Beefy machine? With a modern (Core 2 Duo) processor, a USB file transfer is going to be using less than half a percent of one core. We aren't using 200MHz P-IIs anymore.
The drive rotates at 7200rpm and was preformatted in NTFS. Windows XP SP3 reported 596.2GB of space available. I tested the drive using HD Tach 3.0 from Simpli Software Inc. Using its thorough Long Bench test (which uses 32KB blocks for reads and writes across the entire drive), the drive registered a burst speed of 32 MB/sec., an average read speed of 30.3 MB/sec., and CPU utilization of 23% when connected to the USB port, roughly on par with the Buffalo DriveStation Combo 4 I reviewed in July -- a 1TB external drive with the same three connectors plus FireWire 800.
Using version 2.55 of the HD Tune benchmark test, the drive recorded an average transfer rate of 29.8 MB/sec., an average access time of 15.2 milliseconds, and a burst rate of 24.5 MB/sec., using 13.1% of the CPU.
Performance improved when I connected the drive to the FireWire 400 port. HD Tach reported a higher burst speed (42.1 MB/sec.) and a 32% improvement in read speed (to 40.0 MB/sec.), using just 1% of the CPU. HD Tune results showed similar improvement: an average transfer rate of 38.0 MB/sec., average access time of 15.0 milliseconds, and a burst rate of 35.3 MB/sec. using just 1% of the CPU.
So ....ostensibly Firewirew has given you improved performance even in read/write performance on the same drive and delivered %12 more minimum efficiency.
so explain how USB 3 is going to magically replace
either GigE (for pure p2p)
or firewire (mobile p2p with decent power)....
Eh? Since when did GigE appear in the mix? All I responded to was your asking what could replace FW.
Quote:
especially with USB3's 900 mA (a measly 5 W) current power limitations
So what? It goes 4.8Gpbs. It charges iPods and iPhones just fine. It goes up from 100mA to 900mA. That's plenty right now for external laptop drives.
Beyond that even the 15W on the S3200 is borderline like for the Hitachi 1TB that idles nearly at 15W. Spinup is more than 15W.
If you want to run a HD camera to a USB3 port it can handle the data rates. Or if you want to load the video off the HD camera it'll have the throughput.
Any audio or video capture should be more than handled by the USB 3.0 spec even at 30% of it's rated spec. The biggest problem is that we won't see any devices until 2010.
Quote:
remember that
1. host contolled / star topology requires much more CPU power - and we're already struggling with heat problems in most laptops errr notebooks (Intel's latest air flow heat exchanger 'innovation' is testament to this)... we'll see if USB3 speeds on a star structure simply compounds the problem.
UHCI controllers are CPU intensive. OHCI (open host controller interface) and EHCI controllers have been much less CPU intensive. By 2010 I doubt even handing a 4.8Gbps will be much more than a few % of the CPU load on a single core. Few meaning 3%-5% or so.
Quote:
2. home networks (of which there are going to be a lot in the future) will collapse if you remove the host of a star based network - with tree based they keep on running... not such a problem for desktops (which remain in place) but for notebooks it creates limitations.
Right. This is an issue for the normal laptop user how? When do you remove the host (aka the notebook) and expect to do what with current FW devices?
Quote:
Admittedly No 2 can be solved by using GigE at home (no power limitations since power is readily available)
Again - i'm not arguing for no USB3 - i'd love to have one on the 2009 books for pure data transfer as an upgrade to the out-of-date USB2... but I just can't see how we're actually moving forward if all that extra CPU power we pay for each upgrade is simply spent on speeding up an increasing number of host controller commands.
Do you see USB3 giving us back TDM ?
Not entirely but the most common use of TDM (migration assistant) already works over the network for the MBA.
Besides, how often do you really USE TDM? Once a year? Tops?
Mostly it hurts techs that use TDM to troubleshoot bad machines. BUT the last time I had an issue with my MBP they simply booted using their own drive (and you can boot to USB) to make sure it was a software issue.
Then they reimaged the drive and started the restore process from backup. With USB3 this takes less time than FW400.
Can Apple reimplement TDM over ethernet? Yes...will they? No. iSCSI+Bonjour+Special Sauce could do it but they'd have to write an iSCSI stack for EFI. Probability is zero.
Quote:
Oh and agreed regarding some military technology not reaching consumers... but that's like saying "joe the plumber doesn't buy rocket launchers..."
Except that you're trying to say that the miltiary adoption of FW has any impact on the commercial market or it's technical merits for desktop/notebook use.
Was FW and TDM a nice ot have? For sure.
Was FW and TDM the DEFINING characteristic of a Mac? Heck no. Give me better graphics any day of the week.
FW might come back. But IMHO only likely as 1394c-2006 (RJ45) and only IF someone built a combined 1394c and GigE chipset to go over that single port.
So ....ostensibly Firewirew has given you improved performance even in read/write performance on the same drive and delivered %12 more minimum efficiency.
