Have not had much experience as of yet with the Hardrives with the newer 8Mb cache and was wondering if there is much performance/speed gain with it? Appreciate anyone who has had some dealings with these hardrives with the 8Mb cache explain is it worth buying one or is the 2Mb pretty much as good? thanks for any information concerning this issue!!
Tek-Tips is the largest IT community on the Internet today!
Members share and learn making Tek-Tips Forums the best source of peer-reviewed technical information on the Internet!
-
Congratulations strongm on being selected by the Tek-Tips community for having the most helpful posts in the forums last week. Way to Go!
8Mb Cache compared to 2Mb Cache any real difference?
- Thread starter franky77
- Start date
It's just if you want the very fastest mainstream drive, 8mb cache models are the way to go.
It might only be by a few percent but it's like everything else, getting the very fastest has an accumulative effect.
The best motherboard chipset
The fastest memory with the lowest Cas latencey
The fastest CPU more level 2 cache or faster FSB
All combine for the highest performance
Don't forget Western Digital Raptor 10,000rpm drives are the very fastest for everyday desktop but are v expensive and come in limited sizes
The best "normal" 7,200rpm drive (as voted by Toms Hardware review) is the new Hitachi 7K250 range and SATA is the way to go if you have the interface on your motherboard.
Martin
Replying helps further our knowledge, without comment leaves us wondering.
It might only be by a few percent but it's like everything else, getting the very fastest has an accumulative effect.
The best motherboard chipset
The fastest memory with the lowest Cas latencey
The fastest CPU more level 2 cache or faster FSB
All combine for the highest performance
Don't forget Western Digital Raptor 10,000rpm drives are the very fastest for everyday desktop but are v expensive and come in limited sizes
The best "normal" 7,200rpm drive (as voted by Toms Hardware review) is the new Hitachi 7K250 range and SATA is the way to go if you have the interface on your motherboard.
Martin
Replying helps further our knowledge, without comment leaves us wondering.
Yeah in other words, the 8MB cache drive isn't going to make much difference by itself, especially if other components in your system are sub-par.
Tests have shown as much as 10-15% improvement when reading large files "sequentially". For activities like video/audio editing, defragmenting, or booting it can make a significant difference. However, in everyday tasks with Office applications and web surfing, you won't notice any difference.
So it just depends...with as cheap as they are nowadays, it hardly makes sense not to get it.
~cdogg
[tab]"All paid jobs absorb and degrade the mind"
[tab][tab]- Aristotle
Tests have shown as much as 10-15% improvement when reading large files "sequentially". For activities like video/audio editing, defragmenting, or booting it can make a significant difference. However, in everyday tasks with Office applications and web surfing, you won't notice any difference.
So it just depends...with as cheap as they are nowadays, it hardly makes sense not to get it.
~cdogg
[tab]"All paid jobs absorb and degrade the mind"
[tab][tab]- Aristotle
I found that the SATA Maxtor ATA 150 7200 Rpm drives with 8 Mbyte cache performe exceptionelly well in a Sata raid system. Tests with Sisoft show a transfer of between 87 and 95 Mbytes a second. This compares very well indeed with my other raid array which uses 2 Mbyte caches. It only gives between 58 and 65 Mbytes a second. Greetings
Jurgen
Jurgen
Also keep in mind that RAID arrays are the only way to get such high transfer rates. The type of speeds Jurgen posted are "peak" transfer rates, while the average read and write speeds are closer to 45MB/s in even the fastest drives. This is because write speeds are usually much slower than read speeds.
So unless you're using a RAID array, the difference between a 2MB cache and 8MB cache is not going to be a 50% speed increase. But the point is well made about the potential difference an 8MB cache can make in the right environment.
~cdogg
[tab]"All paid jobs absorb and degrade the mind"
[tab][tab]- Aristotle
So unless you're using a RAID array, the difference between a 2MB cache and 8MB cache is not going to be a 50% speed increase. But the point is well made about the potential difference an 8MB cache can make in the right environment.
