![[H]ardOCP](/images/blank.gif){kind=small}
- Date:
- Wednesday, May 21, 2003
- Author:
- Morry Teitelman
- Editor:
- Kyle Bennett
- Google +1
Intel's Springdale D865PERL
The D865PERL is Intel’s 200 MHz FSB solution for the mainstream user, based on their i865PE chipset. We know many OCers are interested in this chipset as they are hoping it to be the bargain of the summer.
Subsystem Testing
Audio – CPU Utilization
The choice of an audio subsystem is among the most important for a manufacturer when implementing a new motherboard. A poorly implemented audio subsystem can be the make it or break it point for a board, while a quality subsystem can be the crowning touch to a perfect board. By measuring the audio subsystem’s CPU utilization under a variety of situations, you can easily determine the subsystem’s quality. A well designed audio subsystem should have minimal impact on the CPU while operating, thus freeing up the precious CPU cycles for other tasks such as graphical scene rendering. In order to best measure the real world CPU utilization of the audio subsystem, we use Ziff Davis’ Audio Winbench.
Judging just by the synthetic benchmark numbers alone, it seems that Intel’s implementation of the ADI centered sound subsystem is a bit lacking. In most cases, the CPU utilization was measured at around 5%, with it going past 7% in a few cases. This higher than average CPU utilization could negatively impact system performance, especially under situations where the system is under heavy stress. Keep in mind that these results are based on a synthetic benchmark and may not accurately depict what you see on your home system.
Audio – Subjective Listening
CPU utilization measurement may be a good way to give you a ballpark estimate of how a sound solution may perform, but nothing is certain until you sample the sound quality produced by the sound subsystem first hand. A well rounded subjective sound test should include something that will push the envelope on intensity as well as exercise the system’s ability to handle subtlety. I chose the Evanescence album Fallen for this test. The album contains a very modern rock mix of both haunting melodies and in your face rock ‘n’ roll intensity.
I was satisfied with the sound quality produced by the board. It seemed to handle both the high and low frequencies well, and did not seem to stutter or skip during playback of the CD. However, there was a bit of sound crackling detected during playback of certain tracks.
Audio – In Game Testing
Besides music listening, most often enthusiasts use the audio solution for gaming. To test the quality of the audio subsystem during a game type situation, I used a custom designed benchmarking script based on the CPU and Inferno demos developed by our own Brent Justice. The benchmark tool runs through three loops of each demo, one with sound disabled, one with normal 3d sound enabled, and the last with EAX 3d sound enabled.
I was impressed with the in game results. As you can see from the graphs above, the game performance suffered under low quality settings as would be expected, but under normal high quality setting game play, there was no impact from the sound system being enabled. I detected no skipping, crackling or hissing during either of the demo playbacks. In fact, I was most impressed with the quality of the EAX sound, and would gladly use the integrated sound solution over a PCI based sound solution for game play.
USB 2.0 / IEEE 1394
In order to adequately test the capabilities of the on board USB 2.0 and IEEE 1394 connections, we chose to use an ACOMDATA HD060U2FE-72-USB 2.0/FireWire HDD connected first to the USB port and after to the IEEE 1394 port in conjunction with TCD Labs’ HDTach program.
As with previous results encountered during these tests, I am a bit disappointed by the results shown. My disappointment stems from two major facts: one, the maximum theoretical bandwidth of IEEE 1394 is 400 Mb/s and that of USB 2.0 is 480 Mb/s; second, the drive itself has an average read seek time of 8.8 ms, with a maximum transfer rate of 50MB/s according to the manufacturer specifications. Using both connection types, the average read time (RAT) was 19.6ms, a bit longer than the manufacturer specification. The average read speed using IEE 1394 was about 10 MB/s faster than that of the USB 2.0 connection speed, but only came in at just over 30 MB/s. In both cases, the average write speed was just below 15 MB/s. The one truly telling statistic is the average CPU utilization. For USB 2.0 usage, it was a bit over 20%, while for IEEE 1394, the CPU utilization did not even break 1%.
IDE/ATA Performance
The IDE subsystem is among the most critical subsystems for determining a system’s overall performance. Poor IDE performance can lead to unexplained system lags and general system slowdowns, while a highly optimized solution can give a good system an even better performance boost. In order to test the IDE performance of this board, I used TCD Labs’ HDTach program. My test bench currently uses Maxtor 40Gb ATA 133 model 6E040L0 hard drives on the IDE headers. On the SATA headers, I have Seagate 80 Gb Barracuda SATA hard drives installed in the test bench. Multiple SATA drives were used for testing in a RAID 0 configuration. Testing was also conducted using a single SATA drive, and an IDE drive connected in a primary slave configuration.
