[WarDad]

This post has been edited by WarDad: 31 March 2012 - 05:58 PM

[WarDad]

This post has been edited by WarDad: 31 March 2012 - 06:01 PM

[WarDad]

This post has been edited by WarDad: 31 March 2012 - 06:01 PM

[continuum]

Post #3 -- your comments regarding PCI bus bandwidth are overly general to the point they are inaccurate, you may want to considering looking up PCI-e (PCI Express) bus implementations and controller setups as well as PCH (aka "southbridge") communication links.

Post #4-- your statement that Gigabyte NCQ is not functioning-- have you verified this?

Post #5-- should you wish to get further in-depth, you may wish to look up Windows 7 Superfetch and other built-in Windows functions designed to hide disk latency. "Fancy cache" and other non-standard terms may lead to severe confusion with newbies.

Post #6-- your smaller SSDs and slower write times due to 512KB write blocks as a percentage of reserve space is completely inaccurate; see the articles right here on Storagereview relating to the fewer active NAND die per channel and the few active channels for their impact on performance.

Windows 7 and other modern OS's should also be SSD block alignment aware, no need for additional tools (although if you want to elaborate on scenarios where you've seen this break, and hence, additional tools are needed, that would definitely be informative!).

Also to clarify-- Windows 7 is also much, much more SSD friendly than previous OS'es. You can locate your user profile folder and others wherever you need. (see: http://answers.micro...22-4f56aebb296e or http://lifehacker.co...ry-in-windows-7 )

Post #7-- your RAID1 cons about only reading from a single disk are a controller-specific limitation. Plenty of controllers out there interleave reads. A side note-- I wonder if Intel's Matrix RAID is available on current chipsets. Hmm.

Post #9-- RAID5 loses one drive's worth of capacity to parity, not necessarily 1/4th. It is possible (albeit frightening) to build a 16-disk RAID5 if your controller has enough ports.

You should clarify between RAID10 and RAID01-- many low-end RAID controllers do the latter and not the former, which significantly affects the odds of 2nd-failure-survivability. SR's own hardware guide here explains it quite well.


Post #11-- changing SATA drivers/SATA modes in Windows XP is indeed a pain; a handy guide is here: (it relates to changing motherboards, but the concept is the same).
http://arstechnica.c...php?f=8&t=42051

Post #12-- you may wish to clarify compatibility for partitions larger than 2TB, although it would make sense if that is somewhat beyond the scope of what you are trying to accomplish.

[WarDad]

Post 3 - You have a point. I wanted to point them to what is likely a better option.
I did not want to discuss PCI vs PCIe vs PCIe 2.0 or x1 serial channels vs. x4 serial channels, or serial protocol overhead, or RAID card vendors stating the actual x1 BW. I did want to point them to what is likely a better option.

Post 4 - Yea, I have tried several time to get it NCQ function. Set bios to AHCI and loaded the latest drivers. Surfed the web for details. If I missed something, it's because it's hidden, and it shouldn't be. This Gigabyte SATA controller also supports floppy,PATA, and the driver has a check box for Tagged Command Queueing. XP OS. The performance seems low and I don't need it.

Post 5 - I have been reading more about Windows Cache Manager. It's tough to get the details. Most hard documentation originated with Win 2K. I was led astray by some older caching utilies that reported my file cache as 100MB, but of course they did not function or reload on restart.
Small test files do have issues with cache distorting results. It seems the cache to concerned with is in the HDD. Most benchmarks can and shouuld bypass Windows file cache. FancyCache is a utility that uses a different cache method, strategy, whatever, they have a web site.

Post 6 - OK, I'm not sure where I read or drew that conclusion that from. The larger M4 SSD have higher sequential writes. Other brands / models tend to folow this trend.

Block Alignment - My C: to D: was put out of alignment by using ESUS, my favorite partioning program. It heard of the issue to late and don't care to do a clean install. I tried the command prompt queries and the calculate, and tried moving the partition boundries several times. There is software I can buy, but I can also ignore a tiny bit of unallocated space.

Relocating user folders was easy in XP. I did spend hours searching for W7 tips. These tips are work arounds, and suggest or require a clean install. I'm complaining about it, but not hurting with a 256GB SSD. The people using 60GB or less should know they could have issues.

Post 7 - This is Poor Mans Raid and we use what we have on the motherboard. I'm sure there are RAID 1 controllers that interleave reads for a price.

Post 9 - Good point. RAID 5 and RAID 6 capacity loss should be corrected. I had 4 drives to work with.. and did not think beyond that. I only touched on RAID 5 and 10 as they were supported motherboard bios options. Anyway, I made a case for only using two consumer drives. Anything more that that is out of the poor mans budget.

Post 11 - My side step is quick, easy, and low risk. I did not want to bother with registry changes. Tom's Hardware has a sticky with instructions. I saw a link to OCZ fourms for instructions.

Post 12 - It was beyond the scope. I don't have all the GPT details or a 3TB drive to test with. I'm sure there is plenty of info out there.

[continuum]

In reply to post 4-- that sounds like a motherboard/controller specific thing, you might want to clarify. Supporting TCQ makes it sound like an older controller so it may be old, buggy, or both to start with.

Post 6-- Partition out of alignment, I would blame that on ESUS as you stated, not Windows.

Post 7-- I would be curious if AMD chipset RAID does interleave reads or whatnot. Again, it's a blanket statement you've made that I would want to verify is correct before making it. (and yeah, I don't see much in the way of reviews on Google on chipset-level RAID...)

Post 12-- your controller doesn't have to support GPT, but it does have to support sufficiently large LBA's. Now boot off a drive larger than 2TB is indeed a BIOS/UEFI-related issue for the motherboard (and also the OS...). Clarification would be useful even if you think you want to punt this, mostly because 3TB drives are so common.

[WarDad]

I finally got around to revising article, noe that HDD prices are finally coming back down some.
Some corrections accuracy enhancements were made in response to Continumm inputs.
Much of the Idle speculation was removed.
ATTO benchmark is now mentioned, but not shown.

I now clarify the W7 is friendly to the SSD hardware. It's still a problem if you need to relocate folders and files to the HDD. There is a new utility out called Steam Mover which was made to move game folders and leave a link. Some swear by it and say it works for a lot of W7 folders. I have not used it, and I'm not sure how badly I could botch thing up.

"I would be curious if AMD chipset RAID does interleave reads or whatnot. Again, it's a blanket statement you've made that I would want to verify is correct before making it. (and yeah, I don't see much in the way of reviews on Google on chipset-level RAID...)"
CrystalDisk mark shows a normal 133MB/s for large sequential reads with raid 1. IO meter showed 220MB/s with 2 workers for 10M seq. read. Frankly I'm feed up with the marketing types hiding hard engineering data these days. Maybe they deserve and need more blanket statements, slander, and inuendo.

About GPT and large LBA's. I really do want to punt this issue. It's enough that I warn people about this potential issue. There is plenty of good help out there on many different forums.

This post has been edited by WarDad: 31 March 2012 - 06:41 PM