Ad
Ad
Ad
Pages: [1] 2 »   Bottom of Page
Print
Author Topic: From 0 to Win 7 in minus 10 seconds...  (Read 4856 times)
BertramPaul
Newbie
*
Offline Offline

Posts: 30



WWW
« on: March 19, 2010, 06:34:38 PM »
ReplyReply

Great stuff, those SSD's. Too bad they're still so expensive...

Here a small impression how I turbo, no, supercharged my system:

http://bertram-paul.com/RAID/

If you have any questions, I'll be happy to answer.
Logged

εΓΓ♍ P└
feppe
Sr. Member
****
Offline Offline

Posts: 2907

Oh this shows up in here!


WWW
« Reply #1 on: March 19, 2010, 06:44:41 PM »
ReplyReply

Quote from: BertramPaul
Great stuff, those SSD's. Too bad they're still so expensive...

Here a small impression how I turbo, no, supercharged my system:

http://bertram-paul.com/RAID/

If you have any questions, I'll be happy to answer.

Nice setup. I got a OCZ Vertex recently and am very happy with the speed and (lack of) noise as well.

If you have the Vertex, make sure you're running the 1.5 firmware which adds TRIM. Please note that you don't get TRIM with RAID, though - so it might be worthwhile to run benchmarks occasionally to see if you get degradation of performance.
Logged

BertramPaul
Newbie
*
Offline Offline

Posts: 30



WWW
« Reply #2 on: March 19, 2010, 07:22:54 PM »
ReplyReply

Quote from: feppe
Nice setup. I got a OCZ Vertex recently and am very happy with the speed and (lack of) noise as well.

If you have the Vertex, make sure you're running the 1.5 firmware which adds TRIM. Please note that you don't get TRIM with RAID, though - so it might be worthwhile to run benchmarks occasionally to see if you get degradation of performance.

Mine are the Vertex Turbo. They're slightly faster. And yes, I needed to update the firmware. That was funny; when you update firmware for a camera or you BIOS, you can almost have a cup of coffee. But with these drives.... I couldn't even blink my eyes and it was done!

As I understand it, the TRIM should work, by letting the PC idle by logging off. Do you know any more about this?
Logged

εΓΓ♍ P└
feppe
Sr. Member
****
Offline Offline

Posts: 2907

Oh this shows up in here!


WWW
« Reply #3 on: March 19, 2010, 07:31:03 PM »
ReplyReply

Quote from: BertramPaul
As I understand it, the TRIM should work, by letting the PC idle by logging off. Do you know any more about this?

That's correct - OCZ guide says several hours each week which is a lot. I don't plan on idling at all unless benchmarks show marked deterioration.

TRIM doesn't work with RAID, period, and won't in the foreseeable future.

I'm skeptical about the claimed performance reduction over time, but nevertheless I'd keep an eye on benchmarks occasionally.
Logged

Christopher
Sr. Member
****
Offline Offline

Posts: 944


WWW
« Reply #4 on: March 19, 2010, 08:28:59 PM »
ReplyReply

Quote from: feppe
That's correct - OCZ guide says several hours each week which is a lot. I don't plan on idling at all unless benchmarks show marked deterioration.

TRIM doesn't work with RAID, period, and won't in the foreseeable future.

I'm skeptical about the claimed performance reduction over time, but nevertheless I'd keep an eye on benchmarks occasionally.


Well people won't believe it until they see it with their own eyes. Just do some real benchmarks from time to time. ( NOT only throughput, which is probably the least important benchmark. Yes it is great to see big numbers like 500mb/s, but smaller write and read speed is much more important. 4kb for example. Here one will see quite a large loss over time from SSDs (MLC) )

As a few have said before, you can keep your computer in idle for 100 hours it won't change a thing.


RAID = NO TRIM = Performance loss over time --> the only fix there is, is to reformat them completly.
Logged

feppe
Sr. Member
****
Offline Offline

Posts: 2907

Oh this shows up in here!


