Actually it's bridges on the package that make the difference, AMD laser cuts the SMP bridges on Barton core cpus that are intended to be XP's, these can be either failed MP chips, or much more commonly they are just XP's that were never tested for SMP compliance because the extra testing would take time and additional testing equipment. The problems introduced by running SMP XP's are that the quality of connecting the bridges varies greatly and is never as good as the origional connection that was laser cut, and the fact that you may have a core that AMD already rejected because it failed to work correctly in their SMP tests.
windows has it (for all versions of windows that support SMP), and Linux does as of 2.5.8-pre3 and can have it sort of hacked on to earlier kernals via This module.
Mutlimedia, which is the only thing other than games that is pushing the PC at this point will generally get nearly 100% speed improvement from SMP. For instance I run a virtual studio using Reason and have all of my effects and the OS on cpu 0 and the program itself on cpu 1. This allows very low latency and the ability to run more effects without problems. Media encoding (where do you think all those mpeg4 rips come from =) is another task that is frequently done with SMP machines. Other than media though I really can't think of something that a home user would run that would need SMP.
When Novell sold off Unix they gave the trademark to the Open Group, sold the code to SCO, and sold the labs to HP (which ran them into the ground like nearly everything they've done in the last 10 years or so outside of printing). So technically SCO DOES own Unix from the code perspective and they are of course a registered Unix vendor with the Open Group.
Linux as a platform can never be certified as UNIX because there is too much you can change that would break the compatibility. A specific distro with a specific kernal might be, but with the testing and liscensing costs in the 5 figure range I can't see IBM doing it unless a major client says it is a hold item on a contract.
Try NWN, on my GF3 Ti200 I can only get about 25fps at 1024*768 with a bunch of the more hungry options turned off. Of course its a RPG so the fps doesn't matter so much but I definitly wouldn't mind over 40fps.
I used to do this, but with XP and the system restore points I just make a restore point before installing new hardware and roll back if the drivers f' things up. Of course I still do my weekly backups (good enough for a home machine)
What I wish reviewers would start benchmarking is MINIMUM framerate, even if the card averages 120fps if it dips to 20fps for a second it will be highly irritating and noticable. My personal rule of thumb for action games is to find the resolution and settings that average 60fps, then you rarely get below 30fps dips.
Why go with dual MP's at 1.2Ghz when you will get more speed for almost all apps and more memory badwidth out of a modern XP. An Athlon XP 2100+ will be about the same average cpu performance as two MP 1.2Ghz's and will cost the same, you can use DDR400 ram with it, and a SiS746 based board will run you around $65.
hint they are better than professional cards for many things. Trust me we upgraded from some FireGL cards to Geforce 3's and our engineer's SolidWorks performance went WAY up. Actually with 256MB of DDR-2 ram I bet these will be faster than just about any pro card out there (pro cards used to have the advantage with huge oncard texture memory, seems the commercial cards are catching up fast)
Dude I can't get voice recognition to work reliably when fed by a voice cancelling headphone on a 1.2Ghz machine and a couple hours of training, why do you think you would be able to track tons of CMDA signals (this is not trivial due to the way CDMA works), decode the audio or random voices, and process it with voice recognition software? It would cost a lot more than $14K to do this over 100 channels. Now targeted scanning could be a problem, but then maybe I can get my freaking cellular provider to turn on basic GSM encryption (phone supports it but none of the cell sites in the US do AFAIK).
Yes and I think it's a shame that they didn't do a better job. Of course I guess I'm just talking as an old geezer that's been on the net since 1991 or so back when the distinction was imposed through proper reading of RFC's and peer pressure =)
Well the current prototype is the size of the entire PC-Card sleeve for the iPaq so I don't think it's quite ready for even Type 3 PC Card form factor.
The possibilities for software radio are mind boggling.
And the short length of time your batteries will last will boggle the mind even more. Using a general purpose CPU to do all of that comm stuff would use many times more power then dedicated ASIC's. To find out how much this would suck, insert an 802.11b card into the PC Card sleeve on an iPaq, do a constant ping, and run an app that utilized 100% cpu.
modern consumer level cards do not have a seperate Z buffer memory area. The z buffer depth for the Ati 9700Pro is 24bits, the Nvidia Geforce FX line is 32 bits. But z buffer memory is not half of all memory used, due to Z culling and hidden surface removal it is probably a very small % of the total ram used, the majority in most DX8 level games is probably texture space, and on DX9 level games it is probably split between texture and scratch space (need to store the results of the shader and pixel programs somewhere before final display assuming you aren't doing single pass to RAMDAC)
You mentioned Tom's Hardware and reputable in the same paragraph and it wasn't talking about the lack thereof, for shame. (yep it's flaimbait but I have karma to burn and Tom has more bias then a CNN reporter, he just changes loyalties every so often to seem "fair and balanced")
Funny enough I just built a pretty decent system for significantly less then the cost of either of these cards. It's a AMD Athlon XP 2100+ based system using an SIS746 mobo, 512MB of DDR333 ram, CDRW, 400W server class case, lowend graphics card and a pair of 20GB hdd's I had laying around using software RAID1. Total cost $350, around $100 less than just these cards. Sure it can't play the latest games because it has no 3D accelerator, but that could be remedied for around $120 with a Geforce 4 Ti4200 128MB card, it would still come in at about the same cost as the cards alone.
