Software rebuild on primary PC question

Hi everyone

I have just bought a new HDD for my primary (i.e FS external view) pc and am planning to create a custom os install and a separate dedicated drive for FS9.

I have the following:

2x80gig old 1st generation sata seagates (7200rpm) in a raid 0 config. Currently this is partitioned into a c: and an e: drive. C:is the OS, E: is FS plus any other apps (currently fsX and not a great deal else)

I have bought a single 500Gig WD5000AAKS drive, Sata 3 and 16meg cache. According to THG it is competitive with a Raptor on all but seek time, which is not far behind.

I am planning to have two primary partitions (one for the FS OS and one for the non-fs one) and a dedicated drive (Single partition) for the FS9 install and nothing else. This pc has to have a non-fs role as well as it is by far my most capable PC and (just occasionally!) I play other games and

I have two questions:

1) should I use the new 500 gig drive for FS or the older (but RAID 0 ) pair? On that note does anyone know a good benchmarking tool I could compare the two with - and what factors would I look for?

2) The recent MS SQL 2008 Launchwave event scored me a copy of Vista Ultimate w SP1. Should I use Vista or XP for FS9?

My other specs are:

Intel dp35dp mobo
E6600 C2D chip
2Gig of DDR2 667 ram
8800GTS 320 Gig video card
2 monitors, used as two separate views in FS9, 22" lcd for external view, 15" lcd for the instrument panel.
Controls as per my sig below but these are no factor in the decision I need to make..

Thanks in advance.

James

Unless you plan to upgrade to FS10 in the near future, I'd run FS9 on XP or XP Pro.

I'm not upgrading for quite awhile. I have too many add-on's incompatible with FSX, particularly my aircraft. :!:

The new system I have was bought specifically for FS9, and I would not get the performance out of this system with FSX that I'm enjoying with FS9.

Your system specs are comparable to mine with your processor faster, but FSB slower. Your video card is better than mine. You might get acceptable results with FSX, but as with my old system and FS9, I was disappointed with the slide show effect in a metropolitan area.

I too am using two monitors and am adding a third soon to a one computer system as Gwyn has done.

The larger the drive, the more efficient it is since most of the data can be in the outer tracks and the mechanical movement of the heads is minimized. I would go with the large drive myself since mechanical lag has the biggest effect on performance.

I also think that partitioning the drive does not improve FS performance in any way. A good defrag on a large drive is the way to go in my opinion.

Maurice

Partitioning the drive actually slows it down.

Hi and thanks for all the feedback.

I didn`t realise that partitioning slowed things down. I have been faithfully partitioning a drive for FS (and other apps) since about fs2000.

In light of this, given that I want dual boot with a Vista Ult and XP pro + FS installs, should I break the RAID 0 pair into two 80 gig drives (one for each OS) or should I create two partitions on the RAID array?

Redundancy is not an issue as the gaming rig has all data backed up elsewhere.

I have FSX but not Acceleration and and am ambivalent at the moment - I kind of like the flight modelling changes (the default aircraft seem less unstable in the pitch axis) but I REALLY wish you could use the "Previous situation" when restarting. How could they have taken that away??
Also, IMHO, a tricked out FS9 (which I have) is better looking than FSX with a moderate range of addons (of which I have very few)..

Cheers

James

James

This is what Nick N at AVSIM has to say about RAID 0:

"NEVER USE RAID-0 FOR FSX UNLESS IT IS A DEDICATED CARD (or motherboard RAID) THAT CAN SET A 256K STRIPE OR LARGER PRIOR TO WINDOWS/FSX INSTALL

Motherboard and cheap PCI card RAID kills FSX performance because of the lack of ability that format of RAID typically runs. Unless you are using a -real- RAID card, http://www.newegg.com/Product/Product.aspx?Item=N82E16816116042 with dedicated DDR2 memory on the card you are better off on a SINGLE SATAII drive than running motherboard software or a cheap PCI card RAID 0.
In all of my RAID tests, any STRIPE at or below 128K produced scenery load stutters. FSX needs a 256K (or higher) STRIPE RAID
Older systems will magnify this issue because of the high resources FSX pulls. You must understand that RAID has its ups and its downs. The up side you are aware of which is it allows a greater throughput for data to flow, IF and only if set up correctly for the data chunks being accessed. "

The down side is that every drive you install in a system that is accessed at the same time -pulls- from the resource pool so if you are not compensating by optimizing the data flow, you are losing with RAID, not gaining. Slower systems should not try to run 2-4-5 drives in motherboard RAID for FSX, only a dedicated RAID card for such setups.

Motherboard software RAID sucks the life out of a CPU because every drive being polled must get CPU time and every file being called requires cycles. Low STRIPE with FSX will force 3-5x the CPU cycles because the files are broken into far too many parts. The larger STRIPE breaks the files into a reasonable amount and the dedicated HARDWARE RAID card (on a PCIe 4x bus, not a PCI slot) removes the CPU load so therefore you are recovering CPU with a RAID card, the opposite of what motherboard or cheap RAID cards force on the system.

FS9 had an average file size of about 64-128k. FSX average file size is 5-10 times that so if the STRIPE does not grow to compensate, the RAID array will cause stutters from CPU load, and, from the ridiculous access that must take place to call up a single file.

Also, FSX flies and does not fragment a drive if it is installed to a different drive all by itself, and, NEVER EVER PARTITION a OS or GAME performance drive. Always leave them without partitions"



I would suggest listening to his advice as he knows his stuff.

Regards,
Henry

Thanks, Henry

In light of this info, is it okay then to put the OS on say a 40 gig HD and then FS on a much larger HD, both with no partitioning?

Hi Eric

Yes, that is what is recommended.

Here's a very useful link of NickN tips:

http://forums.avsim.net/dcboard.php?...0281&mode=full


Regards,
Henry

Henry

Many thanks for the detailed information. I`m assuming the general OS performance info applies to FS9 as much as FSX as I am running 9 not X?

In a read-only situation such as FS, would there be any benefit to running RAID1 or 1+0 even, assuming the raid controller is a good one? Two concurrent reads of would mean increased data throughput wouldn`t it? Even if seek time were the issue, RAID1 wouldn`t deteriorate seek time would it? Ok, there`s a slight negative hit on the write performance but for FS that wouldn`t matter would it?

If it would be of demonstrable benefit I would happily buy a RAID card.

Once again thanks for all your advice. I will post back in a week or so when I have rebuilt and had a chance to give it a serious thrashing.

Cheers

James

Hi James

Yes, the performance info would apply to FS9 also.

A RAID setup is most beneficial when dealing with large files, but since FS consists of thousands of small bgl's, textures and scenery items I think you would be better off with a raptor drive and forget about RAID. As mentioned previously your best bet would be to have FS on a separate drive with no partitions. You should also defrag at least once a week.

It's also a well known fact that FS (espescially FSX) is CPU bound, graphics is next, then amount of RAM, and so on. The CPU you have now (e6600) is more than sufficient for FS9. But, if you have money to spend, instead of putting it towards a RAID card and extra hard drives (I mean you would need to have 4 HD for RAID 1+0), you should upgrade your processor to a e8400 or e8500 and overclock that sucker and then your system would really fly (pun unintended). You would also be ready for FSX with no problems.

Regards,
Henry

Henry

I really appreciate you taking the time to give me such a detailed reply.

With FS being primarily cpu bound, I will indeed investigate getting a faster CPU and run with the non-raid dedicate drive setup.

Now for the fun part, rebuilding my FS install with its hundreds of mods / tweaks!

Thanks again

James

Hi Henry,

Re: 04-29-08 01:55 PM

I have just read your comments on RAID0 and I have indeed experienced exactly the same with similiar conclusions.
I am now tuning FSX (RAID0) on a very large VISTA ASUS MAXMUS EXTREME QUAD CORE system and I am investigating how to optimise Memory Management.

The article written by Jeff Atwood in http://www.codinghorror.com/blog/archives/000688.html
and the subsequent forum discussion is very informative and well worth reading and investigating.
I have improved my frame rates significantly by understanding what is happening but also expensively by using 4GB of DDR3 1800 overclocked.
The reality is that you are wasting your time and money in a VISTA 32 Bit system by having anything over 3.5GB (approx.) of memory. This is important for people contemplating purchasing more than 4GB of expensive memory - not a good investment!
It is also worth remarking also that "the jury is out" on the advantages of dual channel versus single channel memory in terms of its contribution to performance. My experiments to date demonstrate no observable improvement!

The interesting part of all this is the significant positive difference Super Fetch makes to VISTA performance and also the cache management strategy.

Have you done any experiments in this area as I would be delighted to share
my experiences.

Bill.

Quote:

---

Originally Posted by Prof Bill

Hi Henry,

Re: 04-29-08 01:55 PM

I have just read your comments on RAID0 and I have indeed experienced exactly the same with similiar conclusions.
I am now tuning FSX (RAID0) on a very large VISTA ASUS MAXMUS EXTREME QUAD CORE system and I am investigating how to optimise Memory Management.

The article written by Jeff Atwood in http://www.codinghorror.com/blog/archives/000688.html
and the subsequent forum discussion is very informative and well worth reading and investigating.
I have improved my frame rates significantly by understanding what is happening but also expensively by using 4GB of DDR3 1800 overclocked.
The reality is that you are wasting your time and money in a VISTA 32 Bit system by having anything over 3.5GB (approx.) of memory. This is important for people contemplating purchasing more than 4GB of expensive memory - not a good investment!
It is also worth remarking also that "the jury is out" on the advantages of dual channel versus single channel memory in terms of its contribution to performance. My experiments to date demonstrate no observable improvement!

The interesting part of all this is the significant positive difference Super Fetch makes to VISTA performance and also the cache management strategy.

Have you done any experiments in this area as I would be delighted to share
my experiences.

Bill.

---

Hi Bill

Here's an extremely good post(s) on how to optimize your memory:

http://forums.avsim.net/dcboard.php?...topic_id=37210

I'm still running XP so I can't comment on Super Fetch. Interesting article..... did you experience the same side effect that he had when the game (FSX in your case) initially started up?

Regards,
Henry

Hi Henry,
Many thanks for this link!
I am just going out so I will examine and digest it on my return.
It does look like very familiar territory which is most helpful in terms of what Ihave covered.

Bill.

Hi Henry,

I totally concur with the findings of the article and it would appear that Nick N
has gone further than I in his investigations and indeed has saved me much experimentation and effort!
He has confirmed what I felt about the multifaceted interdependency of CAS, RAS, etc., etc. and DDR3 and how worthwhile it to move there with VISTA if the "price is right"!
I am seeing the improvements that Nick N has experienced and FSX working as it really should - that can only be good news!

I did discover a link that is of real interest in terms of memory performance improvement and is worth a visit in terms of graphical explanation and opportunity to experiment.

http://www.thetechrepository.com/showthread.php?t=195

Many thanks,

Bill.

Hi Bill

Thanks for the excellent link. It pretty much what Nick is saying except that it's much easier for the average person to follow.

It's very unfortunate that most people are only using high frame rates as the only benchmark that FS(9)(X) is running well when there is so much more to look at.

Regards,
Henry