I just ran into this with XboxLive and Open NAT vs Moderate NAT. I have a dual-homed Linux box as a firewall, so UPnP is rather a pain to setup, and I didn't really like the idea of enabling such a beast on my firewall (which filters both inbound AND outbound packets - even on the LAN side).
Turns out UPnP is just an easier way to get XboxLive to work but NOT required. If you have an iptables based Linux firewall, just log all blocked/dropped packets while doing an XboxLive Connection test, then go back and enable those ports both directions (I had to use a combination of state tracking outbound connections, enabling forward rules both ways, and also doing DNAT on some unrelated incoming packets on the INPUT chain). I believe the main ports are 88, 1026 and 3074, mostly UDP. After fixing all blocked/logged packets during the XboxLive test, XboxLive reports my connection as OPEN NAT instead of moderate (but with the firewall still fully functional and filtering both ingress/egress packets). The key for me was to enable incoming 3074 UDP packets using DNAT to my Xbox360 IP and also enable WAN->LAN forward chain for the same ports. My Xbox360 has a staticIP so DNAT is rather easy to configure.
I am much happier with the iptables solution to get Open-NAT status instead of installing/configuring UPnP, or replacing my linux firewall with a Hasbro one (I actually do have a "Hasbro" one behind my linux firewall - mostly to serve as a 802.11g access point). UPnP is also poorly named. Most people think it has something to do with USB or other "Plug and Play" external devices, like HDDs and flash cards. Little do most people know it has to do with punching holes in your firewall, and it's enabled by DEFAULT in Windows (and most routers). There's really no excuse for it - instead of publishing online help such as "Get a router that supports UPnP", Microsoft could just come out and say "open and forward these ports to your Xbox360". A lot of people know how to add port forwarding and such on their routers. It's not rocket science. UPnP is completely insecure if enabled on a router. There's no protection from a rogue UPnP client at all.
Same here - it's better than my VHS version. I was hoping for 16:9 anamorphic but instead it's 4:3 letterboxed. But it's still better than my SE-VHS version, and much better than the very old 2-head mono recording I made off of TV (yes, now that I have the theatrical version on DVD I can erase my 15-year time shifted copy - if it even plays back at this point!)
It's not nearly as pretty as the cleaned up SE anamorphic 16:9, but you know what, even with the inferior picture and 2.0 ProLogic sound, the experience felt more like Star Wars as I remembered it than any of the new versions. Since it's better quality than my VHS copies, and won't deteriorate every time I play it, I feel they are worth the investment and sell my previous trilogy copy to someone who doesn't own the trilogy and doesn't care about the purity of the original script/CGI.
The transfer isn't really that horrible - it's grainy compared to the SE anamorphic, but it's better than my VHS copy, and it's still WAY better than the pathetic transfer of Blade Runner, to which there is still NO decent DVD available for.
I am no fan of what Lucas has done to the original Trilogy, but hey, at least this is pretty darn close and I guess I'm nostalgic for liking the "old feel" which the old transfer + soundtrack evokes (pre-1997 that's the only way I remember Star Wars anyway). I actually enjoyed watching this version, whereas the SE versions make me gag at times (Han stepping on Jabba). Unless you have a LD player, and you want un-edited original, this is the best so far so maybe we shouldn't criticize too much. If this limited edition doesn't sell well that will be a clear sign that there was no reason to release the original in the first place and we might see a lot fewer efforts put forth for releasing original theatrical versions.
Those of us using DLP/Plasma/LCD television that are capable of 1080i/720p/1080p and that are of reasonably large size fully understand the need for a higher quality format to become standard. For us, the DVD looks awful in comparison to what we see on our HD television channels and our Xbox360's, or the output from our PC's. And we are clearly interested in being able to purchase a movie with twice as much data in it, to more accurately represent what you would see from film at the theater.
Except that unless you have a lot of HDMI or DVI-HDCP inputs available after you hook up your satellite receiver, DVR, DVD player, HD-DVD player, Xbox, etc, any of the analog inputs you actually have available will be crippled to only 960x540p or something else barely better than the existing 480p of DVD.
No doubt. This is the number one problem with both HD-DVD and Blu-Ray. Their main advantage is the higher HD resolution.... except that you can't use it on most of the HDTVs out there. Ironically, most of those HDTVs happen to be those owned by the techno-geeks (like most/.'ers) who would otherwise be all over this new HD format stuff if the technology wasn't so blatantly crippled.
Furthermore, most new HDMI equipped TVs have only 1 or 2 digital inputs. What am I supposed to do with my 8:1 Component Video Switch? So not only would I have to replace my perfectly good 1080i HD CRT to use this silly format, but I would have to replace a perfectly good component video switch as well. Not to mention that in some places, you can easily wire your house with mini-Component coax cables, although I've yet to see HDMI home video distribution equipment (and with the DRM, do you think HDMI is ever going to be available for home video distribution?!)
The minority of us who care about the technical advantages of these formats are the ones who are not going to be able to use them. Please vote with your wallet and let these companies know that we want a HD format that works with ALL HDTVs.
I presume you refer to Athlon64, using unbuffered non-ECC RAM. This is one thing I believe is being addressed in the new Venus cores (better memory controller/compatibility with DIMMs). Even so, worst case you run the memory at DDR333 instead of DDR400, and stability should be okay, at the loss of some performance. (Most boards today seem to work fine at DDR400 if all the DIMMs are SINGLE sided).
