Caching software vs Samsung Rapid: Test inside

[ctk1981]

For giggles I elected to do a test. I tried two different caching software programs, only to find that one only worked. I may try primocache also and post results once I check into that program a little more.

Between each benchmark, a trim command was sent to the drive. I had also tested using a larger cache size, but it made no difference between 4GB and 10GB due to the nature of some of these benchmark tests using relatively small data sets for testing.

First up...and stock 830, fresh trim, 25% OP 128GB drive.

Now a 830 with O&O clever cache installed.

Note, the ram usage never really increase. The cache size was showing a meager 150MB. I believe this program to operate more like superfetch or such and really isn't too aggressive on what it caches.

Now an 830 with SuperSpeed supercache V5. Cache size was set to 4GB to mimic the limitation of samsung rapid. No lazy write.

Now we start to see a caching software start to mimic the results of Samsungs Rapid. But the writes aren't looking to great. Lets see if a 5 second lazy write can increase that. Interestingly the AIDA read test suite never improved.

830 samsung with lazy write enabled

An increase in write speed in general, albeit at the risk of data loss in an event of power outage.


But what about regular HDD, can they be improved with this type of software

An old WD 640 black. No lazy write on first test

Now with lazy write enabled.

Now the 840 Pro 256GB. Stock, Rapid, SSLW

edit 2 new notes

After some investigation to rapid mode and lazy writes I'm at the opinion currently that they are of little benefit (writes) and in fact have some drawbacks that I feel outweigh the benefits. Whether or not they can improve the avg write speed to a drive is debatable and I'm still trying to wrap my head around how to even possibly test this. On the other hand, due to the delay that is introduced a lot of times the overall transfer time may be greater than if the feature was turned off entirely. More testing is needed to determine if there are indeed any real world benefits to write caching.

I started to play with primocache and it has some of its own quirks. Immediately I noticed in benchmark testing if write caching is not enabled it is very wild in throughput....5mb-155mb transfer....but this was in AS-SSD. With write and read cache enabled, everything seemed to smooth out...I need more time to see whats going on for sure. I was playing with the setup on the WD 640 caviar HDD I benched above....with primocache it anvil suite, it actually out scored my 840 pro with RAPID enabled. Looks promising....I also want to stick the 830 in as level 2 cache and see how well it works before I invest in my replacement 4TB drive.

I'm due to switch over to an intel system this week or next...which will be another set of tests. My stock 830/840 pro scores haven't exactly been the highest and I think a lot of that has to do with the amd chipset. Will be interesting to see what improvements the newer intel chipset will bring.

 

[Pwndenburg]

Having look at that it does look promising. Still won't deal with the fact that my sammy has 128 GB on it I really want to jump on that MX100 but the company won't even discuss P/E cycles on the drive. Makes me leery.

 

[ctk1981]

I wouldnt worry too much about it. Its MLC nand, albeit a smaller process node but I would expect this drive to be as reliable as the M500/M550. As the anandtech review stated "Even if there is a slight reduction in raw NAND endurance, it's possible to compensate for that by reducing the write amplification through firmware optimizations".

 

[Pwndenburg]

You are this close to selling me. That thing seems to be exactly what so many of us need. I pretty big hold over drive that will last until we see how these beast pci drives help or don't in the coming two years. Man, thank you for helping me with my issues here. As I may have stated before, I'm not nearly as savvy as the average member here. I can assemble my own pc, but have never oced anything for instance.

 

[BonzaiDuck]

Well, that would be good -- to find a software utility with total reliability that adds RAPID-like caching performance to any SATA SSD. The ideal of course would have this software working perfectly for some number of SSDs -- note that Samsung only allows RAPID to work for one drive, even as a system might be configured with several 840's.

The OP tested a Samsung 830. I was thinking that there would be no reason that Magician and RAPID should not be backward compatible with the 830 if RAPID was introduced after that model, but I don't know if it happened that way.

 

[ctk1981]

Testing updated. 840 stock, rapid, supercache with lazy writes posted up.

No, magician software does not enable rapid mode on the 830. Requirements state 840 or newer. However with supercache software (and perhaps others) it is entirely possible to duplicate some of the speed enhancements that rapid offers and not be restricted to just samsung hardware. And as BonzaiDuck pointed out, not limited to one drive is also a plus.