~cdogg
[tab]"All paid jobs absorb and degrade the mind"
[tab][tab]- Aristotle
Cdogg, please post the location of the tests mentioned in:
Tests have shown as much as 10-15% improvement when reading large files "sequentially". For activities like video/audio editing, defragmenting, or booting it can make a significant difference. However, in everyday tasks with Office applications and web surfing, you won't notice any difference.
I couldn't disagree with you more. Sequential reads of large files are going to flush the cache almost immediately, and the cache will not refill until this heavy activity has completed - in other words, a large sequential read's speed is dependent on the actual physical throughput of the drive (non-cached performance, WB99 Disk/Read Transfer Rate - Begin and End benchmarks at storagereview.com), the size of the cache being completely irrelevant.
People in the know in the audio/video editing world have always ignored cache sizes for their streaming drives. As a side note, A/V can also ignore speed of the bus for streaming drives once you hit 66 MB/s for traditional ATA or 80 MB/s otherwise, because no single drive of those types can saturate those channels for any amount of time (applies to single drives only, striped RAID arrays would be a different story of course).
The cache helps you in exactly the places you say it doesn't, when reading and writing relatively small chunks of data, such as standard OS and application activity. When the data you need is small enough to fit in the cache, or mostly fit in the cache, the cache suddenly has huge benefits.
Tests have shown as much as 10-15% improvement when reading large files "sequentially". For activities like video/audio editing, defragmenting, or booting it can make a significant difference. However, in everyday tasks with Office applications and web surfing, you won't notice any difference.
I couldn't disagree with you more. Sequential reads of large files are going to flush the cache almost immediately, and the cache will not refill until this heavy activity has completed - in other words, a large sequential read's speed is dependent on the actual physical throughput of the drive (non-cached performance, WB99 Disk/Read Transfer Rate - Begin and End benchmarks at storagereview.com), the size of the cache being completely irrelevant.
People in the know in the audio/video editing world have always ignored cache sizes for their streaming drives. As a side note, A/V can also ignore speed of the bus for streaming drives once you hit 66 MB/s for traditional ATA or 80 MB/s otherwise, because no single drive of those types can saturate those channels for any amount of time (applies to single drives only, striped RAID arrays would be a different story of course).
The cache helps you in exactly the places you say it doesn't, when reading and writing relatively small chunks of data, such as standard OS and application activity. When the data you need is small enough to fit in the cache, or mostly fit in the cache, the cache suddenly has huge benefits.
Sit back, folks! This could be a long ride...
Here's one location:
"[blue]...we found that the random scores were exactly the same, while the sequential score for the Barracuda was on the higher side. Again, the reason for the higher sequential scores are due to the extra cache, and not the SATA interface...While reading a 200MB image file in Photoshop, the Barracuda completed the task in 60% of the time taken by the IBM Deskstar.[/blue]"
Here's another (IBM 60GXP 2MB vs. WD 1000JB 8MB cache):
"[blue]What's really interesting, however, is that the 60GXP's buffered and random reads are faster, and so is the average access time. However, during the sequential read and write (aka sustained) the WD comes out victorious by a significant margin[/blue]"
Here's an explanation in another forum (lost the link):
"[blue]As far as buffering is concerned, random access reads, such as when reading in many small files, does NOT benefit in speed from larger on-drive caches. That cache is used mostly for multi-block read-ahead and write-back on sequential accesses, and is usually segmented into "areas", where sections of cache memory are linked to sections (cylinder groups) of the platters. The more cache you have on the drive, the less "thrashing" the drive has to do when many disperse files are written to, since it can group many operations into their appropriate "areas" of cache. This doesn't help as much for random reads (like maildir!) as for sequential reads.[/blue]"
Oh yes, and one significant test:
- At the pcextreme site above, you'll see that a WD1200JB outperformed two Maxtors in a RAID 0 array. Its sequential read was 57MB/s while the Maxtor array hit only 47MB/s. You can see here as much as a possible 17% increase in performance (though I still prefer to use a more conservative 10-15%).
Since you're in the know, perhaps you can explain all the misinterpretation people witness on their own through testing. I can't tell you how many sources I've come across over the past year that concur. What you've said about large files that exceed the size of the cache also makes sense, and perhaps in the end I am mistaken about video/audio editing. However, I am confident that normal operations like booting and defragmenting are helped by the larger cache, since both large and small files are involved. I plan on researching it a bit more when I have the time.