It doesn't matter because the featuresets of DX9 and OpenGl 2.0 are about 99% overlapping. A card that can implement one can very easily implement the other.
A substantial portion of the budget of my ISP both in time and capital is devoted these days to blocking and dealing with spam. I am personal friends with the owner of the ISP and he told me that the recent reduction in prices for dialup accounts would have been larger but he can't afford to do that due to the increased costs of dealing with spam, all of his other costs, from machinery to trunk lines, to BRI's have gone down, this is the only area of his business where costs are increasing.
spamhaus is NOT a company, important distinction. Note the.org at the end of the domain, although some companies get stupid and register in.org spamhaus is not one of them.
I believe they get creative and call it a "use" tax on the item, that is they are taxing your ability to use it in your home state, not a tax on the sale of the item which originated outside their state borders and is hence off limits. It's an interstate sales tax by another name and color but it serves the same purpose without technically violationing the interstate commerce clause of the constitution. I ignore them out of following the spirit of the clause and I'm not sure anyone would volutarily add more to their tax burden when the state has no way to track these purchases.
Actually for most peoples desktop machines S-ATA and Ultra 320 SCSI are faster than their PCI busses (33Mhz-32bit PCI is only 133MB/s). Besides, who cares if the bottleneck moves to the physical connection, at least it is away from the slow arse hdd's of today (8+ms access time and slow max transfer vs.5ns access and fast as your bus can handle transfer). I would upgrade in a heartbeat. Plus removing on of the last important physical device would do wonders for improving the reliability of computers (the cpu and case fans are still there). To get the most out of this you might need to redesign things, but I would be willing to bet that the first uses are for retrofits, market inertia tends to work that way.
Slashdot Mirror
Actually it's bridges on the package that make the difference, AMD laser cuts the SMP bridges on Barton core cpus that are intended to be XP's, these can be either failed MP chips, or much more commonly they are just XP's that were never tested for SMP compliance because the extra testing would take time and additional testing equipment. The problems introduced by running SMP XP's are that the quality of connecting the bridges varies greatly and is never as good as the origional connection that was laser cut, and the fact that you may have a core that AMD already rejected because it failed to work correctly in their SMP tests.
windows has it (for all versions of windows that support SMP), and Linux does as of 2.5.8-pre3 and can have it sort of hacked on to earlier kernals via This module.
Mutlimedia, which is the only thing other than games that is pushing the PC at this point will generally get nearly 100% speed improvement from SMP. For instance I run a virtual studio using Reason and have all of my effects and the OS on cpu 0 and the program itself on cpu 1. This allows very low latency and the ability to run more effects without problems. Media encoding (where do you think all those mpeg4 rips come from =) is another task that is frequently done with SMP machines. Other than media though I really can't think of something that a home user would run that would need SMP.
When Novell sold off Unix they gave the trademark to the Open Group, sold the code to SCO, and sold the labs to HP (which ran them into the ground like nearly everything they've done in the last 10 years or so outside of printing). So technically SCO DOES own Unix from the code perspective and they are of course a registered Unix vendor with the Open Group.
Linux as a platform can never be certified as UNIX because there is too much you can change that would break the compatibility. A specific distro with a specific kernal might be, but with the testing and liscensing costs in the 5 figure range I can't see IBM doing it unless a major client says it is a hold item on a contract.
Try NWN, on my GF3 Ti200 I can only get about 25fps at 1024*768 with a bunch of the more hungry options turned off. Of course its a RPG so the fps doesn't matter so much but I definitly wouldn't mind over 40fps.
I used to do this, but with XP and the system restore points I just make a restore point before installing new hardware and roll back if the drivers f' things up. Of course I still do my weekly backups (good enough for a home machine)
What I wish reviewers would start benchmarking is MINIMUM framerate, even if the card averages 120fps if it dips to 20fps for a second it will be highly irritating and noticable. My personal rule of thumb for action games is to find the resolution and settings that average 60fps, then you rarely get below 30fps dips.
