Disk Performance, USB, Firewire and Monitoring
Contents
Summary
Benchmark tests identified performance differences using external disk drives depending on hardware speeds, Windows version, USB or Firewire connection, FAT or NTFS formatting and optimisation choice. Further tests were run copying files from the main drive to the external disk, with Performance Monitor logging. This confirmed benchmark findings but identified other peculiarities. The following results mainly relate to copying a 49 MB test folder containing 857 files.
Formatting and Optimisation - According to timing based on Windows copying progress monitor, Optimise For Performance is faster than Optimise For Quick Removal. However, for the former, completion indicates that the last data has been read into RAM and writing to disk can continue over a relatively long period (like 20 seconds). With Optimise For Quick Removal and FAT formatting, writing time is increased as the Allocation Table has to be frequently rewritten. The best option is probably NTFS formatting and Optimise For Quick Removal.
USB or Firewire - USB has the fastest data transfer speed but with higher overheads, leading to Firewire being faster when copying includes a lot of small files or writing uses small block sizes. Using Optimise For Quick Removal, Firewire was up to 24% faster copying the test folder.
Extra Data Written - Copying the 49 MB test folder, typically with Optimise For Quick Removal, 68 MB was written using FAT (updating table) and 63 MB with NTFS.
Extra Data Read - Using the same test folder, FAT and both optimise options, unlike 32-Bit Vista, 64-Bit Vista and 64-Bit XP appeared to read around 35 MB from the external drive where just writing was anticipated.
CPU Time - CPU time used for copying was higher than might be predicted from disk benchmark results, in one case often equivalent to 100% utilisation of one of the two CPUs. CPU time used for USB tests was higher than that for Firewire.
RoboCopy and XCopy - Using these with the test folder, both were slower than Windows Copy/Paste, using more CPU time. RoboCopy was the slowest but it has the advantage that it only copies changed files and folders, extremely fast when applicable. Unlike XCopy, using 64-Bit Vista, RoboCopy also appears to invoke reading the 35 MB extra data as above.
Windows File Cache - File copy functions transfer data via the File Cache, resulting in much faster speed when tests are repeated without restarting the PC, with data being read from RAM. Some results for copying cached data are provided and these help to resolve other problems. Benchmarks produced faster writing and reading when options to avoid caching were used. These were run, with and without caching, using a USB connected disk, reading existing 243 MB BMP files or writing multiple large files, identifying other caching issues associated with Windows pre-loading data into File Cache. In one case, following reading tests, Windows decided to load 760 MB from the USB disk into the cache. Data generated for writing can displace other cached data and Windows might decide to reload the old data after testing.
To Start
General
Disk benchmarks identified performance differences with the same disk drive connected to a PC via USB and Firewire, also other inconsistencies with usage options. This report provides a summary of the findings, with data obtained using Windows Performance Monitor and results from file copying tests.
DiskGraf Benchmark - This benchmark measures disk writing and reading speeds at different block sizes, also maximum DMA/Bus reading speeds.
DiskGraf Results.htm provides performance details using FAT Format and Optimise For Quick Removal Policy setting via USB and Firewire connections.
Performance can vary quite a lot depending on the speed of the PC and remote adapter electronics. This might be influenced by the version of Windows. With the systems tested, CPU MHz appeared to have little effect.
Firewire writing speed was significantly faster than USB using small block sizes but, depending on the system, the position could be reversed with large blocks. On reading, Firewire tended to be faster at all block sizes.
CDDVDSpd Benchmark - This benchmark can read files from any source and also write and read a combination of one large and 520 small files. In this case, large/small files used varied from 1MB/2KB to 32MB/64KB.
CDDVDSpd Results.htm shows average milliseconds per file for writing and reading small files and minimum milliseconds per MByte from the large file tests. The same disk as above was used plus one of those small laptop type portable drives. Results are given for FAT and NTFS formatting with Optimise For Quick Removal and Optimise For Performance settings.
Format and Optimisation - Using Optimise For Quick Removal, writing NTFS small files is two to three times faster than FAT and the overheads involved can affect writing larger files. With Optimise For Performance, writing time of small files can be further reduced, with FAT formatting often having the edge. The time to read small files or write/read large files seems to be less dependent on formatting and optimisation.