Until then, rest assured that the last line of my first post is undoubtedly true:
"[blue]...with as cheap as they are nowadays, it hardly makes sense not to get it[/blue]"
And I think that is the whole point of this discussion.
![[wink]](/data/assets/smilies/wink.gif "[wink] [wink]")
~cdogg
[tab]"All paid jobs absorb and degrade the mind"
[tab][tab]- Aristotle
Here's one location:
"[blue]...we found that the random scores were exactly the same, while the sequential score for the Barracuda was on the higher side. Again, the reason for the higher sequential scores are due to the extra cache, and not the SATA interface...While reading a 200MB image file in Photoshop, the Barracuda completed the task in 60% of the time taken by the IBM Deskstar.[/blue]"
Here's another (IBM 60GXP 2MB vs. WD 1000JB 8MB cache):
"[blue]What's really interesting, however, is that the 60GXP's buffered and random reads are faster, and so is the average access time. However, during the sequential read and write (aka sustained) the WD comes out victorious by a significant margin[/blue]"
Here's an explanation in another forum (lost the link):
"[blue]As far as buffering is concerned, random access reads, such as when reading in many small files, does NOT benefit in speed from larger on-drive caches. That cache is used mostly for multi-block read-ahead and write-back on sequential accesses, and is usually segmented into "areas", where sections of cache memory are linked to sections (cylinder groups) of the platters. The more cache you have on the drive, the less "thrashing" the drive has to do when many disperse files are written to, since it can group many operations into their appropriate "areas" of cache. This doesn't help as much for random reads (like maildir!) as for sequential reads.[/blue]"
Oh yes, and one significant test:
- At the pcextreme site above, you'll see that a WD1200JB outperformed two Maxtors in a RAID 0 array. Its sequential read was 57MB/s while the Maxtor array hit only 47MB/s. You can see here as much as a possible 17% increase in performance (though I still prefer to use a more conservative 10-15%).
Since you're in the know, perhaps you can explain all the misinterpretation people witness on their own through testing. I can't tell you how many sources I've come across over the past year that concur. What you've said about large files that exceed the size of the cache also makes sense, and perhaps in the end I am mistaken about video/audio editing. However, I am confident that normal operations like booting and defragmenting are helped by the larger cache, since both large and small files are involved. I plan on researching it a bit more when I have the time.
Until then, rest assured that the last line of my first post is undoubtedly true:
"[blue]...with as cheap as they are nowadays, it hardly makes sense not to get it[/blue]"
And I think that is the whole point of this discussion.
~cdogg
[tab]"All paid jobs absorb and degrade the mind"
[tab][tab]- Aristotle
- Thread starter
- #8
Ok, I think I have it but not too sure? by some of the responses, I guess yeah just wanted to know if it made a difference when doing everyday tasks, surfing but also, multi-tasking, Which I do alot of and I do have an older P III computer, so was wondering how it would do multi-tasking, As cdogg pointed out though for the little amount more it costs for one, may as well go for it!! thanks for the info all of you!!
franky777
That's about the long and short of it!
7,200rpm 8mb cache drives are only a few pounds/dollars more than there 5,400rpm counterparts, as top of the range models they do tend to be built more robustly and in my opinion are more reliable.
I know several retailers/builders that simply don't sell/stock 5,400rpm drives any more saying that it just isn't worth it! given the small cost differance.
cdogg! you dog!! :>)
Martin
Replying helps further our knowledge, without comment leaves us wondering.
That's about the long and short of it!
7,200rpm 8mb cache drives are only a few pounds/dollars more than there 5,400rpm counterparts, as top of the range models they do tend to be built more robustly and in my opinion are more reliable.
I know several retailers/builders that simply don't sell/stock 5,400rpm drives any more saying that it just isn't worth it! given the small cost differance.
cdogg! you dog!! :>)
Martin
Replying helps further our knowledge, without comment leaves us wondering.
