Memory Module Analysis. Part 11: Corsair DDR2-533 XMS2 PRO Modules

We proceed with the series of articles, dedicated to the low level analysis of the most important characteristics of DDR2 memory modules using our RightMark Memory Analyzer test package. In this article we are going to examine a pair of 512MB DDR2-533 modules of the XMS2 PRO series, manufactured by Corsair Memory. XMS2 PRO is a new series in the range of DDR2 XMS2 memory modules, manufactured by the company since mid 2003. It's high-performance DDR2 modules in stylish packaging notable for 24 activity indicators.

Manufacturer Information

Module manufacturer: Corsair Memory
Manufacturer of module chips: unknown
Web site of the module manufacturer: http://www.corsairmemory.com/corsair/xms2.html

Module Appearance

Photo of the memory module

The layout of LEDs — activity indicators (2 x 12) can be seen on the photo below (side view). We didn't manage to figure out completely what exactly these LEDs indicate. We are inclined to think that they don't indicate the activities of individual logical banks of a module, but rather the general load of the memory bus in a given module — from the minimal level, indicated by the first four green LEDs, through the middle level (four yellow LEDs in the middle), up to the maximum level (the last four red LEDs).

Module Part Number

The brief technical documentation on Corsair XMS2 PRO memory modules (http://www.corsairmemory.com/corsair/products/specs/CM2X512-4300C3PRO.pdf) provides no information on the expansion of some Part Number elements. Nevertheless, the part numbering system of Corsair modules is rather simple and is easy to decipher on one's own.

Field	Value	Expansion
0	CM	Manufacturer: CM = Corsair Memory
1	2X	Module type: 2X = DDR2
2	512	Module density: 512 MB
3	4300	Theoretical throughput of the module: 4300 MB/s
4	C3	CAS# Latency: 3
5	PRO	Module series: PRO

Out of additional technical characteristics of the modules, specified in the datasheet, we want to note the non-standard 538 MHz frequency, used by the manufacturer for testing the modules, and the recommended and written in SPD 3-3-3-8 timings, which are quite different from typical DDR2-533 timings (4-4-4-11 or 4-4-4-12).

SPD module chip data

Description of the general SPD standard:

JEDEC Standard No. 21-C, 4.1.2 - SERIAL PRESENCE DETECT STANDARD, General Standard

Description of the specific SPD standard for DDR2:

Parameter	Byte	Value	Expansion
Fundamental Memory Type	2	08h	DDR2 SDRAM
Number of Row Addresses on this assembly	3	0Eh	14 (RA0-RA13)
Number of Column Addresses on this assembly	4	0Ah	10 (CA0-CA9)
Number of DIMM Banks	5	60h	1 physical bank
Data Width of this assembly	6	40h	64 bit
Voltage Interface Level of this assembly	8	05h	SSTL 1.8V
SDRAM Cycle time (t_CK) at maximum supported CAS# latency (CL X)	9	3Dh	3.75 ns (266.7 MHz)
DIMM configuration type	11	00h	Non-ECC
Refresh Rate/Type	12	82h	7.8125 ms — 0.5x reduced self-refresh
Primary SDRAM Width (organization type) of the memory module chips	13	08h	x8
Error Checking SDRAM Width (organization type) of the memory chips in the ECC module	14	00h	Not defined
Burst Lengths Supported (BL)	16	0Ch	BL = 4, 8
Number of Banks on SDRAM Device	17	04h	4
CAS Latency (CL)	18	08h	CL = 3
Minimum clock cycle (t_CK) at reduced CAS# latency (CL X-1)	23	00h	Not defined
Minimum clock cycle (t_CK) at reduced CAS# latency (CL X-2)	25	00h	Not defined
Minimum Row Precharge Time (t_RP)	27	2Dh	11.25 ns (3, CL = 3)
Minimum Row Active to Row Active delay (t_RRD)	28	1Eh	7.5 ns (2, CL = 3)
Minimum RAS to CAS delay (t_RCD)	29	2Dh	11.25 ns (3, CL = 3)
Minimum Active to Precharge Time (t_RAS)	30	1Eh	30.0 ns (8, CL = 3)
Module Bank Density	31	80h	512 MB
Write recovery time (t_WR)	36	3Ch	15.0 ns (4, CL = 3)
Internal write to read command delay (t_WTR)	37	1Eh	7.5 ns (2, CL = 3)
Internal read to precharge command delay (t_RTP)	38	1Eh	7.5 ns (2, CL = 3)
SDRAM Device Minimum Active to Active/Auto Refresh Time (t_RC)	41, 40	37h, 00h	55.0 ns (15, CL = 3)
SDRAM Device Minimum Auto-Refresh to Active/Auto-Refresh Command Period (t_RFC)	42, 40	69h, 00h	105.0 ns (28, CL = 3)
Maximum device cycle time (t_CKmax)	43	80h	8.0 ns
SPD Revision	62	10h	Revision 1.0
Checksum for Bytes 0-62	63	0Dh	13 (true)
Manufacturer’s JEDEC ID Code	64-71	7Fh, 7Fh, 9Eh	Corsair
Module Part Number	73-90	—	CM2X512A-4300C3PROMA
Module Manufacturing Date	93-94	00h, 00h	Not defined
Module Serial Number	95-98	00h, 00h, 00h, 00h	Not defined