WWW
« Reply #5 on: March 19, 2010, 08:40:29 PM »
ReplyReply

Quote from: Christopher
Well people won't believe it until they see it with their own eyes.

Indeed. But there's so much FUD about SSDs that I've become pretty skeptical about any such claims - the most prominent being the limited number of read/write cycles which is a non-issue for 99.9999% people, including swap usage.

I don't run mine in RAID, but wanted to warn about potential performance degradation with RAID setups due to lack of TRIM.
Logged

BertramPaul
Newbie
*
Offline Offline

Posts: 30



WWW
« Reply #6 on: March 20, 2010, 05:08:37 PM »
ReplyReply

Quote from: feppe
Indeed. But there's so much FUD about SSDs that I've become pretty skeptical about any such claims - the most prominent being the limited number of read/write cycles which is a non-issue for 99.9999% people, including swap usage.

I don't run mine in RAID, but wanted to warn about potential performance degradation with RAID setups due to lack of TRIM.

If you don't run your in RAID, then how do you know?
Intel just came out with Rapid Stor version 9.6 with TRIM support for RAID 0 (well, RAID...)
This is what I got: about an increase of 20%:

[attachment=20980:Capture.JPG]

No HD can get these numbers, especially not the lower 4.
And with a MTBF of 1.5 million hours.... I prefer speed over the "security" of a HD.
Logged

εΓΓ♍ P└
feppe
Sr. Member
****
Offline Offline

Posts: 2907

Oh this shows up in here!


WWW
« Reply #7 on: March 20, 2010, 05:24:14 PM »
ReplyReply

Quote from: BertramPaul
If you don't run your in RAID, then how do you know?
Intel just came out with Rapid Stor version 9.6 with TRIM support for RAID 0 (well, RAID...)
This is what I got: about an increase of 20%:

[attachment=20980:Capture.JPG]

No HD can get these numbers, especially not the lower 4.
And with a MTBF of 1.5 million hours.... I prefer speed over the "security" of a HD.

No reason to get defensive, we're on the same boat. Just because I don't run RAID doesn't mean I don't know about its features.

At the time I wrote my previous post SSDs didn't have TRIM with RAID. Quick googling shows it was released hours ago, and there's no confirmation it does what it says on the tin, though. Glad to see it's out as that is one of the last shortcomings of SSDs in addition to price.

No need to show benchmarks, as should be apparent from my earlier posts I'm a huge advocate of SSDs and am fully aware of their speed and very skeptical about their claimed shortcomings which are mainly FUD.
« Last Edit: March 20, 2010, 05:36:54 PM by feppe » Logged

BertramPaul
Newbie
*
Offline Offline

Posts: 30



WWW
« Reply #8 on: March 21, 2010, 10:16:22 AM »
ReplyReply

Quote from: feppe
No reason to get defensive, we're on the same boat. Just because I don't run RAID doesn't mean I don't know about its features.

At the time I wrote my previous post SSDs didn't have TRIM with RAID. Quick googling shows it was released hours ago, and there's no confirmation it does what it says on the tin, though. Glad to see it's out as that is one of the last shortcomings of SSDs in addition to price.

No need to show benchmarks, as should be apparent from my earlier posts I'm a huge advocate of SSDs and am fully aware of their speed and very skeptical about their claimed shortcomings which are mainly FUD.

Not really defensive; just wondering how you know. I'd like to get my facts from things I have experience with. People talk a lot  

Now I got to figure out if TRIM is working and how to check that....
Logged

εΓΓ♍ P└
Jack Flesher
Sr. Member
****
Offline Offline

Posts: 2595



WWW
« Reply #9 on: March 22, 2010, 10:47:41 AM »
ReplyReply

Quote from: feppe
Indeed. But there's so much FUD about SSDs that I've become pretty skeptical about any such claims - the most prominent being the limited number of read/write cycles which is a non-issue for 99.9999% people, including swap usage.

