![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
unhappy_mage said:
DougLite said:Clicky, check out the bar graph towards the bottom of Eugene's post. SR will be posting a pretty comprehensive article on RAID, hopefully soon. The graph is a selection from the results obtained for that forthcoming article.
Malogato said:The problem with this graph, DougLite are many fold. But I'll bring up the two most important.
1) You are showing a single, NEWer technology drive versus older drives.
2) There is no performance data showing what *1* wd740 does.
SHould I be surprised that a brand new drive that costs $300 performs as well as, if not better than, 2 years ago's technology that costs less? No, I shouldn't, and If you are, you shouldn't be modding a STORAGE SYSTEMS forum.
Now.. Put the WD1500 Raptors in Raid0x2 and Raid0x4 and show me that graph again, and I have a sneaking suspicion that the raid 0x2 and raid 0x4 will outperform Single drive performance.
Yes, Raid0 has a severe stability issue, but anyone with half a brain keeps an active backup of their Raid0 array.
unhappy_mage said:
I said this before, but apparently I must repeat it. There's a LARGER benefit from two independent disks than from Raid 0. Raid 0 is therefore slower. Turning 2 disks into a Raid array therefore gives no performance benefit.PaHick said:
No, it doesn't. It merely reduces the damage done. You still lose all data since the last backup, plus all the time to reinstall OS and apps. If you're a gamer, thats not a big deal maybe...but for me, I'd lose three full days.
For Raid-0, it's simply the mean of each drives MTBF divided by the array size. If you allow for possible interdrive synch failures, then its somewhat higher. Much worse than a single drive.
Yes there is data on what a single WD740GD, but it's not on the graph. Once again, a single WD740GD scores 833 IO/s in FarCry and 671 IO/s in WoW, which neither of the RAID-0 setups are even close to. A single WD740GD earns 641 IO/s in the SR High-End DriveMark 2006, remarkably close to the 2x RAID-0 and within striking distance of 4x RAID-0.Malogato said:The problem with this graph, DougLite are many fold. But I'll bring up the two most important.
1) You are showing a single, NEWer technology drive versus older drives.
2) There is no performance data showing what *1* wd740 does.
SHould I be surprised that a brand new drive that costs $300 performs as well as, if not better than, 2 years ago's technology that costs less? No, I shouldn't, and If you are, you shouldn't be modding a STORAGE SYSTEMS forum.
Now.. Put the WD1500 Raptors in Raid0x2 and Raid0x4 and show me that graph again, and I have a sneaking suspicion that the raid 0x2 and raid 0x4 will outperform Single drive performance.
Yes, Raid0 has a severe stability issue, but anyone with half a brain keeps an active backup of their Raid0 array.
DougLite said:
DougLite said:
masher said:
masher said:
masher said:
DougLite said:
DougLite said:
unhappy_mage said:
Honestly it sounds good and ill check back tomorrow. Thanks.PaHick said:
E4g1e said:
. Here they are anyways. Time to create (MB/s) Time to copy (MB/s)
Raid 1 4466 (5.73) 6296 (3.97)
Raid 0 2235 (11.45) 4489 (5.56)
Seperate disks 2237 (11.43) 5871 (4.25)Lol are you actually suggesting 2 drives at Raid-0 would be faster than 2 independent drives transferring from one to the other? Good god man, stop embarrassing yourself. You don't even need a benchmark to disprove this...just a little common sense. Two disks have the same raw bandwidth as a 2-disk Raid array...but the array has to read and write TO THE SAME DRIVES. So instead of a nice long sequential read pattern on one disk and a nice long write pattern on the other...you get BOTH drives jumping around reading AND writing.PaHick said:
But with Raid-0, you'll have to reinstall everything TWICE AS OFTEN. We've explained this too you many times; you still refuse to accept the facts.
unhappy_mage said:Well, the first round of tests are finished. These are rather disappointing; maybe I should try a controller that's not 33 mHz
The next round of tests should be a little more, shall I say, modern. One interesting point from this test, though: the raid 0 created the file only 2 seconds faster than the single disk. So the extra potential throughput isn't helping, even in this completely linear-write scenario.Code:Time to create (MB/s) Time to copy (MB/s) Raid 1 4466 (5.73) 6296 (3.97) Raid 0 2235 (11.45) 4489 (5.56) Seperate disks 2237 (11.43) 5871 (4.25)
PaHick said:
That'd be copying the file from the array to itself, no?PaHick said:
ben_johnson said:
See that this perception still persists? See that we must continue to debunk the myth that RAID-0 improves desktop performance across the board? See that we must continue to debunk the myth that running two inferior drives in RAID-0 will allow them to match the application level performance of a single superior drive? See that these misconceptions still exist? See that drive marketing people are absolutely giddy about selling twice as many drives to even a handful of customers on nothing more than hype? Nothing against these members that ask these questions...noone has been there to tell them the truth. Until now. I will simply not allow such myths to be perpetuated without a fight on this forum.iownyou14 said:
Isn't that just showing the cache performance?just2cool said:
DougLite said:
Frosteh said:
From the perspective of someone who grew up in a country where using less to do more has been a credo for ages, I see the power consumption point. Your RAID array will draw more power from the PSU, but that is not the real problem. The problem is that you draw more power from the outlet -> more power needs to be generated. While you would certainly be right to say that the difference is negligible at any point, in the long-term you are paying more for electricity. Most gamers do not take this hidden cost into account.Frosteh said:
This information has been added independently by DL. If you looked more at SR, you would notice that all their benches are done on the same hardware base, therefore the scores are comparable across tests.Frosteh said:
once agian, DL has clearly stated that on page 1 of this thread. I actually like his analogy a lot comparing the argument for RAID-0 to video cards, where people are trying to get the most performance for the money.Frosteh said:
If you really want me to, I can do the financial calcuation and explain to you that buying 2 740GDs and RAID-0-ing them does not -and did not- make sense, even for you.Frosteh said:
Lol, we're comparing Raid 0 to having 2 independent disks...and now you've thrown a THIRD disk into the mix? What lengths will you go to?PaHick said:
masher said:
You sure will. Of course, you just introduced four disks into the picture, not two. Quadrupling your cost, complexity, and power draw. And quadrupling your chance of failure too.drizzt81 said:
By your logic, you're saying that we should all chain-smoke and expose ourselves to carcinogenic chemicals...since our risk increase is "just statistics".
masher said:
drizzt81 said:
this comes down to another trend from the VC forum: do you play HDTach all day long?just2cool said:
