The new SSD Thread (Benchmarks, Brands, News and Advice)

No it's software based. During computer shutdown i do two things, first run an automatic script to backup the ramdrive image. After that is done the ramdrive image is saved to a file that is automatically loaded when the windows restarts and the ramdrive is remade. This is a ll very behind the scenes, very automatic, and easy to do.

This makes it persistent and you can even load applications on it, like I run google earth and and it's disk cache from it. I could even install the 37MB firefox install from there and run it if I desire. I'm just selective what applications i run from ramdrive, not many currently. Space is at a premium after all and I use for some files that can grow like temp folder and pagefile (still a few apps require a pagefile).

 

i use it for all of this. and i espencially need reliability. one thing i am using it for is for djing. i could not have one stutter in hours of gigging, imagine several seconds audio loss in front of the party croud. they wouln't like it.. that can happen because of any cheap hw, not just cheap ssd's.

i currently see an 64gb slc ssd from samsung for 250$ on ebay..

 

Yes, a good ram drive is very nice. Auto load at startup and auto save at shutdown (and at regular intervals if you so desire).

 

My OCZ Solid hasn't been a hassle. And like you said, $399 is much more than $116!

 

this is actually what superfetch should handle: auto load stuff at startup into ram and then run it from there. hm, and it does..

 

but it is for most. i prefer to spend money on something i know it won't be a hazzle than on something that may, or may not. russian roulette for that price? too much.

 

yeah ok, you got it. Except superfetch is pretty lame, although it uses up a bunch of RAM making you think it's doing alot. One day it should be better I hope, but microsoft still wants people depending on a pagefile so who knows.

 

haha sorry but yes, on xp i quite used ramdrives prior to ssds. (and still do at work. no ssd there, and no vista there). it can help really much.

but with vista superfetch sort of replaces the ramdrive, and now with the ssd's i don't even need superfetch anymore (i switch from time to time and see different boot behaviour, but that's about it.. ).

here: Samsung 64GB on Ebay i'd take it but i don't have any use for it right now.

 

it has some issues yes. i even reported them to microsoft. but changing to ramdisk has really no measurable benefit anymore with my systems (well, measurable, it has.. f.e. defragging of that drive is faster than anything including ssd's because of the small write issues but not measurable in day-to-day-usage)

btw: pagefile is a completely different concept and they both fit well together. a pagefile will always be needed, but only if you really hit any ram limit. and it even reduces your actual ram needs as it can virtualize empty memory parts to disk which an application requested. in a non-pagefile situation, imagine an app reserving 1.5gb of my 2gb in the notebook. the os would have to give 1.5gb to the app, and fit itself and all other apps into the remaining 0.5gb. with a pagefile enabled, the os can just give the real needed memory to the app, and fit the resting pages that are still empty to disk (without even writing them down there, all just virtual mapping stuff). the result is more free memory for your system. only if the app really fills it, it gets it, and then paging to disk may happen.

 

So you run entire applications, and all components, logs, every file read and written from superfetch. You must have Vista Super Ultimate.

While, with my ramdrive loaded apps/drivers (and hooked) it's actually close to instantaneous response as you can get as far as IO goes.

 

i said SORT OF.. but yes, i have close to instantaneous responce for all of my apps and stuff.

 

My superfetch never gave me that experience (since I dual boot), but ok.

 

it's quite a hit and miss thing. espencially on notebook drives (or subnotebook drives like the one i had in the 2710p before the ssd) it was more a miss. terribly blocking the disk at boot time. i don't want to know how that would look like on a jmicron controller based ssd, as it can't handle that much io in parallel.

i have quite some vista systems from friends where i disabled it (and right now i have it disabled on all my systems, too. no need anyways thanks to the ssd) because it blocked their boot process.

but if it works, it works great. quite sad. it would be awesome else. should be a bit more configurable.

 

I don't have my ramdrive snapshot the image at shutdown because i also mount a small pagefile to it and that's something i don't want trashing my drive with. My firefox profile is still on the drive, but my firefox cache is on the ramdisk and i don't mind it getting dumped. History and settings reside in the profile folder, i just don't know what the browser cache does???

 

the cache is for caching loaded pages and images. if you have a disk cache, it stores all our avatars there and never loads them again (can even create issues if some time-stamps are wrong and it never actually loads a page but only loads it from the cache which may or may not be outdated).

if you have ram cache, it caches it as long as the app runs.

ramdisk placed cache like diskcache, but with dataloss on powerout (without save/reload procedure somehow)

 

The corsair S128 has 32 mb of onboard cache. I checked with support on their forum.

