Logo   Banner   TopRight
Gigabyte's GA-8S648FX Motherboard
Date: Aug 14, 2003
Author: HH Editor
Intel Springdale Showdown - Page 1

Gigabyte's GA-8S648FX Motherboard
Bridging the gap with the SiS 648FX

"Burned" in by Robert Maloney
August 13th, 2003


A lot of hype and controversy has surrounded Intel's latest chipsets, the i865 Springdale and the i875 Canterwood, and they are proving to be very capable solutions for the Pentium 4 processor. But the question remains to be asked; what have the other chipset designers been up to? We haven't seen VIA's latest offering make much noise, but SiS has kept forging ahead, looking to improve on their previous offerings. If you recall, we had a look at an MSI board based on the SiS648 back in late October of 2002, and it fared very well when compared to Intel i845PE based motherboards. We even went so far as to label it as one of the fastest DDR-based chipsets for the Pentium 4. Since then, starting with Granite Bay motherboards, dual channel DDR has become the norm, and front side buses have been pushed up to 800MHz. SiS went back to the proverbial drawing board, and tinkered with the 648 chipset to create the 648FX. The 648FX supports CPU Hyperthreading at 400, 533, and 800MHz FSB, and officially supports DDR400, but only in single channel configurations. Obviously, the memory bandwidth is therefore much lower than the Canterwood or Springdale, and they can't necessarily be compared head to head.. Instead, the SiS648FX can be considered something of a "go-between", offering an upgrade in performance over 845-based boards, but falling short of an 865/875. Gigabyte's GA-8S648FX is the first motherboard based on the 648FX that we have had in the labs. Is there enough power in this board to make this Giga bite?  Ouch, that was rough.


Specifications & Features of The GA-8S648FX
What makes this board tick?



  • IntelĀ® PentiumĀ® 4 Processor with Hyper-Threading Technology

  • Socket 478 Pentium 4 Processors

  • Supports 800MHz, 533MHz or 400MHz Front Side Bus


  • SiS648FX Host/Memory Controller - North Bridge

  • SiS963 MuTIOL Media I/O - South Bridge


  • Supports 3 x 184-pin DDR SDRAM

  • DDR266/333/400 support

  • 3GB maximum system RAM (unbuffered)


  • Licensed Award BIOS

  • Supports PnP, ACPI, DMI, STR

  • 2Mbit flash ROM

  • Supports Q-Flash, @BIOS


  • System health status auto-detect and report by BIOS

  • Hardware detecting and reporting for case open, power-in voltage, CPU voltage, and fan speed.


  • Realtek ALC650 v2.2 6-channel sound CODEC

  • Digital audio interface supports 24-bit S/PDIF optical in/out



  • 2 IDE bus master (UDMA33/ATA66/ATA100/ATA133) IDE ports

  • Supports up to four ATAPI devices

  • Supports PIO Mode 3,4 Enhanced IDE and ATAPI CD-ROM

  • Bus mastering reduces CPU utilization during disk transfer


  • 2 USB 2.0/1.1 ports

  • 2 DB-9 serial ports

  • 1 DB-25 parallel port

  • 1 mini-DIN-6 PS/2 mouse port

  • 1 mini-DIN-6 PS/2 keyboard port

  • 3 audio jacks: line-in, line-out,  and Mic-in

  • 1 Game/MIDI Port


  • 2 connectors for 4 additional external USB 2.0/1.1 ports

  • 1 1394 pin header (supports 3 FireWire ports via optional daughter card S1394)

  • 1 front audio connector for external line-out and Mic-in jacks

  • 2 internal audio connectors (CD-in and AUX-in)

  • 1 connector for Surround/Center audio channel

  • 1 S/PDIF output connector

  • 2 IDE connectors

  • 1 floppy connector

  • 2 ATX power supply connectors

  • 3 fan connectors for CPU fan, NB fan, and chassis fan


  • 1 AGP 3.0 slot supports 8X/4X modes (1.5V support only)

  • 5 32-bit PCI 2.2 slots


  • 4 layers, ATX form factor

  • 29.5cm x 20.9cm


Specifications of the SiS648FX
What's happening on the other side of the fence

SiS made a tidy little upgrade to their existing 648 chipset, adding support for 800MHz front side bus Pentium 4 CPUs and Hypertheading while keeping all of the good points intact.  Combined with DDR400 and AGP8x support, and SiS's proprietary MuTIOL interconnect between the North Bridge and South Bridge, the SiS648FX is poised as a high-end, low-cost solution for the mainstream user.  What SiS doesn't bring to the table, however, is dual DDR memory channels, which would have increased overall performance, and native support for Serial ATA or RAID configurations.  You might notice that 2 SATA ports are listed in the chipset schematic above, but these come at the price of a PCI-based SiS 180 Serial ATA controller.  Being PCI-based, this could lead to bottlenecks on the already burdened PCI bus.  From what we have heard, SiS' future plans involve a new SiS964 South Bridge, which should have SATA support built in, and is the logical choice for future revisions of the 648FX and 655FX.