Now, in the Opteron market, where you are using buffered ECC DIMMs, we have machines with 8, 16 and 32GB of RAM and not one problem whatsoever.
Oh yeah, and Dell lost our contract for that server because they wouldn't sell us Opterons (they wanted to sell us Xeon-EM64s).
Actually, it's the other way around. Smaller fabrication geometries are more susceptible to SEUs. This is just basic physics. As the geometries get smaller, it takes less energy to disturb the state of a capacitor or transistor and flip it's intended logic state. As geometries continue to shrink, SEUs will become an issue for all consumer hardware, even at sea-level. This is why all level A certified FAA devices require such things as triple redundant logic with majority voting and scrubbing systems.
Because of the ccNUMA architecture for the 2xx and 8xx Opterons, if what you need is a really fast single processor box, it's best to stick with the Opteron 1xx - since then all the RAM you install is local to that processor, and you do not need a NUMA aware kernel/OS.
However, if you do want multiprocessing, the quad 850s are very very nice - but you need to make sure your OS/kernel supports ccNUMA correctly or the memory performance will hurt. Even with the correct support, however, an 8GB 4-way Opteron 850 will only have 2GB local to each processor - so if you have a big job that requires 4GB of RAM, it will run just slightly slower than a single Opteron 150 with local access to all 4GB of RAM.
The penalty isn't too bad however - I believe the HT links are 3.2GB/sec 1-way (6.4GB/sec total) and the memory bandwidth at the processor is 5.4GB/sec, IIRC.
We have been benchmarking several loaner boxes at work to determine what will be our next purchases for our compute farm. We do primarily ASIC and FPGA design, simulation and verification. We have been in dire need of >4GB boxes, and until just recently, we had been forced to run on Solaris machine to get 8GB.
The day of the Opteron, however, has come at last: All these were run with stock tools in 32-bit mode, no fancy compiler optimizations. These are the same programs that we run on 2GHz P4s.
Agilent 3070 VCL vector conversion Perl program (which I wrote, this is very typical of the Perl programs we run to process large vector files - the benchmark only times data processing in memory, no file IO on read/write): Sun Blade-1000 750MHz: 103.08 sec P4 3.06GHz: 36.93 sec Opt 148 (2.2GHz): 27.01 sec Quad Opt 848: 27.42 sec Quad Xeon64 (3.6GHz): 31.17 sec
Modelsim 5.8c simulation of LogicBIST simulation on 50K Flop ASIC: P4 3.06GHz: 5955 sec Opt 148: 3798 sec 4x Opt 848: 5985 sec (See note below) 4x Xeon64: 4858 sec
Mentor Flextest fault grading using make -j1, -j2 and -j4 (parallel runs, results combined in later step that is not benchmarked): Sun Blade-1000: 7362 sec(-j1) P4 3.06GHz: 2188 sec(-j1) Dual P4 3.06GHz: 2189 sec/1333 sec (j2) Opt 2.2GHz 128: 1493 sec 4x Opt 2.2GHz 848: 1562 sec(j1)/ 779 sec(j2)/ 393 sec (j4) 4x Xeon64 3.6GHz: 1465 sec(j1)/796 sec(j2)/ 879 sec(j4)
Mentor LbistArchitect on 50K ASIC: Sun Blade-1000: 15698 sec P4 3.06GHz: 3877 sec Opt 148: 2845 sec 4x Opt 848: 3534 sec (See note below) 4x Xeon-64 3.6GHz: 2604 sec
Note - the poor performance of the quad opteron box was done on RedHat Enterprise Linux 3 AS-6, and I noticed that the SMP kernel did NOT have CONFIG_K8_NUMA set to y, so it's not fair to judge those numbers until we get a new kernel with ccNUMA support. I have run synthetic benchmarks on them too, and the memory performance on the Quad Opteron was indeed hurt by the lack of CONFIG_K8_NUMA in the linux kernel.
Clearly though, the HyperTransport makes the Quad Opteron box scale very well, whereas the Quad Xeon box choked on 4 threads, probably beacuse the memory bus became saturated and the processors starved for data.
Also, any serious optimizations need to use gcc-3.4.1 - which has specific optimizations for both Opteron and Nocona cores. gcc-3.4.0 does not have specific optimizations for Nocona ("Xeon64") cores. gcc-3.x does not have specific optimizations for Opteron.
Anyway, our decision has been made - we are buying Opteron 150s for all our new compute farm boxes.
Not only that, but picking up CBS through the airwaves allows one to watch March Madness in HDTV, given one has the right equipment. Carrying local channels in HDTV over satellite isn't going to happen for a while because of the bandwidth requirements (maybe the nationwide channels, but not all local affiliates, but then there's the FCC blackout rules).
I also save $2/mth on my DirecTV bill by not subscribing to local channels (not avail here anyway). Instead I am picking up all the locals in HDTV format with a $29.99 antenna I got at Best Buy.
The required kernel modifications are not as difficult as you make them seem to be. Granted, I will agree with you that they are slow in releasing patches, but the patch itself is actually quite small.
