starcrafted Posted March 25, 2012 Share Posted March 25, 2012 Hello, Can anyone explain to me what's the difference between Marvell based Performance PRO and SandForce based Force GT drives? What are the strong points of each and what are the weak spots? How about RAID 0 performance on a p67/z68/z77 chipset? I know with the next release of Intel RST drivers will come support for TRIM in RAID, but until then, which one will perform better? Thank you. Link to comment Share on other sites More sharing options...
Toasted Posted March 25, 2012 Share Posted March 25, 2012 The product page explains the difference. Performance Pro http://www.corsair.com/ssd/performance-pro-series-ssd/performance-series-pro-128gb-ssd-hard-drive.html Force GT http://www.corsair.com/ssd/force-series-gt-ssd/force-series-gt-180gb-sata-3-6gbps-solid-state-hard-drive.html Performance Pro can handle compressed and non-compressible data. Force GT can handle compressed data. Link to comment Share on other sites More sharing options...
FoLmEr Posted March 25, 2012 Share Posted March 25, 2012 In short, if the workload is highly compressible, the GT will do better. If the workload is non-compressible, the PRO will be the better choice. In terms of an OS disk, I believe the average amount of compressible data the disk handles during normal usage is around 46% (as per Anvil's SSD utility). Since the GT is cheaper, I'd deem it the best choice. If, however, the disk is must handle non-compressible workloads such as handling (music/movies/pictures), it may be worth paying a premium for the PRO. Regarding RAID I'm clueless. Link to comment Share on other sites More sharing options...
starcrafted Posted March 25, 2012 Author Share Posted March 25, 2012 Performance Pro can handle compressed and non-compressible data. Force GT can handle compressed data. Yes, thank you. That much I know. I was interested in real life performance, reliability and whatnot. People used to view Crucial M4 as one of the best SSDs, but then Performance Pro was released and among elitists is viewed as the best now and they often use words like reliability when talking about Marvell SSDs. I am interested in typical usage scenario: a system disk, a few games, apps and so on. Also, are the SandForce problems fixed with the latest firmwares or should I expect BSODS and random crashes? Link to comment Share on other sites More sharing options...
FoLmEr Posted March 25, 2012 Share Posted March 25, 2012 Yes, thank you. That much I know. I was interested in real life performance, reliability and whatnot. People used to view ******** M4 as one of the best SSDs, but then Performance Pro was released and among elitists is viewed as the best now and they often use words like reliability when talking about Marvell SSDs. I am interested in typical usage scenario: a system disk, a few games, apps and so on. Also, are the SandForce problems fixed with the latest firmwares or should I expect BSODS and random crashes? In real life you probably won't notice any difference using one disk over the other unless you're setting up an environment that specifically favors one of them. The performance PRO is potentially the fastest of the Marvell based SSDs. And, yes, with the 1.3.3 firmware the SandForce problem is a thing of the past. Check out this thread if you want to get an idea what to expect from a drive in terms of reliability/endurance/performance: http://www.xtremesystems.org/forums/showthread.php?271063-SSD-Write-Endurance-25nm-Vs-34nm/page155 Link to comment Share on other sites More sharing options...
Humbug Posted March 25, 2012 Share Posted March 25, 2012 Both are very fast drives that can read >500MB/s. People still talk about the sandforce reliability being an issue but most likely it's sorted now with latest firmware. For me personally the reason I choose the Performance Pro was that based on the reviews that I read the drive was very consistent. The Force GT had slightly higher peak performance, but the performance pro was very fast in just about every test- regardless of the queue depth, type of data or whether it was sequential/random etc... As FoLmEr said in the real world you would not notice the difference between the two unless you were using a stopwatch. btw since you asked- in a RAID environment definitely choose the pro over the force gt, at least until TRIM is supported. The garbage collection in the pro works almost as well as TRIM. Link to comment Share on other sites More sharing options...
stw500 Posted March 27, 2012 Share Posted March 27, 2012 btw since you asked- in a RAID environment definitely choose the pro over the force gt, at least until TRIM is supported. The garbage collection in the pro works almost as well as TRIM. Corsair claims that, my own experiences are the opposite, read more in the german section or look at the screens, showing a big performance-loss in AS SSD in sequential reads and writes, although there was trim implemented in the encryption all the time. http://forum.corsair.com/v3/showthread.php?t=105492 My Intel postville never had such a big performance-loss throughout two years. Link to comment Share on other sites More sharing options...
starcrafted Posted March 27, 2012 Author Share Posted March 27, 2012 Thanks for your answers. I have found a better explanation of the real life differences here: http://www.xbitlabs.com/articles/storage/display/marvell-ssd_7.html This is where Marvell 88SS9174 based SSDs become the stars of this show. They not only ace the TRIM execution by fully restoring their performance to the original level, but can also boast excellent efficiency of the background garbage collection algorithms, which prove capable of restoring the data-packed SSD practically to its new state. SandForce controllers, as we have already seen many times in our previous tests, don’t get anywhere close to that. Link to comment Share on other sites More sharing options...