According to the SPD data, the modules under review can operate only when the CAS# (CL X) latency is equal to three, with the 3.75ns clock cycle, i.e. 266.7 MHz (in DDR2-533 mode). A memory timing scheme for this case can be written as 3-3-3-8, which completely corresponds to the specifications. So, the low timing values expectedly result in a significant reduction of absolute t_RP and t_RCD values to 11.25 ns (typical value — 15 ns) and t_RAS — to 30 ns (typical value — 40-45 ns). At the same time, in general other characteristics (t_RC and especially t_RFC) are almost no different from the typical DDR2-533 values.

Among the other features of the SPD chip is the lack of data on the manufacturing date and the module serial number, but the availability of information on the Part Number. For some reason, the "MA" lettering is added to the part number on the modules.

Testbed Configurations and Software

Testbed #1

CPU: Intel Pentium 4 3.6 GHz (Prescott E0, 1 MB L2)
Chipset: Intel 925X
Motherboard: MSI 925X Neo, BIOS dated 06/18/2004
Memory: 2x512 MB Corsair DDR2-533 XMS2 PRO
Video: Leadtek PX350 TDH, nVidia PCX5900
HDD: WD Raptor WD360, SATA, 10000 rpm, 36Gb
Drivers: nVidia Forceware 62.01, Intel Chipset Utility 6.2.1.1001, DirectX 9.0c

Testbed #2

CPU: Intel Pentium 4 3.6 GHz (Prescott E0, 1 MB L2)
Chipset: Intel 925XE
Motherboard: ECS PF21 Extreme, BIOS dated 12/07/2004
Memory: 2x512 MB Corsair DDR2-533 XMS2 PRO
Video: Leadtek PX350 TDH, nVidia PCX5900
HDD: WD Raptor WD360, SATA, 10000 rpm, 36Gb
Drivers: nVidia Forceware 62.01, Intel Chipset Utility 6.2.1.1001, DirectX 9.0c

Testbed #3

CPU: Intel Pentium 4 3.6 GHz (Prescott E0, 1 MB L2)
Chipset: Intel 925XE
Motherboard: Gigabyte 8AENXP-D, BIOS F2 dated 01/04/2005
Memory: 2x512 MB Corsair DDR2-533 XMS2 PRO
Video: Leadtek PX350 TDH, nVidia PCX5900
HDD: WD Raptor WD360, SATA, 10000 rpm, 36Gb
Drivers: nVidia Forceware 62.01, Intel Chipset Utility 6.2.1.1001, DirectX 9.0c

Testbed #4

CPU: Intel Pentium 4 3.6 GHz (Prescott E0, 1 MB L2)
Chipset: Intel 925XE
Motherboard: Abit Fatal1ty AA8XE, BIOS 1.2 dated 01/14/2005
Memory: 2x512 MB Corsair DDR2-533 XMS2 PRO
Video: Leadtek PX350 TDH, nVidia PCX5900
HDD: WD Raptor WD360, SATA, 10000 rpm, 36Gb
Drivers: nVidia Forceware 62.01, Intel Chipset Utility 6.2.1.1001, DirectX 9.0c

Testbed #5

CPU: Intel Pentium 4 3.6 GHz (Prescott E0, 1 MB L2) at 3.73 GHz (266 MHz x14)
Chipset: Intel 925XE
Motherboard: ECS PF21 Extreme, BIOS dated 12/07/2004
Memory: 2x512 MB Corsair DDR2-533 XMS2 PRO
Video: Leadtek PX350 TDH, nVidia PCX5900
HDD: WD Raptor WD360, SATA, 10000 rpm, 36Gb
Drivers: nVidia Forceware 62.01, Intel Chipset Utility 6.2.1.1001, DirectX 9.0c

