SR TestBed4 Article is up
- Thread starter Will Rickards
- Start date
Dissapointed that there wasn't much in the way of a discussion on the TB4 results. Perhaps a follow up article is forthcoming.
I wonder about the audio testing too. I remember reading a post on SPCR a while back by someone who slammed the way SR does it -- discussing near field interference etc etc. It seemed reasonable (when I read it) that this person's complaints may have validity.
I wonder about the audio testing too. I remember reading a post on SPCR a while back by someone who slammed the way SR does it -- discussing near field interference etc etc. It seemed reasonable (when I read it) that this person's complaints may have validity.
Tannin
Storage? I am Storage!
They lost the plot with Testbed 3, and, it seems, are utterly determined to maintain the silliness no matter how much evidence stacks up. The strident tone of the intro is in itself a giveaway: if you really have a better way to measure performance, you wouldn't have to shout so much about it.
More unscientific wank. Who cares?
More unscientific wank. Who cares?
Buck
Storage? I am Storage!
Tannin said:
Indeed. We know which drives you (and some of us) will buy anyway.
Buck
Storage? I am Storage!
[quote-"SR"]n many ways, processor speed remains among the least important factors when it comes to assessing storage subsystem performance. This time around, the duty falls to a pair of 3.0 GHz Noconas- beefy enough to handle most tasks thrown at them.[/quote] What a contradiction! 'We don't need CPU speed, so we opted for two 3 Ghz Xeon processors.' Their reasoning is laughable. If you're going to present Enterprise results, than build an Enterprise system and use the appropriate benchmarks. Likewise, have another test bed for typical office/home machines, and perhaps a third for the enthusiast crowd. Basically, SR is compromising all three categories with their hardware configuration.
timwhit
Hairy Aussie
What are you going to do, go rough them up?
Gilbo
Storage is cool
Personally, I think the Testbed is well chosen and that the benchmarks are as good as one can get for general-purpose desktop and server testing.
I'd like to see stop-watched OpenOffice.org Writer launches on linux, analysis with and without pre-linking, stop-watched Photoshop launches on Windows XP, NFS GbE network transfers and probably about a dozen other things, but those results would take a lot of time, and be useless to a lot of people. Most importantly, none of those tests would provide a general overview of the disks' performance. SR's benchmarks aren't perfect, but they're by far the best I've seen. They're the best I can think of in fact, excepting specific timed tests of operations that I personally regularly perform.
In the end of course, disk performance doesn't concern me much anymore. And few things are more insignificant than application startup times. I buy a fast, dedicated paging disk, lots of RAM, and load everything I use regularly at boot, minimize but don't close applications, and then rarely perform a full reboot. Other disk limited tasks like GbE transfers, complete in the background, unattended and could take hours or minutes --it doesn't affect me. Reloading an application, even one that has been entirely paged out to the pagefile, takes a fraction of the time that genuinely starting it does.
I'd like to see stop-watched OpenOffice.org Writer launches on linux, analysis with and without pre-linking, stop-watched Photoshop launches on Windows XP, NFS GbE network transfers and probably about a dozen other things, but those results would take a lot of time, and be useless to a lot of people. Most importantly, none of those tests would provide a general overview of the disks' performance. SR's benchmarks aren't perfect, but they're by far the best I've seen. They're the best I can think of in fact, excepting specific timed tests of operations that I personally regularly perform.
In the end of course, disk performance doesn't concern me much anymore. And few things are more insignificant than application startup times. I buy a fast, dedicated paging disk, lots of RAM, and load everything I use regularly at boot, minimize but don't close applications, and then rarely perform a full reboot. Other disk limited tasks like GbE transfers, complete in the background, unattended and could take hours or minutes --it doesn't affect me. Reloading an application, even one that has been entirely paged out to the pagefile, takes a fraction of the time that genuinely starting it does.
Buck
Storage? I am Storage!
Handruin said:
No thanks Handy. I can pretty much make my own determination what drives I should sell. After building computers for a while, it's pretty easy to decide what sort of drive to use in a given scenario. My rant was just stating the obvious. Eugene is a good guy and puts a lot of himself into his work, which is commendable. It just seems sometimes he takes such a detailed, analytical approach to less important aspects and loses focus of the big picture.
Not at all. I was going to suggest creating an alternative approach to benchmarking drives...here. I have no disrespect or ill-will with SR, I still use them as a resource. But I do find it hard to obtain the results from their website...more along the lines of an interface usability problem.timwhit said:
Will Rickards
Storage Is My Life
I agree there were some contradictions in the explanation of why they chose the components. Like saying building and enterprise class system. Then at the ram explanation saying they opted for only 2GB even though enterprise systems have more.
As far as this build went I think he just didn't want anything else in the machine but the disk to be the limiting factor. I think he achieved that but didn't explain it well.
I liked the results though... of course that is because they put my new baby (Fujitsu MAU series) as the top
As far as this build went I think he just didn't want anything else in the machine but the disk to be the limiting factor. I think he achieved that but didn't explain it well.
