fluffles.net is the personal website and playground of Enlightenment.
You can find news, blogs, articles and other interesting stuff here about computers, science and storage.
Posted by Enlightenment on 17 maart 2008 @ 06:50
AMD has met its promise and starts shipping bug-free "B3" Opteron chips.
Tough luck
Processor manufacturer AMD is in desperate need of good news since the failed Phenom-processor launch in late 2007. That launch was spoiled by lower clock speeds than previously announced and a bug that might crash the system in very specific situations.
The TLB-bug caused customers to stay away from the Phenom chip and made AMD postpone the introduction of the Opteron "Barcelona", the server-version of the Phenom-processor. AMD noted that the bug occurred only at very specific situations and therefore should prove no problem for most customers. Yet, the bugged chips would not sell even at very low prices.
But now, AMD is shipping both server and desktop CPUs with a bug-free B3-revision. That means AMD is finally back in the game. The B3 revision does not change much about its performance level, but at least there is no 'dark shadow' cast above AMD's chips anymore.
AMD's recovery
AMD is in a race to get 45nm production going by the third and fourth quarter this year. The 65nm SOI production process has proved a great struggle for AMD. It both consumed too much power for high Phenom clock rates to become possible and the yields were disappointing. If AMD can stick to its deadlines, they might make a reasonable recovery. Intel is making good progress and selling 45nm products already, although in limited supply. On the CeBit 2008 technology fare in Hannover, Germany, AMD demonstrated a running Phenom 45nm system, proving the company has made progress in its journey towards 45nm.
AMD's 8-core Montreal
In 2009, AMD is said to release a native 8-core chip codenamed Montreal, this means no 2 x 4 (dual quadcore) design will be used, but rather a real native octal-core. Just like Intel's Nehalem architecture
, by the way.
Phenom X4 9150e and 9350e
AMD will also release TLB-corrected quadcore chips with a TDP of just 65W. These are dubbed the 9150e and 9350e, and should arrive in summer this year. I should mention that the X4 9100e never was released to retail channel and that the dates have been pushed for more than once. No word on their actual power consumption, and they haven't been reviewed yet.
Tough luck
Processor manufacturer AMD is in desperate need of good news since the failed Phenom-processor launch in late 2007. That launch was spoiled by lower clock speeds than previously announced and a bug that might crash the system in very specific situations.
The TLB-bug caused customers to stay away from the Phenom chip and made AMD postpone the introduction of the Opteron "Barcelona", the server-version of the Phenom-processor. AMD noted that the bug occurred only at very specific situations and therefore should prove no problem for most customers. Yet, the bugged chips would not sell even at very low prices.
But now, AMD is shipping both server and desktop CPUs with a bug-free B3-revision. That means AMD is finally back in the game. The B3 revision does not change much about its performance level, but at least there is no 'dark shadow' cast above AMD's chips anymore.
AMD's recovery
AMD is in a race to get 45nm production going by the third and fourth quarter this year. The 65nm SOI production process has proved a great struggle for AMD. It both consumed too much power for high Phenom clock rates to become possible and the yields were disappointing. If AMD can stick to its deadlines, they might make a reasonable recovery. Intel is making good progress and selling 45nm products already, although in limited supply. On the CeBit 2008 technology fare in Hannover, Germany, AMD demonstrated a running Phenom 45nm system, proving the company has made progress in its journey towards 45nm.
AMD's 8-core Montreal
In 2009, AMD is said to release a native 8-core chip codenamed Montreal, this means no 2 x 4 (dual quadcore) design will be used, but rather a real native octal-core. Just like Intel's Nehalem architecture
, by the way.
Phenom X4 9150e and 9350e
AMD will also release TLB-corrected quadcore chips with a TDP of just 65W. These are dubbed the 9150e and 9350e, and should arrive in summer this year. I should mention that the X4 9100e never was released to retail channel and that the dates have been pushed for more than once. No word on their actual power consumption, and they haven't been reviewed yet.
There are no replies yet
Posted by Enlightenment on 17 maart 2008 @ 06:21
Intel manages to outclass its competitor AMD on yet another selling-point: idle power consumption has dropped to under 5W for its new 45nm-generation dualcore and quadcore chips.