Considering the features, the 648FX falls somewhere in between the older i845/SiS648 boards and the current crop of Springdales and Canterwoods.  So, on which end of the scale will it end up?  We took one company's SiS648FX board, the Gigabyte GA-8S648FX, and compared it to two boards from our Springdale Showdown to find out.  While doing so, we took one board, the Abit IS7-G, that comes with a PAT-like memory enhancement, and another board that didn't, the Chaintech Apogee 9PJL.  These two boards should give us a good idea of how the performance of the SiS648FX compares.

We take a closer look at the challenger

Intel Springdale Showdown - Page 2

Gigabyte's GA-8S648FX Motherboard
Bridging the gap with the SiS 648FX

"Burned" in by Robert Maloney
August 13th, 2003

The Gigabyte GA-8S648FX Motherboard:


We came away with very good first impressions, when opening the box and removing the GA-8S648FX motherboard.  First, the board is color coordinated, from the plastic used for the ports and sockets to the pins used for the front panel.  Gigabyte was one of the first manufacturers to start color coding these pins, and it helps prevent confusion when setting up the board.  The bright orange CPU bracket and lime green AGP and IDE ports stand out against the teal PCB, which Gigabyte has become accustomed to using.  Light blue is used for all three DIMM slots, as there is no need to have the two tone slots that we have gotten used to with the dual-DDR channel boards.  Each of the corners is rounded off, which is a small touch that DIY builders can really appreciate, especially after getting jabbed by a sharp corner one too many times. 


As for the layout of the components, we didn't have too much to mention, at least not without beginning to sound like a broken record.  For example, yet once again, the AGP slot and DIMM slots are placed too close together.  In fact, this may be the closest we have seen these two in some time.  Take a peek at the picture of the lower left of the board and you can see the DIMM retention clips almost touching the AGP slot.  Now that's close!  There was also a small capacitor to the left of the AGP pull clip and a medium one on the right.  That results in having to squeeze your thumb and finger right between the two to pull out the plug to release the AGP card.  Unless there is a strict need for these to be placed here, Gigabyte should look into better placement for these capacitors.  Other than these two issues, the IDE and ATX connections were placed in good positions, along the edge of the board and away from other components.  The floppy port was placed further down the board, just past the last PCI slot, which may cause an issue with larger cases.  It was placed perpendicular to the front of the case, however, which will prevent the cable from hindering airflow around the case.


One of the first things we noticed when we looked over the components was not so much what we saw on the board, but rather what was missing.  We quickly found the Realtek ALC650 Codec chip used for the on-board audio, a common chip found on many modern motherboards.  Just to the left, however, was a label where a Realtek LAN chip could or would have been.  To save on some of the costs of manufacturing, Gigabyte opted to leave on-board LAN off of the list of provided options.  There also aren't any additional IDE or SATA ports for connecting additional drives, leaving only the IDE1 and IDE2 ports for any and all drives one plans to install.  Obviously, this also rules out using RAID configurations of any kind.  One last omission, and one that we would constantly grieve over while testing, was the apparent removal of the pins needed to quickly clear the CMOS.  Again, the label is there, but, inexplicably, the pins are not.  We are at a loss as to why these would not be present, as a few pins clearly would not add or detract from the cost of making the board. 



The bundle follows the same theory as the board.  There's just enough to get the system running with few extras.  There is a quick installation guide for a visual reference when building the system along with the user's manual, driver/utility CD, and a handy little sticker with the motherboard layout for those of us who often misplace the manual.  So far, so good.  Delving deeper into the box, we only found an IDE cable, a floppy cable, and a bracket with two additional two USB 2.0 ports bringing the total out of the box to four.  Notice we didn't mention an I/O shield.  That's because there wasn't any in the box, and no, this wasn't a mistake.  Looking over the checklist within the manual, one can readily see that this item is simply not on the "provided" list.  Another oddity, as it means that you would have to somehow have a plate that exactly matched the configuration of the external connections of the board.  Needless to say, we didn't, and left the back of the system open while testing.



