Why don't Maxtor or Seagate offer 8MB buffers?

[HepDude]

Hi!

If you look at:

http://www.storagereview.com/comparison.html

on any of the first four comparisons, it's clear that increasing the buffer size for IDE HDs from 2MB to 8MB provides the same performance increase as would be provided by increasing the rotational speed from 7200RPM to 10K.

For example, the improvement in Gaming performance is 22% !

Storage Review says the same thing in their conclusion to the review of the WD 120GB 8MB HD.

So, although it is an additional expense, there is definitely a group of people looking for the best performance (otherwise no one would advertise new top CPU clock speeds), and I would think that Maxtor and Seagate would be interested in a portion of the high-end sales.

 

[Oyeve]

<< and I would think that Maxtor and Seagate would be interested in a portion of the high-end sales >>




Yes, we call that SCSI.

All my SCSI HDs have 8 meg buffers. I think its not on IDE drives because most people who choose IDE over SCSI do so to same money and adding more memory, even only 6 megs, would drive up the cost.

 

[Vegito]

yeah my baracuda 180 has 16mb cache on it..

you get what you pay for...

 

[CQuinn]

<< it's clear that increasing the buffer size for IDE HDs from 2MB to 8MB provides the same performance increase as would be provided by increasing the rotational speed from 7200RPM to 10K. >>



Ah no, it isn't (wasn't) clear. Increasing the buffer size gives an augmentation to the performance gains
already inherent from increasing the density of the platters (which helps in sustained data transfers).
But a boost in rotational speed also lends an improvement in the avg access time of the drive, which
neither of the 8MB assisted WD drives seem to improve upon. In fact the competing IBM and Maxtor
drives in the same category seem to still have an edge in that area.
And if you are comparing that performance against actual 10k drives, you should probably also take
measurements like the CPU Utilization into account.

Up until recently, there was no compelling reason to increase the buffer size beyond 2MB. Older reviews
(done by storagereview) showed no great increase in performance when some early drives went from
512K buffers to 2MB, and it is only because of newer standards like ATA-100, and improvements in drive
density and the drive's on board electronics that there is anything for a larger buffer to take advantage of.
(and the drop in the price of component memory to build in with those electronics).

While the indexes from the SR comparison are impressive, it is still worthwhile to also use the Head-to-Head
comparison feature, to see where eachs drives strengths really lay, and best determine what drive suits
your needs.

Both Maxtor and Seagate make drives with 8MB buffers, in their SCSI lines.
They have more to consider when designing a new drive than just the needs of the IDE using mass-market.
And even WD didn't forsee the demand from the hobbyist crowd for this variation on their existing drives.
And in the past both companies have also held the speed crown, for their own design add-ons.
Each drive company designs and manufactures drives to their own strengths.

I'm sure if adding more memory proves to be a significant "selling point" that Maxtor, Seagate, IBM, Fujitsu and
every other drive maker will jump on the bandwagon. But, while WD holds (for the moment) the crown
for IDE sustained tranfers, Maxtor currently holds on to "largest IDE drive" and Seagate, IIRC is in the
corner with the "quietest drive". No single company holds all the cards, and some are playing with
more than one hand.

 

[JellyBaby]

<< Both Maxtor and Seagate make drives with 8MB buffers, in their SCSI lines. >>

Yep, they don't want to "infringe" on the terrority of their more lucrative scuzzy lines.

 

[Rand]

I suspect in the future we may well be seeing more 8MB buffers on IDE drives, by the timeWestern Digital announced they were doing doing so it was likely to late for other manufacturers to revise their next generation drives to include 8MB buffers.... I wouldnt be at all surprised if the next generation of 7200RPM IDE drives have a few high end models from each manufacturer with 8MB buffers.

A better question IMHO though would be why some manufacturers don't start tuning the firmware on their SCSI drives for typical desktop application scenarios, rather then all SCSI drives being specifically tuned for typical server scenarios in which large file random accesses from distant sectors are the predominant situation for drives and heavily access time bound. This situations a drive is put under in desktop applications is of course vastly different.

It's sad that a clearly inferior Western Digital 8MB buffer 7200RPM drive is outperforming the 10K RPM SCSI crowd in desktop applications. One can only imagine how a SCSI drive in which the formware was better tuned for typical desktop applications might perform, certainly it should logically be able to crush the Western Digital 8MB buffer IDE offerings, just as they already do in the server scenarios in which the firmware is optimized for.

A low end 10K RPM SCSI drive with firmware optimized for desktop applications rather then server usage could grab quite a foothold on the high end desktop market were it ever marketed.
Unfortunately SCSI drives are always optimized for scenarios that are quite the opposite of that seen in desktop application usage.

The much lower CPU utilization of SCSI and the clearly superior hardware would make even a low end SCSI drive an incredible performer in desktop application usage were the firmware properly tuned for it... not that they arent already quite nice in such situations but they could be so much better.
Alas, SCSI=Server insofar as the manufacturers are concerned and hence the firmware is so heavily optimized for the workstation/file/webserver usage they so excell in.

 

[HepDude]

<< The much lower CPU utilization of SCSI >>



Although programmers seem inclined to soak up as much CPU power as is out there, it seems to me that the increase in CPU speed is outstripping either the demands of other hardware devices or the increasing demands of software.

I recently purchased a $105 CPU that is faster than anything available (to consumers) only a year or so ago.

At this point, CPU usage by CPU-powered modems is hard to notice at all, and even software DVD players don't tax the CPU.

So, I think that the CPU utilization of HDs and other devices is less and less an issue.

My original point was simply that people were enthusiastic about new generations of IDE drives with larger platters that provide a 4-5% increase, while the 8MB buffer seems to provide far larger performance increases (20-25%) for about $30 extra selling cost.

 

[Rand]

<<

Although programmers seem inclined to soak up as much CPU power as is out there, it seems to me that the increase in CPU speed is outstripping either the demands of other hardware devices or the increasing demands of software.
>>



Storage Review did some tests not long ago, and the difference was extremely minimal under regular usage but under intensive disk access SCSI consumed almost 30% less processor time then IDE.

 

[GoSharks]

<< Storage Review did some tests not long ago, and the difference was extremely minimal under regular usage >>



yup... i was backing up files from my scsi sub-system to the ide drives on my comp, and i was noticing that it was taking up to 50% of my dual 1ghz cpus. (yes, all the latest drivers are installed, win2k sp2). when i copy the same stuff scsi-scsi, i definitely dont see much cpu usuage, and i MOST definitely dont feel it unlike the ide transfers.

 

[starwarsdad]

Although programmers seem inclined to soak up as much CPU power as is out there, it seems to me that the increase in CPU speed is outstripping either the demands of other hardware devices or the increasing demands of software.

Your statement is very true today, but I am concerned that as enigineers and developers get more reliant on fast CPUs, that with future technologies, we will be back in the position of poorly designed hardware monopolizing CPU cycles.

Most of us here are an extreme sampling of the PC market, but even everyday users now have a machine that requires a bare minimum of a 300W PS. If all of the components we stuff into our boxes are CPU hogs, it doesn't take long to get bound up again. I would like to see manufacturers stick with producing quality components that do most of their work "on board", even if it means paying a little more.

 

[volfan]

The thing I'm looking forward to is seeing stats on the upcoming Serial ATA drives. What rotational speed and Buffer sizes will they have? And when will they be ready? What will their throughput and access times be? I am interested, becuase this is supposedly coming fairly soon...am I right?

 

[downhiller80]

<<
Storage Review did some tests not long ago, and the difference was extremely minimal under regular usage but under intensive disk access SCSI consumed almost 30% less processor time then IDE. >>



Doesn't mean anything though. If the CPU usage of the IDE device is 1% then the SCSI device would be "only" 0.7% based on what you said. Either way it's minimal so like HepDude said, who cares?

- seb

 

[jonmcc33]

Look at the price of those Western Digital drives. They are at NewEgg and are over $200 each.

Most people tend to get 20, 40, or 60GB hard drives...not high end 120GB hard drives with 8MB buffers.

 

[thorin]

My question is why did they jump from 2 to 8? Why was there no 4? And for that matter does that mean the next step will be 512?

(2^3 is 8 .... 8^3 is 512).