Checkmate?
This test is from...January 2003. I'm thinking Pentium 4 Northwood. 13% of a single core 2.4 Northwood P4 (without even HT) is what on a 2.0Ghz Core 2 Duo?
This test is from...January 2003. I'm thinking Pentium 4 Northwood. 13% of a single core 2.4 Northwood P4 (without even HT) is what on a 2.0Ghz Core 2 Duo?
The first linked test was from 2003, but the second link is a test posted this month.
Maxtor OneTouch’s FireWire 400 port actually lagged that of its USB port in the Duplicate File and Low Memory tests, which is atypical in a hard drive.
Plus there doesn't appear to be any measurement of CPU Utilization.
t's no surprise that the access time tests are roughly the same.
What is interesting to see is just how slow this implementation of USB is. Part of the issue is likely to be the high CPU utilization—14% versus 4% for eSATA and 0% for either FireWire configuration. Note that the major difference between eSATA and FireWire 800 lies in write performance and burst throughput; for average reads, FireWire 800 is essentially tied with eSATA.
The first linked test was from 2003, but the second link is a test posted this month.
Missed that.
However:
"Using this configuration, HD Tach reported a burst speed of 100.3 MB/sec. and a read speed more than double that of the FireWire 400 test (81.6 MB/sec.), with just 2% of the CPU used. The HD Tune test was equally strong: an average transfer rate of 76.8 MB/sec., average access time of 15.0 milliseconds, and a burst rate of 61.6 MB/sec. using just 4.4% of the CPU."
Which means what? Apple should have stuck an eSATA port on the MB (or an expresscard slot) and FW400 is still obsolecent for anything besides audio and some camcorders.
As stated before I'd much rather have an eSATA port than FW400 even if my camcorder happens to be FW.
"Using this configuration, HD Tach reported a burst speed of 100.3 MB/sec. and a read speed more than double that of the FireWire 400 test (81.6 MB/sec.), with just 2% of the CPU used. The HD Tune test was equally strong: an average transfer rate of 76.8 MB/sec., average access time of 15.0 milliseconds, and a burst rate of 61.6 MB/sec. using just 4.4% of the CPU."
Which means what? Apple should have stuck an eSATA port on the MB (or an expresscard slot) and FW400 is still obsolecent for anything besides audio and some camcorders.
It looks like you shifted from arguing that USB is almost as good as FW for CPU load to pretending that discussion never happened.
"According to Buffalo, the drive's TurboUSB software gives a boost to performance. My tests bore out the claim. After installing the TurboUSB software and turning the feature on, the read speed rose to 35.7MB/sec., and CPU utilization dropped to 7%. (Burst speed was reported at 838.5MB/sec., a figure consistent over several iterations of the test.)
...
FireWire's advantage disappeared when I copied files from my hard drive to the Combo 4. My copy of a 1GB MPEG video file using the USB connection took 48 seconds with TurboUSB off and 44 seconds with it on. When connected to the FireWire 800 port, the copy took 53 seconds.
My benchmark tests did point out one anomaly: Both HD Tach and HD Tune reported seriously degraded performance between the 100GB to 150GB marks. For example, in the HD Tune test with the FireWire 800 connection, the transfer rate dropped from an average of 53.4MB/sec. to just 12.1MB/sec."
High CPU utilization is not always a requirement (it depends on your drivers) and xCHI CPU load is currently unknown (for USB 3.0).
Likewise, even for FW800, USB vs FW performance is not always THAT different. One drive from one vendor is one thing but there's more than just one from two different vendors.
It looks like you shifted from arguing that USB is almost as good as FW for CPU load to pretending that discussion never happened.
Nope. This is what I wrote:
Quote:
UHCI controllers are CPU intensive. OHCI (open host controller interface) and EHCI controllers have been much less CPU intensive. By 2010 I doubt even handing a 4.8Gbps will be much more than a few % of the CPU load on a single core. Few meaning 3%-5% or so.
Given that xHCI CPU load isn't currently known your guess is as good as mine. I say that Intel is aware of the issue but by the time we get to Nahalem who the hell cares?
That Buffalo has software to reduce CPU load from 15% to 7% and get faster speed means that this is possible with better controller interfaces (and implementation) even with USB2.0.
You can cherry pick implementations where USB 2.0 implementations are bad ( 30.3 MB/sec @ 23% CPU utlilization) or decent (30.5MB/sec @ 15% CPU utlization and 35.7MB/sec @ 7% CPU utilization with TurboUSB).
15% vs 1% is less interesting than claiming 23% vs 1%.
You can also cherry pick FW implementations that are good or bad (like the Maxtor). But even on the Buffalo that tested well in his benchmark tools (HD Tune and HD Tach) failed to outperform USB 2.0 under real world use (1GB MPEG copy from computer to drive).
Want to bet that CPU use was at that 10.8% CPU utlilzation measured at some times with FW800 vs the 1% CPU utlization measured at other times?
