Despite the $4,000 price disparity, it is imperative to highlight the significant difference in workload throughput between these two M5 Ultra Studio configurations. This comparison will elucidate the trade-offs associated with choosing 96GB versus 256GB, and determine whether the additional $4000 is truly justified.
A workload that oversubscribes 96GB will suffer from the extra I/O wait times caused by page-in/page-out (swapping) activity.
Apple claims the M5 Ultra's internal SSD is twice as fast as for the M3 Ultra's SSD. I have the M3 Ultra with 2TB SSD and can say the M5 Ultra's internal SSD data rates should be impressive at ~14 GB/s for reads, and ~20 GB/s.
Example...
For a 146GB workload running on an M5 Ultra's 96GB memory system then 50GB of its memory will need to be swapped in & out at times. Thus swapping out and swapping in 50GB on the M5 Ultra is likely to take ~(50/20 + 50/14) seconds or some (2.5+3.6) or some 6 seconds.
So would the accumulation of these 6 sec swapping times during the lifetime of the 146GB workload be significant ?
These swapping times maybe reduced as some of the workload's memory (maybe some of the 50GB as well) could be compressed.
Regardless, the impressive speed of the M5 Ultra’s internal SSD should alleviate some users’ concerns regarding insufficient memory capacity for their M5 Ultra, and the unfortunate surcharge of $4,000 for selecting 256GB.
Certainly, if you have the financial means, a user could purchase both the 96GB and 256GB systems and utilize Apple’s 14-day return policy to conduct a side-by-side comparison. This approach would enable user to determine whether the $4,000 investment is justified.
I can already imagine that some may contend that oversubscribing memory will impose wear and tear on the internal SSD and consequently affect its lifespan!
A workload that oversubscribes 96GB will suffer from the extra I/O wait times caused by page-in/page-out (swapping) activity.
Apple claims the M5 Ultra's internal SSD is twice as fast as for the M3 Ultra's SSD. I have the M3 Ultra with 2TB SSD and can say the M5 Ultra's internal SSD data rates should be impressive at ~14 GB/s for reads, and ~20 GB/s.
Example...
For a 146GB workload running on an M5 Ultra's 96GB memory system then 50GB of its memory will need to be swapped in & out at times. Thus swapping out and swapping in 50GB on the M5 Ultra is likely to take ~(50/20 + 50/14) seconds or some (2.5+3.6) or some 6 seconds.
So would the accumulation of these 6 sec swapping times during the lifetime of the 146GB workload be significant ?
These swapping times maybe reduced as some of the workload's memory (maybe some of the 50GB as well) could be compressed.
Regardless, the impressive speed of the M5 Ultra’s internal SSD should alleviate some users’ concerns regarding insufficient memory capacity for their M5 Ultra, and the unfortunate surcharge of $4,000 for selecting 256GB.
Certainly, if you have the financial means, a user could purchase both the 96GB and 256GB systems and utilize Apple’s 14-day return policy to conduct a side-by-side comparison. This approach would enable user to determine whether the $4,000 investment is justified.
I can already imagine that some may contend that oversubscribing memory will impose wear and tear on the internal SSD and consequently affect its lifespan!
Last edited: