Your link here
|
|
| | | | | | | |
| | | | | | |
|
|
DDR Memory Analysis. Part 2: Kingston DDR2 Modules
We proceed with a new series of articles devoted to the low level analysis
of the most important characteristics of memory modules using our
RightMark Memory Analyzer
universal test package. I remind you that the purpose of this analysis
is to provide all interested readers with information on compatibility
of a given memory module from a given manufacturer with different
mainboards based on various chipsets. Today the object of our analysis
is 512-Mb DDR2-modules Kingston ValueRAM.
Manufacturer Information
Module manufacturer: Kingston Technology
Chip manufacturer: Elpida Memory, Inc.
Web site of the module manufacturer: http://www.valueram.com/ddr2/
Web site of the chip manufacturer: http://www.elpida.com/en/products/ddr2.html
Module Appearance
Photo of the memory module
Photo of the memory chip
Part Numbering System of Modules and Chips
Module Part Number Expansion
Description of the part numbering system of DDR2 Kingston ValueRAM modules: http://www.kingston.com/literature/MKF_586.pdf
| Field |
Value |
Expansion |
| 0 |
KVR |
Manufacturer: KVR = Kingston ValueRAM |
| 1 |
533 |
Frequency: 533 MHz |
| 2 |
D2 |
Memory type: D2 = DDR2 |
| 3 |
N |
Error correction type: N = Non-ECC |
| 4 |
4 |
CAS# Latency: 4 |
| 5 |
512 |
Module capacity: 512 MB |
Chip Part Number Expansion
Data sheet on 512 Mbit DDR2 Elpida memory chips: http://www.elpida.com/pdfs/E0323E81.pdf
| Field |
Value |
Expansion |
| 0 |
E |
Manufacturer: E = Elpida Memory |
| 1 |
|
Type (n/a, D = monolithic device) |
| 2 |
|
Product code (n/a, E = DDR2) |
| 3 |
51 |
Density/Bank: 51 = 512M/4 banks |
| 4 |
08 |
Bit organization: 08 = x8 |
| 5 |
A |
Voltage, interface: A = SSTL 1.8V |
| 6 |
B |
Die revision |
| 7 |
|
Package code (n/a, SE = FBGA) |
| 8 |
5C |
Speed: 5C = DDR2-533 (4-4-4) |
| 9 |
E |
Environment code: E = lead free |
Note that part numbering of this module, to put it mildly, is different from the official specification provided in the data sheet on this memory chip type. Namely, there are no fields with the device type (monolithic, DDR2) and package (FBGA). Perhaps, it has something to do with the fact that the memory module we got is not a production but an engineering sample. And it most likely uses an earlier revision of Elpida memory chips than the one described in the data sheet.
SPD module chip data
Description of the SPD general standard:
JEDEC Standard No. 21-C, 4.1.2 - SERIAL PRESENCE DETECT STANDARD, General Standard
Description of the SPD specific standard for DDR2:
JEDEC Standard No. 21-C, 4.1.2.10 - Appendix X: Specific SPDs for DDR2 SDRAM (Revision 1.0)
JC-45 Appendix X: Specific PD's for DDR2 SDRAM (Revision 1.1)
| Function |
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 |
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 |
38h |
CL = 3, 4, 5 |
| Minimum clock cycle (tCK) at reduced
CAS latency (CL X-1) |
23 |
3Dh |
3.75 ns (266.7 MHz) |
| Minimum clock cycle (tCK) at reduced
CAS latency (CL X-2) |
25 |
50h |
5.00 ns (200.0 MHz) |
| Minimum Row Precharge Time (tRP) |
27 |
3Ch |
15.0 ns
4, CL = 5, 4
3, CL = 3 |
| Minimum Row Active to Row Active delay (tRRD) |
