Cookies op Tweakers

Tweakers maakt gebruik van cookies, onder andere om de website te analyseren, het gebruiksgemak te vergroten en advertenties te tonen. Door gebruik te maken van deze website, of door op 'Ga verder' te klikken, geef je toestemming voor het gebruik van cookies. Wil je meer informatie over cookies en hoe ze worden gebruikt, bekijk dan ons cookiebeleid.

Meer informatie

Door , , 81 reacties

Conclusions

The test results that Tweakers.net produced show that RAID 0 is far from obsolete in a desktop environment. Using RAID 0 is not 'ignorant', as Storage Review stated, and its former reputation deserves to be restored. Striping does not always increase performance (in certain situations it will actually be slower than a non-RAID setup), but in most situations it will yield a significant improvement in performance. It's not just our benchmark results that support this view: the majority of Tweakers.net readers who at one time or another tried striping, feel that the overall responsiveness of their computer improved when employing RAID 0.

AnandTech and Storage Review should be wise to investigate matters more thoroughy before jumping to quick conclusions. You don't judge a Porsche on its capabilities to carry groceries. A car like that serves a different purpose, and it should be judged on that instead - even if it will never fulfill its true purpose in real life. Power users, tweakers and hardware enthusiasts, the target audience AnandTech, Storage Review and Tweakers.net try to please, use their desktop systems in a different way than the pretty blonde next door who only uses it to check her Hotmail account. What we're trying to say is that you shouldn't assess the performance of RAID 0 with benchmarks that are not made to test the performance of the storage subset. AnandTech and Storage Review's negative verdict on RAID 0 in the desktop environment will likely have a profound influence on the opinions of uninformed users for years to come. A sure loss, since their verdict couldn't stand up to trial.


Door Femme Taken

- Architect

Femme is in 1998 als oprichter met Tweakers begonnen en werkt tegenwoordig als ontwerper in het productteam van Tweakers. In de vrije tijd knutselt Femme fanatiek aan zijn domoticasysteem.

Reacties (81)

Reactiefilter:-181080+145+25+30
Moderatie-faq Wijzig weergave
I believe RAID 0 only fulfills its purpose as a storage subset for large quantities of files. Not for applications or games.

If you want to have redundancy in your operating system use RAID 1. If you want to have redundancy in your files and data use RAID 5.

If you want to have performance in games and applications for rendering, photoeditting, etc use multiple single configured disks and seperate the swap file from the other frequently accessed disk (read: not partitions).
I believe that was the point of this review, to show that rather than dismissing RAID 0 outright, it can be used where it makes sense, depending on usage patterns it just might do that.

What does give RAID 0 in particular a bad name is some of the motherboard implementations (no cache, no hardware acceleration, chipset bandwidth limitations and sometimes shoddy firmware), it would be more interesting to see a comparison between these and some more upscale controllers like those from 3ware.

For me it brought a small increase in performance over two single disks, but I dropped it because of reliability problems (not a good idea to combine with the ill fated Deathstars).
For me it brought a small increase in performance over two single disks, but I dropped it because of reliability problems (not a good idea to combine with the ill fated Deathstars).
Is the theoretical loss in reliability really an issue in practice? I mean, in the past few years of very intensive daily work behind my PC with a single HDD I have encountered at most 2 events of data-loss. If this would increase to 4 events per 4 years (100% up) with two HDDs in RAID 0, I don't really see that much of a problem to get rid of it. Sure, every time you lose your data it's one time too many, but really important stuff is backed-up on external storage devices like ZIPs or MEMkeys anyway. Downloaded movies, music and stuff like that can be downloaded again, right?
believe that was the point of this review, to show that rather than dismissing RAID 0 outright, it can be used where it makes sense, depending on usage patterns it just might do that
If that was the point of the review, they made a very poor job of getting it across.
Also it would mean that they haven't gotten the point of the reviews on Anandtech and Storagereview. Their point was not that Raid0 is useless for every scenario. Their point was that Raid0 is useless for the desktop. Even for powerusers. And that is true.

