34 & 35 & 36 & 37. PC mark 2004 is the next version to be tested and it includes a few GPU tests:

The scores of both processors are really close - only 4% apart - within the margin of error. Athlon XP memory score is better by 9%, so the extra cache helps here; however the graphics score of the Duron is 6% higher than the Athlon and that is repeatable! Something is really fishy here, let's see if the next version of PC mark has something to show us!

38 & 39 & 40 & 41. The final version of PC mark compatible with Windows XP is the 2005 one.

Things are looking really odd! Athlon XP manages an 8% better CPU score which is significant; however the real highlight is the memory: Thoroughbred is 58% better than Duron - a colossal advantage!
Again Duron scores far better in the graphics department: 15% better than Thoroughbred! I repeated the test on both CPU's many times and Thoroughbred graphics score is always worse and memory score is always far better!

42 & 43 & 44. We finally reached game testing and the first is Farcry 1 from 2004, a real gem! I used HardwareOC bench tool to provide accurate and repeatable testing conditions. For the Athlon XP I got a screenshot with the results, for the Duron here are the results as they were saved by the HardwareOC benchmark tool:

Far Cry Benchmark

The benchmark started at 11.03.2025 16:25:40

System Information
Operating system: Microsoft Windows XP
System memory: 2,0 GB
CPU: AMD Duron(tm) processor
CPU speed: 1800 MHz
Sound system: : NVIDIA(R) nForce(TM) Audio
VGA Information
Graphics card: NVIDIA GeForce 7600 GS
Memory: 256.0 MB
Current GPU speed: 375 MHz
Current memory speed: 1200 MHz
Driver version: 6.14.0010.9371 (English)

Resolution: 1024×768
Ultra quality option, Direct3D renderer
Level: Volcano, demo: hocvolcano.tmd
Pixel shader: default model
Antialising: 8×
Anisotropic filtering: 16×
HDR: disabled
Geometry Instancing: disabled
Normal-maps compression: disabled

Score = 44,44 FPS (Run 1)
Score = 43,58 FPS (Run 2)
Score = 44,09 FPS (Run 3)
Average score = 44,03 FPS

Resolution: 1280×1024
Ultra quality option, Direct3D renderer
Level: Volcano, demo: hocvolcano.tmd
Pixel shader: default model
Antialising: 8×
Anisotropic filtering: 16×
HDR: disabled
Geometry Instancing: disabled
Normal-maps compression: disabled

Score = 44,74 FPS (Run 1)
Score = 43,86 FPS (Run 2)
Score = 44,46 FPS (Run 3)
Average score = 44,35 FPS

Resolution: 1600×1200
Ultra quality option, Direct3D renderer
Level: Volcano, demo: hocvolcano.tmd
Pixel shader: default model
Antialising: 8×
Anisotropic filtering: 16×
HDR: disabled
Geometry Instancing: disabled
Normal-maps compression: disabled

Score = 43,80 FPS (Run 1)
Score = 42,56 FPS (Run 2)
Score = 43,33 FPS (Run 3)
Average score = 43,22 FPS

Copyright 2002 - 2006 Zoltan Nemeth - Roadside

Athlon XP 2200

Thoroughbred manages an overwhelming victory with a 17% lead at 1024*768 and 16% at 1280*1024!

45 & 46 & 47. The next game is the legendary Half life 2 - original version form 2004. Because Hardware OC benchmark tool refuses to run I used Fraps version 3.4.7 to test to test a 60 seconds sequence on the first level in front of the train station where Dr Breen is making his infamous speech. Fraps outputs the results to an Excel file and here they are:

Duron 1800

1024*768:
Min: 25
Max: 46
Avg: 35.367

1280*1024
Min: 25
Max: 42
Avg: 32.433

1600*1200
Min: 2
Max: 37
Avg: 27.883

Athlon XP 2200

1024*768:
Min: 24
Max: 60
Avg: 45.983

1280*1024
Min: 32
Max: 56
Avg: 43.383

1600*1200
Min: 30
Max: 54
Avg: 37.717

With a minimum 30% better fps Athlon XP destroys the Duron and demonstrates the importance of cache in Direct X 9 games! The game is not actually playable at maximum settings on the Duron because of the low minimums which makes it very difficult for the player in large firefights and is barely playable on the Athlon XP!