USB and Firewire - It appears that USB overheads are higher than those for Firewire, sometimes producing slower speed data transfers and noticeable on small file sizes.
Note that these benchmark runs avoided data entering Windows’ RAM based File Cache. Caching was enabled for some of the tests below.
To Start
Setting Up Performance Monitor
This is arranged via Start, Run or Start Search with Vista, type Perfmon then press Enter - Vista administrative permission required. The program can also be started via Control Panel, Administrative Tools, Performance or Reliability and Performance with Vista. XP - select Performance Logs and Alerts, Counter Logs. Vista - select Data Collection Sets, User Defined. Via menu Action select New Log Settings or New Data Collection Set, XP - type name and OK then Add Counters. Vista - Create Manually, Next, tick Performance Counter, Next, Add.
Performance Object or Counter select and Add - Processor, % Processor Time - PhysicalDisk, Disk Read Bytes/sec, Disk Reads/sec, Disk Write Bytes/sec, Disk Writes/sec - Memory, Page Reads/sec, Page Writes/sec. Close or OK. [For LAN/network measurements add counters from Network Interface - Bytes Received/sec, Bytes Sent/sec, Packets Received/sec, Packets Sent/sec, possibly Output Queue Length, Discard and Error counters].
For other settings see Perfmon Help. Those used/changed were Sample Interval 1 second, log destination, log type Text Comma Delimited (CSV for spreadsheet), manual start/stop, 10000 samples Vista.
To Start
Performance Monitor - Copying Files
The following show performance monitoring results copying 49.4 MB, using a folder containing 857 files. Copying was carried out with “Optimise For Quick Removal” (OQR) and “Optimise For Performance” (OP) settings.
Copying (pasting) time when the Windows Progress Monitor was visible were 9.3 and 6.6 seconds respectively, where this can be two to three seconds longer than the main disk and CPU activity.
The main observations are that, with OP, writing to disk continues for more than 10 seconds after completion is indicated but less overall CPU time is used.
In order to try to identify maximum writing speed, the tests were repeated without rebooting so that data is copied from Windows RAM based File Cache. The total activities are shown at the bottom of the table. With OQR, copying time was 7.7 seconds with writing speed varying between 11 and 18 MB/second over 5 seconds. Using OP, writing again continued for more than 10 seconds after the indicated 4.4 seconds copying time.
System - Core 2 Duo 2.4 GHz, 64-Bit Vista, From main disk to USB disk, NTFS Format
Optimised For Quick Removal Optimised For Performance
Read Write Page Read Write Page
Secs KBytes Reads KBytes Writes % CPU Reads KBytes Reads KBytes Writes % CPU Reads
0 0 0 12 3 2 0 0 0 8 2 0 0
1 2507 59 2768 83 11 59 595 15 12 3 3 16
2 11118 194 11854 252 19 194 14563 301 3749 182 21 275
3 10133 235 11048 295 34 235 13464 356 5099 61 31 353
4 7340 240 8476 418 40 241 23830 540 4831 45 38 540
5 13183 182 14060 144 20 182 705 366 5320 127 22 366
6 6867 290 8620 517 34 289 0 0 4786 25 3 0
7 685 348 3882 786 38 348 0 0 6656 43 6 0
8 64 1 188 12 14 2 0 0 4212 221 4 0
9 0 0 316 12 10 0 0 0 3241 8 1 0
Tot1 51898 1549 61223 2522 224 1550 53158 1578 37915 717 130 1550
10 0 0 16 3 2 0 0 0 2734 16 1 0
11 0 0 8 2 2 0 0 0 2537 13 2 0
12 0 0 487 71 4 0 0 0 2124 27 7 0
13 0 0 16 4 0 0 0 0 2808 21 2 0
14 0 0 32 8 5 0 0 0 1558 6 0 0
15 0 0 32 4 1 0 0 0 1412 10 2 0
16 0 0 12 3 0 0 0 0 1192 6 1 0
17 0 0 82 23 10 0 0 0 814 45 3 0
18 0 0 12 3 2 0 0 0 1095 5 5 0
19 0 0 48 8 2 0 0 0 1093 5 4 0
20 0 0 64 9 0 0 0 0 996 3 1 0
21 0 0 32 8 0 0 0 0 141 77 2 0
22 0 0 44 11 2 0 0 0 135 97 3 0
23 0 0 8 2 1 0 0 0 205 58 2 0
24 0 0 16 4 0 0 0 0 45 69 2 0
25 0 0 60 13 0 0 0 0 36 59 0 0
26 0 0 8 2 2 0 0 0 33 52 2 0
27 0 0 126 33 4 0 0 0 168 32 5 0
28 0 0 8 2 1 0 0 0 968 6 2 0
29 0 0 11 3 0 0 0 0 8 2 4 0
30 0 0 12 3 2 0 0 0 48 12 0 0
Tot2 51898 1549 62357 2741 266 1550 53158 1578 58067 1338 181 1550
Repeated Copying With Data in Windows File Cache - Tot1 after 8 seconds
Tot1 932 218 60918 2513 182 219 872 218 32275 470 124 220
Tot2 872 218 54074 719 159 220
For the above and other examples, with 1 second sampling, Bytes/Second have been converted to KBytes, similarly for number of Writes and Reads. % CPU is for two CPUs where 50% = 1 CPU fully utilised. For Total CPU Seconds, divide Total % by 50. Page Writes/Second were mainly zero.
To Start
Copying Files USB and Firewire
Following are results from copying the same folder (49.4 MB, 857 files) and using the same disk as above, and later four copies, to demonstrate differences between USB and Firewire. These also include measurements using Optimise For Quick Removal and Optimise For Performance with FAT (F) and NTFS (N) format. PCs used had different versions of Windows, that is 64-Bit and 32-Bit Vista and 64-Bit XP.
Details shown are measured copying time (Secs), identified MegaBytes read/written, number of reads/writes (blocks) and derived CPU seconds (1 CPU). The number of Page Reads is also shown and this is virtually the same as blocks read. As shown later, the Page Reads indicate that data is buffered via Windows RAM based File Cache.
Optimise For Performance - All data is not written to disk when the copying progress indicator is closed. An extra line is shown to identify estimated overall elapsed time to when copying is really finished.
Data Written and Read - These tests were run a number of times to minimise the inclusion of background disk activity. The slowest copying is on using Optimise For Quick Removal and FAT formatting. Here, the amount of data written (to external drive) is increased with frequent updating of the File Allocation Table and this frequency might depend on the version of Windows.
On the first system, with XP x64, nearly 90 MB is read with FAT formatting and a series of tests showed that this was consistent. The last entry, for 4 x 49 MB, again shows the extra reading. Later tests shown, using the PC with 64-Bit Vista, also indicated reading nearly 90 MB.
This extra reading was not apparent on a PC using 32 Bit XP. The laptop with 32-Bit Vista often indicates reading around 40 MB. This might be because Windows pre-loads some data into File Cache.
USB v Firewire - With Optimise For Quick Removal, and these particular files, Firewire is shown to be faster by 15% to 24% with FAT formatting and 5% to 21% with NTFS, influenced by lower CPU time. The difference in measured copying time is much less using Optimise For Performance but, with later lazy writes, overall activity time can be less using USB.
Block Size - Derived block sizes, as recorded by Windows, are the same using USB and Firewire connections.
Using Optimise For Quick Removal, average block sizes for writing to the external drive are similar for all versions of Windows, at 25 to 32 KB using NTFS and 9 to 10 KB with FAT, influenced by saving the Allocation Tables.
On the other hand, there are wide variations on reading block sizes (33 to 80 KB FAT, 23 to 58 KB NTFS), in association with the variable amount of data being read.
With Optimise For Performance, block sizes on writing are similar with NTFS and FAT but, along with reading, there are large variations using the different systems.
CPU Time - Firewire uses a little less CPU time than USB. The laptop with 32-Bit Vista appears to consume equivalent to 83% utilisation of one CPU but obtains some of the fastest copying times.
Copying Overheads - At least for the laptop with 32-Bit Vista, comparing times for a 49 MB test and 4 x 49 MB, there appears to be an overhead of 2 to 3 seconds for the same data transfer speed.
