It looks a though it will plug into a standard PCI/ISA passive backplane. This has a 32bit PCI and a 16bit ISA slot lined up.
Each single board computer generally runs it's own copy of an OS. (This company seems to be biased towards NT). The backplane can then be used to talk to PCI cards these could be anything from DSPs to multi modem cards. These things generally find there way into industrial servers that have to have multiple redundant cpu boards etc.
It would be possible for the computers to talk to each other using MPI over the PCI bus. Multiple boxes could then talk to each other using SCI to make a large MPI cluster.
Scali http://www.scali.com make systems based on 4 way SMP boxes connected together with SCI.
In practice Backplane systems tend to use higher performace proprietary designs such as Sun's "Gigaplane" I believe that someone makes a systems based upon backplane Ultrasparc systems connected by SCI. Now that would really rock. However it still wouldn't be any good for Unreal.
Supercomputers are for compute intensive tasks that parallelize easily they don't have any benifit for texture maped games:)
IP was designed to "self heal" in the event of failure of part of the system. However the cost of maintaining redundant links is high. For example the whole of the UK academic network "JANET" is connected to the UK internet at 1 point (LINX) in the UK (it's has its own links to the states).
These days whole countries can be cut off by one or two links failing
This would be very difficult given the current design of the PS2. The reason that RAMBUS was chosen was that it provided the required memory bandwidth (remember this thing only has to render at a fixed resolution that is known in advance) at the right space, thermal and pin count costs.
If the PS2 had been manufactured using SDRAM it would required a 6 layer board instead of the 4 layer one that it actually uses and the board would have had to be bigger to accomodate the traces. For Sony the increased cost of the memory was irelevent compared to the saving in other areas. I think that RAMBUS will not die but it will find a niche in things like the PS2. It is doomed to forever be a low volume product. It is the wrong technology for high performance architectures at the present moment.
It will be interesting to see how the Alpha 21364 performs, putting the controler on the chip might sort out some of the latency issues, but something will still need to be done about the heat (perhaps a shrink).
Unfortunately the best hardware like from manufacturer like: DAL, MOTU, Digidesign and other pro-audio manufacturers will probably never have decent linux support.
Since programs like Cubase VST, Cakewalk, Sadie and ProTools sell very well under Windows there is no incentive to port to Linux. It is a catch 22 without these programs hardware support is pointless and vice-versa.
I have used Slab under Linux and it has a long way to go before it gets anywhere near Cubase VST.
So the wheel should be reinvented on a regular basis then? This is what happens when you don't have at least one Compsci graduate on the team. Compsci teaches lots of basic algorithms (most of the course is what _not_ to do) and also how to estimate how fast new algoithms will run. If the game in your example is CPU bound then a Athlon 850 + DDR Geforce would probably bring it up to a playable 30-40 frames a second. However if it I/O bound then you'll have to wait for a change in memory architecture for a speed up. I do not rate Gates as a programmer _all_ of his code was scrapped in the MSDOS 3.3 rewrite.
Not really. All benchmarks should be taken with a hefty pinch of salt, especially TPC-C
Linus would be the first to admitt that Linux is not all things to all people. However fine grain locking in 2.4 should make SMP suck slightly less.
The ProLiant systems were built using 3 clustered boxes each with 32 550MHz(2MB L2 cache) processors. The Alpha tested is not representative of the current state of the art, they were only 21164's at 612MHz. The lastest is a 21264 at 700Mhz with 16MB L2 cache! This which would smoke the Intel based box. Sun is a generation behind in the processor stakes, Ultra Sparc II max out at 400MHz and do not have as much cache so it is not suprising that the system is slower. This should be corrected when the Ultra Sparc III is brought out. It is a testement to Solaris that they are competative at all.
Slashdot Mirror
It looks a though it will plug into a standard PCI/ISA passive backplane. This has a 32bit PCI and a 16bit ISA slot lined up.
