64 vs 128 Bucket Chia Plotting

[Soviet.Tech]

Can I plot more faster with 64 buckets and 8192MB RAM than 128 buckets and 4096MB RAM ?

[WolfGT]

64 buckets vs 128 buckets

Well, have ended testing. Last test was to mix 10 plots between 32B, 64B, and 128B. I gave them all 14GB memory, all 4T so there was no QS for memory reasons. All running together in the mix, they ALL ran nearly consistent, similar times. So as far as I can tell, less buckets buys u nothing, and uses considerably more memory/plot. Fewer buckets may take less from ur ssd, but not enough to matter to plot times it seems. So 4GB, 3T, 128B is what I’m using now. Of course, YRMV.

[naq1]

I can also confirm there is very little benefit going from 128 to 64 at least. (on my 2 cpu xeon)

I ran 10 plots with u64 and b8000 then 10 plots with u128 and b4000.
4 threads each
Staggered like this:

Plot:	1	2	3	4	5	6	7	8	9	10
PID	2308	5184	5976	6336	17568	12192	2676	2456	10948	17604
Buckets	64	64	64	64	64	64	64	64	64	64
Ram	8000	8000	8000	8000	8000	8000	8000	8000	8000	8000
Temp Drive	F	F	F	L	L	L	L	L	L	L
Dest Drive	X	Y	X	Y	X	Y	X	Y	X	Y
Start Delay	0	900	1800	2700	3600	4500	5400	6300	7200	8100

F drive is 1TB nvme on numa 0 and L is 2TB nvme on numa 1 so thread affinity was also set and each CPU has 64GB RAM on its numa node.
Dest drives X and Y are slow hdds.

So u64 showed a 2% improvement on phase 1 and 4% improvement on phases 3 and 4.
This is so small it may be due to external factors like trim or copy clashing to the slow drives, certainly not worth the extra ram.

Maybe it would be a different picture if using multiple parallel hdd as temp…

[NKhet]

I have a 2X e5 2697V2 with 12 cores 24 threads (total 24/48) and at 8.5hrs per plot on parallel load. I have 128gb ram so using bucket of 64, with 2 firecuda 2tb nvme. I went for 12+ hr per plot to 8.5hr with just changing the bucket size. This is with 6832 ram buffer, 12 max concurrent, 6 phase 1 total and 6 threads.