Bottom line is that for the most part, it doesn't really matter in day to day use. How do I "know" that? Because the market has spoken and chosen USB 2.0.
Bottom line is that for the most part, it doesn't really matter in day to day use. How do I "know" that? Because the market has spoken and chosen USB 2.0.
Of course it matters when performance is monitored. Musicians run firewire because even a %7 hit means you cannot run as many tracks or power is gone for some effects.
For burst data that doesn't require realtime or guaranteed delivery USB is fine but I find it absurd that people have "chosen" USB. With Intel creating USB and shoveling it into every computer there never really was a choice.
Of course it matters when performance is monitored. Musicians run firewire because even a %7 hit means you cannot run as many tracks or power is gone for some effects.
For burst data that doesn't require realtime or guaranteed delivery USB is fine but I find it absurd that people have "chosen" USB. With Intel creating USB and shoveling it into every computer there never really was a choice.
Which in 2010 will be a non issue for the data rates you want with USB3. Because the sustained rates will be far above either USB2 or FW400 burst rates and it won't be even a 7% hit for the rates you're talking about.
Yes, it sucks today for one segment of the MB demographic.
On the plus side the graphic improvements are a benefit for the larger MB demographic of students. The new MB is a great little machine. I'm waiting to see what the mini will look like. If it gets the graphics update but loses FW I'll still buy it and just get a new camcorder.
And if you're THAT CPU bound for your effects then you should get the 2.4Ghz model vs the 2.0Ghz model for a 20% boost. That should cover the USB overhead although it doesn't help the fact that you still can't run your FW devices over USB.
But for the most part, Jobs is right. USB2 is good enough. There appear to be enough sucky FW implementations out there that I don't automatically assume that USB2.0 is such a horrid tradeoff anymore.
But for the most part, Jobs is right. USB2 is good enough. There appear to be enough sucky FW implementations out there that I don't automatically assume that USB2.0 is such a horrid tradeoff anymore.
Why is it an either or proposition?
Why can't we have USB AND FW?
It's not like we get anything in return for loosing the FW port. No more USB ports, just two. That's really borderline adequate when you have iPods, iPhone and external drives plugged in often.
The MBs are very appealing in many respects. But without FW, what should be an easy decision, buy a MB!, is now more difficult for existing Mac users because we bought into FW when we started using Macs.
Eh? Since when did GigE appear in the mix? All I responded to was your asking what could replace FW.
GigE has been in the equation since it's the only p2p connection left on the macbook
...plus you mentioned it a few times in your posts.
Quote:
Originally Posted by vinea
So what? It goes 4.8Gpbs. It charges iPods and iPhones just fine. It goes up from 100mA to 900mA. That's plenty right now for external laptop drives.
Beyond that even the 15W on the S3200 is borderline like for the Hitachi 1TB that idles nearly at 15W. Spinup is more than 15W.
FW is up to 45 W...
Quote:
Originally Posted by vinea
If you want to run a HD camera to a USB3 port it can handle the data rates. Or if you want to load the video off the HD camera it'll have the throughput.
Any audio or video capture should be more than handled by the USB 3.0 spec even at 30% of it's rated spec. The biggest problem is that we won't see any devices until 2010.
UHCI controllers are CPU intensive. OHCI (open host controller interface) and EHCI controllers have been much less CPU intensive. By 2010 I doubt even handing a 4.8Gbps will be much more than a few % of the CPU load on a single core. Few meaning 3%-5% or so.
UHCI is for USB 1.1
EHCI is the standard for USB2 which we both agree is slow... unless you're saying there's been an update?
again for files / photos / low res vid cam USB is fine - we can all do with a 10 min coffee break...
Quote:
Originally Posted by vinea
Right. This is an issue for the normal laptop user how? When do you remove the host (aka the notebook) and expect to do what with current FW devices?
in an ideal world you'd remove the 'host' every morning and take it with you - leaving your peripherals still functioning, connected to whatever network you want operating for whatever you need
- downloading on offpeak periods, buying stocks, keeping your fridge running,
monitoring your solar panels, keeping your animals fed & warm,
allowing your HDD files to be accessible, backing up data - i don't know, use your imagination -
the future is open! (and it's bigger than pulling files off a hard drive)
but I certainly don't want to buy and keep a full PC connected for those light loads
(which could be easily programmed on a small firmware chip but updated/optimised/read from my mac) and I don't want it to all fall in a heap when I walk out the door with my one notebook under my arm
its an issue because not everyone wants to own a desktop at home and a notebook to run around with - i know i would rather have just one machine if it could serve both purposes (no matter how much i love my mac pro i'd rather be able to take it with me)
and i think the future is transportable - which means that keeping CPU loads as light as possible means the fastest graphics possible - which you mentioned is one of your interests...
either way i'm looking for a p2p type interface on a nice light mac (again why i keep mentioning GigE, also because of the possibility that FW could use the port)
and USB3/4/5/6 and Esata will not be able to provide p2p...
