VOGONS


Socket A: Nvidia vs Via - battle of the platforms!

Topic actions

Reply 900 of 1055, by nd22

User metadata
Rank Oldbie
Rank
Oldbie

16. The next test is carried out at 1280*1024 32, bit color depth, 32 bit textures.
Sempron 2800
Sempron 3000

Reply 901 of 1055, by nd22

User metadata
Rank Oldbie
Rank
Oldbie

17. And finally the 4k of the retro universe: 1600*1200!
Sempron 2800
Sempron 3000

Reply 902 of 1055, by nd22

User metadata
Rank Oldbie
Rank
Oldbie

Finally Barton manages a clear victory! Being 8-12% (depending on the resolution) faster than Thoroughbred makes it the undisputed leader in 3dmark 2000!

Reply 903 of 1055, by nd22

User metadata
Rank Oldbie
Rank
Oldbie

18. 3dmark 2001, the de facto standard for testing socket A is next. We start with 1024*768 default settings.
Sempron 2800
Sempron 3000

Reply 904 of 1055, by nd22

User metadata
Rank Oldbie
Rank
Oldbie

19. We increase the resolution to 1280*1024 and set everything to max!
Sempron 2800
Sempron 3000

Reply 905 of 1055, by nd22

User metadata
Rank Oldbie
Rank
Oldbie

20. At 1600*1200 Barton still manages to outscore Thoroughbred by 6%!

Reply 906 of 1055, by nd22

User metadata
Rank Oldbie
Rank
Oldbie

Barton wins again! So cache size does matter! At default settings the 10.84% lead over Thoroughbred is significant and can not be dismissed.

Reply 907 of 1055, by nd22

User metadata
Rank Oldbie
Rank
Oldbie

21. Futuremark's 3dmark 2003 is next. A directx 8 & 9 benchmark it is the only 3dmark version to include 3 sound tests.
1024*768 default settings
Sempron 2800
Sempron 3000

Reply 908 of 1055, by nd22

User metadata
Rank Oldbie
Rank
Oldbie

22. 1280*1024 max settings
Sempron 2800
Sempron 3000

Reply 909 of 1055, by nd22

User metadata
Rank Oldbie
Rank
Oldbie

23. 1600*1200 also max settings
Sempron 2800
Sempron 3000

Reply 910 of 1055, by nd22

User metadata
Rank Oldbie
Rank
Oldbie

The differences between the processor with 256kb of cache and the one with 512kb varies between 0.8 to 3.8%. Practically the results are too close to call! Here the extra cache makes no difference!
24 & 25 & 26. In the 3 tests with sound the differences are equally minuscule!
Barton has no advantage over Thoroughbred in 3dmark 2003.

Reply 911 of 1055, by nd22

User metadata
Rank Oldbie
Rank
Oldbie

27. The final 3dmark version, 2005, is extremely GPU-dependent.
Here are the scores for the Sempron 2800 at 1024*768, 1280*1024 and 1600*1200.

Reply 912 of 1055, by nd22

User metadata
Rank Oldbie
Rank
Oldbie

And the results for Sempron 3000 at the same resolutions and settings:

Reply 913 of 1055, by nd22

User metadata
Rank Oldbie
Rank
Oldbie

There is nothing to see here so please move right along onto the next benchmark!

Reply 914 of 1055, by nd22

User metadata
Rank Oldbie
Rank
Oldbie

30. The one and only GUI test is Tom2d. Here, again, the extra cache does not make a difference as there is only a tiny 2.5% better result for the Sempron 3000.

Reply 915 of 1055, by nd22

User metadata
Rank Oldbie
Rank
Oldbie

31 & 32 & 33. We test now general system performance with PC mark. The 2002 version is first.
Sempron 2800
Sempron 3000

Reply 916 of 1055, by nd22

User metadata
Rank Oldbie
Rank
Oldbie

Both Sempron 2800 and 3000 score very close in the CPU department; the real highlight is the memory score: 13% better on the processor with 512kb of cache! So the extra 256kb of cache of the Barton helps!

Reply 917 of 1055, by nd22

User metadata
Rank Oldbie
Rank
Oldbie

34 & 35 & 36 & 37. PC mark 2004 is next. This benchmark integrates graphics tests with a corresponding result.
Sempron 2800
Sempron 3000

Reply 918 of 1055, by nd22

User metadata
Rank Oldbie
Rank
Oldbie

All results from the 2 processors are within 2% of each other. There is no winner here and again the extra cache fails to have an impact!

Reply 919 of 1055, by nd22

User metadata
Rank Oldbie
Rank
Oldbie

38 & 39 & 40 & 41. The last version of PC mark to be tested is the 2005 one. It is the last PCmark compatible with Windows XP.
Sempron 2800
Sempron 3000

Last edited by nd22 on 2025-05-24, 06:54. Edited 1 time in total.