28 |
1Eh |
7.5 ns
2, CL = 5, 4
1.5, CL = 3 |
| Minimum RAS to CAS delay (tRCD) |
29 |
3Ch |
15.0 ns
4, CL = 5, 4
3, CL = 3 |
| Minimum Active to Precharge Time (tRAS) |
30 |
2Dh |
45.0 ns
12, CL = 5, 4
9, CL = 3 |
| Module Bank Density |
31 |
80h |
512 MB |
| Write recovery time (tWR) |
36 |
3Ch |
15.0 ns
4, CL = 5, 4
3, CL = 3 |
| Internal write to read command delay (tWTR) |
37 |
1Eh |
7.5 ns
2, CL = 5, 4
1.5, CL = 3 |
| Internal read to precharge command delay (tRTP) |
38 |
1Eh |
7.5 ns
2, CL = 5, 4
1.5, CL = 3 |
| Minimum Active to Active/Refresh Time (tRC) |
41, 40 |
3Ch, 00h |
60.0 ns
16, CL = 5, 4
12, CL = 3 |
| Minimum Refresh to Active/Refresh Command Period
(tRFC) |
42, 40 |
69h, 00h |
105.0 ns
28, CL = 5, 4
21, CL = 3 |
| Maximum device cycle time (tCKmax) |
43 |
80h |
8.0 ns |
| SPD Revision |
62 |
10h |
Revision 1.0 |
| Checksum for Bytes 0-62 |
63 |
E1h |
225 (true) |
| Manufacturer’s JEDEC ID Code |
64-71 |
FFh, 98h,
FFh, 00h,
..., FFh |
Kingston |
| Module Part Number |
73-90 |
- |
Not defined |
| Module Manufacturing Date |
93-94 |
04h, 19h |
year 2004, week 25 |
| Module Serial Number |
95-98 |
67h, 28h,
53h, 83h |
83532867h |
According to the SPD chip data, this module has the 64M x 64 organization type and can operate at three different CAS latencies (tCL) of 5, 4 and 3. It's interesting to note that the first two values of tCL, CL X = 5 and CL X-1 = 4, correspond to the same 3.75 ns clock cycle, that is operating in normal DDR2-533 mode at 266.7 MHz. In this case module timings can be written as 5-4-4-12 and 4-4-4-12 correspondingly. The last supported CAS latency value (CL X-2 = 3) corresponds to the 5.00 ns clock cycle, that is operating at 200.0 MHz in DDR2-400 mode. The corresponding timings are 3-3-3-9. Another distinctive feature of this module is a relatively large minimum refresh to active/refresh command period (tRFC) – 105 ns, which corresponds to 28 memory bus cycles at 266.7 MHz and 21 cycles at 200.0 MHz.
Testbed Configurations and Software
Mainboards based on the chipsets of Intel 915 series
Testbed #1
- CPU: Intel Pentium 4 2.8 GHz (Prescott core, 1 MB L2)
- Chipset: Intel i915G
- Mainboard: Intel D915GUX,
BIOS 1028 dated 06/29/2004
- Memory: 2x512 MB Kingston DDR2-533
- Video: Leadtek PX350 TDH, nVidia PCX5900
- HDD: WD Raptor WD360, SATA, 10000 rpm, 36Gb
- Operating system: Windows XP Professional Service Pack 2
- Drivers: Intel Chipset Utility 6.0.1.1002, nVidia Forceware 61.77
Testbed #2
- CPU: Intel Pentium 4 2.8 GHz (Prescott core, 1 MB L2)
- Chipset: Intel i915P
- Mainboard: Foxconn 915A01-P-8EKRS, BIOS 6.00 PG dated 06/10/2004
- Memory: 2x512 MB Kingston DDR2-533
- Video: Leadtek PX350 TDH, nVidia PCX5900
- HDD: WD Raptor WD360, SATA, 10000 rpm, 36Gb
- Operating system: Windows XP Professional Service Pack 2
- Drivers: Intel Chipset Utility 6.0.1.1002, nVidia Forceware 61.77
Testbed #3
- CPU: Intel Pentium 4 2.8 GHz (Prescott core, 1 MB L2)
- Chipset: Intel i915G
- Mainboard: Foxconn 915M03-G-8EKRS2,
BIOS 6.00 PG dated 05/29/2004
- Memory: 2x512 MB Kingston DDR2-533
- Video: Leadtek PX350 TDH, nVidia PCX5900
- HDD: WD Raptor WD360, SATA, 10000 rpm, 36Gb
- Operating system: Windows XP Professional Service Pack 2
