iXBT Labs - Computer Hardware in Detail

Platform

Video

Multimedia

Mobile

Other

Memory Module Analysis. Part 15: Corsair XMS2-8000UL Modules

September 29, 2005




We proceed with the low-level analysis of the most important characteristics of high-performance DDR2 memory modules using our RightMark Memory Analyzer. Corsair XMS2-5400UL modules have recently broken another record, to be more exact — two records: in timings, 3-2-2 in DDR2-667 mode, as well as in performance — these are the first modules tested in DDR2-800(!) mode. Besides, they demonstrated an excellent timing overclocking potential for this case — 4-3-2. In today's article we shall review another offer from Corsair, again from the XMS2 series — not DDR2-800, as you might have expected, but the 500(!) MHz modules, that is capable of DDR2-1000 mode (even if it's not the official specification, approved by JEDEC).

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 Exterior

Photo of the memory module




Module Part Number




The manufacturer's web site does not publish the DDR2 Part Number expansion of XMS2-series memory modules. The datasheet runs that the product under review is a set of two DDR2-667 512 MB modules, based on 64M x8 chips. The manufacturer guarantees 100% stable operation in DDR2-1000 module with high-performance motherboards, 5-4-4-9 timings, and 2.2V voltage (higher than the nominal voltage by 0.4V, i.e. by nearly 25%). According to the manufacturer, the timings, written in SPD, are also 5-4-4-9.

SPD chip data

Description of the general SPD 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 (tCK) at maximum supported CAS# latency (CL X) 9 25h 2.50 ns (400.0 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 30h CL = 5, 4
Minimum clock cycle (tCK) at reduced CAS# latency (CL X-1) 23 37h 3.70 ns (270 MHz)
Minimum clock cycle (tCK) at reduced CAS# latency (CL X-2) 25 00h Not defined
Minimum Row Precharge Time (tRP) 27 32h 12.5 ns
5, CL = 5
~3.4, CL = 4
Minimum Row Active to Row Active delay (tRRD) 28 1Eh 7.5 ns
3, CL = 5
~2, CL = 4
Minimum RAS to CAS delay (tRCD) 29 32h 12.5 ns
5, CL = 5
~3.4, CL = 4
Minimum Active to Precharge Time (tRAS) 30 2Dh 45.0 ns
18, CL = 5
~12.2, CL = 4
Module Bank Density 31 80h 512 MB
Write recovery time (tWR) 36 3Ch 15.0 ns
6, CL = 5
~4.1, CL = 4
Internal write to read command delay (tWTR) 37 1Eh 7.5 ns
3, CL = 5
~2, CL = 4
Internal read to precharge command delay (tRTP) 38 1Eh 7.5 ns
3, CL = 5
~2, CL = 4
SDRAM Device Minimum Active to Active/Auto Refresh Time (tRC) 41, 40 37h, 00h 55.0 ns
22, CL = 5
~14.9, CL = 4
SDRAM Device Minimum Auto-Refresh to Active/Auto-Refresh Command Period (tRFC) 42, 40 4Bh, 00h 75.0 ns
30, CL = 5
~20.3, CL = 4
Maximum device cycle time (tCKmax) 43 80h 8.0 ns
SPD Revision 62 12h Revision 1.2
Checksum for Bytes 0-62 63 97h 151 (true)
Manufacturer’s JEDEC ID Code 64-71 FFh, 7Fh,
9Eh
Corsair
Module Part Number 73-90 CM2X512-8000UL
Module Manufacturing Date 93-94 00h, 00h Not defined
Module Serial Number 95-98 00h, 00h,
00h, 00h
Not defined

The SPD contents look utterly non-standard as well as the DDR2-1000 mode. First of all, according to the SPD, the maximum mode is DDR2-800, not DDR2-1000. This mode corresponds to the first of the supported values tCL = 5. The full timing scheme can be written as 5-5-5-18 — it's far from the official 5-4-4-9, admit it, especially in DDR2-1000 mode. It's not clear what to make out of the second reduced tCL = 4. It corresponds to the 3.7 ns cycle time, which corresponds to 270 MHz — i.e. sort of the DDR2-540 mode. I have absolutely no idea about the reasons. Of course, the non-standard cycle time resulted in the main timing values presented by rational quantities — which can be rounded to 4-3.4-3.4-12.2 in the best scenario. In the worst case (that's what the majority of motherboards are likely to do) — to 4-4-4-12. Like XMS2-5400UL, these modules lack information on the manufacturing date and the serial number — not quite a good thing for brand high-end modules.