Uh, sorry to burst your bubble, but swap is precisely the type of disk I/O that causes the clogs on SSD.  So before you make such claim, I suggest you do as suggested earlier --- run a RAID, then run a benchmark on it when new, then run the same benchmark say 4 months later after a lot of heavy CS processing that actually tagged scratch.  Then share those results along with your claim that swap is a non-issue...
Logged

feppe
Sr. Member
****
Offline Offline

Posts: 2907

Oh this shows up in here!


WWW
« Reply #10 on: March 22, 2010, 01:47:02 PM »
ReplyReply

Quote from: Jack Flesher
Uh, sorry to burst your bubble, but swap is precisely the type of disk I/O that causes the clogs on SSD.  So before you make such claim, I suggest you do as suggested earlier --- run a RAID, then run a benchmark on it when new, then run the same benchmark say 4 months later after a lot of heavy CS processing that actually tagged scratch.  Then share those results along with your claim that swap is a non-issue...

You misread my post. I wasn't referring to clogging up the SSD, but to the limited read/write cycles of them.

And it was me who suggested running benchmarks to verify the FUD. I won't be running them as (again) I don't have RAID.
Logged

John.Murray
Sr. Member
****
Offline Offline

Posts: 893



WWW
« Reply #11 on: March 23, 2010, 12:44:16 PM »
ReplyReply

A nice article from the Windows Engineering Blog:

http://blogs.msdn.com/e7/archive/2009/05/0...drives-and.aspx
Logged

Steve Weldon
Sr. Member
****
Offline Offline

Posts: 1441



WWW
« Reply #12 on: March 28, 2010, 11:24:06 PM »
ReplyReply

Quote from: BertramPaul
If you have any questions, I'll be happy to answer.
Hi -  This is my first post in this forum so be gentle..

I carefully looked over your link and everything you posted and have some questions that I hope will help me sort out some questions about my own system.

You're showing a seq read speed of over 400mbps..  What is the cap on bandwidth of your RAID card and what RAID card are you using?  If you're simply using the motherboards RAID chip and with SATA II ports then maybe this answers my questions below.

The 400mpbs speeds give me pause after seeing your Win7 Index score of 7.7.  My single SSD (Crucial C300) running through a 6g/ps SATA III port is benchmarking at an average of 354mbps with peaks of up to 390.. and is scoring 7.9 on the Win7 index score.

I've always wondered what the Windows Index is actually measuring in each area, and in this case there appears to be something wonky in the Index if you're RAID is truly pushing over 400mbps seq reads and getting 7.7, and mine is pushing significantly less and getting 7.9.

Also kind of curious if mine is 'just' 7.9, or some score over the max it displays.  I suppose we'll have to wait for Win8 to find out.. ;o)


About some other posts in the thread.  Be careful with running  benchmark utilities too often.  The rapid rate of reads/writes is exactly the sort of activity that slows down SSD's over time, but benchmark utilities will accelerate this process faster than the wear leveling TRIM can keep up.  (from the mouths of Crucial and Intel tech support).  Which also makes me hesitant to enable my page files on the SSD.. which I'm not sure are necessary any longer in a system with 10gb's+ of RAM and a SSD..

Overall I'm very happy with my C300, but I caution potential adopters to not expect too much.  After all, any decent SATA II HDD is at the 120mbps mark, many of the entry level SSD's are barely reaching the same speeds.  The top SSD's are only 3-4 times faster.. so a program that takes 2 seconds to load now takes .5 secs.. you've gotta ask yourself if its worth the cost.  Older SATA and SATA II drives (those purchases over 2 years ago) are in the 40-50mbps range.. so even a simple HDD upgrade would give bargain performance gains.
Logged

----------------------------------------------
http://www.BangkokImages.com
Christopher
Sr. Member
****
Offline Offline

Posts: 944


WWW
« Reply #13 on: March 29, 2010, 03:56:12 AM »
ReplyReply


Well most people just have a Sata II interface only very new boards actually have SATA III, the same goes for RAID cards. The only ones I know of are SAS6 controller, which can deal with SATA3 SSDs. Windows Maxes out at 7.9, so I'm not really sure how helpful it really is.