Idle power consumption
While AMD is having a hang-over from a party gone awry, Intel is catching up in some areas AMD traditionally has a better record. One of those areas is idle power consumption, when the system is not using any significant amount of processor power. Since processors have become so fast, most of the time their processing power isn't required nor useful so becomes unused. Modern chips have advanced power-saving techniques, which can significantly lower the power consumption under these circumstances.
And the good thing is: most of the time, a computer is idling! Whether you are leaving your computer on between lunch or work on it, your system is being so lowly utilized that power saving will kick in. Emailing, chatting, listening to music, writting in office applications are all examples of common tasks which require relatively few processing power, allowing the processor to save a lot of power.
Power hogs
The real power hogs are 3D-games, since their inefficient endless-loop design will always consume 100% CPU (singlecore) even though the framerates are well beyond optimal levels (>75fps). Watching a 1080p High Definition movie with very high resolution and compression might also require too much processing power to let your CPU continue to run in power saving mode.
So idle power usage is more important than maximum power usage?
Yes! So do not rely on TDP for your "power consumption" but look at idle consumption instead. Besides, the maximum consumption is often well below the TDP value, as you can read here. For multicore chips it might be interesting to see the power savings when only one core is fully utilized.
How does the power saving work?
When the CPU is idling, power savers like AMD's Cool'N'Quiet and Intel's EIST will lower power consumption by reducing the voltage and frequency of the processor. The result is a drastic decrease in current leakage, resulting in lower power consumption. Hurray!
But what if I start an application and need processing power again? Well once the CPU is beyond a threshold utilization (say 65%) it will "ramp up" to a higher voltage and/or frequency, meaning the CPU will be faster but consumes more energy. It will continue to throttle until it reaches its maximum rated frequency and voltage, when required. The throttling goes so fast that performance decreases caused by the power saving techniques is minimal. So it shouldn't do any harm.
Future technology to save power
AMD's upcoming Turion Ultra
chip has seperate power planes, allowing both cores to work at a different voltage and frequency. In the future, that would mean a chip with 64 or 128 cores will only need a few active when idling, causing hefty power savings. Plus, it combines both performance with power-efficiency.
How to enable power saving on my PC?
If you're using Windows, set your power scheme to "Minimal power saving", which you'll find in Control Panel -> Power options. Also, install the CPU driver for your processor, i.e. AMD. If you're on FreeBSD, make sure the cpufreq driver is loaded in kernel or via module and enable the power daemon powerd.
Why should i care about low-power?
you pay less on your electricity bill, resulting in savings that can be up to €150 a year for a single system.
your system runs cooler, because all consumed energy is converted into heat. Cool systems are more stable and potentially have a longer lifespan.
your systems runs quieter, since less cooling in terms of air movement is required to dissipate the generated heat. Some systems even allow operation without any active cooling, so no fans and zero noise!
you save money because valuable materials used by cooling such as copper and aluminium, will not be needed in large quantities. Modest heatsinks will be sufficient.
low-power components often have a lower failure rate; a CPU drawing 140W may cause material degradation, thereby shortening its lifespan.
you help save the environment by consuming less energy, thus producing less greenhouse gasses. Are you Co²-neutral yet?
Reducing power consumption is not only in the best interest of our environment, but also common business sense. The biggest challenge in modern cpu design is reducing power consumption; for it limits the scaling of frequency and thereby performance. A simple die-shrink (producing on new 45nm process instead of 65nm) should enable either higher performance by higher clock frequency or lower power drain if it keeps the same frequency. Power is everything(tm).
Intel versus AMD in idle power consumption
Up until now, Intel hasn't had much success in lowering the idle power consumption. The Pentium 4 turned into a power hungry underperformer (remember Prescott?) and the Core2Duo Conroe still consumed a hefty 25W when doing absolutely nothing!
AMD did much better, thanks primarily to the Cool'N'Quiet technology which lowers voltage and frequency on-the-fly. Most Athlon 64 processors do under 10W while being idle. Some single-core models even go as low as 1W, while the dualcore BE-series do about 4-5W. Phenom, the new but plagued quadcore chip from AMD, also will have a low-power version dubbed the 9100e (B2) and 9150e (B3, TLB bug corrected) which will launch in summer this year. It is still unclear how much energy this chip uses when idling.
