as the title suggests, i am wondering which aspects of hard drives are relevant to gaming, which ones actually make an impact on gaming performance, obviously as far as hard drives play into gaming performance, which is kind of at the root of my question. what specific aspects of hard drives (performance points) are important to consider when choosing a hard drive solely for gaming purposes

which reminds me, i was wondering what you guys thought about SSDs. i realize they are way more expensive then HDDs, but if they are that much better and will be half or a third of their current cost by the end of the year, then they might be worth the wait. i came across this article by tom's hardware "http://www.tomshardware.com/reviews/ssd-memoright,1926.html". give it a read, it is interesting regardless of my question, but more to the point, as an application of my aforementioned question, what specific performance points that the SSD has over the traditional HDD would actually make a difference in gaming performance? (access times, drive interface bandwidth, read transfer rates, write transfer rates, I/O performance, etc)

I really don't know if the HD will make that much of a difference in the performance of the game, maybe the loading of new missions/scenes yes, but I don't think the actual performance will be affected.

Before I had a WD1600JD, 8 MB Cache 7200 RPM. Then I switched to a Raptor X http://www.youtube.com/watch?v=0E8twKoHNag

I played the same games again and the only thing I noticed was a faster install of the game, faster initial loading of the game, faster loading of missions/videos but the performance itself was pretty much the same. Other people may experience different things. My system has an Opteron 165, 8800 GTS 640 MB and 2 GB of CAS 2 RAM in dual channel. I'd say a typical SATA drive with 7200 RPM and 32 MB of Cache should do you very good, for better gaming performance, pay attention to:

- Fast RAM, low timings, dual channel setup
- Good video card
- Good motherboard
- Very efficient power supply
- Good cooling/airflow techniques
- Modern processor (Core 2 Duo, Athlon X2 etc.)

A faster hard drive will help games load faster but performance during gameplay will not really be affected. Stick to 7200 RPM drives (or faster), get one that has a nice big L2 cache (16MB+), and also one that uses SATA 3.0Gb/s. Also pay attention to platter size, my 320GB drive has a single 320GB platter, which is much faster then 320GB drives that have dual 160GB platters. Manufacturers often don't tell you about the platter sizes, but you may be able to find out in reviews and such.

When a game is running, it is loaded from your hard drive into your memory.

So during game play, you will not notice it, but as Adam said, a faster drive will cut your loading times down in between levels etc

Adam Kolak said:

A large cache will make very little difference, apart from maybe a really un-noticable decrease in loading times. If you have constant access, or transferring large files, then yes, it will help, but games (even level loading) not much.

SATA 3.0Gbps? There is no need. Sure, it might be compatible with the SATA 3Gb/s interface, but the drive will not be any faster than an SATA 1.5Gb/s model.

Typical tranfer rates for any 7200rpm drive(s) are in the region of 70-80MB/s (560-640Mb/s), in other words, a little over half of the Serial ATA 1.5Gb/s specification. Even PATA drives run at those rates easily (though there are some newer features in the newer drives - NCQ / hotplugging / staggered spin up etc).

So really, just get any decent quality drive, from a good manufacturer, and you're good to go.

One of the main things I look for in a drive is noise (as they're all similar enough in performance - apart from the 10K drives)

Yes the performance difference is minimal, but getting a SATA 3.0Gb/s and higher cache drive usually tends to mean that it is a newer drive and therefore probably faster. Also the price difference is very minimal. My WD 320GB has a 3.0Gb/s SATA connection, has 16MB of cache, and has a single platter. It costed around $65. I think it was maybe $2 to $5 more than a few older models that had two platters, less cache, and/or the slower SATA connection. Plus it gives Mark's RAID-0 setup a run for its money in many industry benchmarks.

Increasing your DRAM has a much higher impact on in game speeds than the HDD spin rate or transfer rate. Having said that, if you have minimal amounts of DRAM and the virtual/swap space is used often, then a faster drive is better.
Large cache sizes are useful when loading large files, but not a lot of very small files.
ATA type devices actually load small files quicker than SCSI drives at the same spin speed due to the additional overhead associated with setting up a read/write occurance on a SCSI bus, but for very large file loads, SCSI beats ATA devices by offloading the workload from the CPU to the Host Bus Adapter.
In a system with maybe one HDD, and one Optical Drive, SATA 1.5 is more than enough, but if you are running multiple HDDs and Optical Drives, then SATA 3.0 is the better bet as you could run into a bottleneck on the SATA controller at the slower speeds.
Though SDD pricing IS dropping (the first one I worked on back in 1988 cost around $250K for 16MB, they are not dropping fast enough to keep up with the gains being made in standard HDD technology. Unless you're talking about a "Hardened" Laptop, I'd still stear clear of them as the price to performance ratio isn't supportive of the upfront cost.

Gerritt

If RAIDing, then maybe yes.

If running JBOD, then SATA1.5 is more than enough, heck PATA 133 is well more than enough.

The fact is, the SATA 3.0 spec was designed to last WELL into the future, and it will. It will be a LONG time before we see devices capable of anything near 3.0 Gb/s (at least in the mainstream)

It's really quite simple. The only times hard drives are accessed is during loading/saving be it the initial boot up, level changes, etc. When it comes to hard drive performance, these are the areas most effected by high(er) performance hard drives.

RAID0 helps with massive sequential read operations (e.g. loading a very big map into the memory). Otherwise, it may actually offer a slight decrease in performance as the drives have to constantly sync data (imagine zipping a zipper...if one side is too slow, the other side has to wait until it catches up).

