The SiS 648 Chipset Debuts
A Comprehensive Look at a Pre-Production Shuttle AS45GT/R

By, Marco Chiappetta
August 12, 2002

OVERCLOCKING: Unfortunately, our pre-production AK45GT/R wasn't quite up to par in the overclocking department.  We experienced random lock-ups and re-booting at default clock speeds, and when we tried to overclock the system our problems were magnified.  We used multiple sticks of RAM, and different power supplies, but no matter what we tried, when the system was overclocked, it simply wasn't stable enough to complete any benchmarks.  It turned out the problems we saw were caused by the early A1 revision of the chipset installed on this pre-production board.  We spoke with Shuttle about these problem and they informed us that they have a board with the newer A2 revision of the SiS648 installed available that addresses the issues we saw.  We'll revisit this section once we get our hands on one of the updated boards. TESTING METHODOLOGY: Due to this fact that we, and we're sure you, have seen significant variations in benchmark scores from one site to the next, we feel it is necessary to explain exactly how we configure our test systems before running any benchmarks. When testing these boards, the first thing we did was enter the system BIOSes and set each board to "Load Optimized Defaults". We then configured the Memory CAS Latency and other memory timings to be set by the SPD (2.5-3-3-1T).  We ran the memory at 200MHz (DDR400) and 166MHz (DDR333) on the Shuttle SiS648 based AK45GT/R in "Ultra Mode", and at 533MHz (PC1066) on the Asus P4T533. The hard drive was then formatted, and Windows XP Professional was installed. After Windows was completely installed, we hit the Windows Update site and downloaded all of the available updates with the exception of the ones related to Windows Messenger. Then we installed all of the necessary drivers, and disabled then removed Windows Messenger from the system.  Auto-Updating and System Restore were also disabled, and we setup a 768MB permanent page file. Lastly we set the Visual Effects to "best performance", installed all of the benchmarking software, defragged the hard drive and ran all of the tests at the CPU's default clock speed.  Now, for our results... The Hot Hardware Test Systems Lots of Intel Hardware Here!   Intel Pentium 4 2.4GHz (2400MHz) 533MHz FSB   Shuttle AS45GT/R (SiS 648) 512MB Corsair PC3200 NVIDIA GeForce 4 Ti 4600 (28.32 Drivers) On-Board NIC On-Board Sound IBM 7200RPM 30GB HD Creative Labs 52X CD-Rom Standard Floppy Drive Windows XP Professional (DirectX 8.1) SIS AGP Driver v1.10     Intel Pentium 4 2.4GHz (2400MHz) 533MHz FSB   Asus P4T533 (i850E) 512MB RDRAM RIMM4200 NVIDIA GeForce 4 Ti 4600 (28.32 Drivers) On-Board NIC On-Board Sound IBM 7200RPM 30GB HD Creative Labs 52X CD-Rom Standard Floppy Drive Windows XP Professional (DirectX 8.1) Intel Chipset Drivers v4.00   Performance Comparisons with SiSoft SANDRA Synthetic Action

SANDRA (the System ANalyzer, Diagnostic and Reporting Assistant) is an information and diagnostic utility put out by the folks at SiSoftware.  Besides benchmarking, it provides a host of other information about your hardware and operating system.  We began our testing with the built-in memory sub-system tests that are part of the SANDRA 2002 suite.

In both the Integer and Floating point tests Intel's i850, coupled with PC1066 RDRAM, was able to out run the SiS648 and transfer over 3.3GB of data per second.   Currently, the i850's memory throughput remains about 16% higher, but in time, we feel this performance advantage will be reduced (we'll explain why a little later).  Running at DDR400 speeds, the SiS648 based Shuttle AK45GT/R was able to transfer over 2.8GB of data per second.  At DDR333, the SiS648 was able to break the 2.5GB per second mark.  These scores are very good, in fact, they are the highest we've seen from any P4 DDR chipset while running at default clock speeds.

Performance Comparisons with PC Mark 2002 CPU, Memory and Hard Drive Testing

Next up, we have MadOnion's PCMark 2002 benchmarking suite.  We like testing with PCMark 2002 because it is very simple to run, and produces repeatable results.  We ran PCMark's "CPU" and "Memory" performance modules.  The CPU module incorporates the following tests:

CPU Test:

• JPEG decompression

• Zlib compression & decompression

• Text search

• MP3 Audio Conversion

• 3D Vector Calculation

As you can clearly see, the SiS648 chipset does not hinder our CPU's performance at all, as it performed just slightly better than the i850.  These type of minuscule performance differences don't equate to any "real world" performance though, so let's move on to something more meaningful, shall we?

Memory Test Technical details: (Quoted)

Raw read, write, and read-modify-write operations are performed starting from a 3072 kilobytes array decreasing in size to 1536 KB, 384 KB, 48 KB and finally 6 KB. Each size of block is tested two second and the amount of accessed data is given as result. In the STL container test a list of 116 byte elements is constructed and sorted by an integer pseudo-random key. The list is then iterated through as many times as possible for 2 seconds and the total size of the accessed elements is given as result. There are 6 runs of this test, with 24576 items in the largest run corresponding to a total data amount of 1536 KB, decreasing in size to 12288 items (768 KB), 6144 items (384 KB), 1536 items (96 KB), 768 items (48 KB) and 96 items in the smallest run corresponding to 6 KB of total data.

As was the case with the SiSoft SANDRA tests, the i850 pulled ahead of the SiS648 at both DDR400 and DDR333 memory speeds.  The performance differences with PCMark 2002 were much more pronounced though, with the i850 / PC1066 combo outpacing the SIS648 running at DDR400 speeds by 22% and by 33% at DDR333.