Quote:
Originally Posted by vinea
Except that you're trying to say that the miltiary adoption of FW has any impact on the commercial market or it's technical merits for desktop/notebook use.
i find it ironic that you're using the internet, started/invented by the US military - to claim that the military has no impact on the commercial market...
So ....ostensibly Firewirew has given you improved performance even in read/write performance on the same drive and delivered %12 more minimum efficiency.
Why are there so many people ready and willing to jump to Apple's defense when it is very clear that the removal of FireWire from the new MacBook was an outstanding piece of lunacy simply because the FW port was not replaced with anything ...
A: similar in function and speed. ... and ...
B: simply not replaced! USB had already been included on the previous models.
It is so obvious that a hell of a lot of people still use and want FW. No point side stepping the debate by saying 'Well, that's Apple progress'.
It seems to me that so many are making excuses for this nonsense by 'bigging up' USB. Well, as many have said, USB2 doesn't cut it! And don't give me that USB3 crap. It is not on there yet! We're missing a FireWire port. Where is the 'replacement'?
I really do wonder just what some people's agendas are. Who do you work for? What are your interests? Why do you want to see an inferior technology supercede a superior one? You call that progress? A solution is needed, not excuses.
I really do wonder just what some people's agendas are. Who do you work for? What are your interests? Why do you want to see an inferior technology supercede a superior one? You call that progress? A solution is needed, not excuses.
What were the PC makers' "agendas?" Obviously to marginalize and eventually eliminate an interface too closely identified with Apple. Well, they succeeded! Why wasn't everybody all up in their faces then? If only some of this rage could have been harnessed when it could have done some good!
I really do wonder just what some people's agendas are. Who do you work for? What are your interests? Why do you want to see an inferior technology supercede a superior one? You call that progress? A solution is needed, not excuses.
PROTIP: When you start seeing conspiracies in the motives of people you disagree with, it's time to log off the internet and go outside for a while.
EHCI is the standard for USB2 which we both agree is slow... unless you're saying there's been an update?
Nice twisting of words. We don't both agree is slow unless you also agree that today FW400 isn't all that speedy.
Quote:
in an ideal world you'd remove the 'host' every morning and take it with you - leaving your peripherals still functioning, connected to whatever network you want operating for whatever you need
- downloading on offpeak periods, buying stocks, keeping your fridge running,
monitoring your solar panels, keeping your animals fed & warm,
allowing your HDD files to be accessible, backing up data - i don't know, use your imagination -
Right. Show me any FW system today that does any of this.
Quote:
its an issue because not everyone wants to own a desktop at home and a notebook to run around with - i know i would rather have just one machine if it could serve both purposes (no matter how much i love my mac pro i'd rather be able to take it with me)
and i think the future is transportable - which means that keeping CPU loads as light as possible means the fastest graphics possible - which you mentioned is one of your interests...
GPU load <> CPU load. And the current MacBook is WAY faster for graphics than the old MacBook even if you're transferring files and chewing up 13% of the CPU.
Quote:
i find it ironic that you're using the internet, started/invented by the US military - to claim that the military has no impact on the commercial market...
Actually, while DARPA built the protocols adopted by the NSF, it was NSFNET that later became the internet. I was on ARPANET, BITNET and NSFNET in those early days. Today ARPANET lives on in NIPRNET (formerly MILNET)...not the internet. Ultimately it was the research and subercomputer centers that spawned internet as we know it. Notably CERN and HTML and NCSA...who developed Mosaic.
But you're also talking ancient history. Even at the birth of the internet in the mid 80s, DARPA's research was lagely complete in the 70s. Where the military leads today is in material science. Steath materials, armor, polymers, high temp metals, etc are all technologies where they lead commerical industry.
Computers and computer protocols is not where they spend their real R&D money. Even things like custom chips are derived from existing commercial baselines (like PA Semi with ARM).
Comments
Beefy machine? With a modern (Core 2 Duo) processor, a USB file transfer is going to be using less than half a percent of one core. We aren't using 200MHz P-IIs anymore.
http://www.extremetech.com/article2/...1152200,00.asp
Check
The drive rotates at 7200rpm and was preformatted in NTFS. Windows XP SP3 reported 596.2GB of space available. I tested the drive using HD Tach 3.0 from Simpli Software Inc. Using its thorough Long Bench test (which uses 32KB blocks for reads and writes across the entire drive), the drive registered a burst speed of 32 MB/sec., an average read speed of 30.3 MB/sec., and CPU utilization of 23% when connected to the USB port, roughly on par with the Buffalo DriveStation Combo 4 I reviewed in July -- a 1TB external drive with the same three connectors plus FireWire 800.
Using version 2.55 of the HD Tune benchmark test, the drive recorded an average transfer rate of 29.8 MB/sec., an average access time of 15.2 milliseconds, and a burst rate of 24.5 MB/sec., using 13.1% of the CPU.
