It's the reason I moved from Maxtor to Seagate (aside from the fact the failure rate I was getting on Maxtor drives was nearing 100% within 3 years). Now I'm playing around more with WD, Hitachi and others since I got a 1.5TB Seagate drive that kept having lockup issues due to bad firmware. I questioned their quality control since then and have actively avoided Seagate since without regret.
I can agree and disagree with you to some extent. I disagree to the point that I use CID spoofing using Skype.
Why? I use my cell phone for business all of the time, and about 95% of the time I'm using my cell phone for business, I'm at my computer. I will generally opt to use Skype at 2.4 cents per minute instead of my cell phone at stupid long distance rates, but I want to make sure that clients know it is me calling. I have Skype spoof my cell phone CID so that when I call someone regardless if it is by phone or Skype, they know it is me, and I can get away with significantly cheaper long distance rates. Especially when I'm roaming outside of the country.
Re:I actually kind of miss the old combat system
on
Review: Mass Effect 2
·
· Score: 3, Funny
Go to an arcade where it is a popular game and you'll find out real soon.
What I am kinda hoping is that the updates fix some bugs in the drivers. I've had a number of occasions where the existing drivers caused things like BSOD and crashing. Most notably, I received a full hardware crash while using the webcam within Skype. There are some room for improvements in the driver, I just hope this update addresses them.
This has been the argument that I've seen to justify getting a GeForce over a Quadro in CGI. A few points:
1) The memory system on the Tesla/Quadro is much more rigorously tested and held to a much higher standard of quality than the GeForce. There is plenty of research evidence to prove this, and I have had plenty of anecdotal evidence to prove this point as well. NVIDIA doesn't give a crap about the memory in a GeForce because a miscolored pixel in 1/60th of a second doesn't matter. A soft/hard error in GPU memory for scientific calculations can be catastrophic. This is also a reason that NVIDIA is applying ECC memory to the Tesla C2050 and C2070 GPUs.
2) Some GeForce GPUs will have major threading errors after a few minutes of hard running. I've experienced this with a dgemm torture test with a Tesla and 2 GeForce GTX 285 GPUs in a single system. Give the test about 5 minutes on all 3 GPUs simultaneously and the GeForce GPUs will crash out at nearly the same time. The Tesla will continue the test until completion (which can be about a day or so)
3) Bandwidth starvation is a term to indicate that the cards are getting less bandwidth than they should be getting. On this FASTRA machine, only a few slots are full x16 Gen2, which end up being shared across 2 GPUs, making it effectively x8 Gen2 to each GPU. For other slots, it is even worse when you have a x8 Gen2 link going in. That has to be shared between a pair of GPUs. Technically, you can run the Tesla GPU in a x1 Gen1 slot if you had the right adapter, but the time it will take to transfer data from host memory to GPU memory may end up negating any performance benefits you might see out of using the GPU, unless you are using very heavy computational algorithms that are almost completely compute bound.
I couple years ago, I had a compute rig using 6 Tesla C870 GPUs, and even that setup was starting to get bandwidth starved as all GPUs were using a single x8 Gen1 link being aggregated to 6 x4 Gen1 slots (using adapters). I had to up the output data frame size on an MD simulation in order to have all cards performing equally. With smaller frame sizes, the first 4 GPUs were finishing their computations before the last 2 GPUs got their data.
It is kind of unfair to generalize commercial clusters vs homebrew in that manner. Many institutions that purchase commercial clusters from HP/Dell/SGI/etc opt out of the use of InfiniBand or 10GbE. The logic behind it is when the vender says that for $100,000 they can upgrade to IB, the purchaser goes back and says, "For $100,000, I'll just get more cluster nodes instead." This is probably a big reason that gigabit takes up 52% of the Top500 list of supercomputers.
If this is achieved for a personal aircraft, I'd be very much on board with this. My only beef is the addition of things like parachutes and air bags. I don't really care too much for those features, as I might be able to get TKS de-icing systems installed for similar weight for those IFR flights in the great white north. Or if I don't have a TKS system, maybe a little extra payload capacity so I can actually fit 4 passengers and fuel without going over gross weight.