I think people make one big mistake, they think that seq read and write speed are important. Well they aren't really, or at least not as much as people think. For a larger Workstation I don't see any sense in a SSD RAID. I would say a 120GB SSD as system drive and another one for image temp files, is great, but for everything else HDs are still better. (Sure you can use SSDs if you can spend 1000s $ for drives)

I mean if you only need 100GB for all your images you could get away witth a SSD raid, but if I take myself for example I need at least 2 TB. Which is way to expensive to achive with SSDs.

Another thing are scratch disks. I mean yes you can use RAIDs for that, BUT I mean my current scratch disk has 5x500GB HDs which give me nice  seq read and write speed of 550mbps and peaks up to 650mbps. Now sure I don't have the access times as a SSD has, but it has been shown often enough that PS doesn't care to much about access time of it's scratch disk. The biggest point is that I payed for all 5 drives together 170EUR.
Logged

AndreasSchmidt
Newbie
*
Offline Offline

Posts: 19


WWW
« Reply #14 on: March 29, 2010, 04:58:16 AM »
ReplyReply

As I plan to build a new computer within next weeks: what's about data on SSD when upgrading the firmware? I think to remember that I've read that all data is lost then - OK, I always have backup or image, but it's additional work... So any info about this is welcome.

Andreas
Logged
feppe
Sr. Member
****
Offline Offline

Posts: 2907

Oh this shows up in here!


WWW
« Reply #15 on: March 29, 2010, 05:06:17 AM »
ReplyReply

Quote from: AndreasSchmidt
As I plan to build a new computer within next weeks: what's about data on SSD when upgrading the firmware? I think to remember that I've read that all data is lost then - OK, I always have backup or image, but it's additional work... So any info about this is welcome.

Andreas

OCZ Vertex 1.3 (latest) firmware update doesn't lose data don't know about other manufacturers or versions. But as always with fw updates you should have backups.
Logged

Steve Weldon
Sr. Member
****
Offline Offline

Posts: 1441



WWW
« Reply #16 on: March 29, 2010, 06:46:37 AM »
ReplyReply

Quote from: Christopher
Well most people just have a Sata II interface only very new boards actually have SATA III,

SATA III is a minimal investment.. The Asus U3S6 SATA III/USB 3.0 board is going for about $30.. but its probably more important to ensure you have the PCIe bandwidth available in your motherboard (PCIe v2) before hoping for too much.  I've heard from more than a few disappointed people who are finding out their PCIe bus can't handle speeds above the 250mbps limit of their PCIe v1 motherboards.

And sure, cost per gigabytle HD's are better.. no argument there.  But "better" entails more than cost.  Performance wise SSD's win in most cases no matter what the use.  It just doesn't make good fiscal sense to use SSD's for any sort of data storage.  Working files sure, storage no.

And I'd agree there are other numbers to consider than read/write.. but that hardly makes R/W speeds unimportant.  But its a good point to consider all the numbers.  Though as a general rule with SSD's, the other numbers pretty much go up/down with the read/write speeds since all the numbers depends on the quality of the NAND chips being used, and the controller..  HDD's have more variables since they're mechanical devices.

Scratch disks I'm not worried about.  With RAM so inexpensive these days it's cost effective to have enough RAM so a scratch disk isn't an issue.

Logged

----------------------------------------------
http://www.BangkokImages.com
Jack Flesher
Sr. Member
****
Offline Offline

Posts: 2595



WWW
« Reply #17 on: March 29, 2010, 09:36:36 AM »
ReplyReply

Quote from: SteveWeldon
Scratch disks I'm not worried about.  With RAM so inexpensive these days it's cost effective to have enough RAM so a scratch disk isn't an issue.

