7200.8s - Are you ready to RUMBLE?!! (Don't answer that)

acaurora · Post by **acaurora** » Thu Nov 18, 2004 5:03 am

http://www.techweb.com/wire/hardware/53200125

As soon as my store (Best Buy) has it, will post results.

BTW, is NCQ supported natively by SATA I ? Just curious.

Mats · Post by **Mats** » Thu Nov 18, 2004 6:53 am

No.

Read my posts in this thread about which chipset you need. Or, use a controller card with NCQ.

.....OR, just wait and see how good NCQ is first!

I mean, I can buy a 7200.8 but I'm not sure it's because of the NCQ...

acaurora · Post by **acaurora** » Thu Nov 18, 2004 11:25 am

I got a DFI LanParty Ultra II B, which uses Silicon Image's Sil 3114 SATA controller. According to StorageReview.com (browsed the forums), it appears that Silicon Image makes two different models of the Sil 3114. However, both do support NCQ.

So, I guess that means that i do have the hardware that supports NCQ, now all I have to do is wait for Best Buy to recieve and start selling the Seagate 7200.8 400 GBs w/SATA/NCQ so that I can unlease and utilize the power of "Employee Discount" ;P (that really is mainly the only reason why I work there).

Trip · Post by **Trip** » Thu Nov 18, 2004 11:45 am

NCQ is only helpful for servers, right?

storage review conclusion

grandpa_boris · Post by **grandpa_boris** » Thu Nov 18, 2004 11:53 am

Trip wrote:NCQ is only helpful for servers, right?

NCQ serves the same role as tag queueing in SCSI: if you have a lot of concurrent access requests to your disk, NCQ will make the disk perform significantly better for the aggregate throughput on all I/Os.

where will this matter to you? if you are editing photos or assembling a film while listening to MP3s on the same system, you'll get some of the benefit. but not enough to spend a lot of extra money on it, else you'd have seen everyone use SCSI disks instead of IDE for the last 15 years.

Trip · Post by **Trip** » Thu Nov 18, 2004 1:29 pm

That's true even despite Storage Review's conclusion?

Storage Review wrote:2. Command queuing is meant to assist multi-user situations, not single-user setups. With the recent release of Intel's 9xx chipsets, pundits and enthusiasts everywhere have been proclaiming that command queuing is the next big thing for the desktop. Wrong. As evidenced by the disparities between the FastTrak S150 TX4 and TX4200 (otherwise identical except for the latter's added TCQ functionality), command queuing introduces significant overhead that fails to pay for itself performance-wise in the highly-localized, lower-depth instances that even the heaviest single-user multitasking generates. It is becoming clear, in fact, that the maturity and across-the-board implementation of TCQ in the SCSI world is one of the principal reasons why otherwise mechanically superior SCSI drives stumble when compared to ATA units. Consider that out of the 24 combinations yielded from the four single-user access patterns, one-to-four drive RAID0 arrays, and RAID1/10 mirrored arrays presented above, the non-TCQ S150 TX4 comes out on top in every case by a large margin. TCQ is only meant for servers, much like the technology mentioned just below.

from previously linked article.

grandpa_boris · Post by **grandpa_boris** » Thu Nov 18, 2004 3:01 pm

Trip wrote:That's true even despite Storage Review's conclusion?

an incompetently written driver or incompetent, immature firmware can defeat even the best technology. an incompetent implementation can turn the best protocol spec into a poorly performing product.

you will see no improvement if your workload consists of discrete, non-concurrent requests. what most users do with windows and the way most software on windows is written, you simply can't get the kind of i/o regime that would benefit from tag queueing. i don't know anything about the workload used by the review, but it sounds from the comments that it's predominantly a single task at a time affair. i see nothing about running multiple concurrent i/o streams or, for that matter, the use of a competent SMP OS to be able to generate and consume the multiple independent i/o streams.

i've seen configurations where the same multi-threaded, multi-streamed benchmark was on a scsi drive with the queue depth set to 1 and queue depth set to something meaningful like 128. the difference in performance is drammatic. but these are benchmarks that measure a medium to large scale file servers, not a desktop PC.

these days all disk drivers are written to take a bunch of requests, sort them using the so-called "elevator algorithm", trying to match the discrete I/O scheduling as closely as possible to the disk geometry. this results in a probabilistic external approximation of the task sorting and scheduling the disk firmware could do on its own. there is research that shows how this scheduling can be pushed beyond imaginable limits to realize nearly the full bandwidth of the disks. the point of NCQ/TCQ is that if you provide enough loading on the disk and have even the most primitive priority differentiation, you can greatly improve throughput without the monumental efforts host software would expend.

but as i said before, this is largely irrelevant to the typical desktop environments, unless you are running multiple disk-intensive tasks.