Because I can appreciate and judge great cuisine, doesn't mean I can make it, yet the feedback these judges provide is the cornerstone for a chefs continuous improvement. People who use and judge interfaces in the field are usually a great resource to find ways to improve it. If it is a hit to the interface designers ego that some interface element isn't where the users would like it to be, suck it up. Make it better, always improve it.
I like the comparison you make here, but in reality, it is even worse. For the most part, a $150,000 plane can barely take my family (2 adults, 2 small children) and some luggage with a full load of fuel (legally). I also burn about 22 Gallons of fuel or so on a 206nm flight.
- Support for ics calendar files in Mail.app - More Bluetooth functions, like sending contacts to other phones via bluetooth, or being able to interact with other peripherals - Support for unlocking functions, like many other sane GSM phones. I want to use my AT&T sim card in the US, but I'm locked to my Rogers SIM in Canada.
There are a number of schemes I ended up using in naming systems at my workplace. There really wasn't a rhyme nor a reason to how I named our machines, I just went with what sounded cool, but it also seemed that I had a tendency of having at least 2 system names related.
For example, Excalibur and Dragoon. Genesis and Revelation. Those I guess were the only two system pair that were somewhat related. One system was named Severn simply because I recall a Redhat distro being named Severn and thought it sounded cool. Another, now dead, system was named Velocity because not only did it sound neat, it was also a reflection of the type of acoustic work it was designed to perform.
Excalibur I think was the only system that I had reason to call it that. Being one of the coolest and most sought after swords of legend, it only seemed fitting to call the most powerful workstation in the office that (which at the time was dual Opteron 250's). The next step when I ended up getting my PowerMac Quad G5 was simply to call it Titan. That name ended up succeeding into one of our product names for my HPC startup company.
Because the aeronautical information is public information doesn't necessarily mean it doesn't cost. Jeppesen is one of the primary providers of aviation databases and is used in pretty much all Garmin handheld and panel mount GPS units. I can't say that it justifies cost as much as FAA certification, but the databases alone do cost every 56 days should you choose to update them that frequently.
I don't necessarily fly that often, but I have a $1200 handheld GPS that costs about $50 to update the Jeppesen aviation database, an additional $50 to update obstacle databases, and $150 to update the terrain database (which is not anywhere near as often as the others). For IFR GPS units, I think it is about $150 to update the Jeppesen database.
NVIDIA themselves don't make the personal supercomputer. They partner with companies like ours to build and test systems, like our Slipstream S4 that can be sold and branded as a Tesla personal supercomputer. The system I linked in particular will still give you 4 Tesla C1060 GPUs, but for $8,495. Shameless plug? Sure, but it certainly isn't 10k, and is still certified by NVIDIA.
They have hit the Top500. They just recently made 29th position with 170 Tesla S1070 systems in tandem with Xeon cluster nodes. Their overall performance in LINPACK was about 77.48 TFLOPS. This installation was done by Tokyo Tech.
You're right about the difference pretty much being the DVI port, but it is a pretty expensive DVI port. Compute professionals didn't want the GeForce series because of lack of support, and they didn't want Quadro because it was too expensive, so the Tesla was NVIDIAs middle ground.
Trust me on this one, just because the article only has an implied no, from someone that wasn't actually the CEO, doesn't mean there wasn't an actual no somewhere.
If the author of the article had decided to show up at the CEO's emerging companies summit Q&A session, the quote would have read more like, "No, absolutely not.", then append the the pretty much the jist of the senior VPs quote on the end of that. Not much else to read into it unless you think the guy is lying, rather than batting around the truth.
I haven't done much of the development of the software myself. We have a developer we hired to work with CUDA. From what I've found, the documentation that is available on NVIDIAs site for CUDA is excellent and their developers are active on their forums.
For VMD, it was necessary to have 1 CPU core per GPU. We tested 6 GPUs with 4 cores and we could only spawn 4 threads for GPU processing. The guys at Evolved Machines told me they can use multi GPU off of a single core. If so, I have no idea how. NVIDIA even tells me 1 core per GPU, so that is the gospel I'm following by. Acceleware for some of their stuff even use 2 cores per GPU, but they have their own libraries outside CUDA for GPU stuff, so who knows.
I haven't come across any books on CUDA other than the support manuals, but since it isn't a very mature API, it is only a matter of time.
