Hard drive questions

[rob-the-oil-nut]

I am a Seagate fan. Anyone else? Anyone had bad (recent) experiences with Seagates, particularly Barracuda 7200.7 and 7200.8 models? I did, a 7200.8 toasted on me after just 3 months. But don't get a brand disloyalty just yet; I sure haven't!
I have been waiting, since their June 14 '04 press release, for a BIG drive with 16 mb cache, NCQ and SATA. Seagate has given excuses, has told me "I'd start looking at the competition if I were you", has given OK explanations, but there is no sign of 16 mb cache anywhere on the horizon. No hints, rumors, nothing. That is unless you want to count their assertion that they'll have 9th-generation (7200.9) Barracudas with 16 mb cache on time for Christmas. Once lied to, twice shy

Related matter: Any serious real-world tests comparing 16 mb cache to 8? Especially if it shows that 16 will have little or no real-world effect? Especially if the tests are done with SATA drives with NCQ? Of course the best such test would be one Seagate against another, everything else being equal, except 8 vs 16 mb cache. But I'll settle for Maxtor, WD, whatever.
Application is a new Opteron-based system so I'm not interested in a new Corvette with any used spark plugs, know what I mean?
THANKS!
Rob

 

[04SpecV]

I don't but I think most of their new drives have 5year warranties so they seem good. I generally use Western Digitals, which havent given me any trouble in years. I've had bad luck with Maxtors I think every Maxtor I ever had or known someone else to have has failed.

I'm not sure why everyone is skimping on the cache. Memory isn't that expensive these days. I'd put 128MB of cache on a drive myself

I know some of the higher end SATA raid controllers can accept SDRAM...

 

[pitzel]

I've lost enough hard drives (mainly Seagate SCSI's) that I refuse to consider anything except a pairs of drives in RAID-1 configurations for any of my desktops.

Considering the incremental cost of going RAID-1 is around a hundred bucks, and RAID-1 improves performance significantly, it is pretty much a no-brainer to me.

If you are going Opteron, definitely spring for a RAID-1 setup. Not only will RAID-1 give you some reliability cushion, but it will also improve your performance.

As for cache, who cares, it doesn't seem to be a huge deal in the benchmarks anyways. Save the money you would spend on the extra cache drive and use it for a RAID-1 setup.

 

[kenw]

as long as you insist on staying on the bleeding edge of technology, expect a much higher than normal failure rate.

When my job was establishing warranty costs for new computers, the single most important factor was new technology, including that in the HDD. The more new tech, the more we held back for warranty repairs. More tried and true stuff in the unit, the less warranty reserve (and more profit).

 

[eljefino]

I had a western digital drive that got real slow suddenly.

WD had a diagnostic program I could download, and it confirmed the drive was defective.

Noble to the end, the drive let me transfer all my data off, slowly.

WD let me fill out an RMA online. The drive had a manufactured date 23 months old and it was under a 2 year warranty. (I could not find my receipt and did not need it.) WD shipped me a new drive of the same size at a lower, prorated price (as it was somewhat obsolete). When they got my old drive in the mail a week later they gave me my money back.

Hats off to Western Digital.

 

[kenw]

it's amazing what a great customer service department can do, even for a company that sells junk.

We had a competitor that always got high marks in customer service, the highest in the industry. However, their reliability (as measured by independent labs) was awful. They put their $$ into great customer service instead of R+D (and their CEO bragged on it).

They are still in business, my former company is gone....

 

[Matt_S]

Having been in IT for 8 years, I'll say that Seaagate is the brand I've had the most failures with, Followed by Fujitsu. I've had much better luck with Western Digital and Maxtor.

 

[ediamiam]

I was curious on how hot current drives get. The last Maxtor I had, about 10 years back, ran real hot and basically fried. Its been about 4 years since I put a system together, using an IBM drive, surface temperature warm to the touch. I typically have been using cooling fans on the drives - Is this needless insurance these days ?

 

[rob-the-oil-nut]

One humble opinion but I think USB or FireWire EXT-ernals are different. Internal drives need fans. If they are not built-in, they are needed. Every time I read a comparison of one drive to another, one chart always covers temperature, and it's no news that lower is better. A fan that mounts directly onto the HDD and moves air over it is $15 to $20. Have a drive fail and then tell me how important saving $20 is.

I don't dispute that the fancy new drives have high failure rates. The thing is, they are rated for 50,000 stop / start cycles and something like 1 million hours as the mean time to failure (MTTF). Are they fudging this data, or is it in some way un-representative of reality? There has got to be more than meets the eyes on those life-expectancy tests. No car company comes out with a fancy new high-tech engine and then it turns out no one shy of a BITOG fanatic can get 80,000 miles out of it. Or is that a totally unfair analogy?
THANKS
Rob-the-oil-nut

 

[keith]

quote:

---

Originally posted by pitzel:
I've lost enough hard drives (mainly Seagate SCSI's) that I refuse to consider anything except a pairs of drives in RAID-1 configurations for any of my desktops.