Trip · Post by **Trip** » Fri Nov 19, 2004 8:21 pm

that's good news. I'll look forward to future tweaks then.

Mats · Post by **Mats** » Sat Nov 20, 2004 1:32 am

grandpa_boris: I see that you have a 2.5" Seagate, tell me what you think about that one. I've heard two different ones and I think they are really low noise, no one here agree though. I have a feeling that ALL 2.5" Seagates got a worse reputation than they deserve. What do you say?

grandpa_boris · Post by **grandpa_boris** » Sat Nov 20, 2004 5:01 am

Mats wrote:grandpa_boris: I see that you have a 2.5" Seagate, tell me what you think about that one. I've heard two different ones and I think they are really low noise, no one here agree though. I have a feeling that ALL 2.5" Seagates got a worse reputation than they deserve. What do you say?

the one i have is horrible. it has a piercing high pitched keening whine. i suppose if i had a fan in that system, it would drown out the drive. but there is nothing in that box to conseal the momentus' squeal. i i recently had a debilitating filesystem corruption on that system, so the box is off. when i reconfigure and restart it, it will be running off a flash card (running m0n0wall, most likely). i don't blame seagate for the fs corruption, i blame linux' ext3 and its fragility, as well as its braindamaged fsck.

so that's where i stand today. if i were building this system today, i'd be using a hitachi drive that gets much better reviews. for that matter, a 1GB compact flash costs about the same as an 80GB 2.5" drive (in the $110-120 range) and for what i need, 1GB is plenty of space.

Mats · Post by **Mats** » Mon Dec 06, 2004 9:02 pm

I really like reading this:

Subjectively the three platter [ATA] 400GB is about as quiet as my old two platter Barracuda IVs or one platter DM+8s--very quiet with faint chattering on seeks. A lot quieter than the five platter 400GB 7k250. AAM may be gone for good post-lawsuit, but it is still quiet with fast seeks too.

...yeah maybe he's deaf/insane/not SPCR/blabla... I still like it!

From Storagereview.

Any reviews somewhere? Strange....

jimveta · Post by **jimveta** » Tue Dec 07, 2004 2:52 am

acaurora: are you sure that the 3114 supports NCQ? i spoke with a colleague at work who has worked on drivers for it and he said it's just a 4-port version of the 3112 with no ncq. the linux sata status report page also seems to confirm this.

btw, i agree with grandpa_boris and to be honest, perhaps because of my own biases, but i would say more from my own experiences, i've lost some respect for storage review after reading their conclusions -- i don't doubt their results at all, it's the very generalized sweeping conclusions they (illogically) derived from that. the ONLY thing one can conclude is that for *windows*, with the current state of its i/o framework and hba drivers, that would be the case. but for them to imply in general, that there's no way a "single user" can generate enough i/o load to take advantage of ncq/tcq is pretty ridiculous as most certainly i can prove that wrong on my *nix boxes by running a few scripts with tcq turned on and off.. all the while as a "single user"

i must say though, i wish i could apologize to some folks who i didn't quite believe at the time, who said that for them, supposedly "faster" scsi drives (that they used) felt significantly slower such as on boot up, or opening docs, copying files, etc. as i have not much experience with scsi on windows,, but now i have a better understanding why.. as i've also recently run into some of the older scsi drives which have poor (sequential) transfer rates and _only_ have tcq going for it, which under these circumstances again doesn't seem to really be taken advantage of and would be probably be hard pressed to in windows.

however, i have hope though that the situation will change as sata ncq becomes more widespread especially in the desktop market; maybe a driver solution, a kernel i/o scheduler solution or both is forthcoming.

acaurora · Post by **acaurora** » Tue Dec 07, 2004 3:09 am

1st Party DMA support for SATA II Native Command Queuing

By cold_c

Near the end.

Also noted here, officially.

jimveta · Post by **jimveta** » Tue Dec 07, 2004 8:11 am

hmm, curious way of stating support for it though.. and as for the two versions of the same model that support ncq, you weren't perchance thinking of the 3124 instead were you? because that's the one i know does support it and comes in two flavors.

see their product pages here:
http://www.siimage.com/products/storage.asp#semic

the 3114 description doesn't have the "1st party dma support for.." statement there from the press release; however, the 3124 explicitly states ncq and older style tcq support for the two flavors

in any case, thanks for the pointer because i could be wrong (maybe the 3114 was revised to support it, or i just heard wrong) and as i'm no expert in this area, i'm gonna ask around since i'm looking into purchasing an ncq controller as well.