-end PCs are finally using DDR2 memories. Right now the standard DDR2 speeds are 533 MHz, 667 MHz and 800 MHz, but Corsair was one of the first manufacturers to release 1,066 MHz models. These models are not only targeted to both Intel and AMD overclockers but also to regular users that have an Intel CPU running externally at 1,066 MHz and want to match the memory with the CPU external speed to achieve the maximum performance your system can deliver. Corsair has released a 1 GB kit (TWIN2X1024-8500C5, 2x 512 MB modules) and a 2 GB kit (TWIN2X2048-8500C5, 2x 1 GB modules), both with 5-5-5-15 timings. Let’s see how the 2 GB kit performs.

Figure 1: Corsair TWIN2X2048-8500C5 2 GB DDR2-1066/PC2-8500 memory kit.

The main problem with DDR2-1066 memories today is compatibility, as only a few motherboards are able to use them to their full potential. On AMD side, as socket AM2 CPUs only support up to DDR2-800, these memories can only be used for overclocking, so there is no point of buying these modules for a socket AM2 CPU if you are not going to overclock it. For the regular user that wants to achieve the maximum system performance but without overclocking there are better solutions on the market for this platform, i.e., DDR2-800 memories with lower latencies –TWIN2X2048-6400C3 also from Corsair is a good example, as it is a DDR2-800/PC2-6400 kit where the memory runs with CL3 instead of CL5.

For the regular user that does not overclock his/her system, the main advantage of DDR2-1066 memories would be using them with Intel CPUs running at 1,066 MHz externally – for example, the new high-end Core 2 Duo and Core 2 Extreme CPUs. If you use DDR2-800 memories with these CPUs, the communication between the CPU and the memories will be taken only at half of the maximum speed it could be used. The problem, however, is that only very few socket LGA775 motherboards accept DDR2-1066. On almost all socket LGA775 motherboards your DDR2-1066 modules will run as if they were DDR2-800, even on the latest ones based on the new Intel 975X chipset.

Of course you can use your DDR2-1066 memory modules for overclocking your Intel CPU: even if your motherboard recognizes it as DDR2-800, you will have memory modules that are guaranteed to run up to 1,066 MHz.

But, just like AMD CPUs, if you don’t plan overclocking your Intel CPU and your motherboard doesn’t support DDR2-1066 but you want to have the fastest memory available, we’d recommend you to pick a low latency DDR2-800 memory, like the abovementioned TWIN2X2048-6400C3 kit from Corsair.

We decided to base our review on Intel side using a Core 2 Xtreme X6800 CPU and an ASUS P5B motherboard (Intel 965 chipset), one of the few available today supporting DDR2-1066 memories, in order to simulate the two scenarios: a regular user wanting to use DDR2-1066 instead of DDR2-800 together with a 1,066 MHz external bus (FSB) CPU and an overclocker trying to achieve the maximum clock rate with his/her memory module.

Even though the reviewed modules have official timings of 5-5-5-15-T2, you can lower these timings in order to achieve a higher performance – if your motherboard provides this option, of course. According to Corsair, you should be able to run these modules at 4-4-4-4-T1 if your run them at 800 MHz and even 4-4-4-12-T2 up to 1,080 MHz.

Unfortunately the motherboard we used didn’t provide this option (even though ASUS may release a new BIOS version enabling this feature).

Corsair provides a lifetime warranty to all their memory modules. This warranty, however, is only valid if you bought your modules directly from Corsair or from authorized resellers that are listed on their website.

Let’s now take a closer look at Corsair TWIN2X2048-8500C5 memory kit.

[nextpage title=”A Look Inside”]

As mentioned, this kit is based in two 1 GB DDR2-1066/PC2-8500 memory modules (2x CM2X1024-8500C5) programmed with 5-5-5-15-T2 timings. We decided to remove the heatspreader of one of the modules to take a look. Please don’t do this at home, as you may break your module or one of the chips, as happened to us once, and you will also void the memories’ lifetime warranty.

Figure 2: One of the modules without its heatspreader.

The memory chips used are from Micron, D9GMH to be more exact – Micron’s full part number for these chips is MT47H64M8B6:D, where “D” is the revision code. The funny thing is that officially these chips are rated DDR2-667, so Corsair handpicks the chips they buy from Micron to see which ones can run at 1,066 MHz or above.

As each chip has a 512-Mbit density, 16 chips are necessary on an 1 GB module (512 Mbits x 16 = 1 GB). On 512 MB modules, eight chips are used.

Figure 3: Micron D9GMH (MT47H64M8B6:D) chips.

The printed circuit boards (PCBs) of the modules are manufactured by Brain Power and are labeled as “BP ML E186014 94V-0”, as you can see in Figure 4.

Figure 4: The printed circuit boards are manufactured by Brain Power.