Humbug Posted March 27, 2012 Share Posted March 27, 2012 Corsair claims that, my own experiences are the opposite, read more in the german section or look at the screens, showing a big performance-loss in AS SSD in sequential reads and writes, although there was trim implemented in the encryption all the time. http://forum.corsair.com/v3/showthread.php?t=105492 My Intel postville never had such a big performance-loss throughout two years. Really? Well that sucks. I personally have not tested my drive in a non-trim environment, my opinion was simply based on the test results I had read thus far from other people. e.g. http://www.hardwarecanucks.com/forum/hardware-canucks-reviews/49706-corsair-performance-pro-256gb-ssd-review-11.html p.s. what do you mean there was TRIM encrypted in the encryption, I am not 100% clear on how your drive degraded, don't speak German unfortunately. Link to comment Share on other sites More sharing options...
stw500 Posted March 27, 2012 Share Posted March 27, 2012 Really? Well that sucks. I personally have not tested my drive in a non-trim environment, my opinion was simply based on the test results I had read thus far from other people. e.g. http://www.hardwarecanucks.com/forum/hardware-canucks-reviews/49706-corsair-performance-pro-256gb-ssd-review-11.html p.s. what do you mean there was TRIM encrypted in the encryption, I am not 100% clear on how your drive degraded, don't speak German unfortunately. I encrypted the SSD fully with diskcryptor. Diskcryptor supports trim: http://diskcryptor.net/wiki/Main_Page/en and there under changelog of version 1. The performance-loss also in sequential reading shows, that there is a big fragmentation and the ssd has to do so much accesses that the sequential performance is slowing down (from 230 MB/s (Sata2-Limit) in the beginning to 160-170MB/s now, AS SSD measuring). With a better garbage collection this wouldnt have happened. I am really disappointed because i buyed this ssd especially because of the claimed garbage collection, which should have been better than other ssds. Now i have to realize, that it is worse than that of my old postville. Unbelievable...the test by hardwarecanucks is - as most reviews - no long term test. My SSD has been encrypted since December 2011, so this was a "test" for three month. Link to comment Share on other sites More sharing options...
parsec Posted March 27, 2012 Share Posted March 27, 2012 I encrypted the SSD fully with diskcryptor. Diskcryptor supports trim: http://diskcryptor.net/wiki/Main_Page/en and there under changelog of version 1. The performance-loss also in sequential reading shows, that there is a big fragmentation and the ssd has to do so much accesses that the sequential performance is slowing down (from 230 MB/s (Sata2-Limit) in the beginning to 160-170MB/s now, AS SSD measuring). With a better garbage collection this wouldnt have happened. I am really disappointed because i buyed this ssd especially because of the claimed garbage collection, which should have been better than other ssds. Now i have to realize, that it is worse than that of my old postville. Unbelievable...the test by hardwarecanucks is - as most reviews - no long term test. My SSD has been encrypted since December 2011, so this was a "test" for three month. My Marvell controller based SSDs (not Corsairs) have not had any sequential speed reduction over time. Even when used in RAID 0, that has not happened, although I manually over-provision the volume so it has more free space to work with. I would suspect the encryption software before anything else, it may support TRIM, but it might affect the GC in a negative way. SandForce controllers have built in encryption, and you know what the tests show about the affect of TRIM in recovering performance on those SSDs. That test is done on a completely full, tortured SSD, so is also a worst case situation. How full is your SSD? Link to comment Share on other sites More sharing options...
stw500 Posted March 27, 2012 Share Posted March 27, 2012 The same encryption software was running two years on my Intel postville without any performance-loss. Three month after decryption, cloning the System to the Corsair Performance Pro and reencrypting, the performance-loss was noticed by me, dont know exactly when it happened. Your Marvell-based SSD is at first an older revision of the controller and second has another firmware, because corsair build its own. So a simple comparison is not possible. On the Corsair Performance Pro there was always free space between 10 and 15 GB, but all the space was encrypted, the same as with the postville. Link to comment Share on other sites More sharing options...
stw500 Posted March 27, 2012 Share Posted March 27, 2012 I would suspect the encryption software before anything else, it may support TRIM, but it might affect the GC in a negative way. Garbage collection is - in my opinion - a controller based feature which can not be controlled by software. So the encryption software has no way to influence when and how often or intensive the ssd starts garbage collection. The first samsung 830 firmware also had an error in firmware which led to immense performance loss in sequential speeds when writing over long term. With the second firmware release this problems where solved by samsung. This thread - and my own in the german sector - should help to start corsairs technical support team over improving the firmware, so no performance loss can take place in the future. I dont know how, this is not my task, i can only display the settings which led to the p-loss. All used software, as ssd, diskcryptor, etc. is free. Link to comment Share on other sites More sharing options...
m.oreilly Posted March 27, 2012 Share Posted March 27, 2012 they raid just fine, and hold up, i.e., keep above the low performance 'steady state' that SF controlled drives hit once all nand has been written to, and they don't need any over provisioning, which is a must on SF drives. Link to comment Share on other sites More sharing options...
Humbug Posted March 28, 2012 Share Posted March 28, 2012 Garbage collection is - in my opinion - a controller based feature which can not be controlled by software. This is correct. However as you say if this is full disk encryption it would probably write dummy data all over the disk with the actual data buried in the middle as a security precaution. As far as the controller is concerned this means no free space and it would seriously hamper the garbage collection process would it not? Link to comment Share on other sites More sharing options...
stw500 Posted March 28, 2012 Share Posted March 28, 2012 I am not quite sure, but this should be solved by the trim command. The OS "knows" over the encryption software driver which nands are without data (written "0" in nand or tagged as empty for the trim command), even when the ssd is full encrypted. Because of this the nands can not only be emptied before writing again (trim), the controller also knows which nand is empty and can start garbage collection to enhance performance. This should be the same like in any non encrypted System. For sure it is true, that the garbage collection works worse with less free space, but the transparency through the encryption driver should work. Well at least it worked in my postville. Lets see what the next longer "test period" shows for a result. As for now the sequential speeds didnt slow down the last days after encrypting again (remember sata2-limit and little encryption performance-loss of 20-30MB/s). Link to comment Share on other sites More sharing options...
Recommended Posts
Archived
This topic is now archived and is closed to further replies.