Idle power consumption
While AMD is having a hang-over from a party gone awry, Intel is catching up in some areas AMD traditionally has a better record. One of those areas is idle power consumption, when the system is not using any significant amount of processor power. Since processors have become so fast, most of the time their processing power isn't required nor useful so becomes unused. Modern chips have advanced power-saving techniques, which can significantly lower the power consumption under these circumstances.
And the good thing is: most of the time, a computer is idling! Whether you are leaving your computer on between lunch or work on it, your system is being so lowly utilized that power saving will kick in. Emailing, chatting, listening to music, writting in office applications are all examples of common tasks which require relatively few processing power, allowing the processor to save a lot of power.
Power hogs
The real power hogs are 3D-games, since their inefficient endless-loop design will always consume 100% CPU (singlecore) even though the framerates are well beyond optimal levels (>75fps). Watching a 1080p High Definition movie with very high resolution and compression might also require too much processing power to let your CPU continue to run in power saving mode.
So idle power usage is more important than maximum power usage?
Yes! So do not rely on TDP for your "power consumption" but look at idle consumption instead. Besides, the maximum consumption is often well below the TDP value, as you can read here. For multicore chips it might be interesting to see the power savings when only one core is fully utilized.
How does the power saving work?
When the CPU is idling, power savers like AMD's Cool'N'Quiet and Intel's EIST will lower power consumption by reducing the voltage and frequency of the processor. The result is a drastic decrease in current leakage, resulting in lower power consumption. Hurray!
But what if I start an application and need processing power again? Well once the CPU is beyond a threshold utilization (say 65%) it will "ramp up" to a higher voltage and/or frequency, meaning the CPU will be faster but consumes more energy. It will continue to throttle until it reaches its maximum rated frequency and voltage, when required. The throttling goes so fast that performance decreases caused by the power saving techniques is minimal. So it shouldn't do any harm.
Future technology to save power
AMD's upcoming Turion Ultra
chip has seperate power planes, allowing both cores to work at a different voltage and frequency. In the future, that would mean a chip with 64 or 128 cores will only need a few active when idling, causing hefty power savings. Plus, it combines both performance with power-efficiency.How to enable power saving on my PC?
If you're using Windows, set your power scheme to "Minimal power saving", which you'll find in Control Panel -> Power options. Also, install the CPU driver for your processor, i.e. AMD. If you're on FreeBSD, make sure the cpufreq driver is loaded in kernel or via module and enable the power daemon powerd.
Why should i care about low-power?
you pay less on your electricity bill, resulting in savings that can be up to €150 a year for a single system.
your system runs cooler, because all consumed energy is converted into heat. Cool systems are more stable and potentially have a longer lifespan.
your systems runs quieter, since less cooling in terms of air movement is required to dissipate the generated heat. Some systems even allow operation without any active cooling, so no fans and zero noise!
you save money because valuable materials used by cooling such as copper and aluminium, will not be needed in large quantities. Modest heatsinks will be sufficient.
low-power components often have a lower failure rate; a CPU drawing 140W may cause material degradation, thereby shortening its lifespan.
you help save the environment by consuming less energy, thus producing less greenhouse gasses. Are you Co²-neutral yet?Reducing power consumption is not only in the best interest of our environment, but also common business sense. The biggest challenge in modern cpu design is reducing power consumption; for it limits the scaling of frequency and thereby performance. A simple die-shrink (producing on new 45nm process instead of 65nm) should enable either higher performance by higher clock frequency or lower power drain if it keeps the same frequency. Power is everything(tm).
Intel versus AMD in idle power consumption
Up until now, Intel hasn't had much success in lowering the idle power consumption. The Pentium 4 turned into a power hungry underperformer (remember Prescott?) and the Core2Duo Conroe still consumed a hefty 25W when doing absolutely nothing!