The RAID 0 performance difference seen between the 128k block size and the 16k block is like night and day. The smaller block size seems to get better read performance, beating the 128k block size array by more than 25 MB/s. This major performance leap can be attributed to the more optimized seeks and transfers done when the system has less physical information to parse through in a single block. The write performance was a bit surprising, with the 128k block size array beating out the smaller block size array by 5 MB/s, though both make a good showing with write performance over 40 MB/s. In both cases, the RAT hovered just above 13ms, while the CPU utilization was a bit higher for the 16k block size, almost 8% compared to the 3.5% of the 128k array. It just goes to show that the system has to work a bit harder with the smaller block size array. One of the most astounding statistics is the 16k array RBS of 140 MB/s. This is one of the highest measurements I’ve seen with a SATA based array yet. The 128k array RBS is a bit disappointing, coming in even with ATA/100 IDE performance at just above 82 MB/s.
One thing I continue to be annoyed with is the SATA single drive performance especially in light of the IDE primary slave test results. In both cases, the RBS comes in right around 85 MB/s, which is very good for an ATA/100 IDE drive, but not so good for a 150 MB/s based SATA drive. The read and write performance of the SATA drive are mediocre, with the IDE slave drive beating it out by 10 MB/s in both cases. The SATA drive does shine with a CPU utilization of < 2%. However, the IDE slave drive is no slouch with a utilization of < 4%.
Network Utilization Tests
Hagel Technologies’ DU Meter software was used in conjunction with Windows Task Manager to measure the performance of the onboard Intel PRO/1000 NIC. DU meter was used to measure bandwidth, with Windows TaskMan to monitor the CPU utilization on the test system. For the test itself, a 750Mb archive file containing various sized .WMA audio files for the large file transfer test and a 750Mb worth of various sized .WMA audio files for the small files transfer test were used in conjunction with an Intel Gigabit NIC on the host system, and a crossover cable to connect the host system to the test system. A crossover cable was used to rule out any possible bandwidth losses due to hub or switch passage.
I continued to be astounded by the Gigabit Ethernet performance offered with the ICH5R Southbridge. The NIC performed superbly in both the upload and download tests, with the upload transfers a bit faster than the download speeds. The MB/s rates recorded are approaching that of hard drive speeds. However, the CPU utilization for both tests was a bit high, maintaining a utilization average of 30%.
The large file transfer results mimic the superb results of the small file transfer test, with the upload and download sustained averages coming in a bit faster for the large file transfers. The integrated Gigabit Ethernet solution is blazingly fast, with a small but tolerable price in the CPU utilization numbers.
Test Systems
The following system configurations were used for the Sandra memory benchmark graph, as well as all graphs listed under the Application and Gaming Benchmark sections:
Intel D865PERL (i865PE) - Intel Pentium 4 3.0 GHz CPU (clocked at 15x200) and Intel Pentium 4 2.53 GHz CPU (clocked at 19x133) - 2 x 256Mb Corsair XMS3200 - ATI Radeon 9700 Pro w/ ATI Catalyst 2.3 drivers - 40 GB Maxtor ATA133 HDD - Allied 400w PSU - WindowsXP w/SP1
NOTE: For all benchmark tests done on the Intel D865PERL motherboard, the optimized default BIOS settings were loaded for the benchmark tests. The following BIOS settings were also engaged during the tests: CAS Latency Time = 2; RAS Precharge delay = 2; RAS-to-CAS delay = 3; Active Precharge delay = 2.
ABIT IC7-G (i875P) - Intel Pentium 4 3.0 GHz CPU (clocked at 15x200) - 2 x 256Mb Corsair XMS3200 - ATI Radeon 9700 Pro w/ ATI Catalyst 2.3 drivers - 40 GB Maxtor ATA133 HDD - Allied 400w PSU - WindowsXP w/SP1
Intel D875PBZ (i875P) - Intel Pentium 4 3.0 GHz CPU (clocked at 15x200) - 2 x 256Mb Corsair XMS3200 - ATI Radeon 9700 Pro w/ ATI Catalyst 2.3 drivers - 40 GB Maxtor ATA133 HDD - Allied 400w PSU - WindowsXP w/SP1
ABIT BH7 (i845PE) - Intel Pentium 4 2.53 GHz CPU (clocked at 19x133 for 2.53 GHz speed) - 2 x 256Mb Corsair XMS3200 - ATI Radeon 9700 Pro w/ ATI Catalyst 2.3 drivers - 40 GB Maxtor ATA133 HDD - Allied 400w PSU - WindowsXP w/SP1
ASUS A7N8X Barton (nForce2): AMD AthlonXP 3000+ CPU (clocked at 13x166); 2 x 256MB Corsair XMS3200; ATI Radeon 9700 Pro, onboard nForce2 10/100 NIC; 40 GB Maxtor ATA133 HDD, Zalman 300w PSU. WindowsXP w/SP1, ATi Catalyst 2.3.
Graphs are labeled as follows: Motherboard - CPU Clock - FSB Clock - Memory Clock
SiSoft Sandra Memory Bandwidth Benchmark
The D865PERL is no slouch, but seems unable to keep up with either of the i875P based solutions, even with the memory running at the same speed. It is obvious that Intel did a bit more than speed binning to increase the i875P’s performance (Or cripple the 865's?). As expected, the board beats the performance of the BH7 with the memory running at 333MHz, but not be that big of a margin. Note that all results above, with the exception of those from the BH7 board, were obtained running in Dual Channel mode.