FSB Overclocking Results

With the very complete selection of overclocking options on the Gigabyte, we had very high expectations for the overclocking performance of the board. What we didn't know, and still don't know, however, is what kind of total overclocking range will be available to FX users.

Front Side Bus Overclocking Testbed
Default Voltage Maximum Overclock
Processor: Athlon64 FX51
2.2GHz
Athlon64 FX51
2.2Ghz
CPU Voltage: 1.5V (default) 1.6V
Cooling: AMD Stock Athlon64 FX51 Heatsink/Fan AMD Stock Athlon64 FX51 Heatsink/Fan
Power Supply: Enermax 465W Enermax 465W
Maximum OC: 2409MHz (+10%)
219FSB
?

The above overclocking setup at DEFAULT voltage allowed us to reach a stable FSB of 219MHz. Running our standard Quake3 bench at 2.4GHz yielded Q3 test scores right at 500fps. The maximum overclock, unfortunately, remains a mystery. As you will see in the benchmark suite, the standard test performance with the new F1 BIOS is outstanding. However, there are still problems with multipliers and FSB settings on this motherboard. 11.5X worked until we hit 215FSB; then, mysteriously reset itself to 210, no matter what we set. The 12.0X multiplier worked at 200, and would start generating mysterious FSB and multiplier values above this setting. We even tried downclocking the multiplier, but could not set higher FSB settings that would be retained. We suspect Gigabyte will quickly fix this because this is a very promising board, but for now the Maximum overclock simply cannot be reported. We will revisit this in the near future when we receive a BIOS update.

For a quick comparison, we ran the FX51 in our nVidia Reference board. In that board, by setting the memory to 333, we were able to boot our FX at a CPU setting of 230 at 1.6V. We suspect it may do even better, but with multiplier and FSB settings not completely working, it would be unrealistic to report a maximum overclock at this time. We still believe that with this Gigabyte board, it will be possible to reach the maximum overclock our FX51 chip can achieve. The Gigabyte is remarkably stable in all we can do with it so far, and hopefully a BIOS update will fix the overclocking issues that have surfaced.

Gigabyte K8NNXP-940: BIOS and Overclocking Gigabyte K8NNXP-940: Stress Testing
Comments Locked

35 Comments

View All Comments

  • Anonymous User - Friday, October 10, 2003 - link

    64-bit tests running Linux and hand-compiled programs would be:

    a) Really time consuming
    b) Artificial
    c) Not relevant to the real world
  • Anonymous User - Friday, October 10, 2003 - link

    It is odd that NO 64-bit tests has been made. Why don't people fire up Linux and compile a few programs like MPEG encoding, video/divx processing etc etc?
  • Anonymous User - Friday, October 10, 2003 - link

    Has anyone tried decreasing both the memory speed and the LDT speed when overclocking an athlon 64 board via the fsb?
    The reason I ask is that being able to set the memory, and hypertransport ratio's, may make an independant CPU multiplier adjusment redundant.
    (obviously it would be nice to rule CPU frequency out of such a test)
  • PrinceGaz - Friday, October 10, 2003 - link

    ...almost forgot, why was the P4EE 3.2 not included in the benchmarks?
  • PrinceGaz - Friday, October 10, 2003 - link

    Very very nice board and CPU, and impressive benchmarks throughout (you can't expect it to match the P4 for encoding). But next year's 939-pin FX is definitely the one to wait for.

    #4- QDR is just as unlikely as RDRAM but for different reasons, a key point of the A64/FX is the on-die memory-controller but that means you can't just add another couple of memory-channels to it without a total socket re-design (and for QDR a ridicoulously high pin-count). DDR2 is the way forwards in the future rather than more channels.

    I'd really expected the fastest CPU nearly two years after getting my XP1700+ to be more than just 85% or so faster than it (the Barton 3200+ is barely over 50% faster, and the A64 3200+ about 70% faster). Unfortunately I can't justify an upgrade until its over 2x, preferably 3x as fast so I'll wait into next year and see what speed increases the shift to 90nm brings.
  • Anonymous User - Friday, October 10, 2003 - link

    Very impressive board, We've used Gigabyte boards almost exclusively for the past few years at our computer shop and they just keep getting better with every revision. What I'm looking forward to is what they're going to have coming out early next year for the FX, by that time, having an FX system will become a reality for those of us who can't pay an arm and a testes.
  • Anonymous User - Friday, October 10, 2003 - link

    I want to know whats up with Gunmetal. Otherwise, great review. I just hope that the prices come down, alot, by spring for my upgrade.
  • Anonymous User - Friday, October 10, 2003 - link

    If anyone had doubts about the A64 and FX performance there should be no questions now!

    FX is intended to satisfy the extreme demands of power users who want the best and they want it now. A64 is a more cost effective solution for those who want outstanding performance at a consumer price point.

    As A64/FX ramp all prices will drop as is normal. You'll likely find that the FX series is quite affordable to the enthusiast market and a Helleva value as things ramp.

    And there are some more goodies on the way from AMD and it's partners to make all consumers very happy. Stay tuned!

  • Wesley Fink - Friday, October 10, 2003 - link

    #5 - Regular Opterons are locked - at least that is what we found in the 2 we tested. The FX is unlocked.

    #6 - Yes, this is the first 1394b 800mb/sec Firewire board.
  • mcveigh - Friday, October 10, 2003 - link

    is this the first PC board with firewire800?

Log in

Don't have an account? Sign up now