It won't across the board. The few benchmarks available show so. It's better in reads, of course - but in wirtes and mixed use - nope.
Quote:
Originally Posted by Helmore
Printable View
It won't across the board. The few benchmarks available show so. It's better in reads, of course - but in wirtes and mixed use - nope.
Quote:
Originally Posted by Helmore
Where is that review, I can't find it. I googled and looked around on this forum, but I can't find the review that you are talking about.
Here's the most thorough (though IMO missing a lot of actual tests, especially in the write department):
http://www.nextlevelhardware.com/storage/battleship/
http://www.nextlevelhardware.com/sto...ip/chart12.jpg
An interesting thing is this also:
http://www.nextlevelhardware.com/sto...ip/chart13.jpg
4s to copy 1GB on 9xMtron R0? It takes me less with 8 crappy Seagate 7200.11 drives in RAID5.
Now, there could be a few issues, i.e. I am not sure if Dominick used 1231ML throughout the entire test series, or just in some tests.. and how he cabled the drives.
EDIT: Might wanna take a look at this also:
http://www.anandtech.com/showdoc.aspx?i=3064&p=6
Those are not reviews of the drive that I was linking to :p:, they are about the older drives.
Oh well I guess I can't find any reviews for the drive I linked to, so I can't disprove your standpoint and thus you are right (for the moment).
I think you're right, and here's something I dug up in the spare time:
http://www.storagereview.com/php/ben...3=348&devCnt=4
Puzzling: average random write access time: 6.8ms for the Mtrons?! That would explain the write IOPS.
Sorry - I presumed the tests I mentioned previously _were_ 7000 Pro series, seems they weren't.
Also, test of MemoRight GT:
http://www.tweaktown.com/reviews/143...d_0/index.html
It shows MemoRight GT in more light finally.. (I didn't notice this review yet) previous reviews only showed HDTach/Atto scores, which are pretty much useless.
MR GT seems to be holding the top crown at the moment. Still there's no direct comparison test between MR GT, Mtron 7000, Raptors and VRs.
P.S. here's why TechReport results are fishy at best:
http://www.storagereview.com/WD3000BLFS.sr?page=0%2C2
Like I said - there's not a chance in this world Seagate ES2 drives would outperform even old Raptors much less VRs.
Edit: I jumped the gun abit early, but i leave the rest in anyways...Quote:
Originally Posted by alfaunits
What is very good with ssd (and also raptors abit) is access time. I bet your random access on thoose 8drive raid5 array is like 13-14ms or so(maybe higher, im not exactely shure on how raid5 affects accsess times)? Not the 0.1ms(probably even lower)
of the ssd. Look at vista boot up and shut down time. There is no improvment on boot up with raid over a single drive and no improvment in shut down with more than 2 in raid.
Your sub 4s copy....Wich file(s) did you use?
Also (correct me if im wrong, i only red briefly through the test) it seemes he is limited by his controller card with 9 drives. It doesnt scale nearly as well as the other tests with fewer drives.
I was talking about the Mtron PRO 7500 series, they are more of a response to the Memoright GT and they are a little faster than the Memoright according to their specs.Quote:
Originally Posted by alfaunits
@eXa, RAID should not have any effect on access times, but this can be the case if you have a poor controller.
the Raptor 300 is too thick for most laptops, but some Toshibas and HP's have laptops that are designed for thicker high-capacity drives. In these systems, the raptor 300 will drop right in.
RAID always makes access times poorer, if only slightly. It simply takes more time to get all of the drives coordinated and accessing the data.Quote:
Originally Posted by Helmore
And there always is a little variance in access times across drives and access time will be limited by the slowest drive in the set-up. The RAID-controller should not ad much latency though, not even from coordinating the drives and the extra latency brought in by the controller should be <0,1 ms. That's assuming it's a proper RAID controller and they cost some serious money unfortunately.Quote:
Originally Posted by Speederlander
yah and burn ur legs off when u start up the laptop... read WDs warnings... even if it magically happens to fit into a laptop you dont do it because of heat, vibration, and power usage... maybe the 74gb will be able too, im not sure but if anyone does put it in there laptop... one word.. pics... PICS OR IT DIDNT HAPPEN lolQuote:
Originally Posted by MaxxxRacer
Hoping so. I'm travling to Korea mid-July, BTW ^^Quote:
Originally Posted by Helmore
I'm saying something's up, not that the SSDs are bad, with the copy critic.Quote:
Originally Posted by eXa
Like I said - it seems something not right.Quote:
Also (correct me if im wrong, i only red briefly through the test) it seemes he is limited by his controller card with 9 drives. It doesnt scale nearly as well as the other tests with fewer drives.
The StorageReview numbers are promosing however.
Striped Raid will always add higher latency to access times, but as a trade off for much higher sequential read / write speeds. You can live with 15-20ms latency over 4+ drives, its barely noticeable. Even Raid 0 latency increases as you increase array span. Consider the same way DDR3 has much higher latency over DDR2, its a trade off...as you increase the read / write throughput you need to delay the access to buffer the data, and then stream it. For this extra latency you gain much improved data transfer speeds, and as long as the increase of data transfer speed scales more than the latency for that speed, you are still ahead.Quote:
Originally Posted by Helmore
Raid 5 has quite decent latency for access time, a 8 disk array will generally have around 20-25ms on 15K 3.5" SAS drives. 7200rpm drives fare much worse, even raptors do. Raid 5 has additional overhead of parity calculation, this is where a HW based controller has the advantage, but Raid 5 also has much better random access times over Raid 0, one area it excels highest. No matter how much you spend on a HW Raid Controller you will never see only 0.1ms increase in access time, the laws of science will always apply on platter based hard drives.
