Overclock.net › Forums › Components › Hard Drives & Storage › RAID Controllers and Software › PERC 5/i RAID Card: Tips and Benchmarks
New Posts  All Forums:Forum Nav:

PERC 5/i RAID Card: Tips and Benchmarks - Page 282

post #2811 of 7150
Quote:
Originally Posted by DuckieHo View Post
The stuttering is due to latency issues caused by small writes. It absolutely is a result of the controller. The better controllers have a few MB of cache to buffer these writes. I am not sure if the PERC is "smart enough" for SSDs. I'm not saying it won't help but I'm not sure optimized to fully utilize SSDs.
If you look at the Anandtech initial review of the Vertex, you will see that the initial Vertex version (one tuned for high sequential reads) stuttered almost as badly as the JMicron based controllers. This was fixed by better firmware, utilising the cache to better cache writes - ie the Vertex controller still stutters if it does not properly buffer writes. Therefore I am convinced it is specifically a cache issue, not a controller problem. I have also read reports of using such SSDs on RAID cards without any stuttering issues.

Now I am not saying that the JMicron based drives didn't suck when used on their own. But I am saying it is not really the controller specifically but more the fact that it had no cache - if you take the cache away from the Indilix (sp?) controller it will stutter just as badly. Add external controller cache and the problem goes away.


Quote:
Originally Posted by DuckieHo View Post
What's wrong with 3x RAID0 especially with some SSDs?
The guy didn't ask about RAIDing SSDs - he asked about Caviar Blacks.


Quote:
Originally Posted by DJZeratul View Post
I disagree. I have my OS drive sitting on a 5-disk RAID 0 with 2 partitions, one MBR for boot, and another GPT for scratch. Yes, the scratch is mainly for sequential writes. The read speed also adds up to the fact that I can transfer files from the scratch disk to my PERC RAID5 at high speed (sustained real world transfers of 350+MB/sec) which is nice for large uncompressed video files. The OS disk is quite fast as well, with faster boot times and game loading times than 2 WD Raptors in RAID0 on an onboard controller.
I have doubts as to whether your RAID0 is really as fast as you think. The increased access times mean RAID is rarely a good choice for OS use. For more examples of this, see here.

You will also hurt performance badly by using the same set of disks in 2 partitions - you force the heads to seek a long way each time you read or write to each partition, thus vastly reducing speeds. You would be better separating your 5 disks out, maybe a 3 disk and a 2 disk set, or even a single drive and a 4 drive array.


Quote:
Originally Posted by DuckieHo View Post
Why would seperated drives be better?
If you use your scratch disk for sequential reads/writes, then RAID0 is absolutely the way to go, but for random reads it is not so good (as disks in general are crappy at this. By using separate disks for Pagefile, you allow Windows to choose the least busy disk for each write, and hence separate disks can perform faster.

(apologies if I have screwed up the quoting here - I am kinda in a hurry & tried to copy & paste things about to make my post more concise)
Edited by the_beast - 7/2/09 at 1:54pm
post #2812 of 7150
Thread Starter 
Quote:
Originally Posted by the_beast View Post
If you look at the Anandtech initial review of the Vertex, you will see that the initial Vertex version (one tuned for high sequential reads) stuttered almost as badly as the JMicron based controllers. This was fixed by better firmware, utilising the cache to better cache writes - ie the Vertex controller still stutters if it does not properly buffer writes. Therefore I am convinced it is specifically a cache issue, not a controller problem.

Now I am not saying that the JMicron based drives didn't suck when used on their own. But I am saying it is not really the controller specifically but more the fact that it had no cache - if you take the cache away from the Indilix (sp?) controller it will stutter just as badly. Add external controller cache and the problem goes away.
Then again... the ver2 of the JMicron controller with higher clock speeds did lower latency issues. There's no doubt the the cache is what helps. However, the controller does have an effect on performance (i.e. Samsung vs Intel vs Indilinx)



Quote:
The guy didn't ask about RAIDing SSDs - he asked about Caviar Blacks.
I was talking in general terms...