For what it's worth I've just installed an 8mb 120gb Seagate alongside an 80gb 7200rpm Seagate and a 30gb 5400gb Seagate all at udma5
Nero 6 (sorry nothing more informative than that) reports averages of 2mb/s faster read times than the 80gb drive and over 12 mb/s than the 30gb drive. This is on an aging Gigabyte GA-7DX mobo with AMD 2000XP and 640mb PC2100. The system is not in the least bit optimized!!
Nero 6 (sorry nothing more informative than that) reports averages of 2mb/s faster read times than the 80gb drive and over 12 mb/s than the 30gb drive. This is on an aging Gigabyte GA-7DX mobo with AMD 2000XP and 640mb PC2100. The system is not in the least bit optimized!!
CDogg,
First, I agree with your statement:
"...with as cheap as they are nowadays, it hardly makes sense not to get it"
That being said.....
As far as some of the comparisons go, I think people might be comparing drives which have different uncached transfer rates, then attributing the differences to the cache (in error). For each link you listed I tried to verify this theory (using the WB99 Disk/Read Transfer Rate - Begin benchmark from storagereview.com), but the articles either do not list exact model numbers, or use models which are not in storage review's database.
In this quote:
Oh yes, and one significant test:
- At the pcextreme site above, you'll see that a WD1200JB outperformed two Maxtors in a RAID 0 array. Its sequential read was 57MB/s while the Maxtor array hit only 47MB/s.
Please look at that link again - the exact opposite is true, the RAID array has the 57 MB/s sequential read, the single drive has the 47 MB/s read. The article goes on to say:
Even though the RAID array beat the WD1200JB in buffered and sequential read times....
Again because we don't have an exact model number (we only know they are 7200 RPM) of the Maxtors (or the model number of the RAID device) we can't make this comparison meaningful by comparing unbuffered transfer rates and other attributes of the drives -- although I will grant you that the similar benchmarks are very suspicious considering the Maxtors are in a RAID 0 array!!!!!! That being said, I think it rash to attribute the benchmark differences to the onboard cache.
For a test which really compares based on cache only, check this out:
These are 2 WD drives from the same line, which should differ only by the size of the cache (8 MB in the JB, 2 MB in the BB). In the WB99 Disk/Read Transfer Rates - Begin and End, the only benchmarks worth noting for streaming drives in an A/V setup, the differences are small enough to be irrelevant. The big differences are where I predicted them to be: day-to-day OS/app. use (represented by SR Office, SR Bootup, SR Gaming, individual application tests, etc.). The 10 MB/s. improvement in Sound Forge 4.0 at first seems to contradict what I'm saying, until you realize that Sound Forge is a simple stereo editor, not a multitrack audio editor that has huge streaming requirements - that is, its disk usage is going to mock typical applications more than a true heavy duty A/V application.
This quote:
"As far as buffering is concerned, random access reads, such as when reading in many small files, does NOT benefit in speed from larger on-drive caches. That cache is used mostly for multi-block read-ahead and write-back on sequential accesses, and is usually segmented into "areas", where sections of cache memory are linked to sections (cylinder groups) of the platters. The more cache you have on the drive, the less "thrashing" the drive has to do when many disperse files are written to, since it can group many operations into their appropriate "areas" of cache. This doesn't help as much for random reads (like maildir!) as for sequential reads."
is very interesting, but A/V should not be writing to "disperse areas" or "thrashing" very often. Cakewalk Sonar, for one, lines up audio tracks in a given project (at least as they are recorded) to best optimize multitrack playback - rather than stack each track up by itself on the disk, all tracks are split up and chunks are stacked together by their absolute time in the project - meaning the disk thrashes very little when playing all tracks of a song from beginning to end. The editor over at Prorec.com keeps telling people NOT to defrag their audio drives for this reason, and people want to argue with him all the time (defragging the OS disk is great, but doesn't apply to the audio drives when using a program that lays the tracks down correctly). Now his stock response is "take it up with the developers at Cakewalk."
First, I agree with your statement:
"...with as cheap as they are nowadays, it hardly makes sense not to get it"
That being said.....
