3 ms·
There has to be a memory bandwidth limit where the number of cores is rendered moot for quad-core arm7 to be slower than dual-core arm7. Either that or the tes
by ck2 14y ago
There has to be a memory bandwidth limit where the number of cores is rendered moot for quad-core arm7 to be slower than dual-core arm7.
Either that or the test is somehow being fooled, I noticed that benchmark code is updated frequently.
- wmf 14y agoTegra is known for low memory bandwidth.
- zurn 14y agoGoogling yields a 6.4 GB/s figure for the single channel LPDDR3 supported by Tegra 3. It doesn't sound like the most likely bottleneck for a CPU benchmark given that those A9 cores plod along at about 1/10 the perf of a desktop chip. What sort of bandwidth do other mobile chips provide? edit: Looks more like 1/3 than 1/10, looking at the tests in geekbenchscores. I had the 1/10 figure from looking at Sunspider results (http://forum.beyond3d.com/showpost.php?s=502c8933367f315a17f882b45e70fb17&p=1630202&postcount=16 http://forum.beyond3d.com/showpost.php?s=502c8933367f315a17f...).
- wmf 14y agoThat's still less bandwidth than any other recent SoC and keep in mind that the GPU is also sharing that bandwidth.
- jfpoole 14y agoWhile Geekbench is updated frequently, the actual benchmark code is virtually unchanged since 2.0.0 was released.