First, there is the mki-adapter patch, and since it is fairly orthogonal to the kernel (adds non-existing functionality), it will have almost no chance of colliding with anything during a patch. The kernel patch itself can usually be applied to a fairly close kernel number, or with very few changes (if you know how diff/patch work). Since it deals with the memory management/paging, if there wasn't a major difference between the kernel you are compiling and the patch version, the patch will probably succeed. I'm using the 2.4.20 Win4Lin kernel patches with the 2.4.21 kernel right now, and it works okay.
Now, 2.5.x kernels? Well, really. Can you expect them to have patches for a development version of the kernel? Especially when it so closely connected to memory management? Since most customers of Win4Lin are probably in a SOHO or production environment, I'm not sure how many people are clamoring for 2.5.x patches from Netraverse. I'd say that if you want to run Win4Lin you have to live with a stock 2.4.x kernel. That's acceptable, i think. You can always have more than one kernel (dual-boot kernels!)
NeTraverse DOES supply vanilla patches for the stock linux kernels that you can download from ftp.kernel.org. You do not have to use their prebuilt RedHat/Debian/WhatEver kernels. I'm using a plain vanilla 2.4.21 kernel on RedHat 7.3.
But yes, if you are a 2.5.x user, you'll have to skip on Win4Lin. It's a little hard to ask them to support a moving target that the 2.5.x series kernels are. I wouldn't use 2.5.x in a production environment anyway.
Do what my company does: One big ass Dell machine that runs Linux for engineering work. One additional big ass Dell machine running WindowsXP for the "corporate network" and EMail. Two monitors/keyboards/mice even, no KVMs. Very efficient, not! =-)
At least in our own little engineering group, we've started to use Wine when we can for certain vendors that refuse to port their apps to Linux. But for EDA tools, most of them are supporting Linux finally. But you'd think since IBM is such a big Linux backer, why we can't get Lotus Notes for Linux? That is primarily why I have a 2.4GHz WindowsXP box.
Also, I can simply back up the windows related directories (which are installed in my home folder on linux) and then I never have to install windows again, I can just install Win4Lin and restore the directories complete with programs, file associations, and serial numbers!
I forgot to mention that in one of my previous posts. I LOVE this feature. Since it resides totally inside of the linux filesystem, I can completely backup my "windows" install during my regular linux tape backup. And if i hose the registry? Just restore that directory, from Linux. No longer the chicken-egg problems I had to trying to back up a real Windows partition to tape and actually be able to restore from tape without the damn registry still being corrupted or some other nonsense. Maybe I didn't try hard enough, but i could never get just a file-based restore to ever get a Windows partition fully working again. Win4Lin let's you backup your entire Windows system purely at the file level (and with full Unix permissions too!).
Perhaps I do have a dual-boot system, for playing Games, but when I'm in linux and I really need to open Office2000 I can do that without shutting down all my Linux apps, rebooting, work on the Excel file, then reboot to Linux, reopen all my files...
Furthermore, Win4Lin is more than just a nice hack. They way they integrated it into the linux kernel (for memory management support) and the fact that it just sits on top of your linux filesystem means that it integrates into a linux box fairly well. It uses your linux system for sound, and uses linux's networking. The only thing you can't get is direct hardware access except for COM and LPT1. (But that is nice for being able to run Palm Desktop directly in Linux and sync my Palm, since so many palm apps have conduits that only work in Windows). I don't want to reboot into Windows just to install some Palm program that the programmer only ships in a self-extracting.EXE file and requires Palm Desktop to be installed before installing just a.PRC nonsense (they do exist).
Yeah, Wine does most of this too, and I started with that, but since I already own a Win98 CD, I tried Win4Lin out, and found it's memory requirements and sheer speed (tight integration to linux subsystems) make it worth it. (And ZERO compatibility problems since you are in fact, running a real Windows install. I had a lot of problems installing some apps into Wine, but I'm sure Wine has improved since 3 years ago).
And lastly, probably a bizarre reason I like Win4Lin on my Linux desktop: it's a testimony to the good ol days of running OS/2 and Windows 3 apps seamlessly in windows on the same desktop. Any windows app i need i can bring up in its own little window, quickly, and i don't have to reboot the box just to read some Excel file that KSpread chokes on.
Ack. My Bad. It supports direct serial and telnet, not SSH. Still, it's the best I have found. I also use it to connect to my own machine dialup with PPP, so I don't care about SSH since I'm not all that worried about someone hijacking my CSD GSM connection. If only I could get GPRS in all areas...
When I first got my Tungsten T I wanted a good telnet/SSH app but couldn't find one. All I could find was MochaTelnet, ptelnet, and TGSSH, all of which either were incompatible with OS5, weren't compatible with SSH2, or just had such ugly fonts that I couldn't see what I was doing.
It supports telnet and SSH. It supports file download with X-Modem, Y-Modem to VFS. And most important, it supports HIRES, so you can do a full 80x24 terminal mode and stil read it. On a fast enough connection (say GPRS), using VIM is totally usable.