- Drivers: Intel Chipset Utility 6.0.1.1002, nVidia Forceware 61.77
Testbed #4
- CPU: Intel Pentium 4 2.8 GHz (Prescott core, 1 MB L2)
- Chipset: Intel i915P
- Mainboard: MSI 915P Neo2, BIOS V1.3B0 dated 09/08/2004
- Memory: 2x512 MB Kingston DDR2-533
- Video: Leadtek PX350 TDH, nVidia PCX5900
- HDD: WD Raptor WD360, SATA, 10000 rpm, 36Gb
- Operating system: Windows XP Professional Service Pack 2
- Drivers: Intel Chipset Utility 6.0.1.1002, nVidia Forceware 61.77
Testbed #5
- CPU: Intel Pentium 4 2.8 GHz (Prescott core, 1 MB L2)
- Chipset: Intel i915G
- Mainboard: MSI 915G
Combo, BIOS 080011 dated 07/14/2004
- Memory: 2x512 MB Kingston DDR2-533
- Video: Leadtek PX350 TDH, nVidia PCX5900
- HDD: WD Raptor WD360, SATA, 10000 rpm, 36Gb
- Operating system: Windows XP Professional Service Pack 2
- Drivers: Intel Chipset Utility 6.0.1.1002, nVidia Forceware 61.77
Testbed #6
- CPU: Intel Pentium 4 2.8 GHz (Prescott core, 1 MB L2)
- Chipset: Intel i915G
- Mainboard: ASUS P5GDC-V, BIOS 1003.003 dated 08/18/2004
- Memory: 2x512 MB Kingston DDR2-533
- Video: Leadtek PX350 TDH, nVidia PCX5900
- HDD: WD Raptor WD360, SATA, 10000 rpm, 36Gb
- Operating system: Windows XP Professional Service Pack 2
- Drivers: Intel Chipset Utility 6.0.1.1002, nVidia Forceware 61.77
Mainboards based on the chipsets of Intel 925 series
Testbed #7
- CPU: Intel Pentium 4 2.8 GHz (Prescott core, 1 MB L2)
- Chipset: Intel i925X
- Mainboard: Gigabyte 8ANDXP-D,
BIOS F1 dated 06/07/2004
- Memory: 2x512 MB Kingston DDR2-533
- Video: Leadtek PX350 TDH, nVidia PCX5900
- HDD: WD Raptor WD360, SATA, 10000 rpm, 36Gb
- Operating system: Windows XP Professional Service Pack 2
- Drivers: Intel Chipset Utility 6.0.1.1002, nVidia Forceware 61.77
Testbed #8
- CPU: Intel Pentium 4 2.8 GHz (Prescott core, 1 MB L2)
- Chipset: Intel i925X
- Mainboard: Intel D925XCV,
BIOS 1259 dated 08/19/2004
- Memory: 2x512 MB Kingston DDR2-533
- Video: Leadtek PX350 TDH, nVidia PCX5900
- HDD: WD Raptor WD360, SATA, 10000 rpm, 36Gb
- Operating system: Windows XP Professional Service Pack 2
- Drivers: Intel Chipset Utility 6.0.1.1002, nVidia Forceware 61.77
Testbed #9
- CPU: Intel Pentium 4 2.8 GHz (Prescott core, 1 MB L2)
- Chipset: Intel i925X
- Mainboard: ASUS P5AD2, BIOS 1004.007 dated 07/02/2004
- Memory: 2x512 MB Kingston DDR2-533
- Video: Leadtek PX350 TDH, nVidia PCX5900
- HDD: WD Raptor WD360, SATA, 10000 rpm, 36Gb
- Operating system: Windows XP Professional Service Pack 2
- Drivers: Intel Chipset Utility 6.0.1.1002, nVidia Forceware 61.77
Testbed #10
- CPU: Intel Pentium 4 2.8 GHz (Prescott core, 1 MB L2)
- Chipset: Intel i925X
- Mainboard: MSI 925X Neo, BIOS 6.00 PG dated 06/18/2004
- Memory: 2x512 MB Kingston DDR2-533
- Video: Leadtek PX350 TDH, nVidia PCX5900
- HDD: WD Raptor WD360, SATA, 10000 rpm, 36Gb
- Operating system: Windows XP Professional Service Pack 2
- Drivers: Intel Chipset Utility 6.0.1.1002, nVidia Forceware 61.77
Testbed #11
- CPU: Intel Pentium 4 2.8 GHz (Prescott core, 1 MB L2)
- Chipset: Intel i925X
- Mainboard: Foxconn 925A01-8EKRS, BIOS 6.00 PG dated 08/28/2004
- Memory: 2x512 MB Kingston DDR2-533
- Video: Leadtek PX350 TDH, nVidia PCX5900
