PDA

View Full Version : AMD SB750 vs PERC 5/i


SuperMiguel
07-28-09, 06:15 PM
Hardware Used
Biostar TA790GX (http://www.newegg.com/Product/Product.aspx?Item=N82E16813138130) board (AMD SB750)
PERC 5/i RAID Card (256MB)
3 WD5000AAKS (7200 RPM) (SATA)


AMD SB750 Raid-0 test: (Stripe Block 64 KB) (Using 3 WD5000AAKS)

http://img513.imageshack.us/img513/3393/raid0hdtach.jpg
http://img252.imageshack.us/img252/1941/raid0hdtune.png

PERC 5/i Raid-0 test: (Stripe Block 64 KB) (Using 3 WD5000AAKS)

http://img168.imageshack.us/img168/8082/raid0.jpg
http://img249.imageshack.us/img249/8562/hdtunebenchmarkdellperc.png

AMD SB750 Raid-5 test: (Stripe Block 64 KB) (Using 3 WD5000AAKS)

http://img35.imageshack.us/img35/9998/raid1hdtach.jpg
http://img253.imageshack.us/img253/1518/raid1hdtune.png

PERC 5/i Raid-5 test: (Stripe Block 64 KB) (Using 3 WD5000AAKS)

http://img168.imageshack.us/img168/7242/raid0n.jpg
http://img228.imageshack.us/img228/8562/hdtunebenchmarkdellperc.png

SuperMiguel
07-30-09, 10:36 AM
Hardware Used
Biostar TA790GX (http://www.newegg.com/Product/Product.aspx?Item=N82E16813138130) board (AMD SB750)
PERC 5/i RAID Card (256MB)
2 ST936751SS (15K RPM) (SAS)

PERC 5/i Raid-0 test: (Stripe Block 64 KB) (Using 2 ST936751SS)

SuperMiguel
08-18-09, 02:02 PM
btw the settings on the perc, where set to write-back..

visbits
08-18-09, 02:11 PM
did you try enable read advance?

Neuromancer
08-18-09, 02:15 PM
Wow. I would have thought the Perc would dominate the SB750. AFAIK SB750 is one of the worst raid chipsets available...

Badbonji
08-18-09, 02:20 PM
Wow, I always thought hardware raid was a rip-off unless you needed lots of drives and more reliability...

bchur83
08-18-09, 02:30 PM
Do you know by any chance what PCIe speed the perc was running? I had similar results on my DFI NF4 Ultra-D when I had my Perc in the 2nd 16x slot vs having it in the 1st slot. In the 1st slot I have around 300MB/s with my 5x 640GB WD Blue array. In the 2nd slot I had around 100MB/s. This was because the 2nd slot defaulted to 1x even though it was a physical 16x slot. I see your mobo states 2-16x PCIe slots, but only 2-8x in CF. So are they running 8x all the time or is there a switch to make them both 8x? It almost looks like it is running in 1x mode.

SuperMiguel
08-18-09, 02:32 PM
did you try enable read advance?

did that got this:

http://img517.imageshack.us/img517/8562/hdtunebenchmarkdellperc.png

SuperMiguel
08-18-09, 02:32 PM
Do you know by any chance what PCIe speed the perc was running? I had similar results on my DFI NF4 Ultra-D when I had my Perc in the 2nd 16x slot vs having it in the 1st slot. In the 1st slot I have around 300MB/s with my 5x 640GB WD Blue array. In the 2nd slot I had around 100MB/s. This was because the 2nd slot defaulted to 1x even though it was a physical 16x slot. I see your mobo states 2-16x PCIe slots, but only 2-8x in CF. So are they running 8x all the time or is there a switch to make them both 8x? It almost looks like it is running in 1x mode.


any idea on how to check?

ChanceCoats123
08-18-09, 02:41 PM
Gpu-z and cpu-z will tell you the link speed of your graphics cards so if that shows as x8 you are probably in x8 in both slots.

SuperMiguel
08-18-09, 02:44 PM
well changed perc card to my primary pci-e (using integrated video)

SuperMiguel
08-18-09, 02:50 PM
changed card to pci-e 1 and got:

http://img146.imageshack.us/img146/6157/hdtunebenchmarkdellpercw.png

thideras
08-18-09, 03:01 PM
Try more settings for the Perc 5; different read aheads, etc. This is a pretty unfair comparison. You need to do read/write tests otherwise that card is barely being stressed.

Something also is off, why is the Perc 5 using 20% of your CPU?

bchur83
08-18-09, 03:05 PM
Yeah, something doesnt look right with those results. The 1st 70% of the graph looks like the controller is being link limited as in like its running in 1x or 2x. There should be a constant dropping line across the board, not a flat line then a drop on the last 30%. I would look into some of the other settings for the read ahead and other read settings.

Neuromancer
08-18-09, 03:27 PM
1x mode should not matter, as that is still 500MB/s

Theoretically anyway...

did you make raid array in windows or after POST? (Ctrl+J or whatever the KB shortcut is to get into perc utility) That could make a differnce (shouldnt) but might help explain the high CPU usage (as in its actually a windows software raid that was set not a hw perc raid)

thideras
08-18-09, 03:32 PM
1x mode should not matter, as that is still 500MB/s1x is 250mb/sec (http://en.wikipedia.org/wiki/Pci_express), which is exactly in line with the results.

visbits
08-18-09, 04:39 PM
Capacity Per lane:

* v1.x: 250 MB/s
* v2.0: 500 MB/s
* v3.0: 1 GB/s

16 lane slot:

* v1.x: 4 GB/s
* v2.0: 8 GB/s
* v3.0: 16 GB/s

My understanding of this is that the x16 socket supports 4GB/s at 1x or is it that pci-e 1.0 in x16 slot supports 4gb/s?