Precisely. There is literally decades of research on the design and testability of synchronously clocked designs, whereas there is very little on asynchronous designs. All the EDA tools available for testing chips today (processors, ASICs, what have you) are all based on synchronous design principles. To change to asynchronous design requires an entire paradigm shift, from functional design, to testability, to producing vendor tools to work the flow. Synchronous designs have been based on stuck at fault testing for decades, and greatly simplifies the task of proving 10 million transistors are doing what they are supposed to. One basically only has to check for a 1 or a 0 in the right place during the right clock cycle. Asynchronous designs basically require verifying the timing and path delays from every gate in the chip to every other gate. There is no predictable time when things will happen, or when things will get there, since timing delays are dependent on fabrication process and variation.
For now, asynchronous logic works best in small pieces of a much larger, synchronous design, and where it makes sense - interfacing with things that are asynchronous (UARTs, ADCs, RF receivers, etc). Usually, one can verify with great confidence of achieving over 99% fault coverage on the synchronous portion, while resorting to just functional tests to see if the async logic works right. Writing functional tests for an entire chip these days however is almost insurmountable, unless you have all the time and money in the world to burn. Because synchronous designs have more structure and follow rigid design criteria, structural testing is far far easier.
The drawbacks of synchronous designs is of course, clock tree synthesis and controlling skew. Power can be dealt with these days by gating off clocks to unused regions, using lower power FFs, etc. However, controlling the clock skew on a single clock chip can be the largest hurdle during layout and fabrication. Even so, it is not an impossible task, and verifying full scan synchronous designs via ATPG and/or BIST outweighs most benefits of purely asynchronous designs.
A fully asynchronous processor or chip will not become economically feasible until more research is done. A chip that cannot be tested is worth nothing at all.
My Panasonic SVHS VCR does this and it is quite a nice feature. Coming home after having a power outtage to find my microwave blinking, my clocks blinking, but my VCR knows what time it is. Ahh, progress. =)
Perhaps it means it can now change the desktop size too. Perhaps I'm also displaying my ignorance, but whenever I define multiple resolutions in my XF86Config file and switch between them, I always end up with a lower resolution screen with a virtual desktop the size of my maximum resolution. The CTL-ALT-+ trick seems to change the screen resolution, but not the desktop resolution. I have never figured out how to change both from inside XFree86.
Do what I do. Uninstall every single RPM on your RedHat box pertaining to KDE, goto www.kde.org and download the latest KDE tarballs and compile it from source. The way RedHat has KDE setup, I don't even recognize it as KDE or find it nearly as pleasant to use as a vanilla KDE install.
RedHat's KDE feels like Gnome, and I prefer KDE's look and feel without RedHat's paws all over it. Not to mention that, but I have had far fewer problems installing 3rd party KDE apps on KDE installed from source, because of the aforementioned code changes RedHat has made which apparently breaks more things than it fixes. Just ask the Kapital developers how much fun it is supporting all the different distros.
Re:Not a troll, just a question ...
on
AMD's Athlon XP 2700+
·
· Score: 2, Informative
You're thinking too much along the lines of a home-user running Windows. There's a very real presence of Intel (and unfortunately, not so much AMD) CPUs in low cost workstations for business use. Linux is certainly part of the picture as well, as almost all EDA vendors have or are releasing Linux versions of their tools.
What do I use my 80x86 cpu for? Well, I work in a hardware engineering group which does ASIC and FPGA design. We have a CPU farm of about 30 machines with Intel P4s running RedHat 7.2. (May see AMD Hammer chips in the future - we are excited about this possibility). We run everything from RTL and gate level VHDL and Verilog simulations, to chip synthesis, to test insertion and fault grading simulations. One of the last chips I worked on required such a large set of ATPG vectors (and the design was just so huge), that it required breaking the test vectors into ten groups, and even then, just one file (10% of the total) required an 8GB Sun box to convert the vectors to the fabs tester format, and the gate-level simulation took 10 days. PER FILE. Yes, that was total of 100 CPU days of simulation time for one chip just for ATPG vectors. And these were running on 1.8GHz Pentium 4s with 3GB of RAM. Not surprisingly, leading edge tools in this field are starting to look at distributed simulation over high-speed backplanes (read: not ethernet).
Tomorrow's technology is designed and verified on today's hardware. Every generational step in every sector of industry leapfrogs like this. You can't design next year's high performance video card using 80286s. Definitely for ASIC/FPGA design, there isn't a system fast enough for how quickly we (the engineers) would like a simulation to finish in. Being able to run more simulations overall means a better design out the door. More stuff caught up front. The faster a simulation runs, the quicker it will finish, meaning we can get by with fewer high-priced licenses for our EDA tools. (Licenses are usually in the tens or hundreds of thousands of dollars EACH).
You can never run enough tests before a product is done. How many tests/simulations are run, depends on how long they take to run. Give me the fastest CPU you got, decked out with the most physical RAM it can handle. (Sadly, the 32bit limit on current 80x86 platforms is hurting us badly - go x86-64! go AMD! Capture the workstation market!)
The article refers to the Des Moines Area Community College (DMACC). It's just that, a 2-year community college. The major universities in Iowa are University of Iowa, Iowa State University and University of Northern Iowa. Note they have university in their name, not community college. =)
Paperless is great for many things, but a paperless library???
They'd probably have a better chance of tracking you by triangulating your position using the received signal strength at several cell antennas.
This won't work with many (if not all) digital cell phones. In order to prevent a close-to-tower phone from blocking out a distant weak signal phone, the tower and the cell phone communicate to each other and the tower tells the phone if its too "loud" or not. The phone then regulates its output power such that ALL the phones talking to one tower, irregardless of distance, arrive at the tower with the same signal power. Each incoming signal being at the same energy level is pretty important for making a spread spectrum system work.
