What setup is better for gaming?

[Bop]

I was thinking about my hard drive gaming setup and I was just going to get dual 150GB raptors(one for windows/apps, the other for page file/all my other data) but then I began to think about having one raptor as the boot/games drive and then using dual WD2500KS hard drives in raid 0 as my secondary drive.

Should I just stick with dual raptors or should I try out the other idea?

PS - Space is of no concern to me, I can't even fill up 1 80GB HD. I'm just looking for pure performance.

 

[unhappy_mage]

Pure performance -> raptors. No raid, it hurts latency when you seek. Your original plan is the best way to go.

 

[Asian Dub Foundation]

pure performance? then two raptors on raid 0

 

[TheDoucheMan]

No raid 0 at all. It will only help with large sequential file transfers. Gaming consists of more small file transfers.

What would you do with 500gb of raid 0? Do you have that much data that you could afford to lose?

 

[PaHick]

No raid, it hurts latency when you seek

Not always, depends on the controller.

No raid 0 at all. It will only help with large sequential file transfers. Gaming consists of more small file transfers.

Wrong. Go read some benchmarks at SR. Almost everytime, when a single drive is tested against 2 of the same drives in RAID0, RAID0 setup beats the single in Office DriveMark and High-End DriveMark 2006. Almost all of those apps. deal with small files.

Nonetheless, for pure performance stick with the Raptors.

 

[DougLite]

Wrong. Go read some benchmarks at SR. Almost everytime, when a single drive is tested against 2 of the same drives in RAID0, RAID0 setup beats the single in Office DriveMark and High-End DriveMark 2006. Almost all of those apps. deal with small files.

And you know that the RAID-0 setups go down like a Lead Zeppelin in SR's gaming tests. And that's what the OP asked about.

 

[PaHick]

And you know that the RAID-0 setups go down like a Lead Zeppelin in SR's gaming tests. And that's what the OP asked about.

Been through this chat before DL....lol. Im still with ya on the gaming tests. But,and not to highjack this guys thread, but as far as SR's tests, there are far more games out there than the couple they regularly test. SR needs to look into why the those games test that way in RAID0 when other similar games show postive results. Back to the topic at hand, im not getting rapped into any debates anymore tonight

 

[Justintoxicated]

Raid 0 = slower seek times but faster at loading contigious data.

4 Raptor 150's in Raid 0 would load BF2 maps nice and fast with 2 gigs of ram

 

[defakto]

Raid 0 = lower seek times but faster at loading contigious data.

4 Raptor 150's in Raid 0 would load BF2 maps nice and fast with 2 gigs of ram

I think you mean slower seek times, or Higher seek times.

 

[unhappy_mage]

Not always, depends on the controller.

No. Wrong. It *always* hurts latency. Why, you ask? Rotational latency. The disks do not align themselves to spin in the same position, so that sector #Q on the first drive isn't directly over sector #Q on the second. So when your raid array has to read the data on sector Q of both disks, it has to wait for the drive to spin until Q arrives on whichever's closer (a normal disk has to do this) and then wait for the other drive to spin the rest of the way, too (a normal drive *doesn't* have to do this!). Mikeblas thought that it was silly to have raid controllers hurting seek times (it's just a few gates, latency should be in microseconds!) so he started a thread about it a few months back. He eventually came to this conclusion.

Say what you will about performance, but you *cannot* state that it doesn't hurt latency on a single read. Please, you're welcome to your viewpoint (I don't want it ) but at least state correct facts about it.

 

[Justintoxicated]

I think you mean slower seek times, or Higher seek times.

I typoed the S thanks! thats almost as bad as in office space where he puts the decimal in the wrong place!

So what I think it comes down to is how the game operates and how much memory you have also. If the game loads nearly everything at the beginning before begining gameplay Raid is nice to have. But for example if you do not have alot of memory causing the game to constantly seek while you are playing then raid is not going to perform well at all.

It also depends on how contiguous the data on your HD is, if it is fragmented all over the place Raid can be much slower than Non-Raid. For example if you have 4 fragmented HD's they are not going to read smoothly as smoothly as one drive would.

This is why alot of people don't think Raid-0 is worth the risk, but I do find myself curious what 4 Raptors would do in Raid-0 with my machine when loading large battlefield maps!

 

[PaHick]

Please, you're welcome to your viewpoint (I don't want it ) but at least state correct facts about it.

I am, and I dont care if you want it or not. We actually agree and my point was the latency on most current controllers isnt a factor. Take for example, an access time of 12.8ms for single drive and 13ms for same drives in RAID0. Sure there was a hit, but is anyone REALLY going to notice that? Ill answer that for you,NO. These numbers were pulled out of the air, but most times the increased latency is that small. To make a recommendation against RAID0 because of this(something you WILL NOT notice) is assnine to say the least. The days of going from 13ms non-raid to over 20ms in a raid config are long gone.

