This is why SCSI 160 should help FS2002

[Noel 4,494]

Thanks Ray for the technical insight. Dual channel it is if that's what I go with.The 42% vs 19% was from http://www.storagereview.com/ and it probably had more to do with server installations where there are huge numbers of simultaneous requests. Dunno, but you will see the chart there. Huge numbers of requests should not be an issue I would imagine for me with my PC use. All this technical insight aside I'm simply interested primarily with significant performance/smoothness increases with FS, and in general quicker boot times, quicker application load times, etc. From the few testimonies I've received like Gino's above, the response has been clear that the upgrade to Cheetah drives was definitely worth it. Hard to argue with success. And this is why I keep harping on real world performance in Flight Simulator. Not theroetical transfer rates, not anything else. Besides the typical vast array of home uses for my PC, my main entertainment is with FS and Links 2001. And again, when I ran system monitor and saw the almost constant disk activity represented, it truly makes me deduce this is an area of performance that really needs looked into. Would you not agree that stuttering and less than perfect performance has been a consistent issue with FS users? I would imagine the vast majority of users are also ATA users, agreed? It is conspicuous to me then that the few folks who reported on their Cheetah SCSI systems remarked how much improvement in sim smoothness occurred. This may of course all be relative to other factors in their SCSI systems, or it may be something one can count on.I did chat with a friend with a local PC store who told me that ATA RAID would involve lower CPU utilization because of improved efficiency with the RAID controller over the standard onboard IDE controller. What are your thoughts on that? If SCSI 160 with a dual channel controller and 2 18.6 GB Cheetah X15s performed significantly better for my needs, then I want it for sure. The two drives are not for SCSI RAID, but just for a second drive. For the last 2 years I have never ever used more than about 24GB, and I don't anticipate needing 120GB anytime soon, so drive size is not relavent to me today. As for the relative costs, I can afford a few hundred $$ more. My first system was a 486 33Mhz, 8 mb of ram, 120mb HDD, 4x CDrom, no sound card, no modem, a 15" Mag monitor, and this cost me $2400. You get the drift.Let me know what you know about ATA RAID and CPU utilization if you know . . .Noel

[Guest JonP01]

Noel,I still can't help thinking that top-of-the-line IDE will still provide a satisfactory solution for you, even though SCSI seems to have it on paper and could well be better in FS2002.I looked up Storage Review where you got those figures, and the CPU utilization was measured at 10,000 i/o's per second. imo, this is far removed from the consistent but neverless light i/o rate that occurs when using FS2002. To my way of thinking, those Storage Review figures for light load i/o's would more reflect the scenario that occurs when using FS.Furthermore, my feeling is that a high quality drive such as the WD 120Gb would have the advantage of a high platter density - meaning that if you lightly fill the drive with data your seeks will be faster (less distance for the heads to travel) and the i/o rate will be at it's fastest because all your data is on the outside of the platters where the rotational velocity is fastest.You say that you don't feel high capacity is warranted, but there are possible arguments against this. The less you fill a drive with data (so long as it is properly defragged), the faster on average the read rate will be. For example, if you get the WD 120Gb drive and only fill it with 10Gb, then you are guaranteed maximum read rate 100% of the time. On the other hand, you half fill a small capacity SCSI drive and there is every possiblity your read rate with the SCSI drive could be quite similar - faster yes but not massively so. It is even possible it could be slower. Look at the Cheetah SCSI specs v's WD IDE specs and you will find some overlap. Yes, the Cheetah is faster, no doubt, but then again in the "real world" the head on the Cheetah may consistently be travelling further because you have filled a much larger platter area with data. To my way of thinking you'd have to get a high capacity SCSI drive to beat the WD in FS2002 and then the cost really does become prohibitive.No disrespect to those who have posted glowing testimony regarding SCSI, but from what I gather none of these posters have swapped out their SCSI hardware and dropped the 120GB WD drive in for direct comparison purposes on the same machine. How do they know the WD might do just as good a job in FS2002 (perhaps even better)?Even last night I tried an experiment where I deleted all programs off my hard drive except FS2002. By doing this I reduced my used disk space from 60% to about 10% of my hard drive space and then defragged the drive. As result, FS2002 and CFS2 ran very smoothly, with no "micro-stutters" at all to speak of (and I was looking very hard for them). I sumise this is for the reasons I stated above (ie less seek distance and maximum read rate). And this is a mere 18 month old Quantum Fireball LM that can't compete with the latest WD drives except when it comes to seek rate. So if I bought the latest WD drive and put all my programs on it, I expect it would outperform my current Quantum Fireball with just FS2002 installed on it - such are the differences in outright capacity and platter density.To your other point, this raid looks interesting, but I'm still not really sure how relevant CPU utilisation is at all, given I believe FS2002 represents a consistent but light disk load.I guess it will be very hard to definitively resolve the argument one way or another. As I said before, you could get your SCSI gear and be over the moon with it (I hope you will be). But unless you did a direct comparison with the best in IDE then there is no way of knowing which is actually better in FS2002.

