If the only thing that would be sent over usb is the 80 byte header and the nonces that give a valit share at difficulty 0, you need 80 bytes to the device and 4 bytes from the device on average per share.
The miximum speed for usb 2.0 is somewhere around 35 MB/s so that gives 35M/80 = 437,500 shares/s at difficulty 0.
437,500 shares at difficulty 0 is 437,500 * 2^32 = 1879 THash/s.
The protocol to talk to the asic miner will probably cost a bit more data, but you see, usb isn't holding up anything.
Topic: Is the CPU still not that important when mining with ASIC devices? (Read 1074 times)
October 11, 2012, 08:08:14 PM
CPU's are designed for multitasking, processing user input, streaming multimedia, encoding, virtualization, etcetera & are very generalized devices intended to offer wholesale performance in a wide variety of areas other than mining.
ASICs are specifically designed to achieve high performance generating hashes and nothing else.
The workload differences between a CPU and ASIC vary widely enough that there are performance trade offs involved in designing something to perform well at one task or another.
An ASIC would be horrible at carrying out the functions of a CPU & the opposite also applies.
ASICs are specifically designed to achieve high performance generating hashes and nothing else.
The workload differences between a CPU and ASIC vary widely enough that there are performance trade offs involved in designing something to perform well at one task or another.
An ASIC would be horrible at carrying out the functions of a CPU & the opposite also applies.
October 11, 2012, 04:30:54 PM
The whole point of an ASIC (application-specific integrated circuit) is to remove a specific workload from the CPU and run it on a dedicated processor. For example, in a 3PAR storage array, the ASICs in the storage controller nodes (2 per node, minimum 2 nodes per system) handles all of the XOR calculations for RAID, in-line zero detection/stripping, thin provisioning calculations, and a bunch of other tasks. The CPUs in the controller nodes (2 x quadcore intels) are usually running at close to 0% usage. The only spikes are when an analysis run is triggered to determine movement of data between storage tiers (SATA, SSD, FC disks). Without the ASICs, we'd need a LOT more memory, system cache and data cache, and the CPUs would probably be pegged at 100% all the time, and array performance would start to suck balls. ASICs rock. :-)
Sorry for the 3PAR marketing lesson. I work for HP (who own 3PAR), and specialize in enterprise storage arrays, so it's kinda what I do all day.
Now if i could jsut figure out how to leverage the equipment in my demo lab for mining...
Sorry for the 3PAR marketing lesson. I work for HP (who own 3PAR), and specialize in enterprise storage arrays, so it's kinda what I do all day.
Now if i could jsut figure out how to leverage the equipment in my demo lab for mining...Good lesson, thanks.
October 11, 2012, 08:13:47 AM
The whole point of an ASIC (application-specific integrated circuit) is to remove a specific workload from the CPU and run it on a dedicated processor. For example, in a 3PAR storage array, the ASICs in the storage controller nodes (2 per node, minimum 2 nodes per system) handles all of the XOR calculations for RAID, in-line zero detection/stripping, thin provisioning calculations, and a bunch of other tasks. The CPUs in the controller nodes (2 x quadcore intels) are usually running at close to 0% usage. The only spikes are when an analysis run is triggered to determine movement of data between storage tiers (SATA, SSD, FC disks). Without the ASICs, we'd need a LOT more memory, system cache and data cache, and the CPUs would probably be pegged at 100% all the time, and array performance would start to suck balls. ASICs rock. :-)
Sorry for the 3PAR marketing lesson. I work for HP (who own 3PAR), and specialize in enterprise storage arrays, so it's kinda what I do all day. Now if i could jsut figure out how to leverage the equipment in my demo lab for mining...
Sorry for the 3PAR marketing lesson. I work for HP (who own 3PAR), and specialize in enterprise storage arrays, so it's kinda what I do all day.
Now if i could jsut figure out how to leverage the equipment in my demo lab for mining... October 11, 2012, 05:34:23 AM
Thanks for the answering guys.
October 11, 2012, 03:03:10 AM
USB3 is low market ...
USB2 will work fine - if you need more bandwidth then go to ethernet
USB2 will work fine - if you need more bandwidth then go to ethernet
October 11, 2012, 02:28:29 AM
Then why didn't they develop it for USB3 if it is bottlenecked at USB2. This then would have removed the errors with the cpu?
October 10, 2012, 07:06:15 PM
Will this still be the case when ASIC devices are introduced? How about mining with more than one of these devices, says 3 BFL single SCs on a single PC, will that tax a CPU?
while cpus do have a limit, the bottleneck when found won't be the cpu. flooding of usb 2.0 would probably happen well before any cpu limits arrive. i've not looked but the bfl threads must already have discussed this somewhere.
October 10, 2012, 06:43:25 PM
Hey guys,
I've always read that the CPU is not really important when mining. People using GPUs and FPGAs seem to mine fine with atom chips, single core amd semprons and even raspberry pis.
Will this still be the case when ASIC devices are introduced? How about mining with more than one of these devices, says 3 BFL single SCs on a single PC, will that tax a CPU?
Thanks for any help.
I've always read that the CPU is not really important when mining. People using GPUs and FPGAs seem to mine fine with atom chips, single core amd semprons and even raspberry pis.
Will this still be the case when ASIC devices are introduced? How about mining with more than one of these devices, says 3 BFL single SCs on a single PC, will that tax a CPU?
Thanks for any help.
Jump to: