Athlon XP 2600+ (333MHz) Overclocking December 09, 2002 Brandon Bell

Introduction

Page:: ( 1 / 11 )

When we published our 333MHz Athlon XP 2600+ review we were quite impressed with its performance, especially in relation to its predecessor. Despite running 50MHz slower than its 266MHz sibling (which operates at 2.13GHz), the 333MHz Athlon XP 2600+ finished roughly five percentage points faster than the 266MHz chip. This narrowed the gap between the 2600+ family and Athlon XP 2700+ considerably, the 333MHz CPU offered 97% of the performance of the Athlon XP 2700+ yet it officially costs 15% less. That made the chip a pretty good value in our eyes.

However, one topic we weren’t able to touch on in the original review was overclocking. During the overclocking phase of testing our reference motherboard died a premature death due to what appears to be a corrupt BIOS. With Comdex less than 36 hours away, we were left with no choice but to go with the numbers that we had. As impressive as they were, we still felt compelled to overclock our 333MHz Athlon XP 2600+ processor. Overclocking gives us an indication of how well the Thoroughbred B core the chip utilizes is scaling. Our 333MHz chip came off the line at AMD’s facility in Dresden, Germany just over two months after the initial Thoroughbred B CPUs we’d tested over the summer. We were curious to see if the newer chip may have had a few extra MHz lying under the hood. Plus, lets face it, overclocking is fun! With all that in mind, we hastily setup a new test system.

Overclocking results

For cooling, we chose our copper heatsink manufactured by Vantec. This bad boy does a great job of keeping our Athlon XP CPUs cool, but at the expense of our ears. Every time we crank our system up with the Vantec cooler we feel like we’re on the flight line at Nellis Air Force Base, home of the United States Air Force Thunderbirds (among other things). This cooler is seriously that loud! As a result, we only crank it out when it’s time to overclock our CPUs, and even then we’ll sometimes stick with AMD’s reference coolers.

We paired our Athlon XP 2600+ system up with Gigabyte’s KT400-based GA-7VAXP and a GeForce4 Ti 4600 for video. We’d wanted to use ATI’s RADEON 9700 PRO but until VIA addresses the sluggish OpenGL performance with the KT400/RADEON 9700 PRO combination at low resolutions (where testing CPU performance is most important) we’re sticking with the Ti 4600. VIA is hard at work on a 4-in-1 driver to address this issue, which will hopefully see the light of day sometime soon.

Based on our previous experience with Thoroughbred B chips, we decided to go ahead and crank the bus up to 175MHz immediately. We wanted to know just what our chip was capable of and didn’t feel 175MHz would be too much of a challenge for our processor. Our hunch proved correct, although we did have to crank the voltage up slightly to achieve complete stability.

We slowly marched on in 2MHz increments until we settled for a bus speed of 181MHz. With the chips’ 12.5x multiplier, this resulted in a final clock speed of 2.26GHz, just a bit shy of the 2.3GHz we’d hit with our 266MHz Athlon XP 2600+ and XP 2400+, but roughly the same speed as our Athlon XP 2700+. We needed 1.8V of juice to keep the processor stable in our entire suite of tests. We were actually able to run the chip at speeds in excess of 2.3GHz in many games, but to complete all our testing we had to settle for 181MHz.



SIDEBAR: AMD recently announced that Hammer should debut in February or March of next year

Test Systems

Page:: ( 2 / 11 )

System Setup

AMD Athlon XP 2700+
AMD Athlon XP 2600+ (266MHz FSB)
AMD Athlon XP 2600+ (333MHz FSB)
AMD Athlon XP 2600+ (333MHz FSB) @ 185MHz

Gigabyte GA-7VAXP (KT400)

256MB Mushkin PC3200 DDR SDRAM

NVIDIA GeForce4 Ti 4600
Detonator 40.72 WHQL

30GB IBM Deskstar DTLA 307030 ATA/100 Hard Drive

Windows XP Professional

DirectX 8.1

Benchmarks

3D Mark 2001 Second Edition Build 330 – 32-bit color
Quake III: Arena version 1.17
Serious Sam: The Second Encounter
Unreal Tournament 2003
Jedi Knight II
Business Winstone 2002 ver 1.0.1
Content Creation Winstone 2002
SiSoft Sandra 2003



SIDEBAR: All AMD CPUs produced are now using the Thoroughbred B core.

3D Mark 2001

Page:: ( 3 / 11 )

3DMark 2001 - DirectX 8

Notes

We tested the Athlon XP chips in comparison to Intel’s Pentium 4 line in our last review, so if you’re looking for those numbers, we suggest you go back there. What we’re trying to do now is show the difference the faster bus makes, as well as what kind of potential you can unlock with a 180MHz+ bus.