Testbed #6

CPU: Intel Pentium 4 3.6 GHz (Prescott E0, 1 MB L2) at 3.73 GHz (266 MHz x14)
Chipset: Intel 925XE
Motherboard: Gigabyte 8AENXP-D, BIOS F2 dated 01/04/2005
Memory: 2x512 MB Corsair DDR2-533 XMS2 PRO
Video: Leadtek PX350 TDH, nVidia PCX5900
HDD: WD Raptor WD360, SATA, 10000 rpm, 36Gb
Drivers: nVidia Forceware 62.01, Intel Chipset Utility 6.2.1.1001, DirectX 9.0c

Testbed #7

CPU: Intel Pentium 4 3.6 GHz (Prescott E0, 1 MB L2) at 3.73 GHz (266 MHz x14)
Chipset: Intel 925XE
Motherboard: Abit Fatal1ty AA8XE, BIOS 1.2 dated 01/14/2005
Memory: 2x512 MB Corsair DDR2-533 XMS2 PRO
Video: Leadtek PX350 TDH, nVidia PCX5900
HDD: WD Raptor WD360, SATA, 10000 rpm, 36Gb
Drivers: nVidia Forceware 62.01, Intel Chipset Utility 6.2.1.1001, DirectX 9.0c

Test Results

Performance tests

In the first series of tests we used the standard timing scheme, which was set in BIOS by default (Memory Timings: by SPD) on all our testbeds. As you can see from the table below, these motherboards have not agreed upon the "standards" and the reason is quite obvious — the timings written in SPD are too far from the typical DDR2-533 values. So, only one motherboard, Abit Fatal1ty AA8XE (Testbed #4), sets the manufacturer's default timings (3-3-3-8) correctly. Like all the other Gigabyte motherboards, Gigabyte 8AENXP-D tends to increment t_CL by one resulting in the 4-3-3-8 timing scheme. We found it out before, when tested DDR memory modules "for overclockers". On the contrary, the motherboards from MSI and ECS set a faster scheme — 3-3-3-6. However, it makes absolutely no difference to the memory modules, because they are not an exception from the general "rule", discovered experimentally. According to this rule, t_RAS values in chipset settings are ignored and internal values are used instead (in our case it's obviously 8).

Parameter	Testbed 1	Testbed 2	Testbed 3	Testbed 4	Testbed 5^*	Testbed 6^*	Testbed 7^*
Timings	3-3-3-6	3-3-3-6	4-3-3-8	3-3-3-8	—	4-3-3-8	3-3-3-8
Average memory read bandwidth, MB/sec	5667	5483	5617	5639	—	6919	6785
Average memory write bandwidth, MB/sec	2156	1808	2085	2139	—	2329	2350
Max. memory read bandwidth, MB/sec	6412	6242	6478	6404	—	8173	8117
Max. memory write bandwidth, MB/sec	4266	4255	4287	4261	—	5672	5672
Minimum Pseudo-Random Access Latency, ns	48.5	50.7	49.9	48.6	—	43.2	46.1
Maximum Pseudo-Random Access Latency, ns	55.5	58.2	57.3	55.5	—	50.6	52.8
Minimum Random Access Latency^**, ns	108.1	118.8	110.5	108.4	—	99.8	103.3
Maximum Random Access Latency^**, ns	126.7	136.8	129.1	126.8	—	118.4	121.1
Minimum Pseudo-Random Access Latency, ns (without hardware prefetch)	76.5	78.8	77.7	76.4	—	67.0	71.5
Maximum Pseudo-Random Access Latency, ns (without hardware prefetch)	95.4	99.6	96.6	95.1	—	87.5	89.9
Minimum Random Access Latency^**, ns (without hardware prefetch)	108.9	119.6	111.2	109.2	—	100.2	104.1
Maximum Random Access Latency^**, ns (without hardware prefetch)	128.7	139.8	131.0	128.3	—	119.4	122.3

^*FSB frequency — 266.7 MHz
^**16MB block size

With the FSB set to 200 MHz, the majority of motherboards on Intel 925X and 925XE (Testbeds #1, #3, and #4) demonstrate almost identical results. Strange as it may seem, the only exception is the ECS PF21 Extreme motherboard "for overclockers" (Testbed #2), which is noticeably outscored by the other models both in memory bandwidth (up to 200 MB/s) and in latency (up to 10 ns in random walks). Moreover, when switched to the 266MHz FSB mode, this very motherboard... refused to work with the modules under review at all. The other two models on i925XE (Testbeds #6 and #7) work fine in this mode, demonstrating a typical memory latency drop by approximately 10ns in all cases due to the synchronous mode (FSB:DRAM frequencies = 1:1).