Too bad this isn't really news. I guess it is news if you consider that someone else has had their application accelerated by NVIDIA GPUs. I guess the only other reason that this could be news is by virtue of having 8 GPU cores.
Unfortunately, this setup won't work ideally for a lot of other CUDA based applications. For the past 6 months, I had a system with 6 GPUs (actual physical GPUs). This is the system that I showed at CES. We are easily able to do 8 physical GPUs, and now I've been solely focused on utilizing Tesla.
Given that NVIDIA released the GX2 series, I was not surprised that someone would announce an 8GPU system. I'm surprised it took this long for someone to do it, and almost equally surprised that slashdot took this long to publish any news that is decent in the realm of GPU super computing. I've been cranking out close to 228 billion atom evals. per second in VMD for months now, versus about 4 billion on dual quad core 3.0GHz Xeons.
The problem with the current Folding@Home implementation is that it is comprised of about 50k lines of Brook+ code. While ideal for ATI boards because their compute abstraction layer (CAL) is implemented using Brook+, this makes it a bit difficult for porting to NVIDIA.
From what I've been hearing, while the purpose of CAL and CUDA are the same, the development environments are different enough to require the FAH guys to do a much more rigorous and in depth code modification in order to properly support CUDA. While I believe they are working on it, they are most likely stumbling upon some non-trivial roadblocks along the way.
One could only wish at this point that Rapidmind just abstract CAL and CUDA in their software... then you can support ATI and NVIDIA out of the box.
Liquid cooling can affect the energy costs in a big way depending on how well integrated the system is. As an example, CoolIT systems had developed a server rack with an integrated liquid cooling system that they had shown off at CES this year. The rack essentially used hydraulic fittings to allow you to hot-swap systems from the chassis, while still keeping the cooling centralized.
They had essentially used the radiator from a Honda Accord, which they found to be able to dissipate between 25 and 35 KW of heat. With a system like this centralizing the area where heat is dumped, fluids can be piped out to a radiator sitting outside, so essentially, a large portion of the heat produced from a rack of computers, can be relocated outside of the data center.
Even without moving the heat outside, you can still save on cooling costs. Because you have the capacity to dissipate so much heat, less AC costs are required simply because you can used a forced air system to move the gobs of hot air out and outside air in. This could potentially save up to 30% in cooling costs alone, let alone if you were to just relocate the exchanger to the exterior of the building.
One thing to keep in mind is that more interest in hydrogen vehicles is stemming from greater advancement in production technologies. Hydrogen can be easily produced today from various forms of feedstock in a plasma gasification process. The main advantage in the process today is that newer gasification processes now produce near zero emissions, and can still produce various types of fuels and chemicals.
Slashdot Mirror
It's the reason I moved from Maxtor to Seagate (aside from the fact the failure rate I was getting on Maxtor drives was nearing 100% within 3 years). Now I'm playing around more with WD, Hitachi and others since I got a 1.5TB Seagate drive that kept having lockup issues due to bad firmware. I questioned their quality control since then and have actively avoided Seagate since without regret.
I can agree and disagree with you to some extent. I disagree to the point that I use CID spoofing using Skype.
Why? I use my cell phone for business all of the time, and about 95% of the time I'm using my cell phone for business, I'm at my computer. I will generally opt to use Skype at 2.4 cents per minute instead of my cell phone at stupid long distance rates, but I want to make sure that clients know it is me calling. I have Skype spoof my cell phone CID so that when I call someone regardless if it is by phone or Skype, they know it is me, and I can get away with significantly cheaper long distance rates. Especially when I'm roaming outside of the country.
Go to an arcade where it is a popular game and you'll find out real soon.
What I am kinda hoping is that the updates fix some bugs in the drivers. I've had a number of occasions where the existing drivers caused things like BSOD and crashing. Most notably, I received a full hardware crash while using the webcam within Skype. There are some room for improvements in the driver, I just hope this update addresses them.
I've heard of some long read DNA sequences taking up to a full TB of disk space. Pneumatics would be fantastic... :D
This has been the argument that I've seen to justify getting a GeForce over a Quadro in CGI. A few points:
1) The memory system on the Tesla/Quadro is much more rigorously tested and held to a much higher standard of quality than the GeForce. There is plenty of research evidence to prove this, and I have had plenty of anecdotal evidence to prove this point as well. NVIDIA doesn't give a crap about the memory in a GeForce because a miscolored pixel in 1/60th of a second doesn't matter. A soft/hard error in GPU memory for scientific calculations can be catastrophic. This is also a reason that NVIDIA is applying ECC memory to the Tesla C2050 and C2070 GPUs.