The scenario's Tweakers has used on to show that Raid0 is usefull are not realistic real-life scenario's.
Powerusers opening loads of pictures in photoshop while making a backup at the same time? Yeah right. Power users don't make regular backups. (if at all) It's grasping for straws to create a scenario with high I/O

Sure, I can also create a lot of situations where Raid0 is faster. But none of those are everyday situations for powerusers.
im absolutely not convinced raid-0 is of any use to the desktop user. to the contrary i think the 100% increase in chance of dataloss is a strong indication -not- to use raid-0 at all.

and then there's amdahls law. if you optimize a subsystem you beter take care that the actual use of the subsystem legitimizes that subsystem optimization.

it doesnt make sense to optimize a subsystem that is not the bottleneck for the overall system performance.

look at the numa architecture of the dual opteron workstation. theoretically the improved bandwidth of the numa memorysystem is 100%. but does your typical windows installation use it? hell no! so thats a wasted optimization, clearly violating amdahl law.

the same can be reasoned for raid-0. the 100% extra harddisk benchmark measured with specialized "storage scope" benchmarking utilities -will- show the 100% increase. however this increase wont surface in the overall system benchmark of desktop use. it will evaporate because harddisk througput just isn't the bottleneck in desktop use.

your hard earned bucks can better be invested in additional pc memory. this memory will give the OS more headroom to optimize disk IO by caching and read-ahead strategies.

--
beerbelly
Anand:
I breezed through the pro-RAID article that was /.ed a couple of days ago; I'm not one to directly criticize someone else's work, I'd rather let my work stand on its own and have the readers come to their own conclusions so that is all I will mention on that topic. You all aren't here to read about my review of someone else's review of a review, so I won't bother with something like that - instead I've got a better idea :)

Ever since I posted my RAID article I've been working up ideas to make everyone happy for inclusion in a future definitive desktop RAID guide; the debate died down after the article so I didn't expect to revisit it until later, but with the renewed interest I thought about getting the guide done sometime this week. Reality set in and I don't believe it's physically possible for me to get the type of guide I'm planning done before the week is over, so it will have to wait until September. But rest assured, it will be done. Let me know if there's anything you'd like to see tested in particular, but I'm not planning on leaving any stone unturned for that guide - those of you who know me should know what to expect :)
goede ontwikkeling dat dit soort artikels ook engels worden vertaalt, maakt tweakers toch een stuk 'professioneler'.. zolang de originelen maar in het nederlands worden geschreven :)

is dit eenmalig of gaat die meer met zulke reviews gebeuren?
(translated) is this the first time a review gets translated, and will it happen again in the future?
It certainly isn't 'unique', it happened more than once in the past. Although the Dutch article is prior, we can decide to translate an article if English people might be highly interested in it. In this case we are analysing and criticizing the test results and conclusions of two sites writing in English, so it could be considered as a logical step to translate it.
It would be nice if visitors react in English on this article ;) .
I think he did mean the past simple of react (which is reacted as far as I know). Though he could better say "It would be nice if visitors will react in English on this article"
Gepost door Cookie - zaterdag 7 augustus 2004 - 17:50 - Score: 1 (Informatief)
Only the first line of every post will still be in Dutch ...
Ich sehe nicht ein dass wenn man Beitraege in Englisch uebersetzt, dass dann das Forum professioneller sei.
One last question:

Can anybody please explain to me why companies like Apple, Compaq/HP and IBM (all reasonably 'capable' companies, when it comes to technological innovation and application) don't deliver RAID-0 on their high-end workstations?

Is that because their high-end workstations are not used by the so-called 'powerusers', and for that reason don't require RAID-0? (presuming you buy a high-end system for performance reasons, and not for 'extra' stability)

Or is it because these companies did some research and found that running RAID-0 on a workstation is 'ignorant'.

Or does t.net has knowledge that IBM/Apple/HP does not?

Can anyone please answer this question, without obfuscating it by changing the subject?

I will, for convenience, rephrase the question: Why don't companies like IBM, HP and Apple use RAID-0 (as a standard option) in their high-end workstations?
I think it might have something to do with the bad name raid-0 has. Companies, whether they need performance or not, are likely to favour stability over performance at all times. For that reason the pre-build systems will stand a much better chance of being the product of choice for companies, if they don’t consist of doubtable hardware.
My guess is that if companies like Compaq and IBM sold comparable systems that only differed in having a raid-0 setup, the non-raid systems would sell better.
It might also be true that computers often poses more power than is strictly needed by the end-user. So the extra costs of a good raid-controller won’t weight up to the fact that the performance gain is not that big.
If that was the reason those companies would offer Raid0+1 systems.