Also interested in this because my array speed isn't consistent.

http://hostthenpost.org/uploads/dd3d02d281a075709769670c74cde5d4.jpg

madhatter256
08-18-09, 06:13 PM
i would suggest running a RAID 5 rebuild process on both chips, and also have both PCs' CPUs stressed via Prime95/Orthos while it is rebuilding.

I believe this is what determines a fast RAID chip or not.

SuperMiguel
08-18-09, 06:56 PM
changed the read ahead to adaptive, and got this:

http://img216.imageshack.us/img216/9837/hdtunebenchmarkdellpercn.png
http://img406.imageshack.us/img406/1600/hdwrite.png

SuperMiguel
08-18-09, 07:30 PM
another program

http://img443.imageshack.us/img443/9937/attog.jpg

Neuromancer
08-18-09, 08:25 PM
1x is 250mb/sec (http://en.wikipedia.org/wiki/Pci_express), which is exactly in line with the results.

Not for gen 2 ;) which his mobo has


Capacity Per lane:

* v1.x: 250 MB/s
* v2.0: 500 MB/s
* v3.0: 1 GB/s

16 lane slot:

* v1.x: 4 GB/s
* v2.0: 8 GB/s
* v3.0: 16 GB/s

My understanding of this is that the x16 socket supports 4GB/s at 1x or is it that pci-e 1.0 in x16 slot supports 4gb/s?

Also interested in this because my array speed isn't consistent.

http://hostthenpost.org/uploads/dd3d02d281a075709769670c74cde5d4.jpg


Those listings are broken down to versions 1, 2 and 3 of PCIExpress (3 is not out yet)

The first group is througput of 1x lane in PCIE PCIE2.0 and PCIE3.0

The second group is throughput of 16x Lane in PCIE PCIE2.0 and PCIE3.0


As i said 1x for you is 500MB/s I really doubt that is the problem.

SuperMiguel
08-18-09, 08:43 PM
im going to try tomorrow my 2 15k hd i raid 0 and see what i get

SuperMiguel
08-18-09, 09:02 PM
This result looked a bit better:

RAID 0 [64kb]
HD Tune Benchmark Transfer rate Block Size:

256KB
http://img502.imageshack.us/img502/531/256kb.png

1MB
http://img514.imageshack.us/img514/623/1mbpsm.png

8MB
http://img170.imageshack.us/img170/3803/8mb.png

thideras
08-18-09, 09:03 PM
Not for gen 2 ;) which his mobo hasThe card has to be compatible too.

Neuromancer
08-18-09, 09:19 PM
The card has to be compatible too.

Yup, I thought it was. Just googled it though and I was wrong. *sigh I hate being wrong :p

thideras
08-18-09, 09:20 PM
Muhahahahaahahahaa >:3

SuperMiguel
08-18-09, 09:43 PM
RAID 5 [128kn]
read/write speeds:

http://img502.imageshack.us/img502/5843/raid5r.jpg

and this is my configuration:

http://img502.imageshack.us/img502/2033/settingse.jpg

bchur83
08-18-09, 09:55 PM
Change your IO Policy to Cached IO and re-run HDtune.

SuperMiguel
08-18-09, 09:59 PM
working on it

SuperMiguel
08-18-09, 10:25 PM
Change your IO Policy to Cached IO and re-run HDtune.

about the same results... about 5+ MB more..

SuperMiguel
08-19-09, 01:58 PM
will try the LSI firmware today and see the results

SuperMiguel
08-19-09, 04:58 PM
woot 15k drives are here.

Raid 0 64kb using 64kb bench:
http://img513.imageshack.us/img513/4302/15k1.png

ATTO:
http://img233.imageshack.us/img233/2664/15k2.jpg

visbits
08-19-09, 06:21 PM
200MB/s at 5.5MS is way nice. Windows boot times on that should be real fast.

SuperMiguel
08-19-09, 07:41 PM
u think is worth it getting 2 more? of those 15k?

Vengance_01
08-19-09, 07:58 PM
No. Your access times will go up, and I dough you will notice the difference in day to day stuff. Benchmarks will look alot nicer. :bday:

SuperMiguel
08-19-09, 08:28 PM
No. Your access times will go up, and I dough you will notice the difference in day to day stuff. Benchmarks will look alot nicer. :bday:

so just stay like im :)

visbits
08-19-09, 09:18 PM
No. Your access times will go up, and I dough you will notice the difference in day to day stuff. Benchmarks will look alot nicer. :bday:

Why would the access time go up from running more drives? I've NEVER seen that before in any raid array I've ever setup. The access time will always be the highest single drive.

SuperMiguel
08-20-09, 08:45 AM
woot found another of those 15k hds for $30 brand new :) so i guess ill stay with 3 of them =)

SuperMiguel
08-24-09, 07:30 PM
Installed battery and here are the results/settings:

Perc 5/i settings:
http://img217.imageshack.us/img217/2134/raidsettings1.jpg

Raid 0 using 2 15k rpm hdds:
http://img220.imageshack.us/img220/8670/raid0full.jpg

Raid 5 using 3 7200rpm hdds:
http://img217.imageshack.us/img217/5794/raid5full.jpg

SuperMiguel
08-25-09, 04:41 PM
installed new 15k and :
http://img218.imageshack.us/img218/4614/315k.jpg