 

That drive is MLC basicly the corsair but half the storage...

 

i linked the wrong one. yep i noted that's mlc before posting and found one a little bit more expensive but slc. haven't i posted that link? crap.

 

277$ (or so depending on shipping)

 

I got ripped on mine, should have stalked ebay longer. Oh well. You know, when you're one week before your exams, it's not time to be checking ebay all night to see if you can penny pinch a last second deal.

 

Use an auction sniper. I've been using something called JBidWatcher for the past year now... it really works wonders when you want to bid on something and can't be at your computer when the bidding ends.

 

Boris - isn't sticking a page file on a RAMdisk self-defeating?

 

In most cases yes. While I have spent alot of time without pagefiles completely, there's been some incidents. If every now and then an application requiring one wouldn't crash, then there would be no need at all. I keep a tiny one around just for that scenario. I doubt it's very often when people have no pagefile and an app crashes that they go back and figure out it's the missing pagefile to blame. One such occasion took me a long time, so I learned my lesson. It's rare but it does happen, only with certain apps though, usually older. I look forward to the day the thing can be forgotten completely, we are almost at the point of not looking back. In many case people can exist without a pagefile.

 

I've been running my TC4200 without a pagefile for just over half a year now. On 1GB I'd still end up crashing something with a lot of multitasking going on, but it's been smooth sailing with 2GB. I'm using XP, though, so I dunno how well 2GB without a PF would fly in like Vista...

 

I try to run without one with 4GB. I does speed up the operating system. I do run out occassionally. I haven't had too much problems either. One app I use, a video editing program, will crash a lot. I've heard things like that can be dependent directly upon virtual memory.

I've always thought the whole virtual memory thing was a really slack of Microsoft. It's the opposite direction of performance and encouraged developers to be inefficient with resources. We should all be using ram disks to improve performance, not running out of memory and using the harddrive to pick up the slack.

 

Yep, I've run it for longer. As mentioned, regardless of RAM amounts there are programs out there that actually require a pagefile, just getting rarer and rarer. Most would be fine without it if they got 2GB with basic apps.

 

No, I meant, sticking a page file on a RAMdisk (RAM) in case RAM runs out does not make sense to me. Am I confusing the page file with something else?

 

Finally some more numbers on the Vertex:

http://www.pcper.com/article.php?aid=670&type=expert&pid=14

 

I understand what you are saying, that does sound crazy, and you'd think it should work like that that it wouldn't be used if you still had free RAM. But go ahead and try and make a 10 MB page file. Then open a browser or other application, windows will say it ran out of space and needs to resize it after it reaches the 10MB limit. All along you could have a gigabyte of RAM free. Windows uses the page file even if you have RAM to spare. I didn't design it. When you set a minimum/maximum pagefile, then windows will tell you how much of it's being allocated, it uses it even if you have free memory. That's windows. The ONLY reason I have one at all though is for the 'rare' occasion I need one. But yes it does seem very silly from the surface, it's part of the messed up design. It's recommended numbers are also insane and make no logical sense (mine recommends 5373MB, yeah right), so it needs to be reworked.

 

I think he's trying to say that putting the pagefile on a RAMdisk would effectively be the same as scrapping the pagefile entirely (as opposed to making a small one) and just using your memory?

 

Yeah, it's basically like trying to step up by standing on your own foot. However, in the case where a program requires virtual memory to run, it might be a solution that doesn't cause the performance hit that a disk based page file would.

 

Interesting. How many controllers other than JMicron do we have now? Intel, Samsung, now this new Indilinx... is there anything else coming up? I also count at least 4 different manufacturers making SSDs with 256GB or more disk space. Unless they're are all colluding with each other, the prices should go down very soon.

 

Nice. PcPer is becoming the go-to source for SSD reviews and information.

And guys: can we take the RAMDisk discussion offline or create another thread. It is getting rather annoying to have 3 pages of discussion about off-topic stuff.

 

Sandisk has their own in-house controller being released soon.

 

Intel, Samsung, Mtron... then there are the folks making really fancy stuff like the Fusion IO... I think Memoright might also have their own controller? Not sure about that one.

 

I get it. And sorry for the off-topic discussion - I thought it was a little bit on topic because there is out there a RAM-based SSD that I read about on PCper yesterday actually.

 

So someone more knowledgeable tell me, should we be worried about the Vertex's IOMeter results?