Also a professional user that would really benefit from Raid0 wouldn't mind copying his work package for that day from a third disk to the array and copying it back to that third disk at the end of the day. (or to a server)

Companies like HP, IBM etc might even create some tools for them to make it easier.
I agree with what mtdevries said:
If that was the reason those companies would offer Raid0+1 systems.
Furthermore, they could easily put something about stability issues in a RAID-0 User Disclaimer, or something.

Wouldn't be the first time a big multinational would protect themselves against liability that way ;p

No, I really think the fact that RAID-0 isn't seriously 'pushed' as performance technology for high-end workstations, is simply because the big companies know it doesn't really increase performance like the benchmarks would suggest.

(Of course, if you specifically *ask* for RAID-0, they'll be happy to deliver it. A sale is after all a sale ;p)
That's because the system's they selling are systems for the "beginners" or for busnisse. The beginner doesn't now and don't ned a RAID-0. If you buy a PC for a compagny then the risk is to high for RAID-0. I should use RAID-5 or something for powerstations.
If I buy a high-end HP graphics workstation, I assure you, I am NOT a 'beginner'.

Furthermore, graphics workstations use so called 'scratch disks', which are comparable to '/tmp' directories. Mostly the data on these disks is 'volatile' (it won't reside on the disks for longer then a couple of hours, maybe days).

Running RAID-0 on these disks can, in theory, help a lot in performance. Stability is not a real issue, because of the 'temporary' nature of the data contained on these scratch disks.

Still, companies like Apple (number 1 deliverer of graphics workstations) don't even offer RAID-0 on their systems. If you want RAID-0 on, say, your brand new dual G5, you need to buy a non-apple certified, generic, PCI RAID-0 Adapter. Apple just doesn't offer it.

If RAID-0 was such a blessing for the REAL poweruser, then why don't companies that are famous for their poweruser-base (like Apple with their graphics workstations) 'push' that technology (like they do with dual-cpu systems, etc.), in order to sell more systems?

I find that, to say the least, very peculiar.
My main PC at home has 2 80GB 7200rpm disks in RAID0, and 1 200GB 7200rpm disk in single mode (used for storage and temp. backup).

In normal use (internet, office, some gaming) I don't feel any speed difference, but if I compile a 700000 line Delphi project, resulting in 10MB+ of dll's and exe, RAID0 makes a huge difference!
So, even if you're not a "poweruser" all the time (nobody is), RAID0 can be a nice addition for the moments you can really use the speed improvement.
I am not sure that it will make a huge difference. Because one 200GB 7200 rpm disk could easily reach 40 MB/s. Meaning that a 10 megabyte file can be written in 0.25 seconds.

With delphi projects the cpu speed is way more important and is the determinating factor on how long it will take to compile.

higher RPM drives may improve Delphi's compile time (especially in projects with lots of files which have to be compiled). Because of the faster random access time.

RAID0 is awesome for capturing video (uncompressed, high res). capturing video requires alot of throughput and RAID0 can deliver alot of that.

I would have run RAID0 myself too if not for the fact that RAID0 increases chance of failure by two times. Happens when running 2 disks striped. If one fails the data on the second disk will be useless.

Why is this article english and the responses too? I am curious..
What do we consider a typical 'desktop' PC? I think that's a key question if you want to definitively answer a question like 'is RAID 0 blessing or hype on the desktop?'.

I mean: i find it fairly obvious that a graphics workstation with a lot of RAM, running software that uses a lot of so-called 'scratch disks', can benefit from the extra performance that RAID 0 has to offer. And last time I checked, a graphics workstation was classified under the 'desktop PC' category. So I think I'll have to agree with t.net in this case.

For that same reason, I can understand that a developer workstation that is used to develop software and has to compile millions of lines of code each day, can noticably benefit from the same performance-increase. And a developer workstation is *also* a desktop PC.

But somehow, i find it hard to conceive a situation in which a standard 'office', 'home', or 'gaming' desktop PC (which cover about 95% of the desktop PC market) can noticably benefit from RAID 0. If any of these desktop PC's have performance issues, they are mostly caused by things like the processing power of the CPU, the performance of the GPU or the ammount of available RAM. Equipping these systems with RAID 0 will most likely not do anything positive performance-wise.

Furthermore, I think the 'Porsche'-metaphore that was used by t.net does not completely cover the point. As far as I can see, the point AnandTech and SR made was in the order of: 'It is of little to no use to run RAID 0 in a desktop environment. Spending your money on RAID-0 is ignorant. You can better spend it on a faster (15,000rpm) single disk, more RAM, a faster CPU or just about anything but RAID 0'. Which is kinda analog to saying 'It is of little to no use to use a Porsche for carrying your groceries. If you want to move around groceries, you can better invest in a bigger car (so you can move more groceries around in the same time), or even a truck'.