We also found a small identifier on the PCB labeled “CORSAIR 50-00149B-G”, which can mean that even though the printed circuit boards are manufactured by Brain Power they are only used by Corsair.

Figure 5: PCB identifier.

[nextpage title=”Main Specifications”]

Corsair TWIN2X2048-8500C5 memory kit main features are:

• Total capacity: 2 GB (2x 1 GB modules)
• Modules: 2x CM2X1024-8500C5
• Speed: DDR2-1066/PC2-8500
• Programmed Timings: 5-5-5-15-T2
• Recommended Voltage: 2.2 V
• Memory chips used: Micron D9GMH (MT47H64M8B6:D)
• Warranty: Lifetime
• More Information: https://www.corsair.com/
• Average Price in the US*: USD 450.00

* Researched at Froogle.com on the day we published this review.

[nextpage title=”How We Tested”]

We included in our comparison modules from other DDR2 speed grades, namely DDR2-533, DDR2-667 and DDR2-800 – all from Corsair – for you to have an idea what is the advantage of using the reviewed DDR2-1066 memory kit against regular DDR2 memory modules. The memory modules we included were: 2x CM2X512-4200 (512 MB, 4-4-4-12 timings), 2x CM2X512A-5400UL (512 MB, 4-4-4-15) and 2x CM2X1024-6400C3 (1 GB, 5-5-5-18). The listed timings were the timings programmed on the memory module and used by our motherboard since it was configured at “auto” (our motherboard didn’t provide any timings adjustment), not the lowest timings the memory module could use.

We though it wouldn’t be fair to compare the reviewed kit with the listed kits because they were using different timings and also different capacities. So we decided to configure the reviewed kit as DDR2-533, DDR2-667 and DDR2-800 in order to simulate a memory kit from these speed grades using the same timings and capacity.

During our benchmarking sessions, we used the configuration listed below. Between our benchmarking sessions the only variable was the memory module being tested.
 
Hardware Configuration

• Motherboard: ASUS P5B
• BIOS version: 0211, July 14th, 2006.
• Processor: Core 2 Xtreme X6800 (Dual-Core, 2.93 GHz, 1,066 MHz FSB, 4 MB L2 memory cache)
• Cooler: Gigabyte Neon 775.
• Hard Disk Drive: Maxtor DiamondMax 9 Plus (40 GB, ATA-133).
• Video Card: XFX GeForce 7800 GTX.
• Video resolution: 1024x768x32@85Hz.
• Power Supply: Antec Neo HE 550.

Software Configuration

• Windows XP Professional installed using NTFS
• Service Pack 2
• DirectX 9.0c

Driver Versions

• NVIDIA video driver version: 91.29
• Intel Inf chipset driver version: 8.0.1.1002
• All motherboard drivers

Used Software

• Sandra Lite 2007a 10.98.6.x
• PCMark05 Professional 1.1.0
• Quake 4 – Patch 1.2

We adopted a 3% error margin; thus, differences below 3% cannot be considered relevant. In other words, products with a performance difference below 3% should be considered as having similar performance.

[nextpage title=”Memory Bandwidth”]

We measured memory bandwidth with Sandra Lite 2007. Memory bandwidth, also known as transfer rate, is measured in MB/s. The methodology we used you can check in the previous page.

The results you can see on the chart and table below. On the table we included the maximum theoretical transfer rate for all DDR2 standards, both single channel and dual channel. As you can see, even though the reviewed memory kit achieved the best transfer rate among the selected memory modules, it is far from the maximum theoretical transfer rate. This is normal to occur with Intel CPUs, where the memory controller isn’t embedded in the CPU but located on the motherboard north bridge chip.

Memory Module	Transfer Rate (MB/s)	Difference
DDR2-1066 Dual Channel Max. Theoretical	17,056	201.34%
DDR2-800 Dual Channel Max. Theoretical	12,800	126.15%
DDR2-667 Dual Channel Max. Theoretical	10,672	88.55%
DDR2-1066 Single Channel Max. Theoretical	8,528	50.67%
DDR2-533 Dual Channel Max. Theoretical	8,192	44.73%
DDR2-800 Single Channel Max. Theoretical	6,400	13.07%
Corsair CM2X1024-8500C5 (DDR2-1066)	5,660
DDR2-667 Single Channel Max. Theoretical	5,336	6.07%
DDR2-800 Simulation	5,281	7.18%
Corsair CM2X1024-6400C3 (DDR2-800)	5,280	7.20%
DDR2-667 Simulation	5,137	10.18%
Corsair CM2X512A-5400UL (DDR2-667)	5,106	10.85%
Corsair CM2X512-4200 (DDR2-533)	4,887	15.82%
DDR2-533 Simulation	4,886	15.84%
DDR2-533 Single Channel Max. Theoretical	4,264	32.74%

You should keep in mind that a higher memory transfer rate does not necessarily translates into a higher system overall performance.

