Navigate: Advice Beginners BIOS Guide CPUs Links Mainboards Memory Network Storage Video/Sound Cards Contact Forum SiteMap Sponsors WebNews Home	.	.
Prices: Mainboards ABIT ASUS Chaintech Shuttle Soyo Tyan CPU Intel P4 2.4C-800 P4 2.6C-800 P4 2.8C-800 P4 3.0-800 P4 3.2-800 AMD AthlonXP XP 1700+ XP 2000+ XP 2400+ XP 2500+ XP 2700+ XP 3000+ XP 3200+ Athlon64 Athlon64 3200+ Athlon64 FX-51 Opteron Opteron 240 Opteron 242 Opteron 244 Opteron 246 Memory Corsair Crucial Kingston Mushkin OCZ Search Prices:

LOSTCIRCUITS	SHORTCUTS: Dual Core vs HyperThreading Tech Specs Test Setup Memory Performance WorldBench5 3dsmax CineBench 2003 3D Gaming Power/Temperature Coefficient Power Hogs 3D Rendering Power DOOM3 3DMark2001SE 3DMark'05 Final Thoughts Give Us Some Feedback on this Review
AMD Athlon64 X2 "Toledo" Dual Core on Single Die
(Review by MS May 9, 2005)

Memory Benchmarks

The Venice core scores slightly lower in memory benchmarks such as SiSoft Sandra and Everest than the ClawHammer or NewCastle cores. The difference is in the order of 5%, meaning that on average we found approximately 5770 MB/sec in dual channel mode (2:2:2:8, 1T CMD_Rate) as opposed to some 6050 MB/sec with the Clawhammer and Newcastle cores. The X2 core takes another minor dip in memory bandwidth to approximately 5730 MB/sec. In view of the fact that the memory bandwidth is measured as transfers to the ALU and FPU units, respectively, the performance hit for a dual core is not surprising at all since internal management of the data has to be factored in as well. Since the memory controller is part of the CPU, we were of course, interested in latency differences on the controller level -- Cachemem 2.65 appears to be the right tool to look into this issue.

Cachemem 2.65

Composite graph showing superimposed memory latency plots as function of stride length and transfer block size for the Athlon 64 4000+ (transparent blocks) and the Athlon 64 X2 4800+ (solid blocks). In terms of memory access latencies, the two processors are virtually identical with the exception of a few "runaway" data points.

Since the two graphs shown above are extremely close, we subtracted the two matrices from each other and plotted the differentials. As the graph very convincingly shows, the majority of data points are zero values, meaning that there is no difference between the two cores, the rest of the data are isolated deviations that are consistent with fluctuations between runs to either the positive or the negative. Note that the highest deviation is 6.3 ns

next page: => WorldBench5 =>