Performance improved when I connected the drive to the FireWire 400 port. HD Tach reported a higher burst speed (42.1 MB/sec.) and a 32% improvement in read speed (to 40.0 MB/sec.), using just 1% of the CPU. HD Tune results showed similar improvement: an average transfer rate of 38.0 MB/sec., average access time of 15.0 milliseconds, and a burst rate of 35.3 MB/sec. using just 1% of the CPU.
http://www.pcworld.com/businesscente...eme_speed.html
Checkmate
%13-23 CPU utilization versus %1
So ....ostensibly Firewirew has given you improved performance even in read/write performance on the same drive and delivered %12 more minimum efficiency.
OK you obviously work for intel
so explain how USB 3 is going to magically replace
either GigE (for pure p2p)
or firewire (mobile p2p with decent power)....
Eh? Since when did GigE appear in the mix? All I responded to was your asking what could replace FW.
especially with USB3's 900 mA (a measly 5 W) current power limitations
So what? It goes 4.8Gpbs. It charges iPods and iPhones just fine. It goes up from 100mA to 900mA. That's plenty right now for external laptop drives.
Beyond that even the 15W on the S3200 is borderline like for the Hitachi 1TB that idles nearly at 15W. Spinup is more than 15W.
If you want to run a HD camera to a USB3 port it can handle the data rates. Or if you want to load the video off the HD camera it'll have the throughput.
Any audio or video capture should be more than handled by the USB 3.0 spec even at 30% of it's rated spec. The biggest problem is that we won't see any devices until 2010.
remember that
1. host contolled / star topology requires much more CPU power - and we're already struggling with heat problems in most laptops errr notebooks (Intel's latest air flow heat exchanger 'innovation' is testament to this)... we'll see if USB3 speeds on a star structure simply compounds the problem.
UHCI controllers are CPU intensive. OHCI (open host controller interface) and EHCI controllers have been much less CPU intensive. By 2010 I doubt even handing a 4.8Gbps will be much more than a few % of the CPU load on a single core. Few meaning 3%-5% or so.
2. home networks (of which there are going to be a lot in the future) will collapse if you remove the host of a star based network - with tree based they keep on running... not such a problem for desktops (which remain in place) but for notebooks it creates limitations.
Right. This is an issue for the normal laptop user how? When do you remove the host (aka the notebook) and expect to do what with current FW devices?
Admittedly No 2 can be solved by using GigE at home (no power limitations since power is readily available)
Again - i'm not arguing for no USB3 - i'd love to have one on the 2009 books for pure data transfer as an upgrade to the out-of-date USB2... but I just can't see how we're actually moving forward if all that extra CPU power we pay for each upgrade is simply spent on speeding up an increasing number of host controller commands.
Do you see USB3 giving us back TDM ?
Not entirely but the most common use of TDM (migration assistant) already works over the network for the MBA.
Besides, how often do you really USE TDM? Once a year? Tops?
Mostly it hurts techs that use TDM to troubleshoot bad machines. BUT the last time I had an issue with my MBP they simply booted using their own drive (and you can boot to USB) to make sure it was a software issue.
Then they reimaged the drive and started the restore process from backup. With USB3 this takes less time than FW400.
Can Apple reimplement TDM over ethernet? Yes...will they? No. iSCSI+Bonjour+Special Sauce could do it but they'd have to write an iSCSI stack for EFI. Probability is zero.
Oh and agreed regarding some military technology not reaching consumers... but that's like saying "joe the plumber doesn't buy rocket launchers..."
Except that you're trying to say that the miltiary adoption of FW has any impact on the commercial market or it's technical merits for desktop/notebook use.
Was FW and TDM a nice ot have? For sure.
Was FW and TDM the DEFINING characteristic of a Mac? Heck no. Give me better graphics any day of the week.
FW might come back. But IMHO only likely as 1394c-2006 (RJ45) and only IF someone built a combined 1394c and GigE chipset to go over that single port.
Checkmate.
Wow. That's even worse than I thought.
Besides, how often do you really USE TDM? Once a year? Tops?
Maybe this is just me, but I use TDM a few times a month to quickly grab files I've been working with on my MacBook Pro.
Checkmate
%13-23 CPU utilization versus %1
So ....ostensibly Firewirew has given you improved performance even in read/write performance on the same drive and delivered %12 more minimum efficiency.
Checkmate?
This test is from...January 2003. I'm thinking Pentium 4 Northwood. 13% of a single core 2.4 Northwood P4 (without even HT) is what on a 2.0Ghz Core 2 Duo?
xCHI CPU load is currently unknown.
http://www.macworld.com/article/1318...onetouch4.html
Timed Trials
Copy 1GB file to FireWire 400 0:51
Copy 1GB file to USB 2.0 0:53
Duplicate 1GB file via FireWire 400 1:21
Duplicate 1GB file via USB 2.0 1:19
Low-memory Photoshop: FireWire 400 1:49
Low-memory Photoshop: USB 2.0 1:47
Times are in minutes:seconds
Checkmate?
