4 ms·
This is a use case where persistent memory really shines. Its about 10x slower but you can get 1.5TB or it for less than $10,000. You don’t have to use the pers
by jdsully 7y ago
This is a use case where persistent memory really shines. Its about 10x slower but you can get 1.5TB or it for less than $10,000. You don’t have to use the persistence - you can treat it as RAM.
- basch 7y agowith VROC or the Threadripper equivalent NVME CPU RAID, you can get 7Gbps read speed from a RAID5. Thats DDR2 speed. Unless you truly need DDR4 speeds, I would think you would be better off for your money using CPU Raid and NVME. More drives, more speed. One guy used 8 drives and got 28Gbps. https://i.imgur.com/AuCWq3r.jpg https://i.imgur.com/AuCWq3r.jpg
- termie 7y agoThe throughput is amazing. The screenshot shows 28,375 MB/s — the capital B typically indicates Bytes whereas a lowercase b indicates bits, so your Gb/s translation is a little confusing. The latency is still much higher than main memory or Optane, but that might be fine for many applications.
- basch 7y agoyoure right, I didnt look closely at the buttons, I just assumed bits based on the all caps used all over the interface.
- jdsully 7y agoLinear throughput is the wrong way to think about this. Random access is much more important for DRAM workloads. Persistent memory has a block size of 4 cache lines (256 bytes) vs 4KB for an SSD. This drastically cuts down on read/write amplification.
- AstralStorm 7y agoGenerally burst access and not random is important. A few MB at a time.