Optimised For Quick Removal Optimised For Performance
Read Write CPU Page FAT Read Write CPU Page
Secs MB Blcks MB Blcks Secs Reads NTFS Secs MB Blcks MB Blcks Secs Reads
AMD Athlon 64 X2 2.2 GHz, XP x64 copy 49 MB folder
USB
30.5 88.1 2615 71.7 7854 11.6 2129 F 14.4 88.7 2669 47.2 1395 3.8 2184
24 52.7 1575 3.9
18.1 54.0 2316 46.2 1501 3.4 2317 N 16.0 62.2 2493 37.3 1280 3.2 2444
24 51.6 1640 3.8
Firewire
26.4 87.3 2547 73.1 7844 10.6 2063 F 13.5 88.7 2656 43.9 1336 3.2 2171
24 52.3 1566 3.2
14.9 54.1 2310 41.1 1347 3.2 2317 N 15.6 53.8 2248 41.1 1379 3.6 2248
25 50.1 1573 3.7
AMD Athlon 64 X2 2.2 GHz, XP x64 copy 4 x 49 MB folders
USB
139 385.9 11597 294.7 31840 43.9 9497 F
Core 2 Duo 2.4 GHz, Vista 64 copy 49 MB folder
USB - Average of 3
16.6 88.2 1889 69.6 7135 8.8 1455 F 8.1 74.5 1589 34.8 478 3.4 1097
23 53.6 1183 4.6
9.5 53.1 1547 62.4 2505 4.2 1548 N 6.6 52.3 1539 26.2 366 2.8 1597
22 56.8 946 3.6
Core 2 Duo 2.4 GHz, Vista 64 copy 4 x 49 MB folder
USB
55.1 322.9 7175 276.8 28332 31.2 5047 F
Core 2 Duo 1.8 GHz Laptop, Vista 32 copy 49 MB folder
USB
16.6 39.5 494 68.2 6854 13.8 519 F 8.6 54.2 1512 24.7 335 8.2 1515
22 56.5 991 13.2
11.8 39.8 692 63.0 2543 9.8 719 N 9.0 50.5 1369 37.1 493 7.0 1364
20 56.0 788 9.8
Firewire
14.2 39.7 498 68.3 6832 10.6 504 F 8.3 54.0 1488 30.7 280 5.8 1493
21 54.8 606 8.6
11.1 39.8 691 63.1 2548 7.0 699 N 8.6 50.9 1687 32.4 332 6.6 1693
20 56.5 663 9.0
Core 2 Duo 1.8 GHz Laptop, Vista 32 copy 4 x 49 MB folders
USB
59.5 164.1 2127 278.3 27505 49.0 2114 F 22.7 221.1 6049 162.2 2508 22.8 8031
40 207.6 3641 28.2
38.4 210.2 5658 254.8 10381 31.4 7186 N 26.4 186.2 4428 189.4 3299 27.0 4441
44 229.9 4113 31.2
Firewire
47.9 165.9 2212 274.2 27379 38.0 4298 F 20.7 169.0 2629 153.9 2166 18,4 2635
48 211.4 3732 23.3
36.5 198.4 4363 257.0 10292 28.4 4397 N 25.0 175.2 3791 169.9 2515 18.8 3811
42 228.4 3671 22.4
To Start
Performance Monitor - Reading Unexpected Data 64-Bit Windows
Following are more results for copying 49 MB with 857 files plus some with 36 MB and 7274 files using the Core 2 Duo with 64-Bit Vista and the laptop with slower CPU and 32-Bit Vista. All are for Optimise For Quick Removal and are mainly for reading and writing using the USB disk.
The second lines of results are for re-reading the same data, which will be cached in RAM via Windows File Cache.
The results still show that, with 64-Bit Vista writing FAT files, much more data is read than using 32-Bit Vista. The extra data volume is confirmed by that read when data is cached. It looks as though this is associated with data writing and is from the external drive.
The laptop with 32-Bit Vista can be faster, due to less data being read, but it uses more CPU time that can reverse the effect.
With frequent updating of the File Allocation Table, copying the folder with 7274 files leads to five times more writing than might be expected (at least 175 MB for 36 MB). That with the larger average file size and 49 MB is around 1.5 times more, at up to 77 MB.