Considering the incremental cost of going RAID-1 is around a hundred bucks, and RAID-1 improves performance significantly, it is pretty much a no-brainer to me.

---

RAID 1 is either 2 disks mirrored on a single controller, or less commonly, two disks on two controllers, but in either case the data on both disks is identical. If one fails, the other is good. RAID level 1 offers ZERO performance advantage over a single drive. Sometimes the controller overhead will reduce performance.

Of the other RAID levels, 0 is common and offers no redundancy and is usually configured with striped data across multiple disks, improving performance in typical usage. Basically, the disk I/O bandwidth is multiplied by the number of disks in the RAID 0 array.

I use a RAID 1 array myself and don't worry about disk failures. If one drive fails, I keep working and buy a replacement.

 

[kenw]

Nope, same basic drive best I've been able to tell.

In fact you can buy kits (some with fans) to turn a standard ol' garden variety or the latest bleeding edge HDD, notebook or standard size into a USB/Firewire drive. Done several myself, never bought one with a fan tho.

Now, the HDD maker, if you buy a bare drive, may have something to say about whether it needs air movement or not, I would read the fine print carefully just in case. But I know of external drives with up to 300gb with no fan at all and it is the exact same drive you could buy as a bare internal drive.

if blowing air across the HDD makes you feel better, by all means go for it. Most are sealed pretty well so the blowing dust has little if any chance of getting inside. (now then, does that worry you?)

Now, as I said before, most of my experience is in notebooks, but notebooks tend to be the worst-case application for just about anything especially heat. Smaller sizes just compound thermal issues. Mass is good for thermals....but bad in a notebook.

My current notebook has 2 fans and neither moves any air across the 80gb hdd. It's in it's own little area, all separated from any airflow.

the reliability numbers are really, really complex and unfortunately if you have a failure, your rate is 100% and a 1M hr MTTF is of no use to you. But I do have, as a former insider, pretty good confidence in the mfg's stated MTBF/MTTF etc. And the larger sized drives are infinitely more stable than the notebook drives.

Adding more platters and heads to an existing model is the safest way to go as that's not realy considered "new tech", but new materials, etc are the riskiest.

Problems definitely occur but they tend to be batch related. And woe to any mfg stupid enough to have consistent and repeated failures, the CPU OEMs will shut them off in a heartbeat. And the HDD industry is like most others, so incestuous that everyone has a chance to be the leader or the dog.

And back to the original post, I tend to avoid "the latest and greatest" technology like the plague...25 years in the computer business taught me caution if nothing else.

Early adopters have a role in life, but mostly it's to weed out the junk. Business customers learned long ago to take yesterday's tech: it lasts longer. Bleeding edge is always in the consumer side of the OEM, as Mikey will eat anything...

Add a fan if you want, but definitely back it up!

(edit: sorry for being so long winded!)

 

[pitzel]

quote:

---

Originally posted by keith:

Originally posted by pitzel:
[qb] RAID 1 is either 2 disks mirrored on a single controller, or less commonly, two disks on two controllers, but in either case the data on both disks is identical. If one fails, the other is good. RAID level 1 offers ZERO performance advantage over a single drive. Sometimes the controller overhead will reduce performance.

---

RAID-1, properly implemented, offers a performance advantage during reads since disk read requests can be intelligently directed to the disk that has the heads 'closest' to the data.

In every configuration I have tried, this has meant that RAID-1 gives you a nice performance boost compared to a single-drive configuration. Controller overhead is so insignificant I don't even know why you even bother mentioning it.

RAID-0, of course, is garbage, and decreases your reliabilty substantially.

 

[rob-the-oil-nut]

RAID controllers that accept memory are expensive, like $380 US for starters. Out of my budget range, might not get such a thing if I won a lottery, just seems a bit of a waste.

If I could please steer things back, away from RAID and back to 8 mb cache vs 16. I have some non-Seagates including a Maxtor 300, in 3 partitions including boot. It has 16 mb cache. I am seldom wowed by its speed and often gawk at the hourglass thinking "16 mb cache, 1 GB RAM, all those millions of transistors...is this crazy thing actually WORKING, or what?" Is 16 mb cache faster than 8 mainly in theory and benchmarks that are not real-world? That seems to be the situation but I have no proof, have seen no real-world tests. I HAVE heard that the difference is mainly in theory and under 5%, which 90% of users will never notice (not without a stopwatch, anyway

....some say under 3% which is nothing I'd ever notice. But is this the case with a $400 mobo and SATA and all that fun stuff as factors?