48 & 49 & 50. The last game is Doom 3 - a game which is known to perform better on CPU's with bigger L2 cache.
In order not to overwhelm the poor geforce 7800gs AA is disabled, the rest of the settings are at ultra/maximum.

Duron 1800 at 1024*768, 1280*1024, 1600*1200

Athlon XP 2200 at the same resolutions:

The gap is very large: Thoroughbred has a minimum of 17% advantage over the Duron and scores one more victory !

51. The second to last test is Super PI, heavily dependent on the frequency:
Duron 1800 first
Athlon XP 2200 second

Thoroughbred has a 4.62% lead, within the margin of error - therefore we can not declare a winner here!

52. We reached the final test: Cinebench 2003, again dependent on the frequency first and foremost:
Duron first
Athlon XP second

No winner here as Thoroughbred is only 3.36% faster!

All the tests between Duron 1800 and Thoroughbred 2200 are done! While it was obvious from the very beginning that a CPU with 256kb of L2 cache is faster than one with 64kb of L2 cache, the question is by how much?
10.61% is the advantage that the Athlon XP has over the Duron! This surely can be felt in benchmarks and games; the performance gain is noticeable and cannot be ignored!
So, taking into account the current prices on Ebay, it is far better to opt for a CPU with 256kb of L2 cache.

Now, about some things that can not be observed in the tests.
Duron has a 57W TDP and is very easy to cool and, as I already mentioned, can be used in a system without the auxiliary P4 connector and a modern PSU.
With the same cooler I managed to obtain a near silent system using by same settings in the BIOS with the RPM varying depending on the temperature. The CPU temperature never attained 50 degrees C and the only fans that I could hear were the chipset fan and the geforce 7800gs fan. With the Athlon XP I could clearly hear the CPU fan during tests.
So, should you buy a Duron 1800 or any any model in the lineup?
The answer is NO!
Despite all the advantages my recommendation is the following:
I have a system that meets the criteria below:
- I have a modern PSU with a weak 5V rail
- I have a motherboard without the auxiliary power connector
- I have a motherboard that supports only 266MHz FSB
- I have a motherboard that does not support CPU's with a multiplier of 13 of greater
Buy Athlon XP Thoroughbred 2000. It will work just fine in any socket 462 board that support 266 MHz FSB, does not requires a period correct PSU, it will be faster than any Duron and is extremely cheap!

The only review with tests of the Duron applebred on the net is finished! The processor has potential but is overshadowed by its faster cousins and the current prices on Ebay do not do it any favors!
The last thing I should mention is something that I never did until now: overclocking potential!
As a simple test I set the FSB to 160 in the BIOS and using default voltages it booted into Windows XP from the first try! I did a 3dmark 2001 - and forgot to take the screenshot! - the score increased beyond the score of the Athlon XP 2200!
Not wanting to damage one of my precious Abit AN7 I set the FSB back to 133.
So the 64kb versus 256kb tests are done, it is now time for the final tests!

Final battle on socket A: Barton versus Thoroughbred - 512kb versus 256kb of Level 2 cache.
After testing all CPU generations, all FSB combinations only one more thing remains to be tested in the last comparison for a legendary processor: Athlon XP and socket 462.
Barton represents the final iteration of the K7 family; meanwhile Thoroughbred represents the most numerous processors launched on socket A.

Finding 2 processors with the same frequency, same multiplier, same FSB but with different amount of cache could not be easier!
There are 4 models that correspond to the criteria presented above: both Athlon XP 2700 part number AXDA2700DKV3D and Athlon XP 3000 part number AXDA3000DKV4D have 2167 MHz frequency, 333 MHz FSB, 13X multiplier; the only difference is the amount of cache: 256kb the first one and 512kb the second. I have the Athlon XP 3000 part number AXDA3000DKV4D but not the Athlon XP 2700; therefore this variant will have to wait until I source a cheap 2700.
However on the Sempron line of CPU's for socket 462 we have 2 models that are nearly identical: Sempron 2800 based on the Thoroughbred core has 2000 MHZ, 333 MHZ FSB, 12X multiplier and only 62W TDP; Sempron 3000 based on the Barton core has the exact same specs, including TDP, but comes with 512kb of cache.
I chose to test the 2 Semprons because I bought the 2400, 2800 and 3000 models for 20 euros including shipping on Ebay which is pretty cheap.