Testbed Configurations and Software

Testbed 1

  • CPU: Intel Pentium 4 670, 3.8 GHz (Prescott-2M core, rev. N0, 2 MB L2)
  • Chipset: Intel 955X, 200 MHz FSB
  • Motherboard: ASUS P5WD2 Premium, BIOS 0422 dated 05/27/2005
  • Memory: 2x512 MB Corsair XMS2-8000UL (DDR2-800, single/dual channel)
  • Video: Leadtek PX350 TDH, NVIDIA PCX5900
  • HDD: WD Raptor WD360, SATA, 10000 rpm, 36Gb
  • Drivers: NVIDIA Forceware 77.72, Intel Chipset Utility 7.0.0.1025, DirectX 9.0c

Testbed 2

  • CPU: Intel Pentium 4 670, 3.8 GHz (Prescott-2M core, rev. N0, 2 MB L2)
  • Chipset: Intel 955X, 200 MHz FSB
  • Motherboard: Gigabyte 8I955X Royal, BIOS F7 dated 07/13/2005
  • Memory: 2x512 MB Corsair XMS2-8000UL (DDR2-800, single/dual channel)
  • Video: Leadtek PX350 TDH, NVIDIA PCX5900
  • HDD: WD Raptor WD360, SATA, 10000 rpm, 36Gb
  • Drivers: NVIDIA Forceware 77.72, Intel Chipset Utility 7.0.0.1025, DirectX 9.0c

Testbed 3

  • CPU: Intel Pentium 4 670, 3.8 GHz (Prescott-2M core, rev. N0, 2 MB L2) 3.5 GHz (250 x14)
  • Chipset: Intel 955X, 250 MHz FSB
  • Motherboard: ASUS P5WD2 Premium, BIOS 0422 dated 05/27/2005
  • Memory: 2x512 MB Corsair XMS2-8000UL (DDR2-800, single/dual channel)
  • Video: Leadtek PX350 TDH, NVIDIA PCX5900
  • HDD: WD Raptor WD360, SATA, 10000 rpm, 36Gb
  • Drivers: NVIDIA Forceware 77.72, Intel Chipset Utility 7.0.0.1025, DirectX 9.0c

Testbed 4

  • CPU: Intel Pentium 4 670, 3.8 GHz (Prescott-2M core, rev. N0, 2 MB L2) 3.5 GHz (250 x14)
  • Chipset: Intel 955X, 250 MHz FSB
  • Motherboard: Gigabyte 8I955X Royal, BIOS F7 dated 07/13/2005
  • Memory: 2x512 MB Corsair XMS2-8000UL (DDR2-800, single/dual channel)
  • Video: Leadtek PX350 TDH, NVIDIA PCX5900
  • HDD: WD Raptor WD360, SATA, 10000 rpm, 36Gb
  • Drivers: NVIDIA Forceware 77.72, Intel Chipset Utility 7.0.0.1025, DirectX 9.0c

Test Results

Performance tests

There are only few motherboards taking part in out tests — to be more exact, just two models: ASUS P5WD2 Premium and Gigabyte 8I955X Royal. That's because few models allow the Memory/FSB frequency divider = 2.0 as well as to apply relatively high voltage (2.2 V) to memory modules.

We tested our modules on these motherboards in two modes — DDR2-800 with the standard 200-MHz FSB and DDR2-1000, which required the non-standard 250 MHz FSB. Besides, in each case we tested efficiency of the memory system, both single-channel and dual-channel options. The single-channel tests are necessary, because it's impossible to reveal the full potential of DDR2-800 modules in dual-channel mode with the currently available FSB frequencies. Indeed, the peak memory bandwidth of the dual-channel DDR2-800 (12.8 GB/s) is twice as high as the peak bandwidth of the 200 MHz FSB (6.4 GB/s); and the dual-channel DDR2-1000 memory bandwidth (16.0 GB/s) is twice as high as the bandwidth of the 250 MHz FSB (8.0 GB/s). But the single-channel mode allows to even the peak bandwidths of the FSB and memory.