On which T.net responds: "You shouldn't judge a Porsche on it's ability to carry groceries. A car like that serves a different purpose and should be judged on that." (which is, in the case of a Porsche, a very true statement ;p).

Isn't this the same as saying: "You shouldn't judge RAID 0 on it's ability to perform in a desktop environment. A technology like that serves a different purpose and should be judged on that."?

And isn't that, more or less, the same as saying: "It is of little to no use to run RAID 0 in a desktop environment, since it was designed to perform under heavy loads you rarely find on desktop PC's"?

Anyways, what I would like to know is: for what desktop-enviroment *IS* RAID 0 useful, and when is it more wise to spend your (extra) money on a better CPU, more RAM or a faster (more rpm's) HD? Because this is something *both* reviews fail to inform the reader about, and I think in the end this is one of the few things the readers actually *really* want to know.
Spending your money on RAID-0 is ignorant. You can better spend it on a faster (15,000rpm) single disk, more RAM, a faster CPU or just about anything but RAID 0'.
I think you are missing an important point here: what are the costs of RAID? It's almost free since most motherboards have already RAID functionality and if you want a decent storage capcity buying two disks is cheaper than one large disk. The only real downside of RAID 0 is a higher risk of data loss.
But somehow, i find it hard to conceive a situation in which a standard 'office', 'home', or 'gaming' desktop PC (which cover about 95% of the desktop PC market) can noticably benefit from RAID 0.
I can image a lot of situations where RAID would improve the performance of a 'standard' desktop pc. Think about: starting up windows, copying files, loading games.
I can image a lot of situations where RAID would improve the performance of a 'standard' desktop pc. Think about: starting up windows, copying files, loading games.
I completely agree with this, but:

you didn't include benchmarks of the above with:
1) faster disks
2) More RAM
3) a better CPU

So, in short: RAID-0 *can* (and will) have *some* influence, but to what extent (compared with other solutions)?

The 'abstract' benchmarks that were used in the reviews don't answer any of such questions. They only say as much as: Under the load of our (chosen) set of benchmarks, RAID-0 performs better then a single disk.

But it totally doesn't tell the reader how to translate these benchmarks to actual desktop-PC performance (like the things you mentioned: booting windows, etc.), or how to compare a RAID-0 upgrade to, for instance, extra RAM or a better CPU.
a) you didn't benchmark any of the above
Perhaps you should read the article? All of those things are benchmarked!
b) you didn't include benchmarks of the above with:...
That's outside the scope of this article to include all those things. If you're interessend in that kind of information it shouldn't be hard to find and secondly the article does contain tests with differend hard disks.
The 'abstract' benchmarks that were used in the reviews don't answer any of such questions.
Abstract? They are all based on popular programs and it can't get any more real world than the stop watch tests (and yes, this includes booting windows!).
hehe....my bad; Just after I posted, I noticed that your test in fact did include the above things.

Anyways, my point is that i still don't know whether i can better invest my money in RAID-0 (which effectively means i need to at least buy a second disk), or in extra RAM or a faster CPU.

There is no doubt in my mind that RAID-0 will speed up the boottime of windows, and the load-time of games. But so will extra RAM or a faster CPU.

Now, what can i better spend my 100 euros on: a second HD for RAID-0, a faster (single) HD, a slightly better CPU, a slightly better GPU, or extra RAM? After all this time, this is still the key question, and there is still no answer.

I mean: it is easy to say that RAID-0 is great, and that it can improve your performance, say, 20%. But if you can improve your performance by 30%, simply by spending the same ammount of money on a better CPU, or on more RAM, all of a sudden RAID-0 doesn't seem so great anymore.

Now I'm not saying that RAID-0 isn't any good. I'm just saying that since Anandtech and SR proved one thing, and then you now claim another, things haven't really got more transparent. I mean: anandtech aren't a bunch of fools, and neither are you. But who is right? Both sites claim they have the numbers to support their claims. Strangely enough, they can't *BOTH* be right.