Slashdot Mirror
I just ran into this with XboxLive and Open NAT vs Moderate NAT. I have a dual-homed Linux box as a firewall, so UPnP is rather a pain to setup, and I didn't really like the idea of enabling such a beast on my firewall (which filters both inbound AND outbound packets - even on the LAN side).
Turns out UPnP is just an easier way to get XboxLive to work but NOT required. If you have an iptables based Linux firewall, just log all blocked/dropped packets while doing an XboxLive Connection test, then go back and enable those ports both directions (I had to use a combination of state tracking outbound connections, enabling forward rules both ways, and also doing DNAT on some unrelated incoming packets on the INPUT chain). I believe the main ports are 88, 1026 and 3074, mostly UDP. After fixing all blocked/logged packets during the XboxLive test, XboxLive reports my connection as OPEN NAT instead of moderate (but with the firewall still fully functional and filtering both ingress/egress packets). The key for me was to enable incoming 3074 UDP packets using DNAT to my Xbox360 IP and also enable WAN->LAN forward chain for the same ports. My Xbox360 has a staticIP so DNAT is rather easy to configure.
I am much happier with the iptables solution to get Open-NAT status instead of installing/configuring UPnP, or replacing my linux firewall with a Hasbro one (I actually do have a "Hasbro" one behind my linux firewall - mostly to serve as a 802.11g access point). UPnP is also poorly named. Most people think it has something to do with USB or other "Plug and Play" external devices, like HDDs and flash cards. Little do most people know it has to do with punching holes in your firewall, and it's enabled by DEFAULT in Windows (and most routers). There's really no excuse for it - instead of publishing online help such as "Get a router that supports UPnP", Microsoft could just come out and say "open and forward these ports to your Xbox360". A lot of people know how to add port forwarding and such on their routers. It's not rocket science. UPnP is completely insecure if enabled on a router. There's no protection from a rogue UPnP client at all.
Same here - it's better than my VHS version. I was hoping for 16:9 anamorphic but instead it's 4:3 letterboxed. But it's still better than my SE-VHS version, and much better than the very old 2-head mono recording I made off of TV (yes, now that I have the theatrical version on DVD I can erase my 15-year time shifted copy - if it even plays back at this point!)
It's not nearly as pretty as the cleaned up SE anamorphic 16:9, but you know what, even with the inferior picture and 2.0 ProLogic sound, the experience felt more like Star Wars as I remembered it than any of the new versions. Since it's better quality than my VHS copies, and won't deteriorate every time I play it, I feel they are worth the investment and sell my previous trilogy copy to someone who doesn't own the trilogy and doesn't care about the purity of the original script/CGI.
The transfer isn't really that horrible - it's grainy compared to the SE anamorphic, but it's better than my VHS copy, and it's still WAY better than the pathetic transfer of Blade Runner, to which there is still NO decent DVD available for.
I am no fan of what Lucas has done to the original Trilogy, but hey, at least this is pretty darn close and I guess I'm nostalgic for liking the "old feel" which the old transfer + soundtrack evokes (pre-1997 that's the only way I remember Star Wars anyway). I actually enjoyed watching this version, whereas the SE versions make me gag at times (Han stepping on Jabba). Unless you have a LD player, and you want un-edited original, this is the best so far so maybe we shouldn't criticize too much. If this limited edition doesn't sell well that will be a clear sign that there was no reason to release the original in the first place and we might see a lot fewer efforts put forth for releasing original theatrical versions.
Except that unless you have a lot of HDMI or DVI-HDCP inputs available after you hook up your satellite receiver, DVR, DVD player, HD-DVD player, Xbox, etc, any of the analog inputs you actually have available will be crippled to only 960x540p or something else barely better than the existing 480p of DVD.
No doubt. This is the number one problem with both HD-DVD and Blu-Ray. Their main advantage is the higher HD resolution.... except that you can't use it on most of the HDTVs out there. Ironically, most of those HDTVs happen to be those owned by the techno-geeks (like most /.'ers) who would otherwise be all over this new HD format stuff if the technology wasn't so blatantly crippled.
Furthermore, most new HDMI equipped TVs have only 1 or 2 digital inputs. What am I supposed to do with my 8:1 Component Video Switch? So not only would I have to replace my perfectly good 1080i HD CRT to use this silly format, but I would have to replace a perfectly good component video switch as well. Not to mention that in some places, you can easily wire your house with mini-Component coax cables, although I've yet to see HDMI home video distribution equipment (and with the DRM, do you think HDMI is ever going to be available for home video distribution?!)
The minority of us who care about the technical advantages of these formats are the ones who are not going to be able to use them. Please vote with your wallet and let these companies know that we want a HD format that works with ALL HDTVs.
As long as such a converter is still significantly cheaper than buying a new HDCP-TV, they'll probably be in high demand.
Venus core should read Venice core. Oops.
I presume you refer to Athlon64, using unbuffered non-ECC RAM. This is one thing I believe is being addressed in the new Venus cores (better memory controller/compatibility with DIMMs). Even so, worst case you run the memory at DDR333 instead of DDR400, and stability should be okay, at the loss of some performance. (Most boards today seem to work fine at DDR400 if all the DIMMs are SINGLE sided).