This test is from...January 2003. I'm thinking Pentium 4 Northwood. 13% of a single core 2.4 Northwood P4 (without even HT) is what on a 2.0Ghz Core 2 Duo?
The first linked test was from 2003, but the second link is a test posted this month.
The first linked test was from 2003, but the second link is a test posted this month.
Thank you JeffDM I purposely sought another more recent source.
http://www.macworld.com/article/1318...onetouch4.html
Maxtor OneTouch’s FireWire 400 port actually lagged that of its USB port in the Duplicate File and Low Memory tests, which is atypical in a hard drive.
Plus there doesn't appear to be any measurement of CPU Utilization.
http://eshop.macsales.com/NewsRoom/F...re_hdmerc.html
October of 2006
t's no surprise that the access time tests are roughly the same.
What is interesting to see is just how slow this implementation of USB is. Part of the issue is likely to be the high CPU utilization—14% versus 4% for eSATA and 0% for either FireWire configuration. Note that the major difference between eSATA and FireWire 800 lies in write performance and burst throughput; for average reads, FireWire 800 is essentially tied with eSATA.
0% for Firewire people.
The first linked test was from 2003, but the second link is a test posted this month.
Missed that.
However:
"Using this configuration, HD Tach reported a burst speed of 100.3 MB/sec. and a read speed more than double that of the FireWire 400 test (81.6 MB/sec.), with just 2% of the CPU used. The HD Tune test was equally strong: an average transfer rate of 76.8 MB/sec., average access time of 15.0 milliseconds, and a burst rate of 61.6 MB/sec. using just 4.4% of the CPU."
Which means what? Apple should have stuck an eSATA port on the MB (or an expresscard slot) and FW400 is still obsolecent for anything besides audio and some camcorders.
As stated before I'd much rather have an eSATA port than FW400 even if my camcorder happens to be FW.
And I'd rather than better graphics over either.
Plus there doesn't appear to be any measurement of CPU Utilization.
Because it doesn't really matter that much anymore.
Missed that.
However:
"Using this configuration, HD Tach reported a burst speed of 100.3 MB/sec. and a read speed more than double that of the FireWire 400 test (81.6 MB/sec.), with just 2% of the CPU used. The HD Tune test was equally strong: an average transfer rate of 76.8 MB/sec., average access time of 15.0 milliseconds, and a burst rate of 61.6 MB/sec. using just 4.4% of the CPU."
Which means what? Apple should have stuck an eSATA port on the MB (or an expresscard slot) and FW400 is still obsolecent for anything besides audio and some camcorders.
It looks like you shifted from arguing that USB is almost as good as FW for CPU load to pretending that discussion never happened.
"According to Buffalo, the drive's TurboUSB software gives a boost to performance. My tests bore out the claim. After installing the TurboUSB software and turning the feature on, the read speed rose to 35.7MB/sec., and CPU utilization dropped to 7%. (Burst speed was reported at 838.5MB/sec., a figure consistent over several iterations of the test.)
...
FireWire's advantage disappeared when I copied files from my hard drive to the Combo 4. My copy of a 1GB MPEG video file using the USB connection took 48 seconds with TurboUSB off and 44 seconds with it on. When connected to the FireWire 800 port, the copy took 53 seconds.
My benchmark tests did point out one anomaly: Both HD Tach and HD Tune reported seriously degraded performance between the 100GB to 150GB marks. For example, in the HD Tune test with the FireWire 800 connection, the transfer rate dropped from an average of 53.4MB/sec. to just 12.1MB/sec."
http://www.computerworld.com/action/...icleId=9109539
High CPU utilization is not always a requirement (it depends on your drivers) and xCHI CPU load is currently unknown (for USB 3.0).
Likewise, even for FW800, USB vs FW performance is not always THAT different. One drive from one vendor is one thing but there's more than just one from two different vendors.
It looks like you shifted from arguing that USB is almost as good as FW for CPU load to pretending that discussion never happened.
Nope. This is what I wrote:
UHCI controllers are CPU intensive. OHCI (open host controller interface) and EHCI controllers have been much less CPU intensive. By 2010 I doubt even handing a 4.8Gbps will be much more than a few % of the CPU load on a single core. Few meaning 3%-5% or so.
Given that xHCI CPU load isn't currently known your guess is as good as mine. I say that Intel is aware of the issue but by the time we get to Nahalem who the hell cares?
That Buffalo has software to reduce CPU load from 15% to 7% and get faster speed means that this is possible with better controller interfaces (and implementation) even with USB2.0.
You can cherry pick implementations where USB 2.0 implementations are bad ( 30.3 MB/sec @ 23% CPU utlilization) or decent (30.5MB/sec @ 15% CPU utlization and 35.7MB/sec @ 7% CPU utilization with TurboUSB).
