Poor Seagate 7200.7 performance

[LocutusX]

I've got the PATA version of the Seagate 7200.7 120GB drive (8mb cache), but it seems to be exhibiting poor performance. Here's what HD Tach 2.61 says:

Read speed (max.) : 48632 kps
(min.) : 6258 kps
(avg.) : 36695 kps


Seems a bit too slow doesn't it? What could be causing it? None of the reviews of this drive on the net have a max STR that low; they've all got at least 57MB/s. Note that the really low "min" is due to downward spikes. By the end of the STR graph, it's actually at 20MB/s.

OS: WinXP SP1, KT266A mobo, using mobo IDE controller, fresh install of WinXP.

 

[butch84]

do you have the via 4in1 drivers installed? your ide performance (not to mention other chipset related things) can be adversely affected without it. Also, you could make sure that you have DMA enabled, if you havent already.

good luck

 

[imported_Phil]

That is a low reading, mine tops out at 69.5Mb/sec and averages 55Mb/sec. However, I'm using the nVIDIA IDE drivers that some people say cause instability- they might skew the results a little.

Could you try it in a different computer?

 

[Viper96720]

Acoustic Management Disable
Scores look similar to storagereviews 7200.7 160GB version.

Transfer rate 55.6 begin 32.8 end

 

[LocutusX]

Acoustic Mgmt might be causing it? But there's no way to turn it off, right?

 

[Accord99]

I doubt it's acoustic management, as it only affects the random access times. The transfer rates are unaffected by it.

 

[imported_Phil]

Originally posted by: Viper96720
Acoustic Management Disable
Scores look similar to storagereviews 7200.7 160GB version.

Transfer rate 55.6 begin 32.8 end

I don't think you can disable the acoustic management in the 7200.7 series- I tried it with a wide variety of tools, and none of them would touch it.

[Edit] Here's some benchmarks from the drive, running on an nForce-2 based board; HDTach and ATTO.

 

[nboy22]

Go into your control panel ->Administrative tools -> Computer Management -> Device Manager -> expand IDE ATA/ATAPI Controllers then go to either the primary line, which is probably the one you have you hdds hooked up to.. or your secondary if you have them on the secondary on the mobo. just right click primary or whichever, then hit properties, then go to the advanced settings tab and tell us what UDMA mode you are running in.

 

[LocutusX]

Don't worry, I'm in dma mode 5.

Tried the following:

- tried the drive on both the VIA onboard controller and Promise pci controller:

- using default WinXP SP1a drivers for both VIA onboard IDE and Promise controller, compared them with VIA's latest 4-in-1 and Promise's latest drivers; negligible performance difference between them (Microsoft vs. OEM vendor)

- used Hitachi Drive Feature tool to "check out" the drive. it reports I have 8MB cache, both types of caches are enabled (read-ahead, and the other one), under Acoustic Management it reports something indicating it's unable to access it. Seagate's SeaTools diagnostics gives me a "Drive is fine" report in the Quick Test.


Anything else?

my firmware is 3.76.


Just to reiterate, this is the Seagate 7200.7, plain old ATA/100, 8MB cache, 120GB version.
model #ST3120026A

My HD tach is 2.61.

Gonna try ATTO now, maybe HD Tach is just messed up...
I tried Sandra, but the results for that are partition-specific, and seems a bit messed up.

 

[nboy22]

I can't think of much else.... maybe different cables.. I doubt it's the cables.. but it's worth troubleshooting every little thing u can.. are your cables ATA66 or anything?

 

[IPLaw]

I have the exact same drive hooked up to an AOpen AK86-L, which uses a newer iteration of the VIA IDE controller and the same VIA 4-in-1's. Here are my results with HDTach 2.70:

RAT: 15.7
Read Burst: 87.0 MB/s
Read Max: 60.2
Read Min: 25.3
Read Avg: 47.7

Do you have more than one drive on the IDE Cable? Also, make sure that ANY and ALL programs that may write to the disk, or intercept writes, are closed or disabled. Use the task manager and msconfig to ensure there are no programs slowing you down.

 

[mechBgon]

Bigger picture: you're expecting high performance out of a drive with an average access time that approaches 15ms?

.

It's a quiet mouse-drive. I have one at home. It got even quieter after I unplugged it and put my Cheetah X15-36LP in there :evil: Woohooo, now we're cookin' with gas!