- HDD: WD Raptor WD360, SATA, 10000 rpm, 36Gb
- Operating system: Windows XP Professional Service Pack 2
- Drivers: Intel Chipset Utility 6.0.1.1002, nVidia Forceware 61.77
Test Results
Mainboards based on the chipsets of Intel 915 series
From this time on, memory modules will be tested in two modes. The first series of tests – let's call them "performance tests" – are carried out in normal mode, that is with standard timings written in the SPD chip. The second series of tests is carried out in so called "extreme" mode with maximum possible timings for a given module on a given mainboard. Let's call this test series "stability tests", because the capacity to operate steadily with "harder" timings for a long time can be directly associated with operating stability in normal mode.
Performance tests
DDR2 Kingston modules in DDR2-533 mode suppose the existence of two different timing schemes, 5-4-4-12 and 4-4-4-12. But in the first series of tests we used the 4-4-4-12 scheme, because, firstly, it is closer to values typical of DDR2-533, and, secondly, it was set in BIOS settings by default (Memory Timings: "by SPD") in all mainboards unexceptionally.
As we have already mentioned before, the current generation of processors and chipsets with the FSB frequency of 800 MHz does not allow to reach the maximum DDR2-533 throughput in dual channel mode. Nevertheless, even in this case we can draw certain conclusions about the real memory bandwidth, which can be achieved on various mainboards.
|
| Parameter* |
Testbed 1 |
Testbed 2 |
Testbed 3 |
Testbed 4 |
Testbed 5 |
Testbed 6 |
| Timings |
4-4-4-12
|
4-4-4-12
|
4-4-4-12
|
4-4-4-12
|
4-4-4-12
|
4-4-4-12
|
| Average memory read bandwidth, MB/sec |
4470
|
4469
|
4485
|
4519
|
4467
|
4483
|
| Average memory write bandwidth, MB/sec |
2018
|
2019
|
2010
|
2036
|
2049
|
2015
|
| Max. memory read bandwidth, MB/sec |
6337
|
6346
|
6337
|
6401
|
6321
|
6353
|
| Max. memory write bandwidth, MB/sec |
4267
|
4266
|
4266
|
4324
|
4266
|
4281
|
| Minimum Pseudo-Random Access Latency, ns** |
54.8
|
54.8
|
54.8
|
54.2
|
55.1
|
54.7
|
| Maximum Pseudo-Random Access Latency, ns** |
63.8
|
63.8
|
63.9
|
63.0
|
64.1
|
63.7
|
| Minimum Random Access Latency, ns** |
130.4
|
130.1
|
130.3
|
128.5
|
130.5
|
129.6
|
| Maximum Random Access Latency, ns** |
155.6
|
155.6
|
155.5
|
153.5
|
155.7
|
155.1
|
*the best results are in bold
**Block size – 16 MB
MSI 915P Neo2 based on i915P (Testbed #4) is the leader almost by all parameters (maximum real memory bandwidth, minimum memory latency) among the mainboards based on Intel 915P/915G. Interestingly, it's actually the only leader – all the other mainboards, regardless of the chipset type (P/G), have demonstrated more or less average results. So, the values of real read memory bandwidth are within 4470-4480 MB/sec, maximum memory bandwidth – within 6330-6350 MB/sec, the pseudo-random access latency is within 55-64 ns. By the way, the first place of MSI 915P Neo2 can be comparatively easy explained by its somewhat increased FSB frequency. It results in the increase of maximum real write memory bandwidth to 4324 MB/sec, which is normally limited to the 2/3 of the system bus throughput (4266.7 MB/sec).