:)
Each single board computer generally runs it's own copy of an OS. (This company seems to be biased towards NT). The backplane can then be used to talk to PCI cards these could be anything from DSPs to multi modem cards. These things generally find there way into industrial servers that have to have multiple redundant cpu boards etc.
It would be possible for the computers to talk to each other using MPI over the PCI bus. Multiple boxes could then talk to each other using SCI to make a large MPI cluster.
Scali http://www.scali.com make systems based on 4 way SMP boxes connected together with SCI.
In practice Backplane systems tend to use higher performace proprietary designs such as Sun's "Gigaplane" I believe that someone makes a systems based upon backplane Ultrasparc systems connected by SCI. Now that would really rock. However it still wouldn't be any good for Unreal.
Supercomputers are for compute intensive tasks that parallelize easily they don't have any benifit for texture maped games
-dp
IP was designed to "self heal" in the event of failure of part of the system. However the cost of maintaining redundant links is high. For example the whole of the UK academic network "JANET" is connected to the UK internet at 1 point (LINX) in the UK (it's has its own links to the states).
These days whole countries can be cut off by one or two links failing
-dp
This would be very difficult given the current design of the PS2. The reason that RAMBUS was chosen was that it provided the required memory bandwidth (remember this thing only has to render at a fixed resolution that is known in advance) at the right space, thermal and pin count costs.
If the PS2 had been manufactured using SDRAM it would required a 6 layer board instead of the 4 layer one that it actually uses and the board would have had to be bigger to accomodate the traces. For Sony the increased cost of the memory was irelevent compared to the saving in other areas. I think that RAMBUS will not die but it will find a niche in things like the PS2. It is doomed to forever be a low volume product. It is the wrong technology for high performance architectures at the present moment.
It will be interesting to see how the Alpha 21364 performs, putting the controler on the chip might sort out some of the latency issues, but something will still need to be done about the heat (perhaps a shrink).
-dp
Bull, potato is is debian 2.2 (frozen) woody is the next version after that and if you think it is stable you sould try running it :)
-dp
Unfortunately the best hardware like from manufacturer like: DAL, MOTU, Digidesign and other pro-audio manufacturers will probably never have decent linux support.
Since programs like Cubase VST, Cakewalk, Sadie and ProTools sell very well under Windows there is no incentive to port to Linux. It is a catch 22 without these programs hardware support is pointless and vice-versa.
I have used Slab under Linux and it has a long way to go before it gets anywhere near Cubase VST.
-dp
Oh get a life. //'/ computing
1) Beowulf is for
2) A Jecklin disk is for increasing the accuracy of stereo recording
I fail to see the connection
-dp
So the wheel should be reinvented on a regular basis then? This is what happens when you don't have at least one Compsci graduate on the team. Compsci teaches lots of basic algorithms (most of the course is what _not_ to do) and also how to estimate how fast new algoithms will run. If the game in your example is CPU bound then a Athlon 850 + DDR Geforce would probably bring it up to a playable 30-40 frames a second. However if it I/O bound then you'll have to wait for a change in memory architecture for a speed up. I do not rate Gates as a programmer _all_ of his code was scrapped in the MSDOS 3.3 rewrite.
Not really. All benchmarks should be taken with a hefty pinch of salt, especially TPC-C
Linus would be the first to admitt that Linux is not all things to all people. However fine grain locking in 2.4 should make SMP suck slightly less.
The ProLiant systems were built using 3 clustered boxes each with 32 550MHz(2MB L2 cache) processors. The Alpha tested is not representative of the current state of the art, they were only 21164's at 612MHz. The lastest is a 21264 at 700Mhz with 16MB L2 cache! This which would smoke the Intel based box. Sun is a generation behind in the processor stakes, Ultra Sparc II max out at 400MHz and do not have as much cache so it is not suprising that the system is slower. This should be corrected when the Ultra Sparc III is brought out. It is a testement to Solaris that they are competative at all.