We booted up the system for the first time, hit the 'Delete' key, and felt right at home with the AWARD BIOS.  We quickly enabled and disabled various components and went to optimize the RAM timings, when we realized that there weren't any.  Actually, that's not the whole truth.  They were there, but Gigabyte's policy is to "hide" them, requiring the user to hit CTRL-F1 in order for the Advanced Chipset Features to appear.  We found that our GEIL DDR433 RAM was detected BY SPD as DDR400 with some relaxed timings.  We left these settings for the time being, but set the system performance to 'Enabled' from the main menu.  Unfortunately for us, this resulted in the system being unable to boot (just like the BIOS warning had said might happen, isn't that a hoot).  We rebooted the system a few times, but were unable to get back into the BIOS.  We went to clear the settings, but alas, no pins to short.  So we removed the battery, waited a few seconds and placed it back in, then fired the system up again.

Back in the BIOS, we reconfigured all of our previous settings, and left system performance to default.  This time we got past the POST and finished the initial system setup.  Everything behaving as it should, we went back into the BIOS to tweak those memory settings they way we usually do.  We manually set the RAM to DDR400 with 2-5-2-2 for the timings.  No boot.  OK, back to removing the battery, resetting the BIOS, etc.  This time we tried going halfway, maybe 2.5--5-3-3, but this was a no-go as well.  Back to the routine once again.  Needless to say, this was most frustrating, especially when we finally came to the realization that ANY change to the memory timings would cause the system not to boot.  We searched around the labs and tried throwing in one stick of 512MB Corsair PC3500 and two 256MB sticks of Kingston HyperX PC3500, and they all performed the same.  Faced with this situation, we had no other choice but to run our benchmarks with the RAM set to 3-6-3-3, and system performance left at its default setting.  Gigabyte would do well to take another look at this situation with a BIOS revision, but a lot of time could have been saved had the Clear CMOS pins been present.

Overclocking the GA-8S648FX
Umm...Houston...we appear to have a problem

We would like to tell you that we had a good experience when overclocking this board, but we can't.  To expand this further, we couldn't overclock this board at all.  While we expected that overclocking was going to be limited, due to the lack of available voltage settings, we were thoroughly perplexed to find out that changing the front side bus to anything above 204MHz, resulted in the system not even being able to boot.  Which led to us needing to clear the CMOS.  Which led to us removing the battery.  Which led to us getting frustrated.  You get the picture.  In our experience, SiS based boards just don't do well in this department, so overclockers should probably look elsewhere.

Making some comparisons

Intel Springdale Showdown - Page 3

Gigabyte's GA-8S648FX Motherboard
Bridging the gap with the SiS 648FX

"Burned" in by Robert Maloney
August 13th, 2003

PCMark 2002
Synthetic CPU and Memory Bandwidth Testing

To get another look at the CPU and Memory performance, we turned to PCMark 2002. This benchmark from FutureMark performs a series of CPU tasks such as image compression, text searches, and audio conversion to give us three scores: CPU, Memory, and Hard Disk Drive (HDD).   Although it is a relatively quick test to perform, it is perfect for comparing the performance of two or more systems.  We ran PCMark2002's "CPU" and "Memory" performance modules on all of the test systems, with the CPU clocked at its default speed of 2.40GHz. 

The CPU performance module had the the Abit IS7-G at the top, with the Gigabyte GA-8S648FX at the bottom, but the margin of difference was a mere 83 points, a delta of about 1%, well within the margin of error for such a test.  At the very least, we can see that the CPU is operating within normal operating limits on the Springdale and SiS648FX boards.

Memory Test Technical details: (Quote Taken From Futuremark)

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.

This is another benchmark that separates the men from the boys.  Abit's Game Accelerator technology blows away the other two boards, including the Chaintech 9PJL which also uses dual DDR channels.  The Springdale-based boards have an increased memory bandwidth of 20% for the Chaintech, and almost 40% for the Abit IS7-G.  The single DDR channel architecture of the 648FX simply can't compare, and to be honest, isn't expected to.  The score we got here was more indicative of the performance level we had seen on i845 boards.
Business & Content Creation Winstones
Simulated Application Performance

To find out how the boards relate in "Real World" performance, we used eTesting Labs' Business and Content Creation Winstone 2002 benchmarks.  We'll directly quote ZD's eTestingLabs website for an explanation as to how Business Winstone 2002 derives its score. (Content Creation Winstone 2002 uses the same process, but the scripted activities are comprised of different, more bandwidth hungry applications.):

"Business Winstone is a system-level, application-based benchmark that measures a PC's overall performance when running today's top-selling Windows-based 32-bit applications on Windows 98, Windows 2000 (SP2 or later), Windows Me, or Windows XP. Business Winstone doesn't mimic what these packages do; it runs real applications through a series of scripted activities and uses the time a PC takes to complete those activities to produce its performance scores."