Quote:
I have doubts as to whether your RAID0 is really as fast as you think. The increased access times mean RAID is rarely a good choice for OS use. For more examples of this, see here.
I gotta question those benchmarks... Look at the sequential read graphs. They don't look... right. http://www.anandtech.com/storage/sho...spx?i=2969&p=4
Once again...
(13 items)
 
  
CPUMotherboardGraphicsRAM
i7 920 [4.28GHz, HT] Asus P6T + Broadcom NetXtreme II VisionTek HD5850 [900/1200] + Galaxy GT240 2x4GB G.Skill Ripjaw X [1632 MHz] 
Hard DriveOSMonitorKeyboard
Intel X25-M 160GB + 3xRAID0 500GB 7200.12 Window 7 Pro 64 Acer H243H + Samsung 226BW XARMOR-U9BL  
PowerCaseMouseMouse Pad
Antec Truepower New 750W Li Lian PC-V2100 [10x120mm fans] Logitech G9 X-Trac Pro 
  hide details  
Reply
Once again...
(13 items)
 
  
CPUMotherboardGraphicsRAM
i7 920 [4.28GHz, HT] Asus P6T + Broadcom NetXtreme II VisionTek HD5850 [900/1200] + Galaxy GT240 2x4GB G.Skill Ripjaw X [1632 MHz] 
Hard DriveOSMonitorKeyboard
Intel X25-M 160GB + 3xRAID0 500GB 7200.12 Window 7 Pro 64 Acer H243H + Samsung 226BW XARMOR-U9BL  
PowerCaseMouseMouse Pad
Antec Truepower New 750W Li Lian PC-V2100 [10x120mm fans] Logitech G9 X-Trac Pro 
  hide details  
Reply
post #2813 of 7150
Quote:
Originally Posted by the_beast View Post
I have doubts as to whether your RAID0 is really as fast as you think. The increased access times mean RAID is rarely a good choice for OS use. For more examples of this, see here.

You will also hurt performance badly by using the same set of disks in 2 partitions - you force the heads to seek a long way each time you read or write to each partition, thus vastly reducing speeds. You would be better separating your 5 disks out, maybe a 3 disk and a 2 disk set, or even a single drive and a 4 drive array.
While the random access times increased slightly, it did not hurt performance at all. In fact it helped a lot.

Stopwatch test from POST to windows fully loaded on a clean install of Vista, on 2 Raptors in RAID0 on the onboard ICH10R controller took 45 seconds. On the 5 Disk 7200.10 RAID0 the time decreased to 22 seconds (also on the onboard ICH10R).

It is definitely faster than a single disk.
    
CPUMotherboardGraphicsRAM
Core i7 970 ASUS P6T Deluxe eVGA GTX 680 6x4GB Kingston Hyper X DDR3-1600 
Hard DriveOptical DriveOSMonitor
Boot R0: 2x256GB Corsair SSD; Scratch R0: 3x750GB Samsung 22x LightScribe DVD+/-DL [SATA] Windows 7 x64 Ultimate Dell 2407FPWx2 
KeyboardPowerCaseMouse
Microsoft Sidewinder X6 Antec TruPower Q 1KW Custom Fabricated 18" cube (Mountain Mods UFO) Logitech G9 Laser 
Mouse Pad
Faux Woodgrain (My Desk) 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
Core i7 970 ASUS P6T Deluxe eVGA GTX 680 6x4GB Kingston Hyper X DDR3-1600 
Hard DriveOptical DriveOSMonitor
Boot R0: 2x256GB Corsair SSD; Scratch R0: 3x750GB Samsung 22x LightScribe DVD+/-DL [SATA] Windows 7 x64 Ultimate Dell 2407FPWx2 
KeyboardPowerCaseMouse
Microsoft Sidewinder X6 Antec TruPower Q 1KW Custom Fabricated 18" cube (Mountain Mods UFO) Logitech G9 Laser 
Mouse Pad
Faux Woodgrain (My Desk) 
  hide details  
Reply
post #2814 of 7150
Quote:
Originally Posted by the_beast View Post
I cannot think of a single good reason why you would want 3 or more disks in RAID0 in a home environment. The sequential transfer numbers will look great, but this will not transfer well into real-world use. The only possible use might be as a work/scratch area for photo or video editing, where fast sequential transfers are required. For OS use though it would be disappointing - even if you wanted a fast area to store your pagefile you would likely get better performance from separate drives.

