FSB Overclocking Results

Front Side Bus Overclocking Testbed

Processor:

Pentium 4 Prescott LGA 775
840EE Dual Core 3.2Hz with HT

CPU Voltage:

1.4250v (1.4000v default)

Memory Settings: CPU

4-4-4-15 at 866MHz

Memory Settings: FSB

3-3-3-8 at 732MHz

Memory Voltage:

2.2v

NorthBridge Voltage:

Auto

SouthBridge Voltage:

Auto

Cooling:

Intel 840EE Heat Sink

Power Supply:

OCZ Power Stream 520

Maximum CPU OverClock

260fsb x 16 (4160MHz) +30%

Maximum FSB OverClock

275fsb x 14 (3850MHz) +37%




This board is an excellent overclocker in our initial tests. The board was able to complete the entire benchmark suite at these settings. We will continue to experiment with the board but it was able to boot into Windows at a FSB setting of 300 with the 14x multiplier.

Memory Stress Testing

Memory Tests

Memory stress tests look at the ability of the Gigabyte GA-G1 975X to operate at the officially supported memory frequencies of 667MHz DDR2 at the best performing memory timings the Corsair CM2X512A-5400UL revision 1.3 will support.

Gigabyte GA-G1 975x Deluxe
Stable DDR667 Timings - 2 DIMMs
(2/4 slots populated - 1 Dual-Channel Bank)

Clock Speed:

200MHz (800FSB)

Timing Mode:

667MHz - Default

CAS Latency:

3

RAS to CAS Delay:

2

RAS Precharge:

2

RAS Cycle Time:

8

Voltage:

2.1v



The Gigabyte GA-G1 975X was completely stable with 2 DDR2 modules in Dual-Channel at the settings of 3-2-2-8 at 2.1v.

Test Setup

The Intel 975X chipset fully supports the 820, 830, and 840 dual core Pentium D processors in both stock and overclocked conditions. Dual core really makes a difference in certain multi-tasking scenarios, as was demonstrated in the dual core performance preview. If you are interested in how the various chipsets perform in a real world multitasking setup please take another look at that review.

The board's memory was operated 3-2-2-8 for the benchmarking suites. The Gigabyte GA-G1 975X automatically overclocks the processor system bus up to a setting of 206 so our tests were conducted with a FSB setting of 199. We are very concerned about the inability to lock the FSB in manual mode between the settings of 200 and 205. We set the FSB to 200 in manual mode, the system at post would report a 200 FSB setting, and then the system would report a 206 FSB setting via CPU-Z or Everest once in Windows. We have relayed our concerns to Gigabyte and will report further on this feature and test results in the full article.

Performance Test Configuration

Processor:

Intel Pentium 840EE (3.2GHz, 800FSB, Dual-Core, 2x1MB L2, HT) utilized for all tests.

RAM:

2 x 512mb Corsair CM2X512A-5400UL revision 1.3 Settings- DDR2-667 as noted at (CL3-2-2-8)

Hard Drive(s):

2 x Maxtor MaXLine III 7L300S0 300GB 7200 RPM SATA (16MB Buffer), 1 x Maxtor MaXLine III 7L300R0 300GB 7200 RPM IDE (16MB Buffer).

System Platform Drivers:

Intel Chipset Software - 7.2.2.1006

Video Cards:

1 x XFX 7800GTX OC (PCI Express) for all tests.

Video Drivers:

NVIDIA nForce 78.01 WHQL and 81.85 WHQL as noted

Operating System(s):

Windows XP Professional SP2

Motherboards:

Asus P5N32-SLI Deluxe
Gigabyte GA-8I955X Royal
MSI P4N Diamond

Gigabyte GA-G1 975X: Features General & Gaming Performance
Comments Locked

30 Comments

View All Comments

  • johnsonx - Friday, November 11, 2005 - link

    Gary,

    From an airflow perspective, wouldn't it make more sense to have all the fans blowing the same direction? From the pictures and your description, it appears the while the fans at the backplate are blowing out, as they should, the 2nd set of fans attempt to blow air back at the memory.

    The trouble is that those 2nd fans are trying to pull air against the general airflow direction in the case, and whatever air they do blow over the RAM will be warm from the CPU and other heatsinks in those platic tunnels.

    I think in a real-life case, the 2nd set of fans should blow towards the back of the case, drawing cool air from the front of the case past the DIMMs, thus creating a wind-tunnel effect to more effectively draw the hot exhaust from the CPU HSF to the back fans and out of the case.

    Perhaps if it can be done, you could try swapping the 2nd fans around to get them blowing the other way and do some measurements. I can't tell if it's possible from the pictures; the tunnels and fan mount brackets look like they could be one-piece. I suppose that you could reverse the polarity of the power pins to spin them backwards, but those fans don't look like they'd move much air spinning the wrong way.

    Yes, I do realize that Gigabyte's Engineers must have worked on this, but I've had 'engineers' tell me riduculous things about airflow in the past that I proved wrong by simply flipping the fans around. One that comes to mind from awhile back was a KryoTech Super-G that had the back fan blowing in and the front fan blowing out... their engineers INSISTED that it was supposed to be that way, that they had tested, blah, blah, blah... after feeling the side panel getting warm and nearly burning my hand on the SCSI drive in the bay, I reversed the fans and all was well. And that was a system with NO cpu heat at all!

