hard drives & raid - benchmark and compare!

[wevsspot]

Thanks soundx,

 

This particular rig is my main pc, runs my network and also hosts online games @ novaworld. Mainly trying to understand the best setting for supporting my network and online gaming. I'll let everyone else pipe in with their thoughts. As we all know, benchies are great, but they have to relate to real world performance in the end.

[suspekt]

This particular rig is my main pc, runs my network and also hosts online games @ novaworld. Mainly trying to understand the best setting for supporting my network and online gaming. I'll let everyone else pipe in with their thoughts. As we all know, benchies are great, but they have to relate to real world performance in the end.

My best guess is, that if this PC gets alot of requests to read from the harddrive at the same time, read caching and command queing is for you. It seems both read cache and command queuing has a bit of overhead when used, but if there are plenty of read requests and possibly also write requests, it might speed up things a little.

 

I haven't tested this really, but I would think that without read caching and command queing, every request for read and write gets equal priority (FiFo, first in first served), but with command queing it alters the request queue so that it can read all things in sequel... making some reads faster and some reads slower (as everyone gets different priority in the queue), but it should theoretically make all reads complete faster than with the FiFo queue system, which should be good for a multi-user/multi-application enviorment.

 

Did I make sense? Almost lost myself here hehehe

[wevsspot]

thanks everyone. your explanations make sense. since my main rig is also my network server and online game server i wanted to make sure that my users were getting the best performance.

[red930]

Game Servers will mostly provide file access through the server program its self. This is done by caching the data in RAM. If the server program had to go to the disk every time it needed data, performance would suffer greatly.

 

NCQ and Read Caching perform best in a multi-I/O environment like a database server with many client machines.

 

The mistaken use of I/O Meter as a benchmark on Desktop rigs has furthered this interest in I/O performance as if it has any true relevance to the issue of performance.

 

Put simply, unless and until your rig is performing hundreds or thousands of unique disk access commands per second, NCQ and Read Caching are a non-issue.

[wevsspot]

ExRoadie, you da bomb. So if I understand your post correctly, if synthetic benchmarks show a slight increase in the performance of my array with read cache "off", then thats the way I should leave it?

 

And the onboard cache of desktop hard drives is more for write purposes?

[suspekt]

ExRodie:

 

True. I have no clue what that novaworld is, if it was indeed something really big or just a regular type game. I recon some games can have databases, like those with very large worlds you can walk around in, with currency, trading and all that... ?

 

Heard of one of those games that got a real banking system (IRL) because people in the online game can earn real money, and they can access that gaming money from a real plastic card that they can withdraw cash from the ATM... talk about virtual reality becoming more real

 

And also if they bought some paintings in the game, it was the same art as a painter had made IRL, so not only could they decorate their house in the game with these paintings.. they would also arrive at their place IRL... heh, talk about spooky.

[wevsspot]

Novaworld is the host service for Delta Force BHD and Joint Operations. It's a large server and the maps can be huge.

 

I routinely host coop games where there are 16 to 20 players. Bunches of fun!

[BadWillHunting]

Upon looking at the bench scores (and real-world content creation peformance) of my seagate SATAII 160s, I bought two ST3250624AS spindles from Buy.com.

 

These units were advertised at 16 meg cache and SATAII interface (confirmed by checking the specs on Seagate's support site before purchase), but they bench out at SATA I and read in at the driver description in device mgr as "1st Gen SATA 150".

 

So... what is the issue, since seagate claims this unit is SATAII? Old unit? Old firmware? I researched firmware and supposedly the 7200.9 series is capable of proper SATAII with the version installed, which is Firmware revision 3.AAH .

 

Anyone got any ideas? Should I yell at Buy.com or ask Seagate what the hell is up with the disks?

 

Wouldn't matter to me that much, but that the real-world performance re-linking in dreamweaver 8.0 on a large site is impacted similar to what you see in HDTach and ATTO in terms of write performance, about a 40% hit. If I'd had any clue I wouldn't get SATA300 speeds, I wouldn't have bought these spindles!

 

Is there any limitation on how many Sata II RAID arrays the NF4 interface can handle? Is it only only 300mbs on the first array and 150 on the second?

 

Ideas anyone?

[soundx98]

The NF4 RAID controller can definately run 2 separate RAID 0 arrays at SATA II Speeds.

 

Have you checked the Seagate site. Hitachi drives need to have "Feature Tool" run on them (via DOS) to enable SATA II speeds. Perhaps there is something like that required for the Seagates.

[Angry_Games]

The NF4 RAID controller can definately run 2 separate RAID 0 arrays at SATA II Speeds.

 

Have you checked the Seagate site. Hitachi drives need to have "Feature Tool" run on them (via DOS) to enable SATA II speeds. Perhaps there is something like that required for the Seagates.

it can???

 

I'll have to retest this as the NF4 has only ever run a single RAID array (but up to 4 drives in RAID-0 or RAID-1) for me

[soundx98]

Yes it can.

Cross my heart and eat some flies

 

I think most people are under the same impression that only one can be run.

(they're probably so exhausted from setting up the 1st array)

 

I should qualify that:

I had 2x74G Raptor SATA 1 set as one array and 2x250G WD2500KS SATA II as the other.

But, I was able to reverse the set up order of the array so I don't think SATA speed makes a difference.

 

I had a 2x36G Raptor RAID 0 array plus a 2x74G Raptor RAID 0 array on the Sil Image 3114 on my old NF2 LPB (god I still miss that sista) and just assumed it would work on the NF4 as well and it did. Sometimes my ignorance is bliss :nod:

 

OT, Don't know if anyone has noticed it yet but Western Digital has released additional versions of the 36G and 74G Raptors with 16M Cache. The WD 740ADFD is now the fastest of the 3 sizes and only costs $10 more than the older 8M cache http://www.newegg.com/Product/Product.asp?...N82E16822136033

 

P.S. Nice sig - Happy

[Guest LithoTech]

And also if they bought some paintings in the game, it was the same art as a painter had made IRL, so not only could they decorate their house in the game with these paintings.. they would also arrive at their place IRL... heh, talk about spooky.

 

Now that's what I call convergence!