As far as some of the comparisons go, I think people might be comparing drives which have different uncached transfer rates, then attributing the differences to the cache (in error). For each link you listed I tried to verify this theory (using the WB99 Disk/Read Transfer Rate - Begin benchmark from storagereview.com), but the articles either do not list exact model numbers, or use models which are not in storage review's database.
In this quote:
Oh yes, and one significant test:
- At the pcextreme site above, you'll see that a WD1200JB outperformed two Maxtors in a RAID 0 array. Its sequential read was 57MB/s while the Maxtor array hit only 47MB/s.
Please look at that link again - the exact opposite is true, the RAID array has the 57 MB/s sequential read, the single drive has the 47 MB/s read. The article goes on to say:
Even though the RAID array beat the WD1200JB in buffered and sequential read times....
Again because we don't have an exact model number (we only know they are 7200 RPM) of the Maxtors (or the model number of the RAID device) we can't make this comparison meaningful by comparing unbuffered transfer rates and other attributes of the drives -- although I will grant you that the similar benchmarks are very suspicious considering the Maxtors are in a RAID 0 array!!!!!! That being said, I think it rash to attribute the benchmark differences to the onboard cache.
For a test which really compares based on cache only, check this out:
These are 2 WD drives from the same line, which should differ only by the size of the cache (8 MB in the JB, 2 MB in the BB). In the WB99 Disk/Read Transfer Rates - Begin and End, the only benchmarks worth noting for streaming drives in an A/V setup, the differences are small enough to be irrelevant. The big differences are where I predicted them to be: day-to-day OS/app. use (represented by SR Office, SR Bootup, SR Gaming, individual application tests, etc.). The 10 MB/s. improvement in Sound Forge 4.0 at first seems to contradict what I'm saying, until you realize that Sound Forge is a simple stereo editor, not a multitrack audio editor that has huge streaming requirements - that is, its disk usage is going to mock typical applications more than a true heavy duty A/V application.
This quote:
"As far as buffering is concerned, random access reads, such as when reading in many small files, does NOT benefit in speed from larger on-drive caches. That cache is used mostly for multi-block read-ahead and write-back on sequential accesses, and is usually segmented into "areas", where sections of cache memory are linked to sections (cylinder groups) of the platters. The more cache you have on the drive, the less "thrashing" the drive has to do when many disperse files are written to, since it can group many operations into their appropriate "areas" of cache. This doesn't help as much for random reads (like maildir!) as for sequential reads."
is very interesting, but A/V should not be writing to "disperse areas" or "thrashing" very often. Cakewalk Sonar, for one, lines up audio tracks in a given project (at least as they are recorded) to best optimize multitrack playback - rather than stack each track up by itself on the disk, all tracks are split up and chunks are stacked together by their absolute time in the project - meaning the disk thrashes very little when playing all tracks of a song from beginning to end. The editor over at Prorec.com keeps telling people NOT to defrag their audio drives for this reason, and people want to argue with him all the time (defragging the OS disk is great, but doesn't apply to the audio drives when using a program that lays the tracks down correctly). Now his stock response is "take it up with the developers at Cakewalk."
pedrettig:
Yes, I was up in the wee-bit hours of the morning when I referred to that last link from PCExtreme. And as you said, it was a comparison to two unnamed Maxtors in a RAID array. My apologies...
![[morning]](/data/assets/smilies/morning.gif "[morning] [morning]")
[navy]Going back to your article for a sec...[/navy]
It is apparent that in the Adobe Premier 4.2 benchmark (though severely outdated), the margin or improvement is 9 MB/s which is more than a 30% performance boost. And you've aleady mentioned the Sound Forge anomaly which is around an 18% increase.
Those numbers alone question either the integrity of your analysis or the benchmark itself. I tried to locate specific information about the size of files tested in both the Sound Forge and Premiere benchmark, but still haven't been able to find anything. Also, I would have to question storagereview's decision not to use the HDTach benchmark as one of their tests, which is excellent for showing specifically the sequential and random read averages.
[navy]How the cache plays an important role in A/V editing...[/navy]
It is my understanding that a larger buffer adversely requires less CPU utilization. Therefore when you are rendering an effect on say a 500MB video file in Adobe Premiere, the CPU is able to allocate more of its resources to the rendering process which is mostly CPU-dependent. A smaller cache normally contradicts that theory.