 

[DougLite]

Actually, with rotational latency being above average on an additional 25% of requests, you're looking at an extra 3ms on those requests with 10K drives or 4.2ms with 10K drives. This doesn't show up in seek time tests BTW, as they only request data from individual sectors, requests that are serviced by only one drive in the array.

 

[-ChEM-]

latency on most current controllers isnt a factor. Take for example, an access time of 12.8ms for single drive and 13ms for same drives in RAID0

how does that work? the only thing i can think of is that the drives are kept in the same alignment as each other, is that true?

 

[unhappy_mage]

Magic, duh...

 

[PaHick]

I thought id add some links that were worth looking at. The Hothardware Raptor 150 article has some nice tests that include Raid0, and an article written for Overclockers.com by a forum member in which he tested a single 74G Raptor vs. 2x74 Raptor Raid0. Since most in this forum are against the idea of Raid0, for gaming or desktop use period, it pays to check out opinions from the "other side" of the fence.

 

[unhappy_mage]

Let me state for the record that raid 0 is great for its application. However, here's its application:

• Non-valuable data (scratch) or backed-up data
• Application where sequential transfer rates are most important
• Well-cooled case to deal with extra heat
• Good power supply to deal with extra power draw
• Application where latency is not important
• Application where 4MB or more of I/O is done at a chunk

I can think of one - video capture and editing - and for anything else that would fit the criteria , sure, raid 0 makes sense. Other than that, it's all a single-disk game, or another raid level.

Here are some things that aren't on my list of things to use raid 0 for, and why they're not there:

• Games. In order to even consider a raid 0 array for a gaming machine, you had best have hardware that's top of the line - dual 7900s, at least 2GB of performance ram, an FX or EE chip, you get the idea - because it's a last 5% thing. You get way more performance per dollar out of the second 7900 than you get out of the second Raptor. And a better (lower-latency) internet connection will give you a lot more advantage in online games than loading the map an extra second faster will.
• Audio editing. It's a completely different game from video editing - you lay down a half-dozen tracks, then play back all of those and add more. Repeat until you've got a good set of takes. That means a *lot* of seeking. Audio is best done with a ramdisk, or a scsi disk in the other room.
• Desktop productivity. Do I even need to spell this out? It's all human-bound; Word spends 99.9% of its time waiting for you to push the next key. Nothing you do to the computer will change how long it takes to write an email.

I'm not saying raid 0 doesn't have its place. I'm just saying nobody I know has that kind of money.

 

[PaHick]

Let me state for the record that raid 0 is great for its application. However, here's its application:

• Non-valuable data (scratch) or backed-up data
• Application where sequential transfer rates are most important
• Well-cooled case to deal with extra heat
• Good power supply to deal with extra power draw
• Application where latency is not important
• Application where 4MB or more of I/O is done at a chunk

I can think of one - video capture and editing - and for anything else that would fit the criteria , sure, raid 0 makes sense. Other than that, it's all a single-disk game, or another raid level.

Here are some things that aren't on my list of things to use raid 0 for, and why they're not there:

• Games. In order to even consider a raid 0 array for a gaming machine, you had best have hardware that's top of the line - dual 7900s, at least 2GB of performance ram, an FX or EE chip, you get the idea - because it's a last 5% thing. You get way more performance per dollar out of the second 7900 than you get out of the second Raptor. And a better (lower-latency) internet connection will give you a lot more advantage in online games than loading the map an extra second faster will.
• Audio editing. It's a completely different game from video editing - you lay down a half-dozen tracks, then play back all of those and add more. Repeat until you've got a good set of takes. That means a *lot* of seeking. Audio is best done with a ramdisk, or a scsi disk in the other room.
• Desktop productivity. Do I even need to spell this out? It's all human-bound; Word spends 99.9% of its time waiting for you to push the next key. Nothing you do to the computer will change how long it takes to write an email.

I'm not saying raid 0 doesn't have its place. I'm just saying nobody I know has that kind of money.

This is the kind of posts that you get on this forum, and thats great that you have an opinion. But if you look at the numbers it shows exactly the opposite of what you state as truth. And its always been this way. If you look at those links I provided, theres a rather nice increase in Raid0 almost across the board. The same happened in the thread where DL posted this graph from SR..

You guys against Raid0 for anything but situations that are bound by sequential transfer rates are proven wrong time and again. Tests by even SR and Anandtech, who are also against Raid0 for the same reasons, prove otherwise also. If situations that are bound by sequential transfer rates are the only areas where Raid0 is worth it, then you wouldnt see faster boot speeds,application loading, and faster level loads in certain games. In SR's Office DriveMark 2006 and High-End DriveMark 2006, hard drives in Raid0 routinely win against same drives in single setup. Office DriveMark 2006 is comprise of (copy and pasted from site) A capture of Veritest's Business Winstone 2004 suite. Applications include Microsoft's Office XP (Word, Excel, Access, Outlook, and Project), Internet Explorer 6.0, Symantec Antivirus 2002 and Winzip 9.0 executed in a lightly-multitasked manner. High-End DriveMark 2006-A capture of VeriTest's Multimedia Content Creation Winstone 2004 suite. Applications include Adobe Photoshop v7.01, Adobe Premiere v6.5, Macromedia Director MX v9.0, Macromedia Dreamweaver MX v6.1, Microsoft Windows Media Encoder 9.0, Newtek Lightwave 3D 7.5b, and Steinberg Wavelab 4.0f run in a lightly-multitasked manner. Read over those apps. and tell me they all are bound by sequential transfer rates. You cannot.