Stability tests
Before we proceed to quantitative evaluation of memory bandwidth/latency in the "extreme" mode, let's dwell on memory timings proper, which values were varying "on the run" due to the built-in RMMA function to dynamically modify memory settings supported by the chipset. Memory system stability with certain timings was estimated using a specially developed utility, which will be included into the next RMMA version as an add-on to the existing RAM Stability Test.
So, minimal timings allowed in DDR2 Kingston modules are 4-3-3-8. The memory system is quite stable in this mode on most mainboards used in this test, except for "combined" boards based on i915G supporting both DDR and DDR2 simultaneously – MSI 915G Combo (Testbed #5) and ASUS P5GDC-V (Testbed #6). Minimal timings allowed on these mainboards without compromising the operating stability of the memory system are 4-3-3-9.
|
| Parameter* |
Testbed 1 |
Testbed 2 |
Testbed 3 |
Testbed 4 |
Testbed 5 |
Testbed 6 |
| Timings |
4-3-3-8
|
4-3-3-8
|
4-3-3-8
|
4-3-3-8
|
4-3-3-9
|
4-3-3-9
|
| Average memory read bandwidth, MB/sec |
4502
|
4510
|
4515
|
4577
|
4502
|
4518
|
| Average memory write bandwidth, MB/sec |
2110
|
2110
|
2103
|
2135
|
2114
|
2195
|
| Max. memory read bandwidth, MB/sec |
6411
|
6408
|
6404
|
6498
|
6395
|
6430
|
| Max. memory write bandwidth, MB/sec |
4267
|
4267
|
4266
|
4324
|
4267
|
4282
|
| Minimum Pseudo-Random Access Latency, ns** |
54.6
|
54.6
|
54.6
|
53.8
|
54.7
|
54.5
|
| Maximum Pseudo-Random Access Latency, ns** |
63.5
|
63.6
|
63.7
|
62.6
|
63.8
|
63.5
|
| Minimum Random Access Latency, ns** |
123.9
|
123.7
|
124.0
|
122.2
|
123.8
|
123.3
|
| Maximum Random Access Latency, ns** |
149.5
|
149.6
|
149.6
|
147.4
|
149.8
|
149.2
|
*the best results are in bold
**Block size – 16 MB
You can easily see that the use of a "harder" timing scheme does not change the overall picture of the memory performance, which we got in the previous series of tests. MSI 915P Neo2 (Testbed #4) is still a leader, while all the other mainboards demonstrate approximately the same results.