The one thing I have noticed, with rapid mode I rarely see a large increase in memory usage. With SuperCache V5 I see the ram usage go up considerably (depends on how large you set the cache of course, I've went up to 8GB), so I think it is a bit more aggressive in what it caches.

Next up I think I'll pony up the dough for a primocache license and give it a whirl. I'm interested in their "L2" cache technology which is seemingly similar to what intel does with using a small ssd to cache a larger hdd, but I believe primocache take it one step further and L1 cache is actually the ramcache in front of the L2 setup.

 

[R0H1T]

Testing updated. 840 stock, rapid, supercache with lazy writes posted up.

No, magician software does not enable rapid mode on the 830. Requirements state 840 or newer. However with supercache software (and perhaps others) it is entirely possible to duplicate some of the speed enhancements that rapid offers and not be restricted to just samsung hardware. And as BonzaiDuck pointed out, not limited to one drive is also a plus.

The one thing I have noticed, with rapid mode I rarely see a large increase in memory usage. With SuperCache V5 I see the ram usage go up considerably (depends on how large you set the cache of course, I've went up to 8GB), so I think it is a bit more aggressive in what it caches.

Next up I think I'll pony up the dough for a primocache license and give it a whirl. I'm interested in their "L2" cache technology which is seemingly similar to what intel does with using a small ssd to cache a larger hdd, but I believe primocache take it one step further and L1 cache is actually the ramcache in front of the L2 setup.

Yes & then there's also deferred write (lazy write for Supercache) available on primocache, also if possible try V-locity Endpoint (download link) as it's the best of the lot IMO.

 

[ctk1981]

Thanks for the tip, I'll check into it. I did some beta testing for diskeeper back in the day, around the time they started their hyperfast options for SSD's. It wasn't anything earth shattering back then, but this product has evolved greatly from those days so it may be worth investigating and see what type of improvements they bring with V-locity endpoint.

Come to think of it, I might have even beta tested a v-locity product back in the day. Not overly thrilled about the request a quote option...

 

[R0H1T]

Thanks for the tip, I'll check into it. I did some beta testing for diskeeper back in the day, around the time they started their hyperfast options for SSD's. It wasn't anything earth shattering back then, but this product has evolved greatly from those days so it may be worth investigating and see what type of improvements they bring with V-locity endpoint.

Come to think of it, I might have even beta tested a v-locity product back in the day. Not overly thrilled about the request a quote option...

The thing about v-locity is that it's also supposed to prevent defrag with "intelliwrite" which seems an interesting feature (taken from their defrag apps) & seeing how diskeeper is still one of the best 3rd party defrag utilities out there I wouldn't surprised if it actually enhances the SSD life & improves performance over a period of time, as the drive fills up.

 

[BonzaiDuck]

Testing updated. 840 stock, rapid, supercache with lazy writes posted up.

No, magician software does not enable rapid mode on the 830.

If that's real confirmation (about the 830 and RAPID -- it's good to know. Also, I should warn my friend with the 830 drive so he doesn't waste his time.

Next up I think I'll pony up the dough for a primocache license and give it a whirl. I'm interested in their "L2" cache technology which is seemingly similar to what intel does with using a small ssd to cache a larger hdd, but I believe primocache take it one step further and L1 cache is actually the ramcache in front of the L2 setup.

Hope you don't mean the CPU cache levels, but you refer to "L2" cache technology as a Romex innovation -- what they call it. If I have time, I'll look closer at the implications of what you're saying.

Even so -- no doubt -- absolutely --- I can see myself wanting to try the Romex caching software on some other machine I have. But if you check my post on another thread responding to ctk1981, I think it's all about RAM available. So the best choice for using PrimoCache among LGA775 systems would be the system configured with 8GB of RAM as opposed to 4.

 

[Fred B]

When testing cache it is testing the ram performance , trying out the beta primo cache with ssd raid . Now it is using the intel raid write back cache and primo cache it is double the speed . It is on a 32 bit system that takes advantages of 8 GB ram , the cache software uses ram above the 32 bit os limit.

