My understanding of The Wikipedia for duplex web page And of StackexChange’s responses as this is that in a contemporary ethhernet connection the convey Visitors is meant to be affected by the obtain visitors. However I’m executing some exams between two trendy Mac computer systems and I’m seeing outcomes that make me suppose that there’s something essential that I’m lacking.
The check platform is 2 Mac Studio M2 extremely bins, one with V13.7.4 and one other with Sonoma 14.7.4
To scale back complexity, I’ve linked them straight to one another with a brief copper ethernet cable between its integrated NICs. Once I let the pace that claims 10gbaset mechanically negotiate as anticipated, and for the check I configured it in guide and 1000Baset
-
In 10GBaset, utilizing a replica of the MAC A search engine, copying from the native unit in Mac A to the native unit in Mac B, I get ~ 1200 megabytes/sec as anticipated (solely transmitted)
-
In 10gbaset, utilizing a replica of Finder in Mac A, copying from the native unit in Mac B to the native unit in Mac A, I receive ~ 1200 megabytes/sec as anticipated (simply receiving)
-
In 10GBaset, utilizing a replica of the MAC A search engine, copying from the native unit in Mac B to a different a part of the native unit in Mac B, I receive ~ 650 megabytes/sec (simultaneous transmission/reception, somewhat extra medium pace)
(I obtain related outcomes working two copies on the identical time, one transmitted and one other reception) -
A 1000Baset, utilizing a replica of the MAC A search engine, copying from the native unit in Mac A to the native unit in Mac B, I get ~ 120 megabytes/sec as anticipated (solely transmitted)
-
A 1000Baset, utilizing a replica of Finder in Mac A, copying from the native unit in Mac B to the native unit in Mac A, I get ~ 120 megabytes/sec as anticipated (simply receiving)
-
A 1000Baset, utilizing a replica of the Mac A search engine, copying from the native unit in Mac B to a different a part of the native unit in Mac B, I receive ~ 65 megabytes/sec (simultaneous transmission/reception, once more somewhat extra medium pace)
(Once more, related outcomes that execute two copies on the identical time, one transmitted and one other reception)
Is left over? It isn’t simply fixing, as a result of terminal instructions to repeat do the identical. It isn’t {hardware} of third events of shit, I’m utilizing the constructed -in Ethernet. Even when there have been any {hardware} limitation within the NIC, or an issue with the storage pace, could not it at the least get a full pace of 1 GBE sending/receiving in 1GBE mode?
The rationale we did the check is as a result of we’re seeing the same efficiency utilizing switches and connecting to an actual file server, the check gear was solely to eradicate the variables. Self-negotiation to a single 1GBE change between the 2 check footage didn’t resolve it, and I attempted NFS, which didn’t appear to make a lot distinction. Jumbo Frames (Mtu 9000) helped somewhat, reached roughly 85 MB/sec, however didn’t attain the identical pace because the unidirectional copies, and we can not use Jumbo frames in manufacturing sadly.
