Home Page

Today's News

Ground ZERO
Latest Rumor
Rumor Archive

 

 
 


mesg. board 
 
 
 

2015 © KickAss Gear

 

 

The Testing: The system we used to test the Afterburner on was an Athlon 500 CPU on an Asus K7M motherboard, with 128MB of PC-133 SDRAM and a Diamond V770 Ultra TNT-2 card.  Rather than doing exhaustive benchmark testing, we were more interested in the maximum speed we could achieve while retaining stability.

Multiplier Overclocking: First we decided to the try 650MHz setting at 1.65 volts Vcore.  This means setting the multiplier to 6.5X, and upping the core voltage by 0.05 volts.  The system booted very quickly, and appeared stable in Windows 98SE.  We ran the 3D Mark 99 Max demo overnight, and came back in the morning with it still plugging away nicely. That's stability!  The next obvious test was 7 X 100MHz = 700MHz.  Windows quickly bombed out at that speed, so we increased the core voltage another 0.05 volts to 1.70v.  This improved stability slightly, but not enough.  It is possible that by installing a larger heat sink with multiple fans, and increasing the core voltage further, might have allowed us to squeeze more MHz out of the Athlon.  But we had another idea.

Plus Bus Overclocking: We next wanted to see if bus overclocking could be done in conjunction with multiplier overclocking by using the K7M and Afterburner's overclock features together.  Our first attempt was using the 6.5 multiplier, and a bus frequency of 103MHz, for a final speed rating of 669MHz. We stuck with the Vcore setting of 1.65v. The system was very stable, and ran all benchmarks and demos without any problems! Increasing the bus frequency one more MHz proved too much.  While the system ran all benchmarks at 676MHz (6.5X 104MHz), it crapped out after running the 3D Mark 99 Max demo for awhile.  So with the standard heat sink and fan provided by AMD, the K7M and Afterburner card permitted us to run a 500MHz Athlon processor with excellent stability at 669MHz!  This would save you over $100, even after accounting for the cost of the Afterburner card.

Benchmarks: We ran benchmarks at 650MHz, 669MHz, (the top speed that the system was fully stable at), as well as at 676MHz, where instability first became noticeable.  We then compared these results to the same benchmarks run on the same system with just bus overclocking alone.  The maximum bus frequency that the Athlon would overclock to on the K7M was 107MHz, yielding a final speed rating of 535MHz (5 X 107MHz).

Without Afterburner:
Athlon 500 @ 500MHz (5 X 100MHz, default)

3D Mark 99 Max:

   5175 3D Marks
   9202 CPU 3D Marks

Winbench 99:
  
  47.9 Winbench CPU
    2740 Winbench FPU


Athlon 500 @ 535MHz (5 X 107MHz)
3D Mark 99 Max:

   5331.5 3D Marks (3% increase over default)
   10011 CPU 3D Marks (8.8% increase over default)

Winbench 99:
  
50.175 Winbench CPU  (4.7% increase over default)
   2917.5 Winbench FPU  (6.5% increase over default)



With Afterburner:
Athlon 500 @ 650MHz (6.5 X 100MHz)

3D Mark 99 Max:
    5862 3D Marks (13.3% increase over default)
    11203 CPU 3D Marks (21.7% increase over default)

Winbench 99:
  
  60.4 Winbench CPU (26% increase over default)
    3570 Winbench FPU (30% increase over default)


Athlon 500 @ 669MHz (6.5 X 103MHz)
3D Mark 99 Max:
    5904 3D Marks (14% increase over default)
    11317 CPU 3D Marks (23% increase over default)

Winbench 99:
  
  61.3 Winbench CPU (28% increase over default)
    3650 Winbench FPU (33% increase over default)


Athlon 500 @ 676MHz (6.5 X 104MHz, unstable*)
3D Mark 99 Max:
    5982 3D Marks (15.6% increase over default)
    11381 CPU 3D Marks (23.7% increase over default)

Winbench 99:
  
  61.6 Winbench CPU (28.6% increase over default)
    3690 Winbench FPU (34.7% increase over default)

* All tests were completed normally, but the system hung in 3DMark 99 Max demo.

Bring on the Fans!:  Since the failures at 676MHz and 700MHz were not catastrophic, the next obvious step was to increase the core voltage to the CPU further, and add more cooling.  So we simply put an 80mm fan blowing air across the heat sink fins, and boosted the Vcore to 1.75v (0.15 volts above default). The system booted into Windows, and was completely stable.  At this speed, the CPU heat sink was slightly warm to the touch, and the motherboard temperature reading was 86 degrees Fahrenheit.

Athlon 500 @ 700MHz (7 X 100MHz)
3D Mark 99 Max:

    5911 3D Marks (14.2% increase over default)
    11366 CPU 3D Marks (23.5% increase over default)

Winbench 99:
  
  64 Winbench CPU (33.6% increase over default)
    3840 Winbench FPU (40% increase over default)

Note that the 3D Marks were higher at 676MHz than at 700MHz.  This is because the V770 Ultra's fill rate was limiting at a bus frequency of 100MHz and with a multiplier above 6.5X.  However, by increasing the bus frequency 3MHz, we also increased the AGP slot speed from 66MHz to about 68MHz. This allowed the card to pump a few more pixels per second, due to the boost in clock cycles per second.  Bus overclocking can increase video card fill rates, while multiplier overclocking can not.

  Despite the lack of improvement in Direct 3D benchmarks with the V770 Ultra by going from 650MHz to 700MHz, the increase in raw processing power was evident in the Winbench scores, which were up substantially relative the scores at 650MHz.

  The system even booted into Windows 98SE at 7 X 103MHz = 721MHz, but showed signs of instability. If an L2 cache divider adjustment was available on the Afterburner, we most likely could have overclocked the system further.  Also, with active Peltier cooling, further overclocking probably would have been possible.

Results:  The results are in!  Overclocking the Athlon using higher multiplier settings delivers much higher benchmark scores than bus overclocking.  Bus overclocking resulted in a 3% - 9% speed increase, while Multiplier overclocking resulted in a 14% - 40% increase.  Obviously, multiplier overclocking was far more effective than bus overclocking with the AMD Athlon processor.

Conclusions: Multiplier Overclocking is back! at least with the AMD Athlon.  Intel ain't goin' down that road. If you want an Athlon, and you're an overclocking nut like we are, you want this card.  If you want an overclocked Athlon system, but don't feel comfortable prying the plastic case off your CPU and playing around with system settings, you should buy a pre-configured Athlon system with an Afterburner card already installed in it.  Our Madness system, for example, can be configured with the Afterburner card for you.  

  Multiplier overclocking is just one more reason now to consider an Athlon system over a Pentium III system.

Pros:

  • Affordable
  • Easy to use
  • Overclocks the begeebers out of your Athlon
  • Offers overclock options not available with motherboards
  • Like hand grenades... safe if used properly

Cons: 

  • Could be a little less expensive (costs $80 retail)
  • You need to remove the plastic casing on the Athlon, which voids your warranty
  • No L2 cache divider adjustment

Despite the Cons noted above, this thing is too cool to not give it a top notch rating.

  • Afterburner Rating: 5 + out of 5 smiley faces
    :) :) :) :) :) +
    Availability: Good

Doc's Note: The appearance of the Golden Fingers connector on the Athlon PCB is clearly intended for Overclocking.  The question remains if it was done entirely for in-house use to determine how fast a CPU would run before soldering on the resistors, or was it actually intended for general use? My guess is that AMD did not need this connector for in-house testing, but I could be wrong.

<= BACK   => Home

Copyright 2000, KickAss Gear