Now, in the Opteron market, where you are using buffered ECC DIMMs, we have machines with 8, 16 and 32GB of RAM and not one problem whatsoever.
Oh yeah, and Dell lost our contract for that server because they wouldn't sell us Opterons (they wanted to sell us Xeon-EM64s).
Actually, it's the other way around. Smaller fabrication geometries are more susceptible to SEUs. This is just basic physics. As the geometries get smaller, it takes less energy to disturb the state of a capacitor or transistor and flip it's intended logic state. As geometries continue to shrink, SEUs will become an issue for all consumer hardware, even at sea-level. This is why all level A certified FAA devices require such things as triple redundant logic with majority voting and scrubbing systems.
Because of the ccNUMA architecture for the 2xx and 8xx Opterons, if what you need is a really fast single processor box, it's best to stick with the Opteron 1xx - since then all the RAM you install is local to that processor, and you do not need a NUMA aware kernel/OS.
However, if you do want multiprocessing, the quad 850s are very very nice - but you need to make sure your OS/kernel supports ccNUMA correctly or the memory performance will hurt. Even with the correct support, however, an 8GB 4-way Opteron 850 will only have 2GB local to each processor - so if you have a big job that requires 4GB of RAM, it will run just slightly slower than a single Opteron 150 with local access to all 4GB of RAM.
The penalty isn't too bad however - I believe the HT links are 3.2GB/sec 1-way (6.4GB/sec total) and the memory bandwidth at the processor is 5.4GB/sec, IIRC.
We have been benchmarking several loaner boxes at work to determine what will be our next purchases for our compute farm. We do primarily ASIC and FPGA design, simulation and verification. We have been in dire need of >4GB boxes, and until just recently, we had been forced to run on Solaris machine to get 8GB.
The day of the Opteron, however, has come at last:
All these were run with stock tools in 32-bit mode, no fancy compiler optimizations. These are the same programs that we run on 2GHz P4s.
Agilent 3070 VCL vector conversion Perl program (which I wrote, this is very typical of the Perl programs we run to process large vector files - the benchmark only times data processing in memory, no file IO on read/write):
Sun Blade-1000 750MHz: 103.08 sec
P4 3.06GHz: 36.93 sec
Opt 148 (2.2GHz): 27.01 sec
Quad Opt 848: 27.42 sec
Quad Xeon64 (3.6GHz): 31.17 sec
Modelsim 5.8c simulation of LogicBIST simulation on 50K Flop ASIC:
P4 3.06GHz: 5955 sec
Opt 148: 3798 sec
4x Opt 848: 5985 sec (See note below)
4x Xeon64: 4858 sec
Mentor Flextest fault grading using make -j1, -j2 and -j4 (parallel runs, results combined in later step that is not benchmarked):
Sun Blade-1000: 7362 sec(-j1)
P4 3.06GHz: 2188 sec(-j1)
Dual P4 3.06GHz: 2189 sec/1333 sec (j2)
Opt 2.2GHz 128: 1493 sec
4x Opt 2.2GHz 848: 1562 sec(j1)/ 779 sec(j2)/ 393 sec (j4)
4x Xeon64 3.6GHz: 1465 sec(j1)/796 sec(j2)/ 879 sec(j4)
Mentor LbistArchitect on 50K ASIC:
Sun Blade-1000: 15698 sec
P4 3.06GHz: 3877 sec
Opt 148: 2845 sec
4x Opt 848: 3534 sec (See note below)
4x Xeon-64 3.6GHz: 2604 sec
Note - the poor performance of the quad opteron box was done on RedHat Enterprise Linux 3 AS-6, and I noticed that the SMP kernel did NOT have CONFIG_K8_NUMA set to y, so it's not fair to judge those numbers until we get a new kernel with ccNUMA support. I have run synthetic benchmarks on them too, and the memory performance on the Quad Opteron was indeed hurt by the lack of CONFIG_K8_NUMA in the linux kernel.
Clearly though, the HyperTransport makes the Quad Opteron box scale very well, whereas the Quad Xeon box choked on 4 threads, probably beacuse the memory bus became saturated and the processors starved for data.
Also, any serious optimizations need to use gcc-3.4.1 - which has specific optimizations for both Opteron and Nocona cores. gcc-3.4.0 does not have specific optimizations for Nocona ("Xeon64") cores. gcc-3.x does not have specific optimizations for Opteron.
Anyway, our decision has been made - we are buying Opteron 150s for all our new compute farm boxes.
Not only that, but picking up CBS through the airwaves allows one to watch March Madness in HDTV, given one has the right equipment. Carrying local channels in HDTV over satellite isn't going to happen for a while because of the bandwidth requirements (maybe the nationwide channels, but not all local affiliates, but then there's the FCC blackout rules).
I also save $2/mth on my DirecTV bill by not subscribing to local channels (not avail here anyway). Instead I am picking up all the locals in HDTV format with a $29.99 antenna I got at Best Buy.
No scratches maybe, but he might accidentally delete it to make room for MP3s and then you have to download it all over again!
The required kernel modifications are not as difficult as you make them seem to be. Granted, I will agree with you that they are slow in releasing patches, but the patch itself is actually quite small.