 

[ctk1981]

If that's real confirmation (about the 830 and RAPID -- it's good to know. Also, I should warn my friend with the 830 drive so he doesn't waste his time.



Hope you don't mean the CPU cache levels, but you refer to "L2" cache technology as a Romex innovation -- what they call it. If I have time, I'll look closer at the implications of what you're saying.

Even so -- no doubt -- absolutely --- I can see myself wanting to try the Romex caching software on some other machine I have. But if you check my post on another thread responding to ctk1981, I think it's all about RAM available. So the best choice for using PrimoCache among LGA775 systems would be the system configured with 8GB of RAM as opposed to 4.

No, I was referring to L2 as romex uses it. I currently use a 256GB 840 pro as my boot drive, my 640WD hosts the majority of my applications and games, and a 2TB serves as my media storage drive. What I would like to do, and I'm unsure if its possible or not...is to take a 4TB WD green drive, couple my 830 pro 128GB as a L2 cache, utilize 4GB ramcache in front of this....so basically I'd have a low power, very large capacity HDD that could mimic read speeds of faster SSD.

The big question I have....is can the L2 cache setup be applied to a disk that isn't a boot disk.

 

[R0H1T]

No, I was referring to L2 as romex uses it. I currently use a 256GB 840 pro as my boot drive, my 640WD hosts the majority of my applications and games, and a 2TB serves as my media storage drive. What I would like to do, and I'm unsure if its possible or not...is to take a 4TB WD green drive, couple my 830 pro 128GB as a L2 cache, utilize 4GB ramcache in front of this....so basically I'd have a low power, very large capacity HDD that could mimic read speeds of faster SSD.

The big question I have....is can the L2 cache setup be applied to a disk that isn't a boot disk.

I'll answer that for you, yes it can. What you'll have to do is to select the disk partition(s) to be cached, then select read/write (or both) as the cache strategy & then the amount of RAM to be alloted as a cache.

Now for L2 you need unallocated space, as primocache will delete the partitions if you've selected a whole disk, so you should reduce the size of your C: partition on your SSD & set the unallocated space as L2 cache.

I'll just add that I deleted my OEM recovery partition(~2GB) on my laptop HDD to use as L2 cache & have tried using an external flash drive as well, both the options worked fine but that didn't give me much in terms of performance obviously due to the painfully slow stuff I was using.

 

[ctk1981]

I dont want to reduce my C: partition to cache for the drive, I want to install a separate 128GB sdd for the sole purpose of acting as an level 2 cache for the drive and my 256GB C: be used as normal....but I'm guessing it sounds like either way it will work for what I want.

Thanks

 

[R0H1T]

I dont want to reduce my C: partition to cache for the drive, I want to install a separate 128GB sdd for the sole purpose of acting as an level 2 cache for the drive and my 256GB C: be used as normal....but I'm guessing it sounds like either way it will work for what I want.

Thanks

Yes that'll do but remember if you select a particular device you won't be able to use it for anything else other than an L2 cache.

 

[Deders]

I'm trying to get my head around this, how does the software predict what you are going to want to read from the disk? If you set a 4GB limit then is it a bit hit and miss as to what gets accelerated?

 

[dave_the_nerd]

I'm trying to get my head around this, how does the software predict what you are going to want to read from the disk? If you set a 4GB limit then is it a bit hit and miss as to what gets accelerated?

Read caches are usually populated based on how often particular blocks are accessed, and in what order.

So binaries and data for frequently used applications, mostly.

You could also try getting clever with requests. HDDs are usually written to sequentially, which means data is usually contiguous. Read request for block 51215? Pre-fetch 51216-51218 just in case. They're probably related data, and it's a lot easier to handle a single 4-block read than 4 single-block reads.

If the software is paying attention to application activity, instead of just storage, it could even try being clever by monitoring what you're doing. (Working in Office? Load any locally-stored Word Clip Art into cache. Last ten files you opened were JPEGs? Why not load C:/Users/Me/My Pictures into cache? Playing a movie? Load it into RAM now and let the HDD idle. Scrubbing with be way faster.)

Different strategies, a lot of math. Smart people. *shrug*