Mainboards based on the chipsets of Intel 925 series
Performance tests
As in our previous tests,
on the whole, mainboards based on i925X outscore the mainboards based
on i915P/i915G by all parameters. The leader in this series is again
a mainboard from MSI – 925X Neo (Testbed #10), which is also
achieved by a slightly higher FSB frequency (maximum real memory write
bandwidth – 4327 MB/sec). The second place, almost with the
same result, is shared by Gigabyte 8ANDXP-D (Testbed #7), ASUS P5AD2
(Testbed #9) and Foxconn 925A01-8EKRS (Testbed #11). Intel D925XCV
(Testbed #8) is the last almost by all parameters.
|
| Parameter* |
Testbed 7 |
Testbed 8 |
Testbed 9 |
Testbed 10 |
Testbed 11 |
| Timings |
4-4-4-12
|
4-4-4-12
|
4-4-4-12
|
4-4-4-12
|
4-4-4-12
|
| Average memory read bandwidth, MB/sec |
4640
|
4495
|
4629
|
4682
|
4623
|
| Average memory write bandwidth, MB/sec |
1950
|
1990
|
2078
|
2015
|
1983
|
| Max. memory read bandwidth, MB/sec |
6357
|
6308
|
6343
|
6445
|
6383
|
| Max. memory write bandwidth, MB/sec |
4287
|
4266
|
4282
|
4327
|
4266
|
| Minimum Pseudo-Random Access Latency, ns** |
52.0
|
54.8
|
52.0
|
51.6
|
52.3
|
| Maximum Pseudo-Random Access Latency, ns** |
60.3
|
63.5
|
60.4
|
59.9
|
60.7
|
| Minimum Random Access Latency, ns** |
125.0
|
129.8
|
125.0
|
124.2
|
125.6
|
| Maximum Random Access Latency, ns** |
150.1
|
154.9
|
150.1
|
149.1
|
150.9
|
*the best results are in bold
**Block size – 16 MB
Stability tests
Unlike the first series of tests with mainboards based on i915P/i915G, all tested mainboards based on i925X are doing fine with DDR2 Kingston modules with minimal timings of 4-3-3-8.
|
| Parameter* |
Testbed 7 |
Testbed 8 |
Testbed 9 |
Testbed 10 |
Testbed 11 |
| Timings |
4-3-3-8
|
4-3-3-8
|
4-3-3-8
|
4-3-3-8
|
4-3-3-8
|
| Average memory read bandwidth, MB/sec |
4684
|
4525
|
4676
|
4715
|
4644
|
| Average memory write bandwidth, MB/sec |
2078
|
2089
|
2167
|
2120
|
2099
|
| Max. memory read bandwidth, MB/sec |
6434
|
6367
|
6429
|
6522
|
6438
|
| Max. memory write bandwidth, MB/sec |
4287
|
4266
|
4282
|
4327
|
4266
|
| Minimum Pseudo-Random Access Latency, ns** |
51.7
|
54.6
|
51.8
|
51.4
|
52.4
|
| Maximum Pseudo-Random Access Latency, ns** |
59.9
|
63.2
|
60.1
|
59.6
|
60.8
|
| Minimum Random Access Latency, ns** |
118.1
|
122.0
|
118.1
|
117.1
|
118.8
|
| Maximum Random Access Latency, ns** |
142.9
|
147.4
|
142.9
|
141.9
|
144.1
|
*the best results are in bold
**Block size – 16 MB
And again the use of a "harder" timing scheme did not lead to any considerable changes in the alignment of forces. The first place is taken by MSI 925X Neo, the second place is shared by Gigabyte 8ANDXP-D and ASUS P5AD2, while Foxconn 925A01-8EKRS is a tad behind them taking the third place. Well, and Intel D925XCV is the last in this series of tests.
Results
Thus, the best memory system performance with DDR2 Kingston ValueRAM modules is achieved on the mainboards based on Intel 925X, especially on MSI 925X Neo. Among the mainboards based on Intel 915P/915G, the best performance is also demonstrated by a mainboard from MSI – 915P Neo2. The worst performance of the memory system (among the counterparts) is demonstrated by the mainboards from Intel – D925XCV and D915GUX. What concerns operating stability of the memory system (in "hard" mode), all tested mainboards demonstrate similar results, except for "combined" models supporting both DDR and DDR2 simultaneously – MSI 915G Combo and ASUS P5GDC-V.
|
|
|
|
|
| | | | | | | |
| | | | | | |
|
|
|
|