The CPU time for copying 36 MB on the laptop includes a lot of other background processing and total CPU seconds should be around 90.
Core 2 Duo 2.4 GHz, Vista 64 Core 2 Duo 1.8 GHz, Vista 32
Optimise For Quick Removal Optimise For Quick Removal
Read Write CPU Page FAT Read Write CPU Page
Secs MB Blcks MB Blcks Secs Reads NTFS Secs MB Blcks MB Blcks Secs Reads
Main NTFS to USB FAT
16.6 88 1889 70 7135 8.8 1455 F 16.6 40 494 68 6854 13.8 519
13.1 # 35 504 68 6821 7.3 20
USB NTFS to USB FAT
17.4 87 1554 68 6847 9.3 1510 F 20.6 54 1172 69 6907 15.1 1616
12.9 # 35 504 68 6844 7.4 22 13.9 # 0 19 68 6861 12.3 21
USB FAT to USB FAT
19.0 109 2735 71 7712 10.7 1227 F 19.7 54 1446 72 7731 15.3 1447
14.1 # 35 1214 68 6824 7.4 20 12.8 # 1 68 69 6906 13.3 23
USB FAT to USB NTFS
13.4 75 2451 66 3457 6.3 1429 N 15.8 60 1754 62 2133 10.6 1766
9.3 # 1 929 62 2477 4.4 220 8.5 # 5 259 62 2529 12.6 258
USB FAT to USB FAT 36 MB 7274 files
79.1 84 16803 191 57875 70.7 7943 F 110.0 43 8194 178 54428 179.9 8183
68.7 # 30 8052 176 54128 58.7 581 74.4 # 4 633 177 54253 78.0 584
# Repeat run with cached data
To Start
Performance Monitor - RoboCopy and XCopy
The 49 MB folders were again used to observe system activity using RoboCopy and XCopy via 32-Bit and 64-Bit Vista. The tests employed copying from the main disks to USB FAT and NTFS partitions with Optimise For quick Removal Properties. For comparison purposes, Windows Copy/Paste results are repeated.
RoboCopy speeds are the slowest of the three (note time for 4 x 49 MB), but it has the advantage that it only copies changed files and folders. As with 64-Bit Vista Copy/Paste to FAT, more than 80 MB is read, unlike XCopy. RoboCopy to FAT and NTFS, via both versions of Vista, results in 83 to 87 MB being written, compared with 74 to 79 with XCopy and 70 MB or less using Copy/Paste.
Example Robocopy command in BAT file (Vista needs Run As Administrator):
/COPYALL :: COPY ALL file info
/E :: copy subdirectories, including Empty ones
/LOG+:file :: output status to LOG file (append to existing log)
/TEE :: output to console window, as well as the log file
robocopy d:\newWS1 f:\newWS1 /COPYALL /E /LOG+:d:datalog.txt /TEE
Core 2 Duo 2.4 GHz, Vista 64 Core 2 Duo 1.8 GHz, Vista 32
Optimise For Quick Removal Optimise For Quick Removal
Main NTFS to USB FAT and NTFS Main NTFS to USB FAT and NTFS
Read Write CPU Page FAT Read Write CPU Page
Secs MB Blcks MB Blcks Secs Reads NTFS Secs MB Blcks MB Blcks Secs Reads
Robocopy
19.5 # 34 504 83 10402 13.8 20 F 17.9 # 0 22 82 10350 16.5 18
9.1 # 1 219 86 8346 4.6 658 N 12.4 # 1 210 86 8355 10.2 648
21.1 88 1948 83 10416 15.7 1481 F 22.2 55 1565 83 10437 20.7 1559
11.9 54 1721 87 8349 5.2 2161 N 16.0 42 797 86 8384 12.9 1240
Xcopy
17.4 47 1093 79 9630 12.2 622 F 20.4 47 1298 78 9414 19.2 1943
8.3 51 1310 74 5412 4.3 1749 N 11.7 51 1372 77 5807 11.4 1814
Copy/Paste from above
16.6 88 1889 70 7135 8.8 1455 F 16.6 40 494 68 6854 13.8 519
9.5 53 1547 62 2505 4.2 1548 N 11.8 40 692 63 2543 9.8 719
# Repeat run with cached data
Robocopy 4 x 49 MB - seconds each folder 24, 17, 22, 30
93.0 325 7355 350 44084 56.4 4944 F
Copy/Paste 4 X 49 mb from above
55.1 323 7175 277 28332 31.2 5047 F
To Start
Performance Monitor - Benchmark Reading Large Files
Following are results reading a 243 MB file from an NTFS formatted USB disk drive using the CDDVDSpd benchmark, with Policy Optimised For Performance. This benchmark includes FILE_FLAG_NO_BUFFERING in the CreateFile function so that Windows does not read the data into the RAM based File Cache. A second version is available without this option, when a repeat of the test might find data in memory.