Essentially, it really is not critical unless you have a serious issue staring at that "loading" screen or have a lot of extra cash to spend. 7200 RPM drives (no RAID) are sufficient for most needs. Once you're actually in the game shooting it up, faster hard drives have little to no effect on performance.


DublinGunner said:
Kind of like PCI Express x16. And just like PCI Express 2.0 with x32 support, SATA 6.0 Gb/s is just over the horizon. As is, PCI Express x16 and SATA 3.0 Gb/s are far beyond the needs of most people/businesses today.

They are both mostly just the result of Intel/industry showing off. XD

Both new standards are more or less for servers with ever increasing bandwidth requirements. I'd argue that SATA 3.0 Gb/s only came about due to SAS needing more bandwidth. The same is probably true for SATA 6.0 Gb/s and PCI Express 2.0.

I didnt even touch on the subject of the pending SATA 6.0 spec, as its so ludicrously over specced, its not even funny.

You could have 6 raptors in RAID 0 on 1 channel, and still not come anywhere near to maxing the bandwidth. Stupid if you ask me, with no practical application.

I wish they could develop something like a drive with internal RAIDing (i.e. each platter could be seen as a seperate 'drive' with data stripped across the say, 2 platters, theoretically doubling the throughput, with virtually no loss due to controller overhead). And as they would be using the same read/write heads, there would be virtually no disparities in read / write operations per 'drive', apart from the actual physical locations of the data on the platter.

Maybe I should patent that idea?

DublinGunner said:
It probably already is. The major problems I see in that is it would inherently make a good chunk of the drive's actually space inaccessible (strip size) and a lot of hard drives only have one platter these days thanks to perpendicular recording.. I believe data recovery would also be much more difficult. Ideally, the drive would have 8 platters so that every read would result in at least a single byte. The problem with that is 8 platter hard drives aren't cheap.

DublinGunner said:

increasing the number of platters increases the seek time the best situation is a single platter with a very high ariel desity.

No different than regular RAID (striping), RAID could also be deactivated, allowing the drive to be used as a single drive if wished.



Well you could say that about any current HDD.

Jim,
A single platter wouldnt fit with my idea though!!! You'd need at least 2 for the RAID capability.

(wow this has gotten OT lol)

Ah well, hard drives are going the way of the dodo with SSDs on their way. In 10 years hard drive production will be scarce.

DublinGunner said:

but your idea isn't RAID anyway... RAID = Redundant Array of Independent Disks.. what you're talking is not redundant, nor is it an array, and because of that there is no Independence... Its just "D" lol.

Obviously its its not true RAID (merely striping, but its easier for people to associate that with RAID). Then again, neither is RAID 0.

Touche...............

Redundancy is the key word there. If you're relying on 2 of something to work in order for anything else to work, then you've doubled your chance of a catastrophic failure.

Example: You are flying in an airplane which requires both engines to be working in order for it to maintain altitude. If you lose one engine, you crash. All other things being the same, the probability of one engine failing is half that of the probability of one of the TWO engines failing. So, I'd rather fly in a plane that requires just one engine to stay aloft than in a plane that requires both engines to stay aloft.

However, if you have a plane that requires only one engine to be aloft, but it has two engines, you have redundancy. In this case if one engine fails you will still stay aloft.

DublinGunner said:
Depends how the bits are stored on the platter. In current HDD schemes, most data is stored in a mostly sequential scheme so, if you can find just a little chunk which belonged to what you are looking for, you would likely find the whole lot in the same vicinity.

A platter-level stripping can spread it all over the place but in a somewhat predictable manner so long as you know in advance the specifics of that mannerism. Because every hard drive could have the stripping layout different, it would stand as another obstacle in data recovery.


~Vel said:
I wouldn't be so certain. Flash memory gets very poor marks from me on data survivability. Nothing short of a giant magnet can clear a HDD while it seems that just plugging in flash memory at the wrong time can produce unrecoverable damage to a flash-based medium.

I've had at least two SD cards get damaged and my sister apparently ran into another one. That's a failure rate of over 75% in less than a year of use. Absolutely ridiculous.

FordGT90Concept said: I agree with Ford. I've personal experience with and heard anecdotal evidence from friends and associates that indicates FLASH based memory remains too unreliable for any significant amount of mass storage. In addition, the write cycle lifetime restriction remains unacceptably low.

Now, if you're talking about a solid-state drive comprised of "regular" volatile memory, I've already written about that in another thread.

Dublin, virtual RAID existed in Linux / Unix platforms for a while now. And yeah, even recent versions of OS X, have this feature available. You can setup RAID for two or more partitions in a single hdd, as if they are separate drives. I don't think there is any performance gain for striping in this fashion, since it is the same hdd and performed by the same read/write head. Your idea differ by physically separating the "partitions" into two separate platters, but it's basically the same. Unless you're using two heads to read/write independently on these platters, then it may really make a difference in performance. However, that will make the drive more mechanically complex, and may be more prone to failure. Nonetheless, it's a good idea, though.

Back to topic. Like many said already, once the game is loaded into memory, hdd performance won't matter no more. It is the jobs for CPU / GPU / DRAM at that point. If you need more hdd performance, you can simply setup a RAID 0. It's more economical to do so now, than buying expensive high speed drives like Raptor and any of the SSD at this time. Most mobo these days can setup RAID dependable enough for general use. And really for gaming, you don't really need to worry about redundancy, so RAID 0 is the perfect solution for that purpose.

so hard drives are irrelevant to in-game performance? and thus all the fuss about the raptor from gamers far and wide is unfounded?