Except with very large files even on systems with lots of RAM, some apps are still tagging scratch, even if just to reserve it. Agreed though, with better code utilizing onboard RAM going forward, scratch should be required less often...
« Last Edit: March 29, 2010, 09:37:40 AM by Jack Flesher » Logged

Steve Weldon
Sr. Member
****
Offline Offline

Posts: 1441



WWW
« Reply #18 on: April 01, 2010, 06:35:13 AM »
ReplyReply

I'm trying to get to the bottom of some low benchmarks.  This is somewhat related to this thread because the speeds Paul posted are raising even more questions in my mind.  More specifically the 4k write speeds seem to be very low compared to the seq read/write.


I ran some benchmarks on my C300 the benchmarks Crucial tech support recommended as the most accurate, AS SSD and ATTO.  My 4k writes more than twice what Paul posted, yet three times less than Crucial figures this drive should be performing.  And at the same time my seq reads/writes are less than Paul's RAID.. but my Win7 Performance Index is higher.  Lots of questions why.

Crucial tech support recommended I check two things.

1.  My partition alignment, which can be an issue if you copied a mirrored image to the drive like I did.  After checking my partition is aligned perfectly.  No worries there.

2.  I'm over loading my PCIe bus.  This is a bit more complicated.


I'm running two ATI 5770 cards in a Asus Deluxe V2 MB with a x58 chipset, and my C300 is connected via a U3S6 daughterboard that provides SATAIII/USB3 ports.  The U3S6 is slotted in an available x4 PCIe slot, but not the third x16 PCIe slot because the way my MB is mounted in the case it's not accessible.  The U3S6 card is electrically a x4.. so it should be fine unless there is something specific to this MB where it requires cards to be mounted in a certain order for best performance.  The PCIe bus is V2.. and it's using the x58 chipset, so it should handle the most bandwidth of anything out there.  If anyone knows this not to be true, please tell me.. MB's are relatively cheap and I don't mind upgrading to get the best performance.

So.. could the problem be I'm saturating my PCIe buss because I have the two 5770 cards?  Even when during the benchmarks they're just barely idling?  This is what Crucial says might be happening.  Doesn't sound right to me.

I've checked, uninstalled, and reinstalled the latest drivers related to the C300, U3S6, and 5770's.

Anyone have any ideas?  Online reviews using these same benchmark tools are showing seq reads of 370, seq writes of 190, and the 4k's at near 280 read and 190 writes.  I posted screen shots of my benchmarks.

Any help would be appreciated.

Logged

----------------------------------------------
http://www.BangkokImages.com
Steve Weldon
Sr. Member
****
Offline Offline

Posts: 1441



WWW
« Reply #19 on: April 02, 2010, 02:19:31 AM »
ReplyReply

Quote from: Steve Weldon
The U3S6 is slotted in an available x4 PCIe slot, but not the third x16 PCIe slot because the way my MB is mounted in the case it's not accessible.  The U3S6 card is electrically a x4.. so it should be fine unless there is something specific to this MB where it requires cards to be mounted in a certain order for best performance.  T
After corresponding with Asus, Crucial, and some further forum browsing, all seem to be in agreement there is indeed an issue with using the available PCIe x4 slot even though the U3S6 SATAIII/USB3 card is a PCIe x4 card.  No one can really explain why, but the best guess is that its just the priority given the slots on the Southbridge.

Don't get me wrong, the C300 is anything but slow as is.  The performance difference over my 300g Raptor is huge.  It just seems to be taking an overall 20-30% performance hit from what it should be if everything is optimum.  

Anyway, the only way I can access my third PCIe x16 slot is to relocate the 3 top fans on my Lian Li case to the outside which frankly isn't very attractive, or to get a new case.  Since I've been meaning to pick up a server case to build a server with.. I'll just get a nice mid-tower case for my main system which will allow use of the PCIe x16 slot.. and use the server tower its replacing for my soon to be built server.  Unfortunately here in Bangkok this will probably require a 3-4 week wait..

Logged

----------------------------------------------
http://www.BangkokImages.com
Pages: [1] 2 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad