Yes, it was very disappointing. I was hoping the 8 cores would give a speedup for compiling code, but no dice. On a large Linux build, the BPI-F3 with make -j8 takes exactly the same time as a make -j4 on the VisionFive 2.
Nothing except the Pioneer (and now this and the Milk-v P550) has L3 cache.
BPI-F3 (etc) has only 512 MB of L2 cache for each cluster of 4 cores, vs 2 MB on the VisionFive 2 (etc), and I think this is the biggest issue. Similarly the TH1520 only has 1 MB of L2.
The Pioneer has the same 1 MB of L2 per four cores, but backs this up with 4 MB of L3, and when you're running single-threaded (or just a few) that CPU also has access to the other 60 MB of L3 cache on the chip at still much better than DRAM latency and bandwidth. It's very nice-perofrming machine if you can afford it -- though the P550 should make a better desktop machine. I'll know in a week or two when my Megrez arrives...
BPI-F3 (etc) has only 512 MB of L2 cache for each cluster of 4 cores, vs 2 MB on the VisionFive 2 (etc), and I think this is the biggest issue. Similarly the TH1520 only has 1 MB of L2.
The Pioneer has the same 1 MB of L2 per four cores, but backs this up with 4 MB of L3, and when you're running single-threaded (or just a few) that CPU also has access to the other 60 MB of L3 cache on the chip at still much better than DRAM latency and bandwidth. It's very nice-perofrming machine if you can afford it -- though the P550 should make a better desktop machine. I'll know in a week or two when my Megrez arrives...