Stability tests

Timing values (except for t_CL, which cannot be reduced further!) were adjusted "on the fly" due to the built-in RMMA feature that allows to change dynamically memory settings supported by the chipset. Memory operation stability was evaluated in our specially developed utility, which will soon be released as a stand-alone application and will come shipped with RMMA.

While our motherboards disagree upon the "standard" timing values, the minimum values are selected unanimously. Nevertheless, the mere minimum timing values (allowed by these modules) are much more impressive than this fact, because they look absolutely fantastic for DDR2-533 — 3-2-2.

Parameter	Testbed 1	Testbed 2	Testbed 3	Testbed 4	Testbed 5^*	Testbed 6^*	Testbed 7^*
Timings	3-2-2	3-2-2	3-2-2	3-2-2	—	3-2-2	3-2-2
Average memory read bandwidth, MB/sec	5700	5561	5747	5692	—	7177	6867
Average memory write bandwidth, MB/sec	2302	1868	2245	2284	—	2448	2402
Max. memory read bandwidth, MB/sec	6424	6341	6467	6420	—	8362	8207
Max. memory write bandwidth, MB/sec	4266	4255	4287	4261	—	5679	5674
Minimum Pseudo-Random Access Latency, ns	48.3	50.4	48.0	48.4	—	41.0	45.9
Maximum Pseudo-Random Access Latency, ns	55.2	57.8	55.0	55.3	—	48.0	52.5
Minimum Random Access Latency^**, ns	100.9	109.2	100.2	101.0	—	88.7	96.0
Maximum Random Access Latency^**, ns	119.8	127.6	118.9	119.9	—	107.2	115.1
Minimum Pseudo-Random Access Latency, ns (without hardware prefetch)	76.0	78.2	75.6	76.3	—	63.7	71.1
Maximum Pseudo-Random Access Latency, ns (without hardware prefetch)	94.5	99.5	94.3	95.2	—	82.8	89.8
Minimum Random Access Latency^**, ns (without hardware prefetch)	101.3	109.9	100.6	101.6	—	89.1	96.7
Maximum Random Access Latency^**, ns (without hardware prefetch)	120.9	130.0	119.9	120.9	—	108.3	116.4

^*FSB frequency — 266.7 MHz
^**16MB block size

Overclocking the timings allows to improve the results, obtained on the ECS PF21 Extreme motherboard (Testbed #2), as well as to increase the maximum real memory read bandwidth to 8.2-8.4 GB/s (266 MHz FSB) and reduce the latencies approximately by 10 ns at random access.

Bottom line

Tested samples of Corsair DDR2-533 modules (XMS2 PRO series) provide good compatibility with various motherboards on i925X and i925XE chipsets in standard modes, they demonstrate good performance (memory bandwidth and latencies are typical for high-performance modules operating in DDR2-533 mode). We should also note the excellent overclocking potential of these modules in terms of timings — the previous record (3-3-3), cracked by Corsair XMS2 DDR2-667 (675) modules in DDR2-533 mode, is a nominal value for these PRO modules. Their minimum timings, which still allow stable memory operation (3-2-2), are an absolute record for DDR2-533 modules.

Corsair DDR2-533 XMS2 PRO modules have been provided
by the representative office of Corsair Memory in Russia

Dmitri Besedin (dmitri_b@ixbt.com)
April 4, 2005.

AMD 780G/780V/740G Integrated Socket AM2+ Chipsets

AMD Phenom X4 In Real-Life Applications

Corsair Dominator DDR2-1142 (PC2-9136) 4GB Kit

Foxconn GeForce 9800 GTX / GX2 2x512MB

NVIDIA GeForce GTX 260 896MB

AMD Phenom X3 8750

NVIDIA nForce 790i and Intel X48 Chipsets

NVIDIA GeForce GTX 280 1024MB

x64 CPU Performance Testing Methodology

i3DSpeed, May 2008