 

[IPLaw]

No, he's expecting to achieve reasonable performance in light of the results of others with the same drive.

 

[mechBgon]

Originally posted by: IPLaw
No, he's expecting to achieve reasonable performance in light of the results of others with the same drive.

I suppose so, although I have to wonder if the missing STR is going to make much real-world difference. Anyway... LocutusX, what exact brand/model of motherboard is it, what PCI devices do you have in it (if any), and what slots are they in?

 

[mechBgon]

Heh, well I'll take a guess that it's this system since LocutusX is offline. An add-in disk controller on a KT266A board, huh? WITH a PCI sound card too. And presumably a PCI network card, or else a USB connection to your broadband modem... either of which is equally bad.

I've been down this road and fought this battle with a KT333 board. I was never able to get more than 72MB/sec throughput from my SCSI card (it was 49MB/sec initially), and when I switched to an nForce board that jumped to ~122MB/sec (peak). Can I suggest treating yourself to a shiny new Abit NF7-S with nVidia Hypertransport-based NIC, nVidia Soundstorm audio, Firewire, USB 2.0 and high PCI performance for your Promise IDE card, which will now have the PCI bus to itself (unless you go using the onboard SATA at some point).

 

[LocutusX]

I can't think of much else.... maybe different cables.. I doubt it's the cables.. but it's worth troubleshooting every little thing u can.. are your cables ATA66 or anything?

Yup, tried 2 diff. cables. Rounded 80-conductor.

Here are my results with HDTach 2.70:

Thanks for those results, but I don't think they're comparable with 2.61, and I've heard 2.70 was "notoriously unreliable" (from the StorageReview.com forums)

Do you have more than one drive on the IDE Cable?

Nope, I always moved my Maxtor DiamondMax Plus 60 to the other controller whenever I benched this. Both drives were Master on each respective controller, with the Slave spot empty.

Bigger picture: you're expecting high performance out of a drive with an average access time that approaches 15ms?

Actually, I've been trying to get my drive to perform reasonably similar to the many posted performance results on forums as well the various reviews on the 'Net. One review was done on a P3-800EB and even he got better performance than me (80GB model)!

My paranoid theory: This drive performs almost identical to a Barracuda ATA V, rather than a 7200.7! Could someone be releasing old ATA V's with 7200.7 labels? But how does that explain the drive's firmware/BIOS identifying it as a 7200.7?

Heh, well I'll take a guess that it's this system since LocutusX is offline. An add-in disk controller on a KT266A board, huh?

I would have agreed that the poor PCI performance of the VIA KT266A chipset should be a concern... if not for the fact that even when I use the drive on the "South Bridge" VIA on-board controller (which supposedly bypasses the PCI bus, according to VIA's diagrams) I get the exact same STR performance!

An add-in disk controller on a KT266A board, huh? WITH a PCI sound card too. And presumably a PCI network card, or else a USB connection to your broadband modem... either of which is equally bad

I know what you're getting at, but remember that during benchmarking I wouldn't be playing MP3/WAV files or doing LAN file copies. They're all on different IRQs anyways; I juggled the cards around in the slots to accomplish this, something I did about 2.5 years ago.


At the end of the day, I'm not too deeply concerned about this, unless the drive ends up having physical defects causing bad sectors, etc. When I build my Athlon64 system I plan to buy a fast Raptor and make that my system drive so this 120GB will be relegated to holding media and data.


Oh, and I tried that ATTO benchmark. It's really messed up! It says my DiamondMax Plus 60 is faster than this "7200.7" that I have, while my subjective "feel" of the 7200.7 is better than the DM+60.

 

[mechBgon]

The KT266A's skimpy V-link ...well, I'm not going to embark on another rant here. I hope your A64 rig lays the skeletons to rest in the end.