Three sets of results are provided with PCs using Windows XP x64, 64-Bit Vista and 32-Bit Vista, with three cached and three uncached tests run on each PC.
There can be wide variations on test times, due to other background activity. So, the range of benchmark measured elapsed seconds are shows, derived CPU time seconds (1 CPU) and reading block size from Perfmon results, plus variation in recorded MBytes over the one second samples .
Generally, CPU and elapsed times are higher when file cache is used, where the latter is identified by page reads being about the same as normal reads.
Vista appears to use more CPU time than XP but this does not affect overall run time much. Vista and XP use different block sizes on reading where data is cached. In the case of the laptop with Vista, half of the CPU time appears to be due to background activity.
With a cached test on the latter, there was unexpected further reading of 760 MB from the USB disk over 37 seconds. It seems that Vista decided to load some of the test files into File Cache.
System - Core 2 Duo 2.4 GHz, 64-Bit Vista, USB disk, NTFS Format, Opt Performance
Benchmark measured 28.25 and 28.59 MBytes/second
FILE_FLAG_NO_BUFFERING Cached Reading
Read Write Page Read Write Page
Secs KBytes Reads KBytes Writes % CPU Reads KBytes Reads KBytes Writes % CPU Reads
0 0 0 16 3 2 0 0 0 12 3 2 0
1 21519 23 108 19 11 0 9087 18 32 8 4 19
2 29696 29 87 23 16 2 26622 13 88 6 22 16
3 29694 29 442 36 15 0 30854 18 383 44 38 22
4 28670 28 16 3 1 0 30718 15 16 4 24 15
5 27646 27 16 3 3 0 28670 14 8 2 24 14
6 29693 29 40 8 12 0 30719 15 36 8 8 15
7 29695 29 74 18 7 0 28669 14 8 2 21 14
8 27646 27 28 5 10 0 26624 13 66 17 19 13
M 24574 24 16 3 10 0 31006 23 80 20 28 23
0 17 6144 4 16 3 16 3
0 52 0 28
0 98 0 96
0 408 0 718
Tot 248835 245 1403 118 84 2 249114 147 1575 114 204 154
Blk 1016 1695
MB/s 27-30 26-31
Secs 7.9-8.8 7.9-10.2
CPU 1.8-2.2 1.7-4.1
System - Athlon 64 X2 2.2 GHz, XP x64, USB disk, NTFS Format, Opt Performance
248776 244 434 79 36 0 256494 4193 156 35 61 4171
Blk 1020 61
MB/s 25-32 30-34
Secs 7.8-9.5 8.0-19.0
CPU 0.7-0.9 1.2-2.0
System - Core 2 Duo 1.8 GHz, 32-Bit Vista, USB disk, NTFS Format, Opt Performance
248837 251 313 378 316 13 251839 193 464 156 350 198
Blk 991 1305
MB/s 19-27 19-27
Sec 10.5-16.6 9.3-11.0
After 0-37
CPU 5.6-6.4 6.7-7.3
To Start
Performance Monitor - Benchmark Writing Large Files
DiskGraf also runs using FILE_FLAG_NO_BUFFERING but has an option to ignore this and cache the data. A special version was created to write files then delete them after clicking on a MessageBox. Following are Perfmon statistics for writing sixteen files of 16 MB at 256 KB block size, again to the same USB disk.