Fact of the matter is this..if two drives in Raid0 are compared to same drive in single setup the Raid0 setup is quicker, almost always across the board. Sometimes the difference isnt all that much, sometimes is huge. But as I said to DL once, who are you to tell someone whats an acceptable gain? That decision is left to the end user. The Op stated " I'm just looking for pure performance." and in this case dual Raptors in Raid0 is it.

 

[insanarchist]

...err, not to get involved in this potential (albeit inevitable) shitstorm, but doesn't the graph you have show the 4xraid0 performing slightly faster in office apps (which, as was mentioned before, are spending 99% of the time waiting for you), while performing slower than the 2xraid0, which is performing slower than a single drive?

Or am I reading that wrong? I'm assuming more i/o per second is better?

 

[unhappy_mage]

Where on that graph does raid 0 show a single win? Only in the Office benchmark does 2x74 tie with 1x150; everywhere else there is a lower number of IOPS with the raid 0 setups. And in 2 drive versus 4 drive (adding more drives will tend to accentuate the higher latency/higher STR raid 0 gives) they trade off across the board. And in WoW and Farcry, where people are more likely to be concerned about performance, 4x raid 0 loses to 2x. So adding more drives, and boosting the STR while hurting the latency hurts.

As for the two articles you posted before, the overclockers.com one seems to consist of a metric ton of artificial benchmarks and "it feels faster!!!!". And the hothardware: Sandra is not a banchmark for hard drives!!!, PcMark is artificial, HDTach is artificial and misleading, and they see 14% gains in quake4 level load times.

So what I said in my last post *still applies*! If you were originally a gamer on a budget, and got say a single 7800 and a single Raptor 150 and now have $300 to upgrade your machine, adding a 7800 to your setup gives you 70 or 80% more FPS while you're playing, while another Raptor in raid 0 would give you 15% less time waiting to start playing at your original FPS. I'm not saying raid 0 has *no* place in a gaming machine, just that it's not cost-effective under $3k for the machine, and not with any less than Raptors.

 

[PaHick]

In that graph, you cannot look at the scores of the 150, you can only compare scores from the same drives (single vs Raid0or in this case 2xRaid0 vs 4x Raid0). If you want to compare different generation of equipment, youll never get a straight answer for any type of discussion. And you know that. If that graph showed 2x150 Raptors in Raid, their score would have beaten the single 150. Thats a fact! And WoW and Farcry are not representative of all games. Come on now. Whether the benchmarks listed suit you or not, its what everyone uses to compare hard drives.

If you were originally a gamer on a budget, and got say a single 7800 and a single Raptor 150 and now have $300 to upgrade your machine, adding a 7800 to your setup gives you 70 or 80% more FPS while you're playing, while another Raptor in raid 0 would give you 15% less time waiting to start playing at your original FPS.

They MAY notice 70-80%, it all depends on their monitor. If they are running a lower resolution they just spent all that money to do what? Run benchmarks and brag ? At 1280x1024 on a 19in. monitor I cannnot tell the difference on a number of mid-high end cards. But I sure as hell can tell if im running Raid0 or not.

 

[unhappy_mage]

In that graph, you cannot look at the scores of the 150, you can only compare scores from the same drives (single vs Raid0or in this case 2xRaid0 vs 4x Raid0). If you want to compare different generation of equipment, youll never get a straight answer for any type of discussion. And you know that. If that graph showed 2x150 Raptors in Raid, their score would have beaten the single 150. Thats a fact! And WoW and Farcry are not representative of all games. Come on now. Whether the benchmarks listed suit you or not, its what everyone uses to compare hard drives.

Fine, I agree it's unfair to compare different generations of drives. Especially when it's cheaper to get a single new one that performs better than two old ones. No, wait, that makes no sense. The question is always, what's best for my money? And I think this graph makes it clear, at the very least, that it's not a lot of 74GB raptors - a single 150 outperforms 2 and 4 in 9 out of 10 cases. I have yet to see a raid 0 benchmark of raptor 150s that shows anything relevant to the real world, but there may be a configuration that gets better performance.

And if WoW and FarCry are not representative, how about you post some relevant graphs next time?

Last, there are new games still coming out, last I checked. And while you may be getting 60 fps on HL2 and F.E.A.R, newer games might put you more in the 20-30 fps range with a single card. Then an SLI setup would help a lot more. You might not be able to tell the difference between 60 and 80 fps (I know I can't) but between 30 and 50, I would hope you can tell.