What would you want to use such a system for?
I decided against the idea of RAID 0 and, when I get the funds together I will get a raid 5 setup.
Sandy
(13 items)
 
  
CPUMotherboardGraphicsRAM
I7 2600k @ 4.5Ghz Asus P8P67 Evo PNY GTX 570 8GB G.Skill Ripjaws 
Hard DriveOptical DriveOSMonitor
3x500GB Perc5/i Raid 5 Blu-Ray Windows 7 x64 Samsumg 22" 
KeyboardPowerCaseMouse
Razer BlackWidow Corsair 750W CM690 Razer Death Adder 
  hide details  
Reply
Sandy
(13 items)
 
  
CPUMotherboardGraphicsRAM
I7 2600k @ 4.5Ghz Asus P8P67 Evo PNY GTX 570 8GB G.Skill Ripjaws 
Hard DriveOptical DriveOSMonitor
3x500GB Perc5/i Raid 5 Blu-Ray Windows 7 x64 Samsumg 22" 
KeyboardPowerCaseMouse
Razer BlackWidow Corsair 750W CM690 Razer Death Adder 
  hide details  
Reply
post #2815 of 7150
Anyone got the nailpolish guide/pictures for this card.
Thank you
post #2816 of 7150
Quote:
Originally Posted by elec999 View Post
Anyone got the nailpolish guide/pictures for this card.
Thank you
I believe its in the OP
Quote:
Originally Posted by DuckieHo View Post
PowerEdge Expandable RAID Controller (PERC) 5/i
================================================== ===========================
SMBus Issue with Intel Chipsets

These cards are known to have some compatibility issues with Intel chipsets. However, they are known to work with NVIDIA motherboards fine. The issue stems from a System Management Bus (SMBus) conflicting with the motherboard's memory detection. SMBus is simple signal to provide the motherboad some basic device information and control. Symptoms of the conflict includes improperly reported RAM sizes and POST errors.

The trick is just to physically disable the SMBus signal. It is composed of just two pins B5 (SMCLK, SMBus clock) and B6 (SMDAT, SMBus data). These two pins need to be covered by tape or nail polish. On the top side of the card, they are the 5th and 6th PCIe pins from the left. You can see the pins covered as seen below:


Edited by samstaee - 7/2/09 at 4:35pm
Sandy
(13 items)
 
  
CPUMotherboardGraphicsRAM
I7 2600k @ 4.5Ghz Asus P8P67 Evo PNY GTX 570 8GB G.Skill Ripjaws 
Hard DriveOptical DriveOSMonitor
3x500GB Perc5/i Raid 5 Blu-Ray Windows 7 x64 Samsumg 22" 
KeyboardPowerCaseMouse
Razer BlackWidow Corsair 750W CM690 Razer Death Adder 
  hide details  
Reply
Sandy
(13 items)
 
  
CPUMotherboardGraphicsRAM
I7 2600k @ 4.5Ghz Asus P8P67 Evo PNY GTX 570 8GB G.Skill Ripjaws 
Hard DriveOptical DriveOSMonitor
3x500GB Perc5/i Raid 5 Blu-Ray Windows 7 x64 Samsumg 22" 
KeyboardPowerCaseMouse
Razer BlackWidow Corsair 750W CM690 Razer Death Adder 
  hide details  
Reply
post #2817 of 7150
Quote:
Originally Posted by samstaee View Post
I believe its in the OP
Hope it works with Perc6i
Thank you
post #2818 of 7150
why use this nail ***** and stuff.....just put a small piece of regular tape on it...worked liek a charm for me, it's easily accessible in the house and does not damage the card...

Also, there seems to be a new version of green series 1tb caviars that is 2-platter, so 500gb per platter, but the name is exactly the same as the 3platter one, you need to look at other stuff like the release date...
Seems interesting, I'd liek to know if it runs all the time at 5400rpm or the speed is variable...
post #2819 of 7150
Quote:
Originally Posted by Blinky7 View Post
why use this nail ***** and stuff.....just put a small piece of regular tape on it...worked liek a charm for me, it's easily accessible in the house and does not damage the card...

Also, there seems to be a new version of green series 1tb caviars that is 2-platter, so 500gb per platter, but the name is exactly the same as the 3platter one, you need to look at other stuff like the release date...
Seems interesting, I'd liek to know if it runs all the time at 5400rpm or the speed is variable...
Nail polish doesn't damage the card, you can remove it with some acetone on a q-tip. Its more permanent than tape and if you remove/replace the card a couple times you will find that the tape adhesive heats up and leaves the tape behind, can be an annoyance.

The new Caviar Greens run at variable speeds using IntelliPower, the previous generation were a fixed speed set at random by WD and could have been anywhere from 5400-7200RPM.
    