~cdogg
[tab]"All paid jobs absorb and degrade the mind"
[tab][tab]- Aristotle
Yes, I was up in the wee-bit hours of the morning when I referred to that last link from PCExtreme. And as you said, it was a comparison to two unnamed Maxtors in a RAID array. My apologies...
[navy]Going back to your article for a sec...[/navy]
It is apparent that in the Adobe Premier 4.2 benchmark (though severely outdated), the margin or improvement is 9 MB/s which is more than a 30% performance boost. And you've aleady mentioned the Sound Forge anomaly which is around an 18% increase.
Those numbers alone question either the integrity of your analysis or the benchmark itself. I tried to locate specific information about the size of files tested in both the Sound Forge and Premiere benchmark, but still haven't been able to find anything. Also, I would have to question storagereview's decision not to use the HDTach benchmark as one of their tests, which is excellent for showing specifically the sequential and random read averages.
[navy]How the cache plays an important role in A/V editing...[/navy]
It is my understanding that a larger buffer adversely requires less CPU utilization. Therefore when you are rendering an effect on say a 500MB video file in Adobe Premiere, the CPU is able to allocate more of its resources to the rendering process which is mostly CPU-dependent. A smaller cache normally contradicts that theory.
~cdogg
[tab]"All paid jobs absorb and degrade the mind"
[tab][tab]- Aristotle
Those numbers alone question either the integrity of your analysis or the benchmark itself. I tried to locate specific information about the size of files tested in both the Sound Forge and Premiere benchmark, but still haven't been able to find anything.
I think you hit the nail on the head, we need:
1) Complete details of what exactly these benchmarks test
or barring that,
2) Size of the files being used
I will maintain that the uncached throughput of a drive, the transfer rate it can sustain for five minutes or more on huge files, is the only benchmark that matters for dedicated streaming data drives in a professional A/V setup. That is, not only are sequential rates important, but rather sequential I/O that can be maintained for long periods of time, involving files that would be exponentially larger than any current HD's cache.
Of course none of this applies to an OS drive, so I've taken this thread way, way, out of line.......
Also, I would have to question storagereview's decision not to use the HDTach benchmark as one of their tests, which is excellent for showing specifically the sequential and random read averages.
Sounds neat, do you have any links to databases on different drives using this benchmark? Any other information on this benchmark?
I think you hit the nail on the head, we need:
1) Complete details of what exactly these benchmarks test
or barring that,
2) Size of the files being used
I will maintain that the uncached throughput of a drive, the transfer rate it can sustain for five minutes or more on huge files, is the only benchmark that matters for dedicated streaming data drives in a professional A/V setup. That is, not only are sequential rates important, but rather sequential I/O that can be maintained for long periods of time, involving files that would be exponentially larger than any current HD's cache.
Of course none of this applies to an OS drive, so I've taken this thread way, way, out of line.......
Also, I would have to question storagereview's decision not to use the HDTach benchmark as one of their tests, which is excellent for showing specifically the sequential and random read averages.
Sounds neat, do you have any links to databases on different drives using this benchmark? Any other information on this benchmark?
Here's a link to the benchmark:
A google search will also turn up thousands of hits on websites that have used the benchmark. I haven't personally used it, but I'm sure there's a good tutorial out there somewhere...
~cdogg
[tab]"All paid jobs absorb and degrade the mind"
[tab][tab]- Aristotle
A google search will also turn up thousands of hits on websites that have used the benchmark. I haven't personally used it, but I'm sure there's a good tutorial out there somewhere...
~cdogg
[tab]"All paid jobs absorb and degrade the mind"
[tab][tab]- Aristotle
franky 77.... I'm using Western Digital Raptor Serial
ATA drives on Raid 0. Blindingly fast at around $230
for the two drives. Worth every penny.
Take a look at the spec's.
10,000 RPM 8MB buffer and more.
ATA drives on Raid 0. Blindingly fast at around $230
for the two drives. Worth every penny.
Take a look at the spec's.
10,000 RPM 8MB buffer and more.
- Status
Similar threads
- Locked
- Question