Now tell me, how exactly does this techno-tugg-o-war help us, the readers?
Anyways, my point is that i still don't know whether i can better invest my money in RAID-0 (which effectively means i need to at least buy a second disk), or in extra RAM or a faster CPU.
That's indeed something that isn't aswered in this article. However it is clear that RAID gives a performance advantage and that is what this article is all about.
Now, what can i better spend my 100 euros on: a second HD for RAID-0, a faster (single) HD, a slightly better CPU, a slightly better GPU, or extra RAM? After all this time, this is still the key question, and there is still no answer.
That's simply a question that can't be answered. It depends what kind of tasks you want to perform with your computer and what you want to be fast (and also how you calculate costs). This article gives an idea of the performance advantages of RAID, nothing more and nothing less.
Now tell me, how exactly does this techno-tugg-o-war help us, the readers?
It will lead to the truth about the performance of RAID with desktop usage. Nothing more, nothing less.
No, this article is not about Raid0 giving a performance advantage.

This article is about proving AT and SR wrong that Raid0 is not interesting for desktop usage.

Which is something completely different!
i read a lot of post above where people are trying to devine "a desktop".

IMHO if we speak about a desktop on a site where the knowledge and interest in computers is far above avarege and in a lot of cases also professional, then i think most people will agree that if we say desktop here we also mean computers used for video/audio editing and software/graphics development systems and these are systems that can surely benefit from RAID 0
offtopic:
i dont now if its the englisch language but the general level of discussion in this thread is much higher than in most other threads
thums up for us :P

and sorry for the spelling misstakes ;)
Which tweakers are doing video editing?

Since they don't have a life they don't have anything interesting to put on video anyway. A video about cold pizza's and warm coca cola is kinda boring ;)

Software graphics development? Yeah right. How about they might do that at the office and just play games at home. A more realistic scenario.

Many tweakers think of themselves as power users. But if they would really monitor their systems for a month, they would find that their own machines are not as heavily taxed as they think.

If the editors monitor their own system, then I am convinced that they will discover that 99.99% of the time that they do heavy editing in photoshop, that they are not running a backup in the background.

They best way to define desktop usage, would be to let all tweakers here run some tool, that for one month would monitor their desktop usage. The results would be an eye-opener to many people.
mtjdevries hits the nail on it's head (slaat de spijker op z'n kop ;p) in the above posting. I would have modded him up if I still had mod-privileges in this thread, but now I can do nothing but to fully agree with him.

Most self-proclaimed 'powerusers' are only 'powerusers' because the multiple 24/7 systems they have at home, with all the dozens of fans and watercooling pumps, eat up a lot of 'power'.

However, this kind of 'power' usage has nothing to do with cpu-power or computing power, which are associated with 'heavy computing tasks'.

I really think that everyone that is NOT a company, graphics designer, or a very (and i mean: VERY)active member of some software-development community is also NOT a real poweruser. Not even most of us tweakers (including myself, for that matter).
I still think, that if u have an P4 2.4GHZ.
With raid-0 and an P4 2.8GHZ.
The 2.4GHZ boots games up faster, and give u performance.
I notice it in games and Photoshop.
I do much things at one time, such as watching an movie,
scanning for virusses.
Running an server.

Thats wy i get much benefit from raid-0.

I have an friend of mine, wich has an 36GB raptor.
Wich is sort of effective as an raid-0.
He is always in the game first, doesnt have to wait.
Only at himself.

But, what i dont like @ raid-0, is that it isnt safe.
Thats wy im going to buy me an 74GB raptor.

Wich is expensiver though, but i think its worth it.
Watching a movie isn't such a demanding task for a harddisk.

Running virusscans? Most pople just use some virus shield. Just scanning files when they are being written. Far more efficient, and doesn't put a heavy load on the harddisk.

What kind of server do you run? Webservers usually don't tax a harddisk, and fileservers are usually limited by network bandwith.

Are you still so sure, that your multitasking puts a lot of load on the harddisk that will benefit highly from Raid0?
Well, not too hard, 70.000 views until now, while the dutch version also has 68K and for example the article about the apple mini ipod almost 50k.

So this is not a too big increase in the serverload I guess..

Op dit item kan niet meer gereageerd worden.



Microsoft Windows 10 Home NL Apple iPhone 6s Star Wars: Battlefront (2015) Samsung Galaxy S6 Edge Apple Watch Project CARS Nest Learning Thermostat Smartphones

© 1998 - 2015 de Persgroep Online Services B.V. Tweakers vormt samen met o.a. Autotrack en Carsom.nl de Persgroep Online Services B.V. Hosting door True