CPUMotherboardGraphicsRAM
Core i7 970 ASUS P6T Deluxe eVGA GTX 680 6x4GB Kingston Hyper X DDR3-1600 
Hard DriveOptical DriveOSMonitor
Boot R0: 2x256GB Corsair SSD; Scratch R0: 3x750GB Samsung 22x LightScribe DVD+/-DL [SATA] Windows 7 x64 Ultimate Dell 2407FPWx2 
KeyboardPowerCaseMouse
Microsoft Sidewinder X6 Antec TruPower Q 1KW Custom Fabricated 18" cube (Mountain Mods UFO) Logitech G9 Laser 
Mouse Pad
Faux Woodgrain (My Desk) 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
Core i7 970 ASUS P6T Deluxe eVGA GTX 680 6x4GB Kingston Hyper X DDR3-1600 
Hard DriveOptical DriveOSMonitor
Boot R0: 2x256GB Corsair SSD; Scratch R0: 3x750GB Samsung 22x LightScribe DVD+/-DL [SATA] Windows 7 x64 Ultimate Dell 2407FPWx2 
KeyboardPowerCaseMouse
Microsoft Sidewinder X6 Antec TruPower Q 1KW Custom Fabricated 18" cube (Mountain Mods UFO) Logitech G9 Laser 
Mouse Pad
Faux Woodgrain (My Desk) 
  hide details  
Reply
post #2820 of 7150
Quote:
Originally Posted by DuckieHo View Post
Then again... the ver2 of the JMicron controller with higher clock speeds did lower latency issues. There's no doubt the the cache is what helps. However, the controller does have an effect on performance (i.e. Samsung vs Intel vs Indilinx)
I have not heard of a higher clockspeed JMicron controller. It was my understanding that the second-gen JMicron controllers were actually 2 first-gen controllers bolted together with a cacheless RAID0 controller. As each controller was basically separate, the system hangs were bascially cut in half as the controller could address each half of the disk separately, thus needing half the effective number of writes per bank of flash.

(although I think we may be a little off-topic with this.)

Quote:
Originally Posted by DuckieHo View Post
I gotta question those benchmarks... Look at the sequential read graphs. They don't look... right. http://www.anandtech.com/storage/sho...spx?i=2969&p=4
There is something definately off with some of the graphs, I will admit. But the resulting measured results still kinda speak for themselves.


Quote:
Originally Posted by DJZeratul View Post
While the random access times increased slightly, it did not hurt performance at all. In fact it helped a lot.

Stopwatch test from POST to windows fully loaded on a clean install of Vista, on 2 Raptors in RAID0 on the onboard ICH10R controller took 45 seconds. On the 5 Disk 7200.10 RAID0 the time decreased to 22 seconds (also on the onboard ICH10R).

It is definitely faster than a single disk.
Vista must be much more heavily optimised for sequential reads at sartup than XP. Every time I have set up RAID0 system disk the startup times have been with 10% of each other. So I gave up with it. I have not tried Vista however - if your times are acurate then that is a healthy boost. Does the same apply to other apps? I find it somewhat hard to believe that every test I have seen actually comparing RAID0 to single disks in real-world tests (rather than fully synthetic benchmarks) conclude that performance gains are negligible, yet you report a two-fold speed increase. You figures are impressive nontheless.

Quote:
Originally Posted by samstaee View Post
I decided against the idea of RAID 0 and, when I get the funds together I will get a raid 5 setup.
You would be better off with RAID0 for OS use - RAID5 is not really suited to OS use either (it will suffer from the same problems as RAID0 as far as access times, but also has poor small random write performance - exactly the kinds of writes an OS makes regularly). The effect of the cache does work to mitigate this, but I would still recommend a single smaller OS disk (short-stroked HDD or SSD) and then a RAID5/6 storage array. If you need extra space for your apps or games and cannot afford a larger SSD then an extra 2-disk RAID0 array will be faster than a single disk for some games and roughly equal for others.

This does add up to a lot of disks though. I believe the optimal setup for most users buying disks now would be either a large enough SSD (prob 60/120GB) or single 1TB Caviar Black for OS (but not data - keep that off the main disk). Then have a separate RAID5/6 either locally or on a networked server to store your data.
Edited by the_beast - 7/3/09 at 2:00am
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: RAID Controllers and Software
Overclock.net › Forums › Components › Hard Drives & Storage › RAID Controllers and Software › PERC 5/i RAID Card: Tips and Benchmarks