 

I must have missed it... where did we see IOMeter results for the Vertex? It wasn't in the pcper article, was it?

 

i've got an m1730 with a raid 0 setup using 2 scorpio black 320GB drives. I want to remove one of them to use elsewhere, and thought i'd pick up an SSD for installing Vista + applications, and doing a dual boot with Ubuntu on the SSD drive. I would use the other 320GB drive in the laptop for storage of all my files (music/movies/docs/etc.)...

i looked at my drive and it looks like i would need about 25-50GB for Vista, and then about 15-20GB for Ubuntu.

I think I'll be ok with a 64GB SATA2 SSD, can anyone recommend a reliable brand/model ? is OZC good ? Mtron ? i game with this machine, but i'm not looking for super-high end SSD's, just a pretty good one that will not fail in the next 3-5 years. I use my pc on average 7 hrs a day.

is it possible to do a dual boot with vista/ubuntu on an SSD ?

thanks

 

Sure was: http://pcper.com/article.php?aid=670&type=expert&pid=10

 

It is possible to dual boot on an SSD. It is possible to anything on an SSD that you could do with an HDD. Well almost anything. You can't use the platters from your SSD to make coasters when it dies.

I dunno about the long term reliability of the JMicron MLC drives... I'd still go for either an Mtron or Samsung SLC; daveperman posted an eBay Samsung going for $277 just a few pages back. The Mtron tends to be a bit pricier and runs a bit hotter, but it's also a bit faster. Intel X25-M or the Corsair/Samsung 128GB MLC are probably also good choices.

Ho deng. I'm not an expert on IOMeter... but I think even I can tell that they don't look good?

 

The question is what are the real life implications. Check out posts 70-77 in this thread: http://www.ocztechnologyforum.com/forum/showthread.php?t=51200&page=5

Tony says they're irrelevant but I'm wary of OCZ's word after their initial handling of the Core series, and I don't know enough about IOMeter to form an educated opinion.

 

so has anybody reviewed 2 samsung 256gb ssd's in raid0 yet..? it would be an interesting review for people with two spaces in their laptops..

would like to know how they would compare to 2 x-25e 64gb's and 2 vertex's in raid0....

this year is totaly getting exciting in regards to ssd's...

 

no clue how much raid controllers in notebooks can actually surpass s-ata2 speeds. else, you won't really get much of raid0 with any of them.

 

You know guys, if you want the benefits of an SSD without spending full on the drive itself, just look for future versions of Turbo Memory. Calpella will feature 16GB flash memory with an integrated NAND controller on the PCH that is supposed to bring throughput on the levels of SSD drives.

 

Tony actually has a point.

It's like this. It's like server CPUs vs. PC CPUs. PC users don't understand how server usage is and for them, the huge caches and bandwidth is useless for them. In servers, when you have multiple CPUs and so much data passing through everywhere, memory bandwidth and interconnect speeds become the limiting factor in server performance(which is why Nehalem is such a win, combining Core 2's CPU strength with vastly improved I/O subsystem).

IOMeter was initially designed by Intel for benchmarking hard drives in server usage scenarios. And as you know the X25-M and X25-E has same controllers. The reliability and the IO throughput is optimized more for server usage than PC.

You don't notice freezes in old SLC drives because the minimum IOPS figure to not notice is achieved. >50 IOPS is enough for people to not notice freezing issues.

Maybe the new Intel drives coming in Q4 of this year will have more of segmentation and the PC drives will sacrifice some of that IOPS for better PC performance.

PC applications and usage is latency sensitive.
Server is bandwidth and throughput sensitive.

 

freezing has another chance of problem: namely if the controller can't handle more than one job, and one of the jobs takes up to 20ms, then the rest gets blocked, the os gets stalled.

i can get my mtron down to it's knees, but it still reacts at the same time, as it can handle more than one job. same for the intels. it's mainly the jmicrons that can't work in parallel.

 

Thanks for the great info and confirmation IntelUser! Looks like I can keep looking forward to the Vertex then.

 

This looks good in theory but as I understand it the intel drive has 10 memory chips that are run in a raid 0 configuration. I can't see a single memory chip out performing this. Of course 16GB of ram would work perhaps some one could build an external turbo ram with a small battery so the memory wouldn't be erased. This will cost around $100 a year from now.

 

I don't know that i want to rely on a small battery to keep my data in tact and i'm not very picky about that sort of thing usually... maybe a desktop that would only use the small battery "just in case" but not on a regular basis like laptops being turned off and on constantly for hours/days at a time possibly.