PowerPC Setup Benchmarks (3)

 
Sven Negrassus - 27. März 1999
 

Prozessor MHz Bus Speed Cache Size Cache Speed

Cache Size
Technische Daten
Konfiguration: Prozessor MHz L2-Cache L2-MHz Bus MHz
joeCARD 300/150/50/1MB  300.0 100%  1024k  150.0 2:1  50.0 100%
joeCARD 300/150/50/512k  300.0 100%  512k  150.0 2:1  50.0 100%
joeCARD 300/---/50/0k  300.0 100% 0k ---  50.0 100%
Benchmarks
Macbench 5.0 CPU FPU
joeCARD 1024k  987 100.0%  1002 100.0%
joeCARD 512k  871   88.2%  996   99.4%
joeCARD 0k  364   36.8%  815   81.3%
Real World Tests NeuroNet POVray
joeCARD 1024k  356 sec. 100.0%  245 sec. 100.0%
joeCARD 512k  360 sec.   98.9%  263 sec.   93.2%
joeCARD 0k  387 sec.   92.0%  552 sec.   44.4%
Real World Tests Cinema 4DXL Cinebench SP Index
joeCARD 1024k  691 sec. 100.0%  4.07 100.0%
joeCARD 512k  697 sec.   99.1%  3.96   97.3%
joeCARD 0k  804 sec.   85.9%  2.52   61.9%
Real World Tests PhotoShop Actions (38MB Image) Norton Gfx Bench @1024x768x16bit
joeCARD 1024k  223 sec. 100.0%  217 100.0%
joeCARD 512k  231 sec.   96.5%  213   98.2%
joeCARD 0k  269 sec.   82.9%  180   82.9%
Real World Tests VirtualPC 2.1 / Winbench 97 32 Bit (W95) VirtualPC 16 Bit
joeCARD 1024k  126 100.0%  85.9 100.0%
joeCARD 512k Memory Fault Memory Fault
joeCARD 0k Memory Fault Memory Fault
 
 Memory Throughput
Memory Throughput
 
Fazit: 512k Backside-Cache führen zu einem relativ geringen Performancenachteil gegenüber 1MB Cache (je nach Anwendung 1-7% Einbusse in unseren Beispielen). Ganz ohne Cache verliert der Prozessor aber deutlich an Biss (8-55% Einbusse gegenüber 1MB).
Der CPU-Test von Macbench 5.0 greift hier etwas sehr hoch, 12% (512k) oder gar 63% (kein Cache) Leistungseinbusse konnten wir nicht feststellen.
Round up: 512k of backside cache do not show a very clear performance disadvantage over 1MB of cache (1 to 7% in our examples). Running with no cache the G3 loses its power (8 to 55% slower compared to 1MB).
The CPU test of Macbench 5.0 shows very low performance when running with 512k (-12%) or with no cache at all (-63%) - we did not see that bad performance in our tested applications.
 
Next: Cache Speed * Cache Speed


Letzte Änderung * Last Update:
%LastFileUpdateDate%