Read time for a 2.7 TB plot
-
Hi,
The current read time for a plot with 2.72 TB is ~50 Seconds. Is this acceptable or this is on higher side?
Thanks in advance
-
less than 1 minute time is good i guess. Try Open CL option, it will utilize your internal GPU.
-
This is quite slow reading (USB2?), you can check you general drive speed with e.g. CrystalDiskMark:
http://crystalmark.info/software/CrystalDiskMark/index-e.htmlIf that's @100MB/s or something on sequential reading ... you should maybe think about optimizing you plotfile with:
https://forums.burst-team.us/topic/26/plot-optimizer-v1-6-with-gui-1-0-3-j6jq-win/3
but you would need enough free space to do that (will temp need twice the space of the plotfile)
On creating new plots, you could use https://forums.burst-team.us/topic/58/gpu-plot-generator-v4-0-3-win-linux/3 in 'direct' mode.
-
Its actually a SATA 6.0Gbps drive running on a SATA 3.0 Gbps interface, I will post the benchmark results here
-
These are the results, sequential read is less than 100 MB/s :(
Just read few forums where they say a full drive may perform slowly, my drive is more than 95% full with plots...
-
Just an update, drive reads the 2.7 GB plot in ~26 seconds now after a system restart..
-
Well, not the fastest drive but ok, you need to read 0.025% of plotfile on every round (4096 scoops, One Scoop needs to be read per block) that is in your case ~670MB per round, divided by 26sec. we are at ~25MB/s.
Like mentioned above, you could try optimizing ...
Also check: https://bchain.info/BURST/tools/overlap it will show you needed drive seeks, very useful.
After set stagger in there equal to number of nonces, and you will get an idea what improvements optimizing can be.e.g.
not optimized
324324234324324_0_1000000_100 -> 10000 disk seeks per Scoop (80 seconds @ 8ms)
optimized
324324234324324_0_1000000_1000000 -> 1 disk seeks per Scoop (0 seconds @ 8ms)