[Guest]

Just some quck numbers off my box ;-)I have the slower WD 80 Gig with no raid, just standard ATA-100Running FS 2002 Pro in the FS Genesis Alaska mountain area (figured that was a high throughput area being enhanced mesh)Highest Disk Read: 102,257.022 Average: 7653.960Highest Disk Write: 130,133.364 Average: 1468.733These are in bytes/second (not cache testing, physical disk requests from Windows 2000 Server Performance monitor threads)Since we are dealing with drives that have 8 Meg caches available to them, maximum requests in the 130Kytes/sec on a drive that can sustain about 50 Mbytes/sec obviously should not stress you in any way shape or form ...Max Read : 27,815,718 bytesMax Write : 11,780,154 bytesCopying a huge VOB file from the drive to the same drive, putting the drive to it's maximums shows what the drive can really dish out. As you can see, the utilizations by FS2002 are NOTHING compared to what a moderately high end IDE drive is capable of. My 80 Gig shows itself in the 40 Meg/sec range, You can look up the improvements between my 80 and the current 120 to give you a clear idea of how much better it will be ...The numbers on Storage View are based on a multi-user filesystem with many access attempts at the same time (a server), where SCSI comes into it's own right, with it's multi-threading and command disconnect/queuing that IDE just can not compete against by design. This is not realistic for an end user single tasking desktop computer unless you feel like running Oracle or DB/2 or some other SQL engine in the background while playing with FS2002 ;-)PS: These are real numbers, off my real machine, running FS 2002, not a staged benchmark, so they reflect reality a tad better ...Ray

[Noel 4,494]

John thanks for posting this info. It's so great we can tap into others' knowledge and experience with PCs and FS and everytning else. I agree fully there is no way to really resolve this argument without comparing ATA vs SCSI 160 vs ATA RAID in the same box, in the same situation, etc. Unfortunately I am not quite dedicated enough to try that test. But I wonder how old Bubba W might be?! Maybe I'll move over there and see if I can contact him for this?! Certainly for all the FS users out there, if Cheetah X15s would truly improve performance with regard to disk access impact on stutters I think there would be some definite interest.I think an acid test might be to do the following, and you need John Hnidec's excellent utility to do this: load FS, get to a runway, start FSFlightMax, and turn it on just as you roll down the runway. This sucker loads a hefty db of world navaids etc and it is always amazing how it kills FS into a very slow and erratic slideshow on my machine while the db is loading. It's so bad of course that I do this while parked, not while moving. It's evidence of what large file reads do during FS, and certainly suggests the continuous low-level disk activity could be quite responsible for micro-stutters as you call them. I should see if Gino has FSFmax!Noel

[Guest JonP01]

Ray,Do you have the new 8MB cache WD 80 GB drive?? If so, I'm curious about one spec so that I can directly compare it to the drive in my box. I'm interested in the "Random Access Time" recorded by the HDTach benchmark utility. I'd like to know what this figure is with both the "Advanced Size Check" checkbox checked and unchecked. The reason I am interested in the unchecked figure is that it measures seek time on the outside of the platter (I think just the first 8GB). As my intention is to buy a large capacity drive and lightly fill it (for the reasons I stated above), this figure most interests me. There is no doubt read rates on the latest WD drives blow my drive away, so it is really seek time that matters when I do my comparisons.Thanks a lot if you can test this for me.

[Guest]

Nope, those are the numbers of the older 80 Gig drive, with the much slower performance, compared to the new special edition version with the larger cache and higher platter density. :-lolSorry, but it would appear that HDTach requires registration for Windows NT/2K use :( so I can't test for you ... Unless you care to spring for the fee :-lolRayEdit:Best I can do to give you a reference, heres a hack out of pcbench2002 saved XML file ... if you have some other form of benchmark you want me to test that is not locked down, let me know...Cached file write 17.6136 MB/sUncached file write 25.8478 MB/sCached file read 26.9071 MB/sUncached file read 28.9537 MB/sFile copy 10.511 MB/sStrange numbers eh? guess Windows 2000 & benchmark software don't quite get along :-lol

[Guest JonP01]

Thanks Ray. Pity about HDtach. I guess they expect people with those OS's to all be "professionals". A bit bizarre, but nevertheless.The reason I mentioned HDTach is because I got very interesting seek time results. On my Quantum fireball, if it does random seeks across the entire platter surface, then I get approximately 12 ms access time. This equates to roughly 8 ms seek time after taking into account rotational latency.On the other hand, restrict the seeks to the first 8 GB, and the access time drops to 9.5 ms, or roughly 5.3 ms seek time - right up there with the SCSI drives. That is why I am saying that WD's top drive filled to only 10% capacity could to comparable with a small capacity SCSI drive. I'm sure my drive isn't the only one to experience dramatically reduced seek time when all the data is restricted to a small area of the platter.