First, there is the mki-adapter patch, and since it is fairly orthogonal to the kernel (adds non-existing functionality), it will have almost no chance of colliding with anything during a patch. The kernel patch itself can usually be applied to a fairly close kernel number, or with very few changes (if you know how diff/patch work). Since it deals with the memory management/paging, if there wasn't a major difference between the kernel you are compiling and the patch version, the patch will probably succeed. I'm using the 2.4.20 Win4Lin kernel patches with the 2.4.21 kernel right now, and it works okay.
Now, 2.5.x kernels? Well, really. Can you expect them to have patches for a development version of the kernel? Especially when it so closely connected to memory management? Since most customers of Win4Lin are probably in a SOHO or production environment, I'm not sure how many people are clamoring for 2.5.x patches from Netraverse. I'd say that if you want to run Win4Lin you have to live with a stock 2.4.x kernel. That's acceptable, i think. You can always have more than one kernel (dual-boot kernels!)
NeTraverse DOES supply vanilla patches for the stock linux kernels that you can download from ftp.kernel.org. You do not have to use their prebuilt RedHat/Debian/WhatEver kernels. I'm using a plain vanilla 2.4.21 kernel on RedHat 7.3.
But yes, if you are a 2.5.x user, you'll have to skip on Win4Lin. It's a little hard to ask them to support a moving target that the 2.5.x series kernels are. I wouldn't use 2.5.x in a production environment anyway.
Do what my company does: One big ass Dell machine that runs Linux for engineering work. One additional big ass Dell machine running WindowsXP for the "corporate network" and EMail. Two monitors/keyboards/mice even, no KVMs. Very efficient, not! =-)
At least in our own little engineering group, we've started to use Wine when we can for certain vendors that refuse to port their apps to Linux. But for EDA tools, most of them are supporting Linux finally. But you'd think since IBM is such a big Linux backer, why we can't get Lotus Notes for Linux? That is primarily why I have a 2.4GHz WindowsXP box.
I forgot to mention that in one of my previous posts. I LOVE this feature. Since it resides totally inside of the linux filesystem, I can completely backup my "windows" install during my regular linux tape backup. And if i hose the registry? Just restore that directory, from Linux. No longer the chicken-egg problems I had to trying to back up a real Windows partition to tape and actually be able to restore from tape without the damn registry still being corrupted or some other nonsense. Maybe I didn't try hard enough, but i could never get just a file-based restore to ever get a Windows partition fully working again. Win4Lin let's you backup your entire Windows system purely at the file level (and with full Unix permissions too!).
Oh please.
.EXE file and requires Palm Desktop to be installed before installing just a .PRC nonsense (they do exist).
Perhaps I do have a dual-boot system, for playing Games, but when I'm in linux and I really need to open Office2000 I can do that without shutting down all my Linux apps, rebooting, work on the Excel file, then reboot to Linux, reopen all my files...
Furthermore, Win4Lin is more than just a nice hack. They way they integrated it into the linux kernel (for memory management support) and the fact that it just sits on top of your linux filesystem means that it integrates into a linux box fairly well. It uses your linux system for sound, and uses linux's networking. The only thing you can't get is direct hardware access except for COM and LPT1. (But that is nice for being able to run Palm Desktop directly in Linux and sync my Palm, since so many palm apps have conduits that only work in Windows). I don't want to reboot into Windows just to install some Palm program that the programmer only ships in a self-extracting
Yeah, Wine does most of this too, and I started with that, but since I already own a Win98 CD, I tried Win4Lin out, and found it's memory requirements and sheer speed (tight integration to linux subsystems) make it worth it. (And ZERO compatibility problems since you are in fact, running a real Windows install. I had a lot of problems installing some apps into Wine, but I'm sure Wine has improved since 3 years ago).
And lastly, probably a bizarre reason I like Win4Lin on my Linux desktop: it's a testimony to the good ol days of running OS/2 and Windows 3 apps seamlessly in windows on the same desktop. Any windows app i need i can bring up in its own little window, quickly, and i don't have to reboot the box just to read some Excel file that KSpread chokes on.
Ack. My Bad. It supports direct serial and telnet, not SSH. Still, it's the best I have found. I also use it to connect to my own machine dialup with PPP, so I don't care about SSH since I'm not all that worried about someone hijacking my CSD GSM connection. If only I could get GPRS in all areas...
When I first got my Tungsten T I wanted a good telnet/SSH app but couldn't find one. All I could find was MochaTelnet, ptelnet, and TGSSH, all of which either were incompatible with OS5, weren't compatible with SSH2, or just had such ugly fonts that I couldn't see what I was doing.
Then I found this program:
Online
It supports telnet and SSH. It supports file download with X-Modem, Y-Modem to VFS. And most important, it supports HIRES, so you can do a full 80x24 terminal mode and stil read it. On a fast enough connection (say GPRS), using VIM is totally usable.
It's definitely worth the $30.