Business Winstone Applications:
  • Five Microsoft Office 2002 applications
    (Access, Excel, FrontPage, PowerPoint, and Word)

  • Microsoft Project 2000

  • Lotus Notes

  • WinZip 8.0

  • Norton Antivirus

  • Netscape Communicator

Content Creation Winstone Applications:
  • Adobe Photoshop 6.0.1

  • Adobe Premiere 6.0

  • Macromedia Director 8.5

  • Macromedia Dreamweaver UltraDev 4

  • Microsoft Windows Media Encoder

  • Netscape Navigator 6/6.01

  • Sonic Foundry Sound Forge 5.0c (build 184)

Business Winstone 2002 gave us our first real surprise.  All three boards were tightly grouped, but the Gigabyte board pulled into second place behind the IS7-G this time.  Still, with the difference being only 0.3 stones from top to bottom, we can call this a tie.  Content Creation Winstone 2002 shows a bit more disparity.  Abit's IS7-G clearly ran away from the other two, 1.3 stones above the Chaintech 9PJL, and 2.2 stones over the Gigabyte GA-8S648FX.  Considering the memory handicap, however, the Gigabyte board still puts up a good showing.

Gaming benchmarks and the conclusion

Intel Springdale Showdown - Page 4

Gigabyte's GA-8S648FX Motherboard
Bridging the gap with the SiS 648FX

"Burned" in by Robert Maloney
August 13th, 2003

3DMark 2001 and 3DMark03
Synthetic Gaming

We wanted to get an idea of how the SiS648 would perform in gaming situations, and started with Futuremark's 3DMark 2001 SE and 3DMark03.  Both of these benchmarks render a variety of scenes, using DirectX 8 and 9 vertex and pixel shaders.  We ran both benchmarks at exactly the same settings; 1024x768 resolution and 32-bit color, with all other settings left at their defaults. 

The memory bandwidth came into play with the 3DMark 2001 scores.   The GA-8S648FX came in last, with 11895 points, a full 10% behind the leader, the Abit IS7-G.  3DMark03 was a bit more forgiving, since the scores have less to do with system performance, and are more closely tied to the video card being used.  There was a slight drop off in scores from board to board, in the same order as the previous test.  The difference this time was only about 2%.
Comanche 4 Demo and Quake 3 Arena
Let's get a look at some frame rates

OK, enough mumbo-jumbo about synthetic numbers, lets see some real games in action.  We took two games that have built-in testing modes, which will show us the average frame rate during gameplay.  Comanche 4 is a DirectX benchmark that is highly CPU and Memory bandwidth dependent.  For testing purposes, we ran the Demo version at 800x600x32, but disabled the audio.  Quake 3 Arena is everyone's favorite OpenGL benchmark, used in reviews seemingly since the dawn of time.  Although a bit dated, it still can be used to give reliable comparisons of system performance.  Since the frame rates can get quite high, we maxed out the graphical settings, and ran "demo four" at a resolution of 1024x768 with 32-bit color and textures.

The Comanche 4 and Quake 3 benchmarks gobble up the memory bandwidth that the Springdale boards provide, and help them run right past the SiS648FX.   The Comanche 4 scores might not seem as big of a deal when you only see a difference of 6 frames from top to bottom.  These six frames actually equate to a difference in performance of almost 13%.  This difference can be seen better in Quake 3 Arena where the Abit IS7-G outpaces the Gigabyte GA-8S648FX by almost 40 frames per second.

First impressions aside, it was hard for us to find too much to root for with Gigabyte's SiS648FX-based motherboard.  Missing features, mediocre performance, and the lack of any overclocking ability really put it at a disadvantage, especially for power-hungry PC enthusiasts.  On the other hand, if one was looking for a stable, but cheap solution that doesn't require finding twin-sets of DDR RAM, then the GA-8S648FX may be a perfect match.  At a price point as low as $87, it's not a bad choice when looking to upgrade to 800MHz front side bus Pentium 4 CPUs.  Our general feeling is that it may be better to wait for a 655FX based solution, which will combine the 800MHz FSB with Dual DDR.  As for the Gigabyte GA-8S648FX, we are going to give it a 6 on the HotHardware Heat Meter.



Gigabyte's GA-8S648FX Motherboard Page 5

Content Property of HotHardware.com