Hello,
can you pls help me with my config to use the Intel Optane Memory effective?
I have a Intel Optane Memory 32GB installed and tried a few configs with benchmarks, but it's always slower than my ram or PrimoCache does not use the Optane Memory.
After the L2 format it is 27,25 GB for using, I tried with the full size and a smaller size of it. I let the Optane 100% Read and the ram 100% Write + reverse that, I share both 50/50 Read/Write, but didn't find any good setting. Tried the default gather interval and fast/normal... is the Optane now more a read or a write cache? or both?
Cache Hit Rate is most of the time 99,99% at all HDDs/SSDs, PC freezes sometimes (seams because of the high workload of the drives)
The best benchmark I get without the L2 at the moment, something seams wrong on my config.
L1 is at 12288 MB (32GB) Shared
Block Size 32 KB (smaller does not speed up anymore, it only takes more ram with overhead)
Defer-Write 30 Intelligent + Free Cache on Written
Intel Optane Memory
Re: Intel Optane Memory
The speed of Optane Memory is lower than RAM, so it's not weird that you see it's slower than your RAM.
If you have heavy read and write working load, you may use RAM 100% write cache with defer-write, and optane (L2) cache 100% read for HDD.
And you may also reduce defer-write latency to 10s if you have heavy writing tasks.
Re: Intel Optane Memory
Did I need to install the Optane Software to run the M.2 correct? I have the RST installed version 17.5.0.1017 and the Intel Optane runs with that driver 17.5.0.1017 as Intel(R) NVMe Controller. My mainboard Asus Prime Z270-A can it only run at 2x not 4x.
SmartHDD say the Optane runs:
Performance:
Maximum speed interface - 16,7 MB/s (Ultra DMA - 0)
Current speed interface - 16,7 MB/s (Ultra DMA - 0)
Buffered read speed - 25000 MB/s
Average read speed - 23723 MB/s
Stable read speed - 24038 MB/s
Instability of reading - 1%
This is way more than I get with caching the DDR4 ram... but with AS SSD Benchmark I can see that the Optane is slower.
I mean my main question is, what to do with the Optane Memory? Is it useless? Should I use the Optane for all HDDs and the Ram for the SSDs?
SmartHDD say the Optane runs:
Performance:
Maximum speed interface - 16,7 MB/s (Ultra DMA - 0)
Current speed interface - 16,7 MB/s (Ultra DMA - 0)
Buffered read speed - 25000 MB/s
Average read speed - 23723 MB/s
Stable read speed - 24038 MB/s
Instability of reading - 1%
This is way more than I get with caching the DDR4 ram... but with AS SSD Benchmark I can see that the Optane is slower.
I mean my main question is, what to do with the Optane Memory? Is it useless? Should I use the Optane for all HDDs and the Ram for the SSDs?
- Attachments
-
- Benchmark Optane.png (32.96 KiB) Viewed 5681 times
-
- Cache DDR4 Ram.png (33.75 KiB) Viewed 5681 times
Re: Intel Optane Memory
Optane Memory can be used as level-2 cache to speed up HDDs because it is much faster than HDDs.
Re: Intel Optane Memory
I made now two cache tasks, one for my SSDs only with RAM cache and the second one with all my HDDs cached by the Optane Memony as L2 Storage.
It runs good so far, the L2 with the HDDs has a cache hit rate 99,93%, not sure if ok or I did something wrong. The SSDs runs perfect with 8GB RAM cache, cluster size is on 4KB and the highest hit rate was 38,29%.
A question about the "Prefetch Last Cache" pls, is it the same like the Windows prefetch? On SSDs you deactivate the prefetch service at Windows, that's why I'm a little bit confused about it.
And on "Free Cache on Written" is it with this setting activated better or not, I'm using Intelligent as write mode, but don't understand the setting right. For me it sounds like its good, because it releases RAM but the info say it holds the RAM blocks.
Thanks for your support!
It runs good so far, the L2 with the HDDs has a cache hit rate 99,93%, not sure if ok or I did something wrong. The SSDs runs perfect with 8GB RAM cache, cluster size is on 4KB and the highest hit rate was 38,29%.
A question about the "Prefetch Last Cache" pls, is it the same like the Windows prefetch? On SSDs you deactivate the prefetch service at Windows, that's why I'm a little bit confused about it.
And on "Free Cache on Written" is it with this setting activated better or not, I'm using Intelligent as write mode, but don't understand the setting right. For me it sounds like its good, because it releases RAM but the info say it holds the RAM blocks.
Thanks for your support!
Re: Intel Optane Memory
It can take the L1 a while to get a higher hitrate, depending on what kinds of software you run, if there's space for application data and boot data in the cache, how often you reboot (frequent restarts hurt hitrate usually), and so on.
The prefetch simply retrieves data from the drive as it was in the cache prior to shutdown/restart, so it continues where it left off. It's slightly different from the Windows prefetch, but ultimately far more effective.
I don't use free cache on written. It means that data you send to a read/write cache task starts out in the cache, is then written to the drive, then is -removed- from the cache. So if you then re-access that written data, it is no longer cached and has to be retrieved -again- from the drive and placed back in the cache (if there's space available). It is detrimental to the performance of the L1 cache task, if it's a large enough task. You would typically only use that with things that you write once and don't want in the cache (like if you were encoding large video streams).
The prefetch simply retrieves data from the drive as it was in the cache prior to shutdown/restart, so it continues where it left off. It's slightly different from the Windows prefetch, but ultimately far more effective.
I don't use free cache on written. It means that data you send to a read/write cache task starts out in the cache, is then written to the drive, then is -removed- from the cache. So if you then re-access that written data, it is no longer cached and has to be retrieved -again- from the drive and placed back in the cache (if there's space available). It is detrimental to the performance of the L1 cache task, if it's a large enough task. You would typically only use that with things that you write once and don't want in the cache (like if you were encoding large video streams).