[nextpage title=”Memory Performance”]

We measured memory performance using the memory performance module of PCMark05 Professional program. The results are given in a specific unit of the program. The methodology we used you can check on the “How We Tested” section.

We also ran the system benchmark provided by PCMark05. However, the system performance remained on the same level with all memory modules we installed.

The results you can see on the chart and table below.

Keep in mind that a higher performance of the memory subsystem does not necessarily translates into a higher system performance.

[nextpage title=”Gaming Performance”]

Gaming is one arena where changing the memory module by another with a higher speed grade usually raises the performance of the game. Here we used Quake 4, which is based on the same engine as Doom 3.

We ran Quake 4 multiplayer demo id_demo001 on 1024x768x32 with no image quality settings enabled. We run it four times and the results shown on the chart is an arithmetic average of the collected data. The results are in frames per second. For more information on how to use Quake 4 to benchmark your system, read our tutorial on this subject.

The methodology we used you can check on the “How We Tested” section. The results you can see on the chart and table below.

[nextpage title=”Overclocking”]

With this memory we could increase our CPU external bus (FSB) up to 279 MHz with stability, making the memory modules to run at 1,116 MHz, a 4.69% increase over Corsair’s labeled spec. We could configure our FSB even higher (up to 290 MHz), but the system was unstable. We considered our system stable when we could run PCMark05 and Quake 4 at least four times without crashing.

Since we have an unlocked CPU, we decreased the CPU multiplier from x11 to x10 to make our CPU running at a lower internal clock rate, checking if what was preventing us from pushing the memories even more was the CPU, which could have already reached its overclocking limits. However, the results were just the same, so what was limiting our overclocking was really the memory kit being used.

With this overclocking set, we achieved 116.45 frames per second on Quake 4, a 2.60% increase. Keep in mind, however, that this improvement also reflects the increase on our CPU internal clock rate, since the CPU clock was increased from 2.93 GHz to 3.07 GHz.

The screenshot below was taken with CPU-Z. This program shows the real clock rate for both memory and CPU. Our CPU works externally at 266 MHz transferring four data per clock cycle, thus the labeled 1,066 MHz external clock rate. DDR2-1066 memories work at 533 MHz transferring two data per clock cycle, thus the labeled 1,066 MHz clock rate. On our overclocking, the CPU was running externally at 279 MHz and the memories were actually running at 558 MHz.

Figure 6: Corsair CM2X1024-8500C5 modules overclocked.

[nextpage title=”Conclusions”]

As we mentioned before, Corsair TWIN2X2048-8500C5 memory kit has two targets. First, the regular user that has an Intel CPU running at 1,066 MHz externally and wants to have his/her memory modules also running at 1,066 MHz. The problem here is finding a motherboard that accepts DDR2-1066 memory modules. The new ASUS P5B is one of the few motherboards around supporting both Core 2 Duo CPUs and DDR2-1066 memories.

The second target is, of course, the overclocking community. With these modules Corsair ensures that you will reach at least 1,066 MHz with them – and we were able to push the memory frequency up to 1,116 MHz in our tests – even if your motherboard doesn’t support DDR2-1066. Of course you must use an overclocking-oriented motherboard such as ASUS M2N32-SLI De Luxe, or buying this memory kit won’t make any sense at all.

The main problem for this memory kit, in our opinion, is pricing. While it is quoted around USD 400 at Newegg.com and ZipZoomFly.com, there are some stores charging USD 500 or even more for it. Just to give you an idea, at Newegg.com a DDR2-800 2 GB kit with CL4 also from Corsair is quoted at USD 255, while a DDR2-800 2 GB kit with CL5 from OCZ is quoted at USD 200. Of course on both cases the manufacturer doesn’t guarantee any overclocking at all, but these prices are just for you to have an idea on what is going on at the market right now.

This huge price difference crushes the idea of having average Joe installing this memory kit on their new Core 2 Duo system: as the performance difference between using DDR2-800 and DDR2-1066 was only between 5% and 6% in Quake 4 (you may have a higher performance gain in other games, though), we don’t think paying twice to have such performance gain makes any sense at all – unless you are Richie Rich or a relative of his, of course.

Thus this memory kit is a terrific product for overclockers that can afford it, as Corsair guarantees the modules will ran at least at 1,066 MHz plus they offer a lifetime warranty (if you buy them on an authorized reseller), which is fantastic.

We have a couple of tips for you if decide buying this memory. If you leave the memory configuration on “auto” on the motherboard setup, your motherboard may recognize the modules as DDR2-800 even if it supports DDR2-1066. Also, you can achieve a better performance lowering the memory timings, so we’d suggest you to change the memory timings from “auto” to “manual” and configuring them as 4-4-4-12-T2 instead of the default configuration for a higher performance.