Or maybe more like:

(0.5x4 is 2 ... 2x4 is 8 ... 8x4 is 32)

Thorin

 

[Mem]

<< Why don't Maxtor or Seagate offer 8MB buffers? >>




I wouldn`t be surprised if Maxtor are running some D740X harddrives or other IDE models with 8mb cache in their test labs ,you can bet they read and keep an eye on what`s going on with other harddrive brands & reviews etc.

Come to think of it I wouldn`t mind seeing a review of Maxtor D740X with 8MB up against the WD 120GB 8MB HD .

 

[downhiller80]

I have the 120GB 8MB WD drive and it's lovely

Here in the UK it was only a £10 premium over the 2MB version.

- seb

 

[fkloster]

<< Why don't Maxtor or Seagate offer 8MB buffers? >>



hehe some seagate drives have 16mb buffers lol...

 

[CocaCola5]

<< If all of the components we stuff into our boxes are CPU hogs, it doesn't take long to get bound up again. I would like to see manufacturers stick with producing quality components that do most of their work "on board", even if it means paying a little more. >>





I totally agree on this because lately, my mp3s(all 320kb) are almost worthless because I can't listen to them the way they're suppose to sound.

 

[Rand]

<<

<<
Storage Review did some tests not long ago, and the difference was extremely minimal under regular usage but under intensive disk access SCSI consumed almost 30% less processor time then IDE. >>



Doesn't mean anything though. If the CPU usage of the IDE device is 1% then the SCSI device would be "only" 0.7% based on what you said. Either way it's minimal so like HepDude said, who cares?

- seb >>



In basic regular usage it's "only" 1%, under stress tests the processor usage of IDE jumped all the way up to between 40% which most definitely is NOT a minor amount. SCSI meanwhile only jmped to about 19-20%, and this was under IPeak AnalyzeDisk's Read CPU Utilization with tests standardized on 10,000 requests per second? a load equivalent to reading nearly 160 MB/sec in 16 KB blocks, admittedly that's an extreme example, with basic usage under 1%. But if it can jump as high as 40% under 160MB/sec loads then i can easily see it hitting as high as 15% under more typical desktop intensive usage scenarios... and considerably higher is file servers.
Personally, I would call a 40% processor utilization a pretty damn significant figure.... even the more realistic figure of 15% for regular intensive usage is rather significant.

 

[Om]

<< wouldn`t be surprised if Maxtor are running some D740X harddrives or other IDE models with 8mb cache in their test labs ,you can bet they read and keep an eye on what`s going on with other harddrive brands & reviews etc. >>



You are right, and if you look at the history of the Maxtor company they have always been playing catch-up with other manufactures, mostly with Western Digital. So rest assured, if Western Digital is doing it then Maxtor will follow.

 

[CQuinn]

Playing catchup? Funny that they still beat the WD in access times, and I suppose the acqusition of Quantum
putting them back in the SCSI arena with the fastest 10000rpm-rated drive is playing catch up?

 

[Pariah]

"I wouldn`t be surprised if Maxtor are running some D740X harddrives or other IDE models with 8mb cache in their test labs"

Maybe, but that's where they are going to stay. Maxtor bought Quantum's hard drive division to acquire their highend SCSI drives. Common sense says they want to recoup as much of that money as possible. Releasing an IDE drive that benchmarks similar to those SCSI drives and could shift buyer decisions to cheaper drives does not fit into that equation.

 

[Pariah]

Maxtor is playing catchup to WD in the IDE market. It is highly possible though that WD is the only one who knows they are playing, as Maxtor might not really care.

 

[CrashX]

WD is the only major HD maker that only makes IDE drives. So you can be sure they are going to concentrate hard on their market and that means being one step of the competition if they can.

And all HD makers have had their day in the sun. Maxtor has the biggest 5400RPM drive out, WD had the biggest 7200RPM drive out for several months. Two years ago, IBM had the biggest when they released their 75GXP. They've all had their major advancements. We continue to go back and forth. Before know it, we will see 16MB cache available, but for a premium no doubt.

 

[Vegito]

Well.. after buying so many systems.. i gave up on buying scsi raid mirrors... ide mirror works just as good... and ide performance is good enough and cheap enuff to replace..