|
I'm keeping my eye on Dataplex, or a competitor that sells the software commercially instead of locking it to specific hardware. C:\Windows alone is 25+ GB, and most of the crap in there is hardly ever used. Prime candidate for being on a large, slow medium. PageFile.sys and hiberfil.sys are going to be another 10-12gb (depending on RAM and settings). The files needed to fast-boot windows are going to be a fraction of that, and the most common DLLs not much more. Managed tiered storage is the way to go. How much of CS5 is actually read when it launches? Games are one of the few program installs that actually do use the majority of their data in the hotpath, cutscenes excluded. Also not cool Intel, their SSD cache software is locked to the Z68. There's no hardware support in the Z68, they just decided to require you to buy that chip in order to use SRT.
|
#
¿
Dec 10, 2011 06:23
|
|
|
|
| # ¿ Apr 25, 2024 00:09 |
|
|
Where's my best bet for a 250gb SSD that will work well in linux (with TRIM)? 830s are mostly out of stock at this point, and nobody is discounting them. I see deals on other drives but mostly on the don't-buy list. Honestly, the way I'd do the OP is "Here's a list of drives known not to suck. If you buy ANYTHING not on this list we will laugh at you." There's so many loving rebrands, relabels and 'model number may not correspond to contents of box (I'm looking at you, OCZ)' out there that are all garbage it seems pointless to list them all. Is the following list correct? Buy now: Samsung 830 Intel 520-series Mushkin Enhanced Chronos Corsair Force GT Buy but beware firmware issues: Samsung 840 pro Crucial M4 Never buy: Anything else, unless you are a reviewer endurance testing them Never buy, called out for name by people who can't read OCZ Crucial V4 Samsung 840 non-pro On the firmware issue front: Are the Intel 330/335 in the 'maybe' column now with their updates?
|
|
#
¿
Dec 2, 2012 17:05
|
|
|
Alereon posted:It's in the "do not buy" list and mentioned elsewhere. On Sandforce drives you can disable the RAISE error-correction to go from 120/240GB to 128/256GB (power-of-two sizes). Disabling error-correction is a bad thing if you want a reliable drive for obvious reasons. I'm assuming that only applies to sandforce drives, as other chipsets have different ways of overprovisioning?
|
|
#
¿
Dec 2, 2012 18:17
|
|
|
Alereon posted:The problem is that there's a large number of very equivalent Sandforce drives that are fine, so we can't just say "always buy this drive". There's also OCZ, first-gen Sandforce drives, and drives without RAISE. Thus, the best guidance overall is that a SATA600 Sandforce drive with a non-power-of-two size that is not OCZ is probably fine. Thing is, the reason I said "If it's not listed we'll laugh at you" is because of the sheer amount of lovely race-to-the-bottom sandforce clones there are. Even if you ignore idiots making "branded changes" to firmware that they don't understand, you've got bargain bin "Huh, wonder if this still works?" NAND chips being used because the only consideration is price. I guess a list of "These are model numbers of generic sandforce drives that don't appear to be using the cheapest possible NAND chips." People ask about random brand in here all the time, so if someone does take the time to look into RandomBrand 256GB SSD x-x-x-treme edition, and decide it's using good chips, toss it in the 'probably not complete garbage' clone list in the OP. Harik fucked around with this message at 19:06 on Dec 2, 2012 |
|
#
¿
Dec 2, 2012 18:50
|
|
|
Alereon posted:I don't think there's actually the level of brand-to-brand variability you think there is for Sandforce drives. With the notable exceptions of OCZ, everyone is using flash of about the same quality, and no one's doing custom firmware (aside from the RAISE thing, which isn't custom it's just a config option). The OP does list some recommended drives that will work well, but for Sandforce drives you can largely assume the drive is OK if it's SATA600, has a normal capacity, and isn't OCZ. That's pretty clear - should put that in the OP. Turds in magma posted:Is the Samsung 830 256 GB still probably my best bet for performance/price, if I can get it for around 200 dollars? Is the 840 still "unverified" in terms of reliability? 840 non-pro uses cheaper flash that wears out fast. 840 pro may be as good as the original 830, but multiple review models bit the dust before they even finished testing. Samsung says they fixed the issue with pre-release firmware - it's your data to risk. Edit: Anyone know the difference between the Mushkin Enhanced Chronos and the -Deluxe version? Newegg compare says "SandForce SF-2281 SSD processor with unthrottled - IOPS firmware" But they both have the same 90k IOPS rating - the deluxe is even rated for 5MB/s lower than the cheaper version. Harik fucked around with this message at 23:06 on Dec 2, 2012 |
|
#
¿
Dec 2, 2012 23:01
|
|
|
If windows file copy gets stuck forever, is that a indicator of a faulty SSD? In linux any kind of stall like that would be clearly identified as to what the hell, but in windows it's just hung forever. Crucial M4 256gb, 000f firmware, hooked to the 6gb intel port on a P67 motherboard. Windows 7, freshly installed. Re-doing the copy has always worked, but the copy window sticks forever. I'd rather know if I need to RMA this now or if I should look into driver issues - it's a fresh install as of yesterday so I may have missed some intel chipset driver or installed a wrong version or something. Edit: CrystalDiskInfo tells me no errors, 15 hours powered on. That's about right, I installed it yesterday. Harik fucked around with this message at 05:25 on Jan 9, 2013 |
|
#
¿
Jan 9, 2013 05:17
|
|
|
Rexxed posted:It could be a lot of things, windows file copies aren't very verbose about what's going on. I'd dobule check that you have the right sata (AHCI) or raid controller drivers installed and run a chkdsk to see if anything's wrong with the filesystem. SSD or not sometimes there's just weird errors. AHCI driver was 9.10.0.1021, from early 2010. Ouch. Apparently that wasn't included in the intel chipset driver bundle for my motherboard. I grabbed Intel storage manager 11.whatever from late 2012 and now it's updated. Hopefully that fixes the problem.
