Possible speed boost from using PCI-X RAID SCSI controller

[bambiboom]

RE: Possible speed boost from using PCI-X RAID SCSI controller
Gentlemen?,

Those with time and verbiage patience limitations may skip "Background" and go to "Finally, the Question!"

Background >

I'm using> Dell Precision T5400 (2009) [2X QC Xeon X5460@ 3.16GHz , 16GB DDR2-667 ECC, Quadro FX 4800 (1.5GB), WD RE4 500GB/ Seagate Barracuda 500GB, Win 7 Ultimate 64-bit] with AutoCad, Sketchup, Revit, Solidworks (learning), Adobe CS 4, Corel Techical Designer, MS Office Pro, WordPerfect Office.

As you can see, I have a old, but I think, high quality computer that is also loaded with lots of large and resource hungry applications. Especially after adding the 2nd CPU- yesterday- in Passmark Performance Test 8 Build 1016, this computer now has a respectable CPU performance, the 2D is reasonable- 512 as the Quadro FX 4800 is a specialist in Solidworks 2010 and CS4 as it happens, and all seems well. DDR2 RAM does run very hot and I panicked a bit a few days ago when I saw 93C (197 F).

Currently, there is some factor- possibly a device driver- that is not allowing Performance Test 8.0 Build 1016 to complete the 3D test, but what I've learned so far is that the memory (646) and disk (956) scores are proportionally poor to CPU and 2D scores and way off typical good- not the best- computers. Passmark by the way, has been fantastically devoted to solving my running problem, which apparently affects only 1% of computers that have tried the test. Someone in Sydney has spent most of a couple of days trying to duplicate my trouble. I really recommend having a look at Performance Test and the baseline results- plus there's a free 30-day trial.

The T5400 had a rating of 1623 in it's last successful test, but another T5400, having 2X Xeon E4400 @ 2.83 , Quadro 4600 (768MB), and an OCZ Vertex had a CPU rating of 7663- as compared to mine with 8459, a 2D of 462 as compared to my 512, 3D of 642 -mine = 912, and still an overall rating of 2059. The higher rating of the other T5400 can only be thanks to the OCZ Vertex SSD, the disk score of 6699 where mine scored 956. The higher rating in spite of lower ratings on other tests seems to me to be a test weighting that shows the importance of the disk performance.

The Precision T5400 is the very close cousin of the Dell Poweredge 2950 server-2x CPU's, 8 RAM Slots (I just added +4GB RAM the 5400 out of a 2950)- and has 2X "PCI-X" slots which was a PCI bus devised to double the bandwidth of the PCI for servers. The common use of these PCI-X slot in servers-Dell Precisions, and I think some HP workstations of the day, was to plug in a RAID controller card that was specified to deliver 3GB/s. These PCI-X controller cards could also create drive clusters- and (I think) that would support up to 256 Drives. After- or possibly out of- the PCI-X, came the PCIe we all know and love today.

The thing is, there are piles of obsolete PCI-X controller cards, many made by LSI Logic, which appears to make reliable, high performance stuff- RAID and NAS related. They are making these controller cards still for the PCIe bus- they're 6GB/s and some with 4X or 8X HD connections. If you look, many of the highest rated machines in the Passmark baselines use either one of these or OCZ RevoDrive 3 X 2, which are more or less SSD's "direct data injection" on a PCIe card. The current versions of both of these solutions are generally very expensive,..

Finally, the Question!

So, finally, the question. >

> Would using a PCI-X (not PCIe) SAS RAID controller card, like the LSI SAS3080X, which has 2X SATA connections- , make a useful disk performance increase on the Dell Precision T5400? This would probably be used with the current 2 HD's, but not in RAID.

Still, the T5400 is very RAID oriented (0,1,5,10) and if the performance and system reliability is worth the effort and expense, I'd buy a 2nd WD RE4 500GB and try a RAID 1 (I'm too anxiety prone to consider RAID 0!). The logical thing is to buy an SSD, but I would only use an SSD if I had RAID 1 or 5. I'd need a pair of ~500GB SSD's or 2X 240 SSD plus 2X 500HD- a minimum investment of three to more than five times the value of the T5400. Plus I'm not excited about SSD's after reading so many low reliability/quirky installation/erratic behaviour/poor longevity reports.

If you have a recommendation for a good HD performance tester that reports read, write, access, times & etc., that would be a help as well.

Thanks!


Cheers,

BambiBoom

 

[RealBeast]

You will not get improved disk performance from RAID 1, that just writes everything to each drive for redundancy. RAID 0 will give you increased performance, but at the cost of increased chance of failure.

An SSD is a great performance enhancement, I've been using them for years (since back when 80Gb Intel X-25M's were $500) and have never had any issues with dozens of different models. But it is not worth putting them in RAID. While reads and writes are fast, a bigger advantage is the fast access times. Programs like AutoCad, Solidworks and Adobe CS4 love SSDs. I have a 480GB boot SSD (Intel 520) in my workstation and a 160GB scratch disk SSD (old X-25M) for Adobe, which really improves performance.

Crystal Disk Mark is a good test suite: http://crystalmark.info/software/CrystalDiskMark/index-e.html

 

[bambiboom]

RealBeast,

Reading a few post here makes me realize how astoundingly uninformed I am concerning RAID and servers.

My concept of RAID 1 is that a secondary but equal drive mirrors the first, but this process actually slows the disk speed a bit. But the advantage is that a failure of one of the drives allows the substitution of a new drive in that position and RAID automatically builds a new copy, meaning you don't miss a beat in having to restore system images, reload files and etc.

My thought was to try one of these PCI-X RAID controllers- not in RAID but only connecting two separate drives that would then - if the PR is to be believed- have much higher transfer speeds than a conventional SATA controller due to the PCI-X bandwidth being greater. his $20 solution- as I already have the two HD's- would be the poor man's substitution for the high performance of $800- $3,000 enterprise PCIe drives, things like RevoDrive.

Thank you for commenting on SSD reliability as I was curious if the the many report/complaints about SSD's were due to people being more "expressive" of opinions when it goes wrong, than if it just always works and becomes semi-invisible. The Intel 520 is one I see has among the best reputations for reliability/longevity- Samsung 840 too.

Also, thank you for recommending Crystal Disk Mark, the first run results ( 5X 1000MB) on the 5 partitions which are:

___________________________________________

______________READ> MB/S_____WRITE> MB/S

Sequential______133.6____________130.0
512K___________44.09____________77.35
4K______________0.543____________1.392
4K QD32_________1.255____________1.424

___________________________________________

I'm not sure what these really mean but, as Crystal Disk Mark has green bands under the figures and there was a lot of blank space to the right of the bars, presumably for better performers, I might assume these are not data demon disks,...

Oh well, an SSD might be inevitable if I'm to get another couple of years out of the old banger T5400.


Cheers,

BambiBoom

 

[RealBeast]

That benchmark looks about right for your HDD.

The PCIe controller will not give you any speed boost because the fastest HDD is slower than a SATA II port already so that is not a bottleneck. The drive seek times and read/write speeds are the limitation on you.

As far as high quality SSDs, I would stay with Intel, Samsung, and Crucial. Despite differences in the benchmarks, all of the current SSDs really perform about the same in actual use.