Are SSDs in raid-0 still relevant?

[mathieuturcottetech]

Long story short, I'm starting to need more space on my machine. I'm a gamer, and I cam'T stand long loading screens. With SSDs becoming more and more affoardable, I'm thinking about adding a big SSD to offload my games on. Essentially, I have a small SSD for my OS, and a 7200rpm hdd for my games and data. That disk is getting full.

I checked on amazon.ca, and I found a 500gb WD ssd for 150$. I also found a 240gb WD SSD for 75$. The hamster started to go... What if I was gonna get two of those? would cost me the same (150$), but I would have 2x 240gb in raid 0. So almost the same capacity(480 vs 500), but I would have two disks in Raid-0 instead of a single one.

I do understand the risks of a drive failure in a raid-0 configuration. I do not care in the slightest. I would put Steam, Origin and Blizzard games on there... nothing I can'T simply reinstall in minutes. I also have pretty fast unlimited internet. I don't care having to download everything again. The only things I care about are capacity and performance. This would also means that if I have a drive failure, I have only a 75$ drive to replace, instead of a 150$ drive to replace. Altough the chances of a failure would be greater. I have been using SSDs for the last 6 years and never had any problems, so this is an increased risk I can easily live with.

As an example of a game, I'm a huge fan of flight simulators. I play a ton of X-Plane 11. it takes forever to load from the HDD. I made space on the OS ssd yesterday and transfered the game on it. The loading times were substantially better, and thats on a SSD that is also used by the system, and it's not in a raid configuration. I can only imagine I would have even more gains on a raid-0 array, on disks that are dedicated to running the games.

What is your opinions on this?

Thanks!

 

[J_E_D_70]

“Also an option, but significantly below NVMe speeds and even RAID speeds.”

Let me repeat... it’s not significantly below nvme in real life, only in benchmarks. That 6x faster interface translates into very slightly faster load times and general use. In hindsight, I shoulda gone 1tb sata iii over the 500gb nvme I bought for around the same price.

 

[USAFRet]

Hmm. I think I could do that. I have replacement SSDs just sitting on a shelf. I could disable my RAID, throw one unused SSD in, restore my OS to it, then benchmark it "ceteris paribus". So we can see the difference between 1 SSD and a 4X SSD RAID.
It won't be this weekend, but I'll post results.

user facing performance. Not just benchmarks.

 

[Deleted member 217926]

“Also an option, but significantly below NVMe speeds and even RAID speeds.”

Let me repeat... it’s not significantly below nvme in real life, only in benchmarks. That 6x faster interface translates into very slightly faster load times and general use. In hindsight, I shoulda gone 1tb sata iii over the 500gb nvme I bought for around the same price.

Yep. I replaced a 500GB 840 Evo with this 512GB 950 Pro. Benchmark performance went WAY up. In real life there might be a fraction of a second difference in most load times. Boot times are a few seconds faster. There are situations where super fast storage matters. In daily use, including gaming there's no appreciable difference between a fast SATA III drive and an NVMe drive. And the slightly better OS performance of the NVMe drive is attributable to the faster deep queue speeds and NOT sequential speeds.

 

[Ralston18]

Performance question:

How is "difference" or "significant difference" to be determined? Keyword being "significant".

Something, as stated, that is end user facing significant. Or deemed to be experimentally significant. Maybe both - two measurements..

Need some objectivity beforehand.

 

[USAFRet]

Entry into a new game level or opening/saving a large CAD file:

120 seconds reduced to 90 seconds
25% reduction. User benefit = Huge. Worth actual money.

2 seconds reduced to 1.5 seconds
25% reduction. User benefit = meh.

 

[jdlech2]

Please excuse the longer response time. I'm making goulash for myself and the kids tonight.

user facing performance. Not just benchmarks.

That could be a problem as I have a problem with real life timing of software loads in general. Specifically, people try to time them using stop watches, clocks, counting, etc.. It's all very dependent on the person.
I'm not much of a coder. I used to code quite well back when DOS was king, but not anymore. But I would need a script that starts a timer and then does a bunch of real world file and data moves as well as software startups before ending the timer at the end. Otherwise, the timing is only as accurate as the person using the timer.
So, if anyone wants to whip up a powershell script to do that, I'm sure we can play ball. Assume a Windows 10 home machine.

I recall my first IBM. Yes, it was a true blue 8088 I paid a thousand dollars for. It had a whopping 10Meg hard drive. But I recall literally hitting the power button, then going to make myself a cup of coffee. It would be almost ready when I got back. Apps and data was never more than a few hundred KB back then.
I was taking electrical engineering in college and owned a 386-25. I bought a cyrix mathco for Pspice I was running. I remember throwing a floating point spreadsheet at it. The difference was astounding.
Today, the big difference a gamer is going to see may be a few seconds shaved off your waiting time during scene changes. The hard drive is not the bottleneck during game play. But shaving just 20 seconds off boot up time can be significant when you're looking at a 2 minute boot time.
It doe remind me of something I read, many years ago, about HiFi audio. The engineer who cut THD from 1% to 0.1% was celebrated as a world class hero. The engineer who cut THD from .01% to .001% didn't even make page 11 in any newspaper. Even though both performed the same feat, and the second engineer's task was much harder.

How is "difference" or "significant difference" to be determined?

It's always been somewhat subjective. I prefer to go with percentages, while others might prefer to compare raw numbers. I'm inclined to post the data and let people decide for themselves how they want to parse and judge it.

 

[clarkjd]

It doe remind me of something I read, many years ago, about HiFi audio. The engineer who cut THD from 1% to 0.1% was celebrated as a world class hero. The engineer who cut THD from .01% to .001% didn't even make page 11 in any newspaper. Even though both performed the same feat, and the second engineer's task was much harder.

Both of those feats, while significant in benchmarks, were not significant "User facing" when you consider the fact that loudspeakers, even the expensive ones, had a THD on the order of 5 to 10 percent( I did not leave out the decimal point).
Point being that "User Facing" is the great equalizer here. A user can't feel benchmarks, he can only notice how much better (or worse!) his system performs in his everyday use.

 

[jdlech2]

Both of those feats, while significant in benchmarks, were not significant "User facing" when you consider the fact that loudspeakers, even the expensive ones, had a THD on the order of 5 to 10 percent( I did not leave out the decimal point).
Point being that "User Facing" is the great equalizer here. A user can't feel benchmarks, he can only notice how much better (or worse!) his system performs in his everyday use.

With audio, it's always been my experience that, while a unit with great specs can sound bad, a unit with bad specs never sounds great. I've learned to always "spec it" first and then go on to listening to it from there. It's a bit different with computers, though. Sometimes one does not care about a particular specification - the speed of a drive meant to hold backups, for instance, is largely irrelevant to me. Other things can spec great and still not work well. My wireless keyboard and mouse, for instance. It works great when the usb receiver is in the front of the computer - where I really don't want it. But if I move it to the back of the computer, it loses signal and works poorly. Great specs, but poor performance. The only indication that it might not perform well was the fact that the kb uses 1 AAA battery and that battery lasts about a month - suggesting it generates a weak signal to preserve battery life. Hardly something one can really know before the purchase. If Toms ever did a comparison of wireless keyboards and mice, I've never seen it. (But would love to see one.) I use rechargeable batteries throughout my house, so their efforts to preserve battery life is not appreciated.
Anyways, it's been a great conversation, but I'm slowly derailing it now. Sorry.