    Airflow in an ATX case should always be front to back, bottom to top.
  • Gary Key - Friday, November 11, 2005 - link

    quote:

    From an airflow perspective, wouldn't it make more sense to have all the fans blowing the same direction? From the pictures and your description, it appears the while the fans at the backplate are blowing out, as they should, the 2nd set of fans attempt to blow air back at the memory.


    This was exactly my first thought when I looked at the board, read the engineering information, and then powered on the board. My first reaction was the thermals coming off the MCH would be blown right across the memory slots and then getting swept into the power supply or back over the CPU area creating additional heat. I think I was right to some degree. ;->

    However, Gigabyte is recommending and has provided pictures of their preferred CPU cooling system which consists of a cooler that is designed like their G-Power series that blows air over the entire CPU surface area. According to Gigabyte their reduction in thermals at the memory, MCH, and MOFSETs were accomplished with this combination. I am trying their cooling suggestions currently along with swapping the 40mm fans with others to blow in opposite directions. This is why I was not about to publish their test results until we had a chance to fully review the system and take our own measurements.

    I have not completed testing and still have additional cooling units to try along with additional measurements from the stock heatsink/fan. My initial thoughts and test results favor the Asus 8-phase setup with the fanless heat pipe system although the Gigabyte design actually works well. I will have measurements and a final synopsis in the full article. Until then if I find something of significance I will post an update.

  • bldckstark - Friday, November 11, 2005 - link

    quote:

    The DIMM module slots' color coordination is correct for dual channel setup.


    I am somewhat confused by the above listed statement as it can mean several things to me.

    Is there a specification for dual channel slot color coding, or are you stating your personal preference?

    I know it is nitpicky, but I am curious.

    Thanks for the article, you did a great job, first time or not.
  • Gary Key - Friday, November 11, 2005 - link

    quote:

    Is there a specification for dual channel slot color coding, or are you stating your personal preference?


    There is not a hard standard for the color designation. Most manufacturers have started color-coding the various port connectors and slots but in some cases these colors do not mean anything. With the memory slots it is now customary to color code the slots to signify which slots need to be filled to enable the dual channel memory capabilities of the board. Instead of signifying in the user's manual to utilize slot one and three or two and four for dual channel capability the quick setup sheets can now use a visual color to represent the slot numbers. It is easier to understand this way and helps when you cannot locate the user manual while you are stuck under a desk with low light conditions. ;->

    The main problem is that some boards (earlier review this year) have the first and second slots the same color and slots three and four a different color yet you have to use slot one and three for dual channel. This is confusing to others and me as it is not intuitive and does not follow a "soft" standard that has developed regarding the memory slot color-coding.

  • johnsonx - Friday, November 11, 2005 - link

    Yeah, I've found the color coding of the memory slots pretty useless, as I can never be sure whether they mean to put the two dimms in the two slots that are colored the same, or colored different. I always just figure the first two slots are one channel, and the other pair are the other channel (usually there's a gap between the two pairs, but not always).

    Personally, for me it makes more logical sense to have the slots the same color within a channel, so you put the two dimms into different color slots. In other words, the different color should signify a different channel. This scheme also lends itself well to single-channel boards - their slots should all be one color, signifying one channel.

    That said, I can also understand the logic of making the first slot of each channel the same color, and the second slot of each channel a different color. So then you put each pair of DIMMs in the same-color slots.

    Either way though, they (the MB manufacturers as a group) really ought to pick one method and stick to it. For now, I'll just ignore the colors.

  • rqle - Friday, November 11, 2005 - link

    Can i see a pic of the backplate I/O shield? Old request, probably not even gonna get read. But it just look interesting.
  • Gary Key - Friday, November 11, 2005 - link

    Hi,

    I will see what I can do for you.
  • Marlowe - Friday, November 11, 2005 - link

    Is this the final implementation of the 975x chipset? I heard there were going to be some sort of an upgrade along its lifetime, as the "current" 975x based motherboards does not support the future processors conroe and yonah? (because those two will be pin compitable right?) More PCI-Express lanes as seen in ATI's new RD580 chipset would also be nice.
  • Gary Key - Friday, November 11, 2005 - link

    Good Day,

    I am awaiting word from Intel about the 975X progression path and will report any findings along with additional technical information in the full article. This article was meant as a preview and since the chipset has not officially launched there were details I could not provide at this time. I fully agree that additional PCI-Express lanes would be nice. If I receive additional information from Intel before the full article is released I will update this preview.
  • xsilver - Friday, November 11, 2005 - link

    how is putting a turbo on a scooter going to help?? ;)

    and i guarantee that all 4 of those tiny fans are going to either:
    1) die in 4 months
    2) sound like a turbocharger in 4 months


    I think the next step up should actually be intergration with the case, their is just too much contstriants otherwise (especially working with the oldish ATX structure)

Log in

Don't have an account? Sign up now