It seems Seagates are the least likely to be on sale, one reason suggested is that they really are built to a higher standard, and like a M-B 500SEL, is not ready to be sold at 20% off any time soon

I DID get FIVE 160 GB 7200.7's on sale, $40 after rebates. Got 5 volunteers in the US to receive them on my behalf. The cost however was all on my plastic, BEFORE the rebates...OUCH!

 

[kenw]

drives are mounted in sealed enclosures all the time (look at external US/FW drives). I would say a dedicated fan is overkill but certainly won't hur anything.

My experience is with notebook drives, and I just placed an old 20gb in a USB case as a portable drive. I does get "warm" but not enough so that I would worry. I usually leave it in the soft case because dropping it is more likely than overheating it.

 

[motorguy222]

I have a very old computer with a Seagate hard drive.

They seem to be a very good product.

 

[Matt_S]

quote:

---

Originally posted by motorguy222:
I have a very old computer with a Seagate hard drive.

They seem to be a very good product.

---

Seagate used to be the best there was. More recently, though, their quality has been lacking IMO. Even so, most of their drives will outlast the useful life of the computer, but you'll still get a higher number than usual, compared to other manufacturers, that will fail.

 

[theguru]

You might find that the reason drives are not running 16mb caches is that in most situations it will not be faster than 8mb, probably slower.

It will depend on what size files you are transferring, but the larger cache suits bigger files. If you are transferring many small files you may find that because the cache needs to be flushed it can indeed be slower.

If you look at the tests I think you will find that only in certain tests that the 8mb is faster. And not by much.

I've stuck by Seagate but my experience has been that the 5400 drives seem to run much cooler and I've had few die. I think the newer drives are built to the moden day standard and are throw-awy items. Expect your drive to die.

I don't know about others, but even though the newer drives have built-in error reporting, all my experiences have been that they dead just the same as drives without it.

Raid 1 is a good suggestion. Don't bother buying a Raid card. Many new boards have hardware raid built-in for little or no extra cost.

Personally I go for extra cooling. I'm of the belief that a drive too hot to touch isn't good karma. Is it a coincidence that laptop drives run hot and are unreliable?

 

[rob-the-oil-nut]

Theguru: Thanks, I like that reply. A few questions / points of clarification:
"If you look at the tests I think you will find that only in certain tests that the 8mb is faster. And not by much."
Can you direct me to such tests?

"I've stuck by Seagate..."
So you believe they are generally the most reliable OVERALL? Any tests done, on reliability, involving 7200.7's, and 7200.8's? Tests done in late '04 and better still, 2005?
Their warranty is 5 years but the 4th-year terms, it's no big consolation, and 5th year is like 4 free oil changes if your engine blows

I have NO ISSUES with selling all the drives as soon as one goes on me. Or after 3 years.

"I think the newer drives are built to the moden day standard and are throw-away items. Expect your drive to die."
Does modern-day standard means they are (in your opinion / experience) as good as any other brand, give or take a lousy batch? Or that they are 'premium' drives? Are their clean-rooms cleaner, is quality control better...is the 5-year warranty a marketing ploy?
Looks like I'm buying many anyhow. How long I'll hold onto them may be a different matter.
Backups, backups, backups

ALWAYS have backups....but, of a TB of stuff?
Any way of measuring start / stops or hours on a specific drive? Just curious.
THANKS
Rob

 

[pitzel]

rob-the-oil-nut, making hard drive reliability predictions is fairly futile these days, as defects in new designs often don't really show up for a year or two after they are sold.

As I pointed out above, building some redundancy into your setup is the only way to go, especially considering a second drive can be purchased for under $100, and RAID-1 comes with practically every motherboard and OS available these days at no extra cost (or a nominal cost).

My first hard drive cost over $3000 (in 1987-1988), spun at 3600rpm, held 44 megabytes, and was a Seagate (ST4053M in fact). Yes it still runs, but I now buy $100 drives in pairs for RAID-1 and throw away whatever dies -- far less expensive than trying to buy a 'perfectly' reliable drive.

 

[rob-the-oil-nut]

Thanks very much...When you say "a second drive can be purchased for under $100" what sort of HDD are you talking about? 2 X 120 GB (or smaller) for boot? I'm looking at 300 GB, 400 GB, or larger drives. Isn't it far better to have 2 identical drives in RAID, than to have 2 partitions of the same physical drive? As you might be able to tell, I have no real-world experience with RAID. $8600 + in training taught me little for the real world (I'll give you a hint: I was told to go 3.x track, not 4. In hindsight I should have gone with 4.x ...Hint 2, not NT, I'm talking Clarke IV)
I just give $ to my trusted friend, the system builder, when he advises I need more

I thank him repeatedly, he's the best system builder and super-techie I've ever hired and I'd be a mallard hit with 00 buckshot if it wasn't for him. But WOW does he look like the guy Matt Groening must have used as a basis for Homer!

Thanks Gerry P....thanks everyone!
Rob