Why go with dual MP's at 1.2Ghz when you will get more speed for almost all apps and more memory badwidth out of a modern XP. An Athlon XP 2100+ will be about the same average cpu performance as two MP 1.2Ghz's and will cost the same, you can use DDR400 ram with it, and a SiS746 based board will run you around $65.
hint they are better than professional cards for many things. Trust me we upgraded from some FireGL cards to Geforce 3's and our engineer's SolidWorks performance went WAY up. Actually with 256MB of DDR-2 ram I bet these will be faster than just about any pro card out there (pro cards used to have the advantage with huge oncard texture memory, seems the commercial cards are catching up fast)
Doh, that of course should read noise cancelling headset, maybe 3am is too late to post....
Dude I can't get voice recognition to work reliably when fed by a voice cancelling headphone on a 1.2Ghz machine and a couple hours of training, why do you think you would be able to track tons of CMDA signals (this is not trivial due to the way CDMA works), decode the audio or random voices, and process it with voice recognition software? It would cost a lot more than $14K to do this over 100 channels. Now targeted scanning could be a problem, but then maybe I can get my freaking cellular provider to turn on basic GSM encryption (phone supports it but none of the cell sites in the US do AFAIK).
Yes and I think it's a shame that they didn't do a better job. Of course I guess I'm just talking as an old geezer that's been on the net since 1991 or so back when the distinction was imposed through proper reading of RFC's and peer pressure =)
Well the current prototype is the size of the entire PC-Card sleeve for the iPaq so I don't think it's quite ready for even Type 3 PC Card form factor.
No this uses the huge sleeve addon for the iPaq, it's about 3/4 as thick as the iPaq and the same in all of the other dimensions.
The possibilities for software radio are mind boggling.
And the short length of time your batteries will last will boggle the mind even more. Using a general purpose CPU to do all of that comm stuff would use many times more power then dedicated ASIC's. To find out how much this would suck, insert an 802.11b card into the PC Card sleeve on an iPaq, do a constant ping, and run an app that utilized 100% cpu.
Obviously it does, all of the removed surfaces are removed from memory along with their Z buffer data =)
modern consumer level cards do not have a seperate Z buffer memory area. The z buffer depth for the Ati 9700Pro is 24bits, the Nvidia Geforce FX line is 32 bits. But z buffer memory is not half of all memory used, due to Z culling and hidden surface removal it is probably a very small % of the total ram used, the majority in most DX8 level games is probably texture space, and on DX9 level games it is probably split between texture and scratch space (need to store the results of the shader and pixel programs somewhere before final display assuming you aren't doing single pass to RAMDAC)
You mentioned Tom's Hardware and reputable in the same paragraph and it wasn't talking about the lack thereof, for shame. (yep it's flaimbait but I have karma to burn and Tom has more bias then a CNN reporter, he just changes loyalties every so often to seem "fair and balanced")
Funny enough I just built a pretty decent system for significantly less then the cost of either of these cards. It's a AMD Athlon XP 2100+ based system using an SIS746 mobo, 512MB of DDR333 ram, CDRW, 400W server class case, lowend graphics card and a pair of 20GB hdd's I had laying around using software RAID1. Total cost $350, around $100 less than just these cards. Sure it can't play the latest games because it has no 3D accelerator, but that could be remedied for around $120 with a Geforce 4 Ti4200 128MB card, it would still come in at about the same cost as the cards alone.
It doesn't matter because the featuresets of DX9 and OpenGl 2.0 are about 99% overlapping. A card that can implement one can very easily implement the other.
A substantial portion of the budget of my ISP both in time and capital is devoted these days to blocking and dealing with spam. I am personal friends with the owner of the ISP and he told me that the recent reduction in prices for dialup accounts would have been larger but he can't afford to do that due to the increased costs of dealing with spam, all of his other costs, from machinery to trunk lines, to BRI's have gone down, this is the only area of his business where costs are increasing.
spamhaus is NOT a company, important distinction. Note the .org at the end of the domain, although some companies get stupid and register in .org spamhaus is not one of them.
I believe they get creative and call it a "use" tax on the item, that is they are taxing your ability to use it in your home state, not a tax on the sale of the item which originated outside their state borders and is hence off limits. It's an interstate sales tax by another name and color but it serves the same purpose without technically violationing the interstate commerce clause of the constitution. I ignore them out of following the spirit of the clause and I'm not sure anyone would volutarily add more to their tax burden when the state has no way to track these purchases.
Actually for most peoples desktop machines S-ATA and Ultra 320 SCSI are faster than their PCI busses (33Mhz-32bit PCI is only 133MB/s). Besides, who cares if the bottleneck moves to the physical connection, at least it is away from the slow arse hdd's of today (8+ms access time and slow max transfer vs .5ns access and fast as your bus can handle transfer). I would upgrade in a heartbeat. Plus removing on of the last important physical device would do wonders for improving the reliability of computers (the cpu and case fans are still there). To get the most out of this you might need to redesign things, but I would be willing to bet that the first uses are for retrofits, market inertia tends to work that way.