Parameter Testbed 1 Testbed 2 Testbed 3* Testbed 4*
dual single dual single dual single -
Timings
5-6-6-18
5-6-6-18
5-5-5-15
5-5-5-15
5-6-6-18
5-6-6-18
-
Average memory read bandwidth, MB/sec
5651
5658
5610
5587
6937
6940
-
Average memory write bandwidth, MB/sec
2241
2080
2273
2100
2675
2465
-
Max. memory read bandwidth, MB/sec
6868
6806
6774
6680
8469
8371
-
Max. memory write bandwidth, MB/sec
4314
4313
4287
4287
5340
5337
-
Minimum Pseudo-Random Access Latency, ns
52.3
52.7
49.7
50.7
42.6
42.8
-
Maximum Pseudo-Random Access Latency, ns
61.4
61.3
59.1
59.9
49.9
50.3
-
Minimum Random Access Latency**, ns
103.3
102.7
103.5
103.4
88.3
88.2
-
Maximum Random Access Latency**, ns
126.3
126.1
127.2
126.7
107.5
107.4
-
Minimum Pseudo-Random Access Latency, ns (without hardware prefetch)
80.0
80.2
75.8
75.3
65.2
65.2
-
Maximum Pseudo-Random Access Latency, ns (without hardware prefetch)
106.2
106.9
102.4
104.7
86.5
86.9
-
Minimum Random Access Latency**, ns (without hardware prefetch)
103.3
103.5
105.2
104.3
89.0
88.8
-
Maximum Random Access Latency**, ns (without hardware prefetch)
131.0
130.8
131.5
131.1
110.1
110.0
-

*250 MHz FSB, DDR2-1000 Mode
**16MB block size

As usual, memory timings were set by default (BIOS Setup — Memory Timings: "by SPD"). As you can see, the ASUS and Gigabyte motherboards do not agree on the standard values for the DDR2-800 mode (the 5-5-5-18 timings are written in SPD for this very mode). ASUS sets the maximum 5-6-6-18 scheme and Gigabyte — the faster 5-5-5-15 scheme, resulting in reduced memory access latencies.

Performance results of the modules in dual-channel and single-channel modes are practically the same (somewhat increased maximum real memory bandwidth, obviously exceeding the theoretical limit, has to do with a relatively L2 Cache in a CPU). It means that the modules under review (actually like the other DDR2 modules) are really capable of the memory bandwidth, close to the theoretical limit. It does not come due to the dual-channel mode. Interestingly, pseudo-random and random memory access latencies in dual-channel mode are practically comparable to those in the single-channel mode. They are even a tad lower in some cases.

What concerns the DDR2-1000 mode — alas, there is only one contender here — the ASUS P5WD2 Premium motherboard. The second contender — Gigabyte 8I955X Royal — failed to start up in this mode despite all our efforts (such as applying higher voltage (2.3V) and manually assigning maximum timings). The default timing scheme for this mode on the ASUS motherboard is the same — 5-6-6-18.

Stability tests

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, allowed by these memory modules in DDR2-800 mode without losing stability, turned out the same on both motherboards — 4-3-3 (we again "forget" about tRAS, as it's ignored by these modules as well as by the majority of other modules). Unfortunately, lowing down timings to 4-3-2 (this very record was achieved by Corsair XMS2-5400UL modules in our previous review) resulted in errors.

Parameter Testbed 1 Testbed 2 Testbed 3* Testbed 4*
dual single dual single dual single -
Timings
4-3-3
4-3-3
4-3-3
4-3-3
5-3-3
5-3-3
-
Average memory read bandwidth, MB/sec
5728
5746
5724
5711
7017
7025
-
Average memory write bandwidth, MB/sec
2468
2182
2478
2221
2972
2589
-
Max. memory read bandwidth, MB/sec
6884
6832
6829
6755
8512
8426
-
Max. memory write bandwidth, MB/sec
4314
4314
4287
4287
5340
5339
-
Minimum Pseudo-Random Access Latency, ns
49.0
49.3
49.0
49.3
42.1
42.3
-
Maximum Pseudo-Random Access Latency, ns
58.5
58.5
58.3
58.8
49.4
49.6
-
Minimum Random Access Latency**, ns
98.2
98.0
94.9
94.8
82.7
82.7
-
Maximum Random Access Latency**, ns
121.6
121.5
118.6
118.3
101.6
101.5
-
Minimum Pseudo-Random Access Latency, ns (without hardware prefetch)
74.4
74.4
74.7
75.2
64.7
64.6
-
Maximum Pseudo-Random Access Latency, ns (without hardware prefetch)
101.0
101.3
102.9
103.2
85.6
85.8
-
Minimum Random Access Latency**, ns (without hardware prefetch)
99.0
98.9
95.5
95.2
83.6
83.6
-
Maximum Random Access Latency**, ns (without hardware prefetch)
124.6
124.1
121.8
121.1
104.5
105.8
-