Precisely. There is literally decades of research on the design and testability of synchronously clocked designs, whereas there is very little on asynchronous designs. All the EDA tools available for testing chips today (processors, ASICs, what have you) are all based on synchronous design principles. To change to asynchronous design requires an entire paradigm shift, from functional design, to testability, to producing vendor tools to work the flow. Synchronous designs have been based on stuck at fault testing for decades, and greatly simplifies the task of proving 10 million transistors are doing what they are supposed to. One basically only has to check for a 1 or a 0 in the right place during the right clock cycle. Asynchronous designs basically require verifying the timing and path delays from every gate in the chip to every other gate. There is no predictable time when things will happen, or when things will get there, since timing delays are dependent on fabrication process and variation.
For now, asynchronous logic works best in small pieces of a much larger, synchronous design, and where it makes sense - interfacing with things that are asynchronous (UARTs, ADCs, RF receivers, etc). Usually, one can verify with great confidence of achieving over 99% fault coverage on the synchronous portion, while resorting to just functional tests to see if the async logic works right. Writing functional tests for an entire chip these days however is almost insurmountable, unless you have all the time and money in the world to burn. Because synchronous designs have more structure and follow rigid design criteria, structural testing is far far easier.
The drawbacks of synchronous designs is of course, clock tree synthesis and controlling skew. Power can be dealt with these days by gating off clocks to unused regions, using lower power FFs, etc. However, controlling the clock skew on a single clock chip can be the largest hurdle during layout and fabrication. Even so, it is not an impossible task, and verifying full scan synchronous designs via ATPG and/or BIST outweighs most benefits of purely asynchronous designs.
A fully asynchronous processor or chip will not become economically feasible until more research is done. A chip that cannot be tested is worth nothing at all.
My Panasonic SVHS VCR does this and it is quite a nice feature. Coming home after having a power outtage to find my microwave blinking, my clocks blinking, but my VCR knows what time it is. Ahh,
progress. =)
Perhaps it means it can now change the desktop size too. Perhaps I'm also displaying my ignorance, but whenever I define multiple resolutions in my XF86Config file and switch between them, I always end up with a lower resolution screen with a virtual desktop the size of my maximum resolution.
The CTL-ALT-+ trick seems to change the screen resolution, but not the desktop resolution. I have never figured out how to change both from inside XFree86.
Do what I do. Uninstall every single RPM on your RedHat box pertaining to KDE, goto www.kde.org and download the latest KDE tarballs and compile it from source. The way RedHat has KDE setup, I don't even recognize it as KDE or find it nearly as pleasant to use as a vanilla KDE install.
RedHat's KDE feels like Gnome, and I prefer KDE's look and feel without RedHat's paws all over it. Not to mention that, but I have had far fewer problems installing 3rd party KDE apps on KDE installed from source, because of the aforementioned code changes RedHat has made which apparently breaks more things than it fixes. Just ask the Kapital developers how much fun it is supporting all the different distros.
You're thinking too much along the lines of a home-user running Windows. There's a very real presence of Intel (and unfortunately, not so much AMD) CPUs in low cost workstations for business use. Linux is certainly part of the picture as well, as almost all EDA vendors have or are releasing Linux versions of their tools.
What do I use my 80x86 cpu for? Well, I work in a hardware engineering group which does ASIC and FPGA design. We have a CPU farm of about 30 machines with Intel P4s running RedHat 7.2. (May see AMD Hammer chips in the future - we are excited about this possibility). We run everything from RTL and gate level VHDL and Verilog simulations, to chip synthesis, to test insertion and fault grading simulations. One of the last chips I worked on required such a large set of ATPG vectors (and the design was just so huge), that it required breaking the test vectors into ten groups, and even then, just one file (10% of the total) required an 8GB Sun box to convert the vectors to the fabs tester format, and the gate-level simulation took 10 days. PER FILE. Yes, that was total of 100 CPU days of simulation time for one chip just for ATPG vectors. And these were running on 1.8GHz Pentium 4s with 3GB of RAM. Not surprisingly, leading edge tools in this field are starting to look at distributed simulation over high-speed backplanes (read: not ethernet).
Tomorrow's technology is designed and verified on today's hardware. Every generational step in every sector of industry leapfrogs like this. You can't design next year's high performance video card using 80286s. Definitely for ASIC/FPGA design, there isn't a system fast enough for how quickly we (the engineers) would like a simulation to finish in. Being able to run more simulations overall means a better design out the door. More stuff caught up front. The faster a simulation runs, the quicker it will finish, meaning we can get by with fewer high-priced licenses for our EDA tools. (Licenses are usually in the tens or hundreds of thousands of dollars EACH).
You can never run enough tests before a product is done. How many tests/simulations are run, depends on how long they take to run. Give me the fastest CPU you got, decked out with the most physical RAM it can handle. (Sadly, the 32bit limit on current 80x86 platforms is hurting us badly - go x86-64! go AMD! Capture the workstation market!)
The article refers to the Des Moines Area Community College (DMACC). It's just that, a 2-year community college. The major universities in Iowa are University of Iowa, Iowa State University and University of Northern Iowa. Note they have university in their name, not community college. =)
Paperless is great for many things, but a paperless library???
This won't work with many (if not all) digital cell phones. In order to prevent a close-to-tower phone from blocking out a distant weak signal phone, the tower and the cell phone communicate to each other and the tower tells the phone if its too "loud" or not. The phone then regulates its output power such that ALL the phones talking to one tower, irregardless of distance, arrive at the tower with the same signal power. Each incoming signal being at the same energy level is pretty important for making a spread spectrum system work.