 

[LocutusX]

oh, do tell... the KT266A's v-link was really crapulous I take it?

not as if I'm particularly surprised or anything

hmm, I wonder why I keep buying VIA, as I recall my P3-600E @ 840 build was also on a VIA board and I also had some "fun" with it, heh

 

[VirtualLarry]

Originally posted by: LocutusX
oh, do tell... the KT266A's v-link was really crapulous I take it?

not as if I'm particularly surprised or anything

hmm, I wonder why I keep buying VIA, as I recall my P3-600E @ 840 build was also on a VIA board and I also had some "fun" with it, heh

I have an MSI KT4V-L mobo, with a KT400/8235, and I get only about 75% of the theoretical max burst xfer rates, on both the (non-PCI) IDE, as well as my Promise Ultra100 TX2 PCI IDE. There was an issue with this mobo in the past, supposedly leading to "IDE DMA CRC errors". My guess is that the BIOS workaround implemented, to reduce inter-chipset noise, was to slow down the V-Link bus. Is there any software tool that I can use to actually benchmark/verify the speed of the V-Link bus itself, or possibly a way to read out the chipset registers to determine the V-Link bus speed settings?

I'm really quite curious. Also, my DRAM memory-write benchmarks are significantly slower than published benchmarks for the same system specs (CPU, chipset, DRAM speeds). Read speeds are about the same. That too, seems slightly suspicious, and I wonder if it is somehow related to the slower PCI/IDE performance issues.

As far as the OP's HD in question, this seemly isn't the first that I've heard about a possible Barracuda ATA-V under a 7200.7 label. Google around a bit, I don't think I'm crazy for thinking I remember something similar popping up before. I'm kind of surprised that the starting STR is so much lower than a "normal" benchmark of that drive.

DopeFiend's benchmarks look a bit high, actually, as well. Even with the NV IDE drivers, I don't see how he is getting those scores, unless that drive was built using Seagate's new 100GB platters, and simply not mentioned anywhere. (Part of the production ramp-up to the 7200.8 drives?)

 

[LocutusX]

VirtualLarry,

If I`m not mistaken, DopeFiend has the 200GB version. The 200Gb version is unique in 2 ways:
- 100GB platters
- has NCQ! and its enabled on both IDE and SATA versions.

so his drive will certainly perform much better than the rest of the series.


Back to the OP,

I think the problem I had with my drive is universal or something. I downloaded the 3.0 beta of HDTach from their website, it has a database of common drives and benchmark results. Someone put a seagate 7200.7 120GB IDE in there and... guess what... same performance as my drive! Note that it was NOT me, I have nothing to do with HDTach, and that database entry is dated April 25th. Also interesting that particular drive was tested on an nForce motherboard`s controller!

If you want to see for yourself, go to the official HD Tach forum, the author posted a link to the beta in one of the threads. The link was working as of last night and the program works fine.


If I may make a recommendation here... if you are really into performance, stay away from the "Seagate 120GB 7200.7 Plus". Get the 160GB or 200GB version instead!


I myself will probably be keeping this drive since there doesn`t seem to be anything WRONG with it, and it`ll become my media\data drive in a month or 2 anyways... Got that 3 year warranty too!

 

[LocutusX]

Link to latest beta of HD Tach v3.0:

http://www.simplisoftware.com/Public/download.php?request=HdTach-3.0

 

[IPLaw]

Here are my results with HD Tach v3.0 beta:

Segate ST3120026A 3.06
RAT 14.6
CPU 4%

Read Burst: 88.5
Read Avg: 47.1

Sequential Read (from graph)
Max: 58
Min: 35 (at 120 gb mark)

Interestingly enough, for the exact same Segate 120, the results from an Nforce 2 platform are lower, per the HD Tach database:

RAT 17.1
CPU 3%
Burst 78
Sequential Max: 46
Sequential Min: 27

Interesting results!

 

[LocutusX]

Originally posted by: IPLaw
Interestingly enough, for the exact same Segate 120, the results from an Nforce 2 platform are lower, per the HD Tach database:

RAT 17.1
CPU 3%
Burst 78
Sequential Max: 46
Sequential Min: 27

Interesting results!

Yes, that's exactly what I mentioned in the long post I made earlier today.

The point was that MY Seagate performs almost identical to the one in the HDTach 3 database, as far as STR is concerned anyway (My RAT and Burst is comparable to yours).

What does this mean? Maybe a recently issued firmware or batch of drives suffers from this *reduced* performance. I wouldn't go so far as to call it *poor* since it's still faster than the Barracuda ATA V (previous generation) 120GB.


Anyways I think everyone should grab the HD Tach I linked to and post their results in a new thread; it seems to have most of its kinks worked out, finally!