|
|
#
¿
Jan 9, 2013 17:50
|
|
|
Alereon posted:You don't have to, but there's no reason not to since it results in improved performance. If you don't like them for some reason you can always go back to the Microsoft drivers. I was getting random disk hangs when copying directories to the SSD when using the MSFT drivers. They're just 2 year old copies of the RST drivers anyway, so updating to current fixed the problem. This is on a P67 chipset.
|
|
#
¿
Jan 24, 2013 13:35
|
|
|
What's the eraseblock size on a 840 EVO 250gb? I need to know because I'm using it for bcache, which is designed to journal one eblock at a time for maximum speed. I saw 8kb*192 (due to TLC) but it wasn't clear if that was the 840-nosuffix or the 840 EVO.
|
|
#
¿
Feb 18, 2014 06:42
|
|
|
Alereon posted:Anandtech says 8kb*256 on the 840 evo with 19nm TLC, it was 8kb*192 on the 840 with 21nm TLC. Great, so a 2MB erase block, that's a pretty sane size. Appreciate it, I checked the datasheet but like all consumer goods it's really more a marketing glossy.
|
|
#
¿
Feb 18, 2014 07:09
|
|
|
Quick bcache trip report: 250GB 840 EVO (short-partitioned to 230GB since I'm guaranteed to need that burst write capability) as L1 storage in front of a 4x4TB RAID5 L2. Holy poo poo it's fast: SSD Cache enabled: 4.2 seconds. Dropping the OS block cache in between, of course. For comparison, with all the important blocks in RAM it's 1.8 seconds. Edit: Durp, I only was dropping the pagecache, leaving the inode cache intact. Fixed, now on an idle system: 98 seconds without cache 25 with. 1.9s in-RAM. ~400,000 files. I thought those first numbers were a bit fishy so I redid it. Still, that's a 4x speed boost which is nothing to be sneezed at. That's covering the important parts for interactivity as well, listing files & sizes, tab completion, etc. I may start using this type of setup on more machines - mirror two big drives with a smaller SSD cache in front of them. I've burned 7.2 erase cycles in less than two days, although that's probably due to the fact I'm caching during the initial array copy. I've switched to cache-only-reads for now, so the write-mostly copy shouldn't be eating my SSD so badly. Thanks to Alereon, I'm not getting any write amplification; LBAs written according to disk is very close to what my cache stats show. Harik fucked around with this message at 09:27 on Feb 20, 2014 |
|
#
¿
Feb 19, 2014 21:47
|
|
|
|
| # ¿ Apr 25, 2024 00:09 |
|
|
DrDork posted:The hold on recommendations has little to do with their performance (which is about all Anandtech was reviewing) and more to do with the fact that Crucial has had a bad track-record with reliability. A drive that shits itself at 5,000 hours isn't going to have a problem passing some stress-tests and benchmarking well, but isn't exactly something that we'd want to recommend you purchase, you know? So until a bit of time goes by and more "early adopters" have a chance to find out if the MX100 follows closely in the long line of firmware-based fuckups from Crucial, it's hard to recommend without reservation. Speaking of 5k, was there ever a postmortem on what the gently caress that bug was? It's not a power-of-two overflow that I can see unless they were counting time in 4.2 millisecond increments or something equally "creative". So some kind of periodic sanity check that went horribly wrong because it was never tested?
|
|
#
¿
Jun 10, 2014 07:28
|
|