15% vs 1% is less interesting than claiming 23% vs 1%.
You can also cherry pick FW implementations that are good or bad (like the Maxtor). But even on the Buffalo that tested well in his benchmark tools (HD Tune and HD Tach) failed to outperform USB 2.0 under real world use (1GB MPEG copy from computer to drive).
Want to bet that CPU use was at that 10.8% CPU utlilzation measured at some times with FW800 vs the 1% CPU utlization measured at other times?
Bottom line is that for the most part, it doesn't really matter in day to day use. How do I "know" that? Because the market has spoken and chosen USB 2.0.
Bottom line is that for the most part, it doesn't really matter in day to day use. How do I "know" that? Because the market has spoken and chosen USB 2.0.
Of course it matters when performance is monitored. Musicians run firewire because even a %7 hit means you cannot run as many tracks or power is gone for some effects.
For burst data that doesn't require realtime or guaranteed delivery USB is fine but I find it absurd that people have "chosen" USB. With Intel creating USB and shoveling it into every computer there never really was a choice.
Of course it matters when performance is monitored. Musicians run firewire because even a %7 hit means you cannot run as many tracks or power is gone for some effects.
For burst data that doesn't require realtime or guaranteed delivery USB is fine but I find it absurd that people have "chosen" USB. With Intel creating USB and shoveling it into every computer there never really was a choice.
Which in 2010 will be a non issue for the data rates you want with USB3. Because the sustained rates will be far above either USB2 or FW400 burst rates and it won't be even a 7% hit for the rates you're talking about.
Yes, it sucks today for one segment of the MB demographic.
On the plus side the graphic improvements are a benefit for the larger MB demographic of students. The new MB is a great little machine. I'm waiting to see what the mini will look like. If it gets the graphics update but loses FW I'll still buy it and just get a new camcorder.
And if you're THAT CPU bound for your effects then you should get the 2.4Ghz model vs the 2.0Ghz model for a 20% boost. That should cover the USB overhead although it doesn't help the fact that you still can't run your FW devices over USB.
But for the most part, Jobs is right. USB2 is good enough. There appear to be enough sucky FW implementations out there that I don't automatically assume that USB2.0 is such a horrid tradeoff anymore.
But for the most part, Jobs is right. USB2 is good enough. There appear to be enough sucky FW implementations out there that I don't automatically assume that USB2.0 is such a horrid tradeoff anymore.
Why is it an either or proposition?
Why can't we have USB AND FW?
It's not like we get anything in return for loosing the FW port. No more USB ports, just two. That's really borderline adequate when you have iPods, iPhone and external drives plugged in often.
The MBs are very appealing in many respects. But without FW, what should be an easy decision, buy a MB!, is now more difficult for existing Mac users because we bought into FW when we started using Macs.
Eh? Since when did GigE appear in the mix? All I responded to was your asking what could replace FW.
GigE has been in the equation since it's the only p2p connection left on the macbook
...plus you mentioned it a few times in your posts.
So what? It goes 4.8Gpbs. It charges iPods and iPhones just fine. It goes up from 100mA to 900mA. That's plenty right now for external laptop drives.
Beyond that even the 15W on the S3200 is borderline like for the Hitachi 1TB that idles nearly at 15W. Spinup is more than 15W.
FW is up to 45 W...
If you want to run a HD camera to a USB3 port it can handle the data rates. Or if you want to load the video off the HD camera it'll have the throughput.
Any audio or video capture should be more than handled by the USB 3.0 spec even at 30% of it's rated spec. The biggest problem is that we won't see any devices until 2010.
UHCI controllers are CPU intensive. OHCI (open host controller interface) and EHCI controllers have been much less CPU intensive. By 2010 I doubt even handing a 4.8Gbps will be much more than a few % of the CPU load on a single core. Few meaning 3%-5% or so.
UHCI is for USB 1.1
EHCI is the standard for USB2 which we both agree is slow... unless you're saying there's been an update?
again for files / photos / low res vid cam USB is fine - we can all do with a 10 min coffee break...
Right. This is an issue for the normal laptop user how? When do you remove the host (aka the notebook) and expect to do what with current FW devices?
in an ideal world you'd remove the 'host' every morning and take it with you - leaving your peripherals still functioning, connected to whatever network you want operating for whatever you need
- downloading on offpeak periods, buying stocks, keeping your fridge running,
monitoring your solar panels, keeping your animals fed & warm,
allowing your HDD files to be accessible, backing up data - i don't know, use your imagination
the future is open! (and it's bigger than pulling files off a hard drive)
but I certainly don't want to buy and keep a full PC connected for those light loads
(which could be easily programmed on a small firmware chip but updated/optimised/read from my mac) and I don't want it to all fall in a heap when I walk out the door with my one notebook under my arm
its an issue because not everyone wants to own a desktop at home and a notebook to run around with - i know i would rather have just one machine if it could serve both purposes (no matter how much i love my mac pro i'd rather be able to take it with me)
and i think the future is transportable - which means that keeping CPU loads as light as possible means the fastest graphics possible - which you mentioned is one of your interests...
either way i'm looking for a p2p type interface on a nice light mac (again why i keep mentioning GigE, also because of the possibility that FW could use the port)
and USB3/4/5/6 and Esata will not be able to provide p2p...