AMD did much better, thanks primarily to the Cool'N'Quiet technology which lowers voltage and frequency on-the-fly. Most Athlon 64 processors do under 10W while being idle. Some single-core models even go as low as 1W, while the dualcore BE-series do about 4-5W. Phenom, the new but plagued quadcore chip from AMD, also will have a low-power version dubbed the 9100e (B2) and 9150e (B3, TLB bug corrected) which will launch in summer this year. It is still unclear how much energy this chip uses when idling.
There are no replies yet
Posted by Enlightenment on 29 februari 2008 @ 13:52 - Source: [ FreeBSD ]
Oh yes, time has come! The long-awaited FreeBSD 7.0 operating system is finally released
to the wild. This version incorporates many new features and finishes the work that has been started since FreeBSD 5.0, when a fundamentally new architecture was put in place. Today is the day the fruits of this long journey have become visible to the public, with impressing results!
A quick summary of major changes:
Drastic performance increases, especially in SMP workloads
Database benchmarks show 7.0 to be 15% faster than the best performing Linux-kernel (2.6.24) in realistic workloads
Experimental support for the Sun ZFS filesystem, bringing one of the most advanced filesystems to FreeBSD!
Support for journaling using the new geom_journal module, supports UFS
Auto-adapting network buffers and TSO/LRO support
Many drivers were re-written to be freed of the inefficient Giant-lock
New lagg-driver allows link aggregation (combine two gigabit adapters into one 2Gbps link)
The ULE process scheduler has been updated to version 3, and incorporates all the work done in the SCHED_SMP project.
The libthr threading model is now enabled by default
The freebsd-update utility allows binary upgrades
CD's are bundled with X.Org 7.3, KDE 3.5.8, GNOME 2.20.2.
Get it while it's still hot:
Dutch FTP mirror of FreeBSD 7.0 i386
(32-bit)
Dutch FTP mirror of FreeBSD 7.0 amd64
(64-bit)
Other links:
FreeBSD 7.0 Release Announcement
(quick summary of changes)
FreeBSD 7.0 Release Notes
(full-list of changes)
FreeBSD 7.0 Hardware Notes
(check if it supports your hardware)
Onlamp: What's new in FreeBSD 7.0?
(in-depth article)
to the wild. This version incorporates many new features and finishes the work that has been started since FreeBSD 5.0, when a fundamentally new architecture was put in place. Today is the day the fruits of this long journey have become visible to the public, with impressing results!A quick summary of major changes:
Drastic performance increases, especially in SMP workloads
Database benchmarks show 7.0 to be 15% faster than the best performing Linux-kernel (2.6.24) in realistic workloads
Experimental support for the Sun ZFS filesystem, bringing one of the most advanced filesystems to FreeBSD!
Support for journaling using the new geom_journal module, supports UFS
Auto-adapting network buffers and TSO/LRO support
Many drivers were re-written to be freed of the inefficient Giant-lock
New lagg-driver allows link aggregation (combine two gigabit adapters into one 2Gbps link)
The ULE process scheduler has been updated to version 3, and incorporates all the work done in the SCHED_SMP project.
The libthr threading model is now enabled by default
The freebsd-update utility allows binary upgrades
CD's are bundled with X.Org 7.3, KDE 3.5.8, GNOME 2.20.2.Get it while it's still hot:
Dutch FTP mirror of FreeBSD 7.0 i386
(32-bit)Dutch FTP mirror of FreeBSD 7.0 amd64
(64-bit)Other links:
FreeBSD 7.0 Release Announcement
(quick summary of changes)FreeBSD 7.0 Release Notes
(full-list of changes)FreeBSD 7.0 Hardware Notes
(check if it supports your hardware)Onlamp: What's new in FreeBSD 7.0?
(in-depth article)
| Reply by Blobber on Fri, 29 Feb 2008 @ 18:28 | Quote |
Now that should be exciting news for you Enlightenment (well quite, for the rc's have been available for some time
)
It's a pity that I have based my fileserver on gentoo, maybe I can find some old hardware + time, just to have some fun with Freebsd 7.I am very curious if and how zfs will work
| Reply by p3n1x420 (unregistered) on Mon, 03 Mar 2008 @ 23:40 | Quote |
happy happy joy joy
im using 7 as well. finding it quite plesant
execution time: 225 msec, queries: 20