*250 MHz FSB, DDR2-1000 Mode
**16MB block size

The minimum stable timings scheme for the DDR2-1000 mode is 5-3-3 — we managed to set this mode only on the ASUS P5WD2 Premium motherboard. Good news: this scheme is noticeably lower than the official scheme for this mode — 5-4-4-9. As we haven't carried out memory tests in the non-standard DDR2-1000 mode before, we have here another record achieved by Corsair modules.

Bottom line

The reviewed Corsair XMS2-8000UL memory modules proved themselves as high-performance modules, which can fully reveal the potential of this memory type in the standard DDR2-800 mode as well as in the non-standard DDR2-1000. It's too early to speak of their compatibility with various motherboards, as there are currently only two motherboards supporting such high-performance memory modes. Nevertheless, even in this case the DDR2-1000 mode is available only in one of them — ASUS P5WD2 Premium. It means that such high-performance non-standard solutions are evidently too buggy (it's hard to tell why the modules don't work with Gigabyte 8I955X — whether it's the fault of the modules themselves, the motherboard, the chipset, or the BIOS version). On the other hand, it's hard to imagine real users of high-performance modules (even DDR2-800), considering the 200 MHz and rarely 266 MHz FSB and plenty of dual-channel DDR2-533. So, high-performance modules are interesting as technological innovations rather than mass products. So in conclusion, let's dwell on the technical aspect of our review, overclocking potential of the modules (timings) in particular. This potential is quite good — the modules can operate stably in DDR2-800 mode with 4-3-3 timings (a tad worse than the previous record 4-3-2, achieved by slower Corsair XMS2-5400UL modules, strange as it may seem); in the standard DDR2-1000 mode — with 5-3-3 timings (it's still the absolute record, as we haven't tested DDR2-1000 before).

Corsair XMS2-8000UL modules are kindly provided
by Corsair Memory representative office in Russia

Dmitri Besedin (dmitri_b@ixbt.com)
September 29, 2005.



Write a comment below. No registration needed!


Article navigation:



blog comments powered by Disqus

  Most Popular Reviews More    RSS  

AMD Phenom II X4 955, Phenom II X4 960T, Phenom II X6 1075T, and Intel Pentium G2120, Core i3-3220, Core i5-3330 Processors

Comparing old, cheap solutions from AMD with new, budget offerings from Intel.
February 1, 2013 · Processor Roundups

Inno3D GeForce GTX 670 iChill, Inno3D GeForce GTX 660 Ti Graphics Cards

A couple of mid-range adapters with original cooling systems.
January 30, 2013 · Video cards: NVIDIA GPUs

Creative Sound Blaster X-Fi Surround 5.1

An external X-Fi solution in tests.
September 9, 2008 · Sound Cards

AMD FX-8350 Processor

The first worthwhile Piledriver CPU.
September 11, 2012 · Processors: AMD

Consumed Power, Energy Consumption: Ivy Bridge vs. Sandy Bridge

Trying out the new method.
September 18, 2012 · Processors: Intel
  Latest Reviews More    RSS  

i3DSpeed, September 2013

Retested all graphics cards with the new drivers.
Oct 18, 2013 · 3Digests

i3DSpeed, August 2013

Added new benchmarks: BioShock Infinite and Metro: Last Light.
Sep 06, 2013 · 3Digests

i3DSpeed, July 2013

Added the test results of NVIDIA GeForce GTX 760 and AMD Radeon HD 7730.
Aug 05, 2013 · 3Digests

Gainward GeForce GTX 650 Ti BOOST 2GB Golden Sample Graphics Card

An excellent hybrid of GeForce GTX 650 Ti and GeForce GTX 660.
Jun 24, 2013 · Video cards: NVIDIA GPUs

i3DSpeed, May 2013

Added the test results of NVIDIA GeForce GTX 770/780.
Jun 03, 2013 · 3Digests
  Latest News More    RSS  

Platform  ·  Video  ·  Multimedia  ·  Mobile  ·  Other  ||  About us & Privacy policy  ·  Twitter  ·  Facebook


14

Copyright © Byrds Research & Publishing, Ltd., 1997–2011. All rights reserved.