I liked the results though... of course that is because they put my new baby (Fujitsu MAU series) as the top
Pradeep
Storage? I am Storage!
Why anyone who wants a quiet machine would choose dual Nocona Xeons is beyond me. Those things are like freaking cores of molten metal, they can easily heat a room.
No worries Will, my Atlask 15K II bests your MAU........in power consumption it has no equal
No worries Will, my Atlask 15K II bests your MAU........in power consumption it has no equal
Mercutio
Fatwah on Western Digital
I don't understand why they can't have two testbeds. I realize there are high-end tests where they need PCI-X controllers and the goodies on that Xeon board, but that machine doesn't represent the kind of hardware that 99% of people use from day-to-day.
I'm interested in performance on a vanilla machine with a vanilla disk controller, not something with I/O hard so rarified that only a dual Xeon board even has those options.
I'm interested in performance on a vanilla machine with a vanilla disk controller, not something with I/O hard so rarified that only a dual Xeon board even has those options.
sechs
Storage? I am Storage!
What are they testing?
Are they trying to clear the field so that the only bottleneck would be the object of testing; or are they actually trying to replicate the enterprise environment?
As much as I can tell, they've done a pretty damn good job of neither.
Are they trying to clear the field so that the only bottleneck would be the object of testing; or are they actually trying to replicate the enterprise environment?
As much as I can tell, they've done a pretty damn good job of neither.
Santilli
Hairy Aussie
- Joined
- Jan 27, 2002
- Messages
- 5,278
Looks like they took my machine, and copied it, with slightly newer equipment. :wink:
I think they finally got sick of dog slow computers, and figured they can write it off, so get something that actually moves. It certainly is going to be intresting to see if their iPeak thread(if they still use it) is affected by the duals, which it should be. Also time is money, and they might have figured out the easier, and faster they can do the tests, the more money they make, per hour.
They have also sort of future proofed their machine for awhile, with that motherboard.
It certainly gives their results more validity for me, but, I don't really worry about their stuff anyway. I'm going to stick with Seagate Cheetahs, using 36 gig 15.3's, a 100 dollar lsi controller, scsi 320, and the red cables. I also like the hotswap SCA drives, but I don't think I'd build many machines with them.
Everytime someone comes out with a faster drive then the Cheetahs, it' seems like 6-12 months later they come out with bad sectors, or such.
I guess what really bugs me is the way the market has shaken out.
It's really too bad that desktops don't have faster drive options, without going to scsi, except for the WD Raptors. I really don't see why they couldn't build a 15K ata drive, or SATA for that matter.
I sure wouldn't object to not having to purchase scsi cards and cables, and being able to use a skinny little SATA cable.
But, that's life and computers. Reminds me of the "not a monopoly" gas prices we have in Kalifornia. It sure seems that all those gas stations, and companies move the prices as a united front, except for Costco...
GS
I think they finally got sick of dog slow computers, and figured they can write it off, so get something that actually moves. It certainly is going to be intresting to see if their iPeak thread(if they still use it) is affected by the duals, which it should be. Also time is money, and they might have figured out the easier, and faster they can do the tests, the more money they make, per hour.
They have also sort of future proofed their machine for awhile, with that motherboard.
It certainly gives their results more validity for me, but, I don't really worry about their stuff anyway. I'm going to stick with Seagate Cheetahs, using 36 gig 15.3's, a 100 dollar lsi controller, scsi 320, and the red cables. I also like the hotswap SCA drives, but I don't think I'd build many machines with them.
Everytime someone comes out with a faster drive then the Cheetahs, it' seems like 6-12 months later they come out with bad sectors, or such.
I guess what really bugs me is the way the market has shaken out.
It's really too bad that desktops don't have faster drive options, without going to scsi, except for the WD Raptors. I really don't see why they couldn't build a 15K ata drive, or SATA for that matter.
I sure wouldn't object to not having to purchase scsi cards and cables, and being able to use a skinny little SATA cable.
But, that's life and computers. Reminds me of the "not a monopoly" gas prices we have in Kalifornia. It sure seems that all those gas stations, and companies move the prices as a united front, except for Costco...
GS
Tannin said:
Sad, but basically true. I never would have believed it 7 years ago on the old SR, but hard drives are nearly a moot point. The bottlenecks for me are CPU performance and lack of 64-bit applications.
Santilli
Hairy Aussie
- Joined
- Jan 27, 2002
- Messages
- 5,278
LunarMist said:
It's certainly true that hard drive progress is moving at a snails pace, compared to other computer parts. It kind of looks like the oil companies,
where price fixing has become a way of life. By keeping certain areas
to themselves, they keep certain areas artififically high. 7200 rpm laptop drives, and scsi drives come to mind, along with WD Raptors.
Maximum profit, little retooling, and a more lengthy period to spread out R&D costs...
GS