The lowest access latency will be found on an array of 74gb 2.5" 15K rpm SAS Seagate Savvio drives, each drive has an individual access latency of 2ms, but that comes at a rough cost of $750-800 USD / drive, and the necessity for a PCI-E 8x minimum SAS HW Raid Controller which will set you back at anywhere from 1.5K - 2K USD minimum. Some of the high end Adaptec and 3Ware HW controllers are just amazing. But even after spending nearly $9K USD on an 8 disk savvio Raid 5 array, you will probably be looking at around 12-18ms latency across the array on average, but with the ability to push over 1200MB/sec read and 900-950MB/sec write at the inner most locations on the platters.
I don't think 15-20ms avg latency will bother you much when you have those kind of sequential read/write speeds, or if you will even notice them.
With Raid its all give and take, you configure an array specifically for the task you need it to do. if you are after balanced performance for small and large files, random access and sequential read/write, you lose alot of the advantages Raid has. It's simply a trade off. A raid array using small stripes and formatted using small clusters will have phenomenal performance handling small files, but will show decreased performance and increased fragmentation when handling large files. Much like using larger stripes and larger cluster blocks will show great performance when doing sequential read / writes and handling large files, but will show much worse behaviour with small files, slower access times, and increase space usage. For a 128Kb stripe with 8Kb clusters, any file less than 8Kb and which cant be aligned to 128Kb, will take up at least 8Kb of space and depending on the blocks of data stored in that stripe, it can take upto 128Kb. But then a 2Gb file will nearly never fragment, and handling it in the OS will be a breeze. This basic understanding of Raid goes understood by most, and its pretty much a compulsory requirement to understand the underlying principles to be able to setup any kind of array with performance close to its maximum. Anybody can throw together an array and make it perform adequately, just don't expect miracles I guess I'm saying.
If any of the figures I've given are a little out I apologize and feel free to correct them, I've tried to recall them as best I remember. Too many facts and too much experience adds to too much to remember sometimes!
Quote:
Originally Posted by Kobalt
:rolleyes:
For real...?
Do you honestly sit there watching an install... or better yet, do you INSTALL the game everyday before you play it?
Your points are extremely insignificant and laughable. SSD's own Raptors, even the cheaper ones.
***********
I reformat often, so write speeds are quite important. And your argument about sitting there and watching an install is, should I say, extremely insignificant and laughable? With your logic we should all just quit upgrading our computers because it doesn't matter how long something takes because we can just do something else during that time.Quote:
Originally Posted by Xoulz
False, NCQ.. go learn about that please.Quote:
Originally Posted by alfaunits
I have found threw trial and error that raids are not the best set-up for a gaming rig. As an early adopter to SCSI and raid0, stemming from the early 90s, when it was actually expensive... understanding how you set up your OS and it's I/O subsystems benefits you more than just going with a raid set-up.
It most cases it is better to have your OS on a separate drive than anything else, then have your virtual memory on a separate drive and your games on a third HD. I have found this this set up is actually quicker/faster than a typical RAID0 set up. It a matter of bottlenecking... not speed. (ie duplexing)
Raid0(stripping) read and write fast, but are also hindered by an inherent problem not found in the above recommendations, I/O bottle necks and queuing.
3 cheap perpendicular drives will load up and play games faster than a Raptor raid due to how and what the software is. I see the results every day because my rig is extremely efficient and smother than my friends identical rig with raptors. Both computers were built by me and have the same mobo's, cases, memory and GPU/CPU.
Basically, my HD's can multitask, while the raid set-up has to rely on sheer speed. unless you go with nested levels. Yes RAID0/stripping is fast, but it cannot compete with parallelism.
With the advent of SSD, you can have both your OS and virtual memory on the same drive with no adverse impact. 2 perp drives are faster than 1 Raptor, but are two raptors worth it, over another SSD...?
In 6~7 months that will almost be moot.
***********
BTW.. good post Mikey!
So... basically when your bored you reformat and re-install your os, games, programs 6..7 times a year.Quote:
Originally Posted by Kobalt
Honestly, I find your argument somewhat childish. Installing COD4 once is enough to enjoy the game. Installing 6 or 7 times a year indicates severe neurotic tendencies. Let alone assume that the time it takes to install them matters. Specially when we are talking about a few minutes difference.
All-the-while knowing the gains of an SSD once the game is installed...? Please, don't insult me!
****************
This is Xtremesystems... we change hardware a lot which often requires reformatting.Quote:
Originally Posted by Xoulz
Post count bumper alert ;) nn_step beware :D
So how about: 32Gb ssd for windows and general programs, i-ram for swap, and a Tb disk for games,vids and music.Quote:
Originally Posted by Xoulz
Pretty ideal without emptying ur pocket, no?
Yeah, I've somewhat pondered this. Windows Vista does make use of NAND type devices, etc. But I should see no reason why a 32GB SSD would suffer with it being also the target of virtual memory. I'd would have to play with both to see how well Vista actually works with NAND.Quote:
Originally Posted by eXa
I just don't see a use for Raptors anymore, specially when MP3 or even games don't require speed. A simple Samsung F1 300MB typical yields as good as performance as an expensive raptor.
Dunno, I would like to build both and see.
***********
Quote:
Originally Posted by Whokidmo
psh... who needs warnings. Its power draw isnt that high. Yes, it will get stupidly hot, but thats why you only use it on a desktop replacement with a cooling unit under it.
And where can you find these F1's?Quote:
Originally Posted by Xoulz
On a race track of course.....
Just search for a Samsung F1 320GB.....
Nearly impossible to find this or the 640 in America as far as the samsung F1 goes. You can however bu the 1TB version here. Main issue with these drives is the terrible quality control. WD makes a 640 gig two platter drive that is just about as good, is cheap, and is easy to find. (WD6400AAKS)Quote:
Originally Posted by Helmore