As you can see, the margin between the 266MHz and 333MHz Athlon XP 2600+ is 2% at 800x600x32 in 3DMark 2001 SE, which is hardly noticeable at all. Of course, this is more of a 3D benchmark than CPU test, so lets move on to Serious Sam SE.


SIDEBAR: Our 333MHz Athlon XP was produced nine weeks after our 266MHz chip.

Serious Sam 2

Page:: ( 4 / 11 )

Serious Sam 2 - OpenGL

Notes

The margin between the 266MHz Athlon XP 2600+ and 333MHz chip opens up slightly in Serious Sam, three percentage points separate the two processors at 800x600. This is a little slimmer than the margins we noticed with nForce2, suggesting that it’s truly the superior platform for 333MHz processors.



SIDEBAR: Wondering where the Athlon XP 2800+ is? Retail availability should begin to pick up next month.

Quake III

Page:: ( 5 / 11 )

Quake III - High Quality

Notes

Even in Quake 3 we’re seeing the same 3% gap between the 266MHz Athlon XP and the 333MHz variant, with nForce2 this number was 6%. We know what you’re thinking: “the video cards are different!” So we ran the numbers with the RADEON 9700 and KT400 – the margins didn’t change. Check out the conclusion of this article for those results.


SIDEBAR: Will a benchmark to finally kill Quake 3 please stand up?

Comanche 4

Page:: ( 6 / 11 )

Comanche 4 demo

Notes

Comanche 4 is a little more reliant on the CPU than the previous tests, so we were curious to see how the results played out. The margins actually closed slightly, as the 266MHz Athlon XP 2600+ now trails the 333MHz chip by 2% at 800x600.


SIDEBAR: As of right now, AMD has no plans to move the 333MHz bus to any of its CPUs below the 2600+ level.

Unreal Tournament 2003

Page:: ( 7 / 11 )

Unreal Tournament 2003 - flyby

Unreal Tournament 2003 Botmatch – DirectX 8

Notes

Like Comanche 4, the margins tighten up in Unreal Tournament 2003, the 5 and 6% margins we saw a few weeks ago with nForce2 are now gone. There’s practically no difference between either processor on the KT400 platform.



SIDEBAR: The name for desktop Hammer is now official: Athlon 64.

Jedi Knight II

Page:: ( 8 / 11 )

Jedi Knight II – High Quality

SIDEBAR: AMD recently announced its fourth quarter sales are expected to be 15% higher (35% quarter-over-quarter) than initially projected.

SiSoft Sandra 2003 Memory Bandwidth

Page:: ( 9 / 11 )

SiSoft Sandra 2003 Memory Bandwidth

SIDEBAR: We’re really looking forward to AMD’s Barton core. With 512K cache we can’t wait to see what it will do when paired with nForce2.

Office Productivity

Page:: ( 10 / 11 )

Business Winstone 2002

Content Creation Winstone 2002

SIDEBAR: A patch was recently issued for CC Winstone 2003. Hopefully it will resolve the issues we ran into earlier.

Conclusion

Page:: ( 11 / 11 )

In the process of running our simple little overclocking article, we not only discovered that the newer Athlon XP chips are scaling roughly the same as the first generation Thoroughbred B chips, we’ve also found that the KT400 chipset doesn’t scale as well as nForce2 with the new 333MHz Athlon bus. We always knew nForce2 was faster, now we also see it scales better as well. It will be interesting to see if the performance margins between both chipsets increase as clock speeds scale higher. In theory this is what should take place.

Of course, VIA is hard at work on its KT400A chipset. At Comdex we discovered that it will be a dual-channel implementation. When nForce was originally announced VIA criticized the chipset for its complexity, and even spread a rumor that it would require a six-layer motherboard design. Now they appear to be singing a different tune. The old saying “if you can’t beat him, join him” appears to apply. Here are those KT400/RADEON 9700 PRO numbers we promised earlier, as you can see the margin between the 266MHz and 333MHz CPUs remains the same 3% we saw with the Ti 4600. In comparison, we saw a 6% performance difference between both processors when paired with nForce2 and RADEON 9700 PRO.

As far as the Athlon XP 2600+ is concerned, we see no reason why you shouldn’t purchase the 333MHz chip if you have a motherboard based on the KT333 chipset or better. It costs the same as the 266MHz variant, both officially and among the street prices we’ve seen, and both processors are now readily available on the retail market.

If you’re also picking up a motherboard, we highly suggest you pair it with a motherboard based on the nForce2 chipset. As you’ve just seen, you’ll get superior performance out of that solution at 333MHz. Now if AMD would just release Barton, the fun would really begin on the Athlon platform!



SIDEBAR: What’s on your wish list this Christmas? nForce2? A 333MHz FSB Athlon processor? Speak!