Samsung Magician 4.8 out - RAPID Results

Teizo

Golden Member
Oct 28, 2010
1,271
31
91
Downloads are limited this week, but if you check the Reddit thread, there is a link to TechSpot for the download. It is fully compatible with Windows 10 now. The thing I see now that is most improved is the RAPID software coding. It is now not only faster, but more consistent throughout the entire range. It used to be extremely volatile. Now, not at all.

Performance also improved for me when I realized I could install the Windows 8.1 RST Driver in Windows 10 before Samsung updated Magician, so not sure if having the Intel RST driver installed helped, but the results speak for themselves. Everyone can jump for joy over the expensive Intel 750/Samsung 950 NVMe, but unless I am looking at things wrong, my two year old 840 Pro with RAPID is faster.

Win%2010x64%20Intel%20AHCI%20Samsung%20RAPID_zpsmdiqbfvu.png~original
 

Coup27

Platinum Member
Jul 17, 2010
2,140
3
81
but unless I am looking at things wrong
You are. There is far more to an SSD than maximum sequential read and write figures which is what you are looking at here. People get fixated on these because they go to silly high numbers and think that translates into real world performance. I'm afraid it doesn't. High sequential reads and writes are not anywhere near as useful as you think. Only a few workloads such as professional video / photo editing yield any noticeable benefit from these high numbers.

The purpose of a RAM cache or RAPID in this scenario is to cache the writes to RAM and then write them to the SSD during a period of little activity, so to you, the action completes quicker. If you got hit with a power cut before the writes had been flushed they would be lost (without a UPS / capacitors etc). Also, to flush 8GB of writes to your SSD would still take around 14 seconds it's just the RAM caching is making it seem faster. What ATTO is benchmarking there is the speed of your RAM, not the speed of your SSD.

The 750 and the 950 Pro both use NVMe instead of AHCI which have lower latency and in the correct workload (eg. a heavy one) would yield better performance but for the average user you're not likely to notice any difference.

You should probably do some reading on what makes an SSD fast and what's generally more important (random reads over sequential writes). Oh, and also Alt+Pcn then crop in Paint next time :p
 

Teizo

Golden Member
Oct 28, 2010
1,271
31
91
I'm aware of benchmarks and real world performance. The same can be true of NVMe worshipers. Very little in the realm of day to day computing needs the speeds of what the marketing department of these companies want you to believe. Desktop computers for regular users won't see much if any benefit. Only in the scenarios you mentioned are NVMe needed. Same as doing Raid set ups with SSDs to increase theoretical performance. The point still stands that as this benchmark shows the speeds gained from RAPID are better than what NVMe offers in this benchmark. So, for people not wanting to spend a huge chunk of money on fancy drives, there is proof you can get good performance with Samsung RAPID. And, I live in a first world country. The chances of me losing power are not very high.

Oh, and next time simply enjoy the beauty of my wallpaper/deksktop than gibbering over nonsense. My random read IOPs with RAPID enabled are 140,000 and random write IOPs are 111,000.
 
Last edited:

sxr7171

Diamond Member
Jun 21, 2002
5,079
40
91
I'm aware of benchmarks and real world performance. The same can be true of NVMe worshipers. Very little in the realm of day to day computing needs the speeds of what the marketing department of these companies want you to believe. Desktop computers for regular users won't see much if any benefit. Only in the scenarios you mentioned are NVMe needed. Same as doing Raid set ups with SSDs to increase theoretical performance. The point still stands that as this benchmark shows the speeds gained from RAPID are better than what NVMe offers in this benchmark. So, for people not wanting to spend a huge chunk of money on fancy drives, there is proof you can get good performance with Samsung RAPID. And, I live in a first world country. The chances of me losing power are not very high.

Oh, and next time simply enjoy the beauty of my wallpaper/deksktop than gibbering over nonsense. My random read IOPs with RAPID enabled are 140,000 and random write IOPs are 111,000.

When you see read/write speeds that are in the 4-6GBps range that's clearly the speed of reading and writing to RAM.

What are your Random IOPs without Rapid enabled and at what sizes and queue depths? I'm just wondering what the difference is.
 

Teizo

Golden Member
Oct 28, 2010
1,271
31
91
When you see read/write speeds that are in the 4-6GBps range that's clearly the speed of reading and writing to RAM.

What are your Random IOPs without Rapid enabled and at what sizes and queue depths? I'm just wondering what the difference is.

I'm not sure. Whatever the default is used in Magician. I am not aware of any tests that I can download that I can test it with. I am aware of how RAPID works in relation to DRAM, which is why I mentioned the improvement they have made. It used to be very volatile and inconsistent and not really worth using. They've obviously improved the coding drastically from the onset of releasing it.

The cache set in DRAM has improved the loading time of some apps I use consistently (ccleaner for one), but overall my system is not really that much more snappy. The same can likely be said for NVMe as well. Theoretically a pretty high ceiling. Actual day to day use...not as much as most programs and usage scenarios are not demanding enough to require much beyond SATA III.

I posted this simply to share information because Samsung has made good strides in the consistency of performance in the software and people may see benefit.


Here is the AS SSD score of the drive with RAPID just for the sake of comparison.

Untitled-1_zpsinwmohs5.png~original
 
Last edited:

Rubycon

Madame President
Aug 10, 2005
17,768
485
126
Windows cache already does this.
The benchmarks are written taking this into consideration.
These newer cache programs like RAPID, Fancycache, etc. will benchmark far higher because the (benchmark) programs are in fact benchmarking the ram cache. If they updated them to reflect this you most likely would see similar results with higher cpu utilization and other factors.

And because of this, users don't feel real world differences when using nvme on the desktop with real world everyday tasks.

Disable the Windows cache altogether, and this will change.

Oh and it's most definitely uncharted waters as to what happens to your data if an unexpected system crash or power outage occurs and all that cached data cannot possibly be flushed out of volatile storage!