In most cases, two totals are shown, the first being for the measured seconds and the other to show unexpected disk activity after the tests finished. This is where Windows decides to reload data, displaced from FileCache by the benchmark files. Using Vista, the files might be indicated in Perfmon Resource Overview, Disk. In this case, a 780 MB file from the main hard drive was shown.
Result for and older PC using XP, and where USB is slow, are also shown.
Writing with File Cache enabled is generally slower and uses more CPU time. When avoiding File Cache, writing block size is that defined by the application. Using File Cache XP uses 64 KB and Vista probably 512 KB. Generally, XP again appears to use less CPU time than Vista.
Core 2 Duo 2.4 GHz, Vista 64, NTFS Opt Performance
Benchmark for 16 files measured 25.8 to 26.1 and 25.1 to 25.6 MBytes/second
FILE_FLAG_NO_BUFFERING Cached Writing
Read Write Page Read Write Page
Secs KBytes Reads KBytes Writes % CPU Reads KBytes Reads KBytes Writes % CPU Reads
1 0 0 18973 83 15 0 64 4 26777 204 14 0
2 0 0 26228 115 2 0 0 0 24606 41 10 0
3 0 0 25432 123 6 0 0 0 25615 35 4 0
4 0 0 25893 114 16 0 0 0 24623 36 7 0
5 0 0 25461 131 13 0 0 0 25634 34 17 0
6 0 0 25910 109 7 0 0 0 25615 29 10 0
7 0 0 25886 107 4 0 0 0 24603 30 13 0
8 0 0 25427 116 6 0 0 0 25619 29 6 0
9 0 0 26407 110 15 0 0 0 24786 40 7 0
10 0 0 25453 122 10 0 0 0 26495 35 18 0
11 0 0 11815 51 8 0 0 0 4170 26 4 0
12 0 0 12 3 3 0 0 0 8 2 4 0
13 0 0 44 11 2 0 0 0 8 2 1 0
14 0 0 12 3 2 0 0 0 40 9 2 0
15 0 0 12 3 0 0 0 0 12 3 1 0
16 0 0 12 4 2 0 0 0 16 4 3 0
17 0 0 9 3 2 0 5650 95 38 6 20 95
18 0 0 80 15 4 0 11250 212 8 2 13 231
19 0 0 36 5 1 0 1184 17 40 9 3 56
20 0 0 58 17 1 0 0 0 12 3 0 0
21 0 0 32 5 0 0 0 0 58 18 6 1
22 0 0 8 2 2 0 108 5 16 3 5 4
23 0 0 36 8 2 0 18443 288 16 3 2 297
24 0 0 12 3 2 0 33959 467 64 12 8 468
25 0 0 70 20 1 0 19001 306 40 7 2 319
Tot1 0 0 262886 1181 102 0 64 4 258541 539 112 0
Tot2 0 0 263320 1283 125 0 89659 1394 258918 622 182 1471
Blk 223 16 480
Blk 205 64 416
MB/s 25-26 25-27
Secs 10 11
CPU 2.0-2.1 2.3-2.5
AMD Athlon 64 X2 2.2 GHz, XP x64, NTFS Opt Performance
Tot1 0 0 262589 1111 56 0 0 0 269448 4297 77 0
Tot2 9430 263 269820 4380 81 208
Blk 236 63
MB/s 35-36 34-36
Secs 7 8
CPU 1.1-1.2 1.5-1.6
Core 2 Duo 1.8 GHz Laptop, Vista 32, NTFS Opt Performance
Tot1 368 13 263214 1210 300 0 368 15 262994 452 345 1
Tot2 204262 7171 263939 655 795 7164
Blk 28 218 25 582
MB/s 18-22 18-23
Secs 13-14 13-14
CPU 6.0-6.1 6.9-7.2
Pentium 4 1.9 GHz, XP, NTFS Opt Performance
348 25 263085 1202 164 24 23 6 265556 4276 283 7
Blk 14 219 4 62
MB/s 13-15 13-15
Secs 19-20 21-22
CPU 1.6-1.9 2.8-3.1
To Start
Roy Longbottom August 2009
The new Internet Home for my PC Benchmarks is via the link
Roy Longbottom's PC Benchmark Collection
|