Some Raid 0 results with my new 7811

Well I just recently got a 7811 from a fellow forum member and decided to flash the bios to a newer version with raid and see if it would work with my Hitachi 7200rpm 200gb drive. Here are the quick results, not too shabby.

Here is the first HD tach without writeback enabled in the intel matrix storage manager:


Here is the run with writeback enabled:

 

whats writeback?

 

Ok..... Did you have a question?

 

what is writeback, why does turning it on increase the performance so much.

 

It's pretty much as it sounds: cache memory the system writes to before actually writing the data to the physical hard disk. Cache, being solid state, is always faster than a physical disk which has to move heads and write cylinders, etc.

The downside is that on a desktop system if you have a power failure while your data is in the cache but before it's spun out to the disk you will lose it. Not an issue with a laptop with the battery attached, however.

 

Thanks so much for this: I'd assumed write-back was enabled by default (it isn't)... + rep for you!

 

I just wanted to share my results. I never have done raid with different manufactures of hard drives before.

Tallan,
glad this post helped you, I was worried the first time I saw the results w/o it enabled.

 

so is it better to have it enabled?

 

yeah it boots performance by quite a bit

 

Losing power is not the only issue with this. You also have to take into account the stability of the system. If you run any unreliable applications that hang or crash the system you may have significant data loss that can in some situations even corrupt your OS. Just because you are running a laptop with a backup power source does not mean this feature is 100% safe.

This is especially true in a RAID 0 setup as you could potentially lose your whole array.

 

That is true, but people who run raid 0 typically do so for the extra performance. I've ran raid 0 for years and only had 1 corruption issue when i was trying to find the limits of my processor in my desktop.

 

Yep that is also true

However you gain the most performance benefits from write-back caching in a RAID 5 setup. Real world benefits in a RAID 0 setup are negligible and IMHO not worth taking the risk.

 

says the guy whos running raid 0.


hypocrite!



*puts on flame suit*

 

Is it really that hard to understand two linked sentences. Let me spell it out for you:

When write-back caching is enabled real world benefits in a RAID 0 setup are negligible.

I'd prefer it if you kept your snide comments to yourself next time.

 

In his own defense, your second sentence, that you claimed to be linked to the first, mentioned "not worth taking the risk." There's very little risk difference between Raid 0 with write back caching enabled and Raid 0 without it, since no one is going to be unplugging their internal hard drives while something is being written. So I'm going to have to call you out and say your second sentence was talking about the risks of Raid itself, and not write back caching. Either that, or you've been misinformed about write back caching.

Regardless, can you please tell me what magical software will allow me to use my write back caching in the Fake World? I'm curious because I see a huge increase in my read/write capabilities now that I've enabled it, and, I'm not 100% positive, but I'm pretty sure the world I'm in right now is real...

P.S. I'm pretty sure ER was just playfully mocking you, I think you may have taken it the wrong way

 

Playfully mocking or not I don't appreciate being called a hypocrite.

Enabling write-back caching will greatly increase your chances of data corruption. You don't have to physically unplug a drive whilst something is being written. I think you misunderstand what write-back caching actually is. All you need is a system lock or crash and there is a high chance data could become corrupt due to the loss of un-written data. I've seen it happen countless times.

There's also no need to be facetious about my real world comment. When I said real world I meant not some pointless synthetic benchmark. It's OS load times, application/game load times, shutdown times etc that actually matter. Have you compared those? I doubt it. If you did you would find that write-back caching makes pretty much zero difference in your setup.

 

i was just stating that how you said that the benefits of RAID 0 are negligible but yet you still use it dont you?or are you talking about the risks of using write-back caching

 

If you had both read the entire thread including my previous post maybe you would have understood what I was talking about:

 

reading every post ina thread is for chumps

 

I know all about what write back caching is, trust me. As I said, the only real risk that goes along with it is sudden power loss. And considering even blue screens don't pose any danger to HDDs using write back caching, since they don't stop the power, there is relatively little risk. I've never had a computer simply and suddenly turn off on me. Yes, I've had system crashes, but never does the power simply stop. You would either have to unplug a computer while it has no battery in it AND while its writing to the HDD, or you would have to unplug the battery from a computer that is already unplugged in the same situation, or hit the power button which is normally designed to prevent being hit by accident. And if it's not something you did, but is instead a power failure within the computer itself, you're going to be losing data regardless, but that will be the least of your problems. I think the fact that write back caching (in a non raid array) is enabled by default in windows says that even Microsoft SDs agree that it poses little risk, and also says that if they think it's worth taking even the little risk, they must see some performance increase with it as well. If you've seen it happen that much, you need to stop hanging around novice computer users who ooops spill their drink all over the keyboard....

And don't take my bantering the wrong way, I was just playing around with you. But to answer your question, yes I did, don't assume so quickly. And I'm not talking about benchmarks (even though its obvious that your skepticism towards them is fueled by your current bias towards your argument). I'm talking about me, holding my watch in my hand, and transferring the same file from my external to my HDD in multiple trials with wbc on and with it off while the same processes are running, and seeing a more than significant difference, then starting up and shutting down in multiple trials and seeing a small but noticable difference, seeing a small increase in FPS and level load times or save times in different games, etc. When I said I know about write back caching, trust me, it's because I've just recently done a lot of testing with it, yes I am that big of a dork... so if you could tell me what measuring tool you've been using that shows "zero difference," I'd like to test for myself, and I'd even be willing to post screen shots for you...

 

Intel themselves didn't even include the option to enable write-back caching in IMSM on RAID 0 arrays until recently due to the high risk of data corruption. Also, if the 'enable advanced performance' box is ticked on the volume properties data is not only cached in the memory of the drive, it's also cached by the OS. This is what can lead to serious data corruption on system hangs.

So what are your results with write-back enabled and disabled?

My measuring tool? Funnily enough a clock.

 

Actually that's not true. Either you were misinformed or you're simply speculating, I'm going to guess and say the latter, and I'm not all that sure you know how writeback caching works... If you're in a raid array, it doesn't matter if the OS write back is ticked or not. Likely it will be greyed out regardless and you will not be able to change it. But the msm takes priority. In fact it would actually be impossible when in a Raid array for the OS to take advantage of any hdd caching at all...

If I simply told you my results, would you even believe me? Since you don't trust benchmarks, it's kind of hard for me to take screenshots of my watch.

I would like to see the documentation that places the reasoning behind Intel's only recent addition of the feature as "wariness of data corruption."Just to make sure you're not assuming that as well... But anyway if you're using a clock to test and you're telling the truth that you see no difference, then we must be using very different testing procedures. You ARE testing the difference while the CPU is both in use and not in use, am I correct? Because that's a large part of the difference here if you know how writeback works, and I'm not sure anymore that you do

 

Ummm they had that option 2 years ago when I did a series of tests with dual raptors with and without write-back caching and the difference in game loading times. It did have a significant increase in performance in all games. If I've only had one issue of data corruption in my years of overclocking with raid 0 and write-back caching then I'm not worried about it (countless lockups while trying to find the max clock speeds).

 

Ah, I had suspected he was making that up to fuel his argument...