Except that you're trying to say that the miltiary adoption of FW has any impact on the commercial market or it's technical merits for desktop/notebook use.
i find it ironic that you're using the internet, started/invented by the US military - to claim that the military has no impact on the commercial market...
Checkmate
%13-23 CPU utilization versus %1
So ....ostensibly Firewirew has given you improved performance even in read/write performance on the same drive and delivered %12 more minimum efficiency.
All right, well here's what I was looking at:
Source
That's with an Intel E8500 processor and the ICH10's USB controller. So obviously this depends heavily on the USB chips involved.
One bloody simple question.
Why?
Why are there so many people ready and willing to jump to Apple's defense when it is very clear that the removal of FireWire from the new MacBook was an outstanding piece of lunacy simply because the FW port was not replaced with anything ...
A: similar in function and speed. ... and ...
B: simply not replaced! USB had already been included on the previous models.
It is so obvious that a hell of a lot of people still use and want FW. No point side stepping the debate by saying 'Well, that's Apple progress'.
It seems to me that so many are making excuses for this nonsense by 'bigging up' USB. Well, as many have said, USB2 doesn't cut it! And don't give me that USB3 crap. It is not on there yet! We're missing a FireWire port. Where is the 'replacement'?
I really do wonder just what some people's agendas are. Who do you work for? What are your interests? Why do you want to see an inferior technology supercede a superior one? You call that progress? A solution is needed, not excuses.
I really do wonder just what some people's agendas are. Who do you work for? What are your interests? Why do you want to see an inferior technology supercede a superior one? You call that progress? A solution is needed, not excuses.
What were the PC makers' "agendas?" Obviously to marginalize and eventually eliminate an interface too closely identified with Apple. Well, they succeeded! Why wasn't everybody all up in their faces then? If only some of this rage could have been harnessed when it could have done some good!
I really do wonder just what some people's agendas are. Who do you work for? What are your interests? Why do you want to see an inferior technology supercede a superior one? You call that progress? A solution is needed, not excuses.
PROTIP: When you start seeing conspiracies in the motives of people you disagree with, it's time to log off the internet and go outside for a while.
GigE has been in the equation since it's the only p2p connection left on the macbook
...plus you mentioned it a few times in your posts.
But no where do I suggest that USB replaces GigE but rather GigE + USB fills many of the needs that FW did before.
FW is up to 45 W...
You keep on believing you can get 45W off a laptop FW port. 7 Watts off an iMac.
http://www.apple.com/imac/specs/
UHCI is for USB 1.1
EHCI is the standard for USB2 which we both agree is slow... unless you're saying there's been an update?
Nice twisting of words. We don't both agree is slow unless you also agree that today FW400 isn't all that speedy.
in an ideal world you'd remove the 'host' every morning and take it with you - leaving your peripherals still functioning, connected to whatever network you want operating for whatever you need
- downloading on offpeak periods, buying stocks, keeping your fridge running,
monitoring your solar panels, keeping your animals fed & warm,
allowing your HDD files to be accessible, backing up data - i don't know, use your imagination
Right. Show me any FW system today that does any of this.
its an issue because not everyone wants to own a desktop at home and a notebook to run around with - i know i would rather have just one machine if it could serve both purposes (no matter how much i love my mac pro i'd rather be able to take it with me)
and i think the future is transportable - which means that keeping CPU loads as light as possible means the fastest graphics possible - which you mentioned is one of your interests...
GPU load <> CPU load. And the current MacBook is WAY faster for graphics than the old MacBook even if you're transferring files and chewing up 13% of the CPU.
i find it ironic that you're using the internet, started/invented by the US military - to claim that the military has no impact on the commercial market...
Actually, while DARPA built the protocols adopted by the NSF, it was NSFNET that later became the internet. I was on ARPANET, BITNET and NSFNET in those early days. Today ARPANET lives on in NIPRNET (formerly MILNET)...not the internet. Ultimately it was the research and subercomputer centers that spawned internet as we know it. Notably CERN and HTML and NCSA...who developed Mosaic.
But you're also talking ancient history. Even at the birth of the internet in the mid 80s, DARPA's research was lagely complete in the 70s. Where the military leads today is in material science. Steath materials, armor, polymers, high temp metals, etc are all technologies where they lead commerical industry.
Computers and computer protocols is not where they spend their real R&D money. Even things like custom chips are derived from existing commercial baselines (like PA Semi with ARM).