2) Some GeForce GPUs will have major threading errors after a few minutes of hard running. I've experienced this with a dgemm torture test with a Tesla and 2 GeForce GTX 285 GPUs in a single system. Give the test about 5 minutes on all 3 GPUs simultaneously and the GeForce GPUs will crash out at nearly the same time. The Tesla will continue the test until completion (which can be about a day or so)
3) Bandwidth starvation is a term to indicate that the cards are getting less bandwidth than they should be getting. On this FASTRA machine, only a few slots are full x16 Gen2, which end up being shared across 2 GPUs, making it effectively x8 Gen2 to each GPU. For other slots, it is even worse when you have a x8 Gen2 link going in. That has to be shared between a pair of GPUs. Technically, you can run the Tesla GPU in a x1 Gen1 slot if you had the right adapter, but the time it will take to transfer data from host memory to GPU memory may end up negating any performance benefits you might see out of using the GPU, unless you are using very heavy computational algorithms that are almost completely compute bound.
I couple years ago, I had a compute rig using 6 Tesla C870 GPUs, and even that setup was starting to get bandwidth starved as all GPUs were using a single x8 Gen1 link being aggregated to 6 x4 Gen1 slots (using adapters). I had to up the output data frame size on an MD simulation in order to have all cards performing equally. With smaller frame sizes, the first 4 GPUs were finishing their computations before the last 2 GPUs got their data.
It is kind of unfair to generalize commercial clusters vs homebrew in that manner. Many institutions that purchase commercial clusters from HP/Dell/SGI/etc opt out of the use of InfiniBand or 10GbE. The logic behind it is when the vender says that for $100,000 they can upgrade to IB, the purchaser goes back and says, "For $100,000, I'll just get more cluster nodes instead." This is probably a big reason that gigabit takes up 52% of the Top500 list of supercomputers.
If this is achieved for a personal aircraft, I'd be very much on board with this. My only beef is the addition of things like parachutes and air bags. I don't really care too much for those features, as I might be able to get TKS de-icing systems installed for similar weight for those IFR flights in the great white north. Or if I don't have a TKS system, maybe a little extra payload capacity so I can actually fit 4 passengers and fuel without going over gross weight.
Because I can appreciate and judge great cuisine, doesn't mean I can make it, yet the feedback these judges provide is the cornerstone for a chefs continuous improvement. People who use and judge interfaces in the field are usually a great resource to find ways to improve it. If it is a hit to the interface designers ego that some interface element isn't where the users would like it to be, suck it up. Make it better, always improve it.
My 4 year old can beat almost anyone in a game of Mario Kart for Wii. I'm about to get him online to see how he fares against the rest of the world.
I like the comparison you make here, but in reality, it is even worse. For the most part, a $150,000 plane can barely take my family (2 adults, 2 small children) and some luggage with a full load of fuel (legally). I also burn about 22 Gallons of fuel or so on a 206nm flight.
- Support for ics calendar files in Mail.app
- More Bluetooth functions, like sending contacts to other phones via bluetooth, or being able to interact with other peripherals
- Support for unlocking functions, like many other sane GSM phones. I want to use my AT&T sim card in the US, but I'm locked to my Rogers SIM in Canada.
Uwe Boll?
There are a number of schemes I ended up using in naming systems at my workplace. There really wasn't a rhyme nor a reason to how I named our machines, I just went with what sounded cool, but it also seemed that I had a tendency of having at least 2 system names related.
For example, Excalibur and Dragoon. Genesis and Revelation. Those I guess were the only two system pair that were somewhat related. One system was named Severn simply because I recall a Redhat distro being named Severn and thought it sounded cool. Another, now dead, system was named Velocity because not only did it sound neat, it was also a reflection of the type of acoustic work it was designed to perform.
Excalibur I think was the only system that I had reason to call it that. Being one of the coolest and most sought after swords of legend, it only seemed fitting to call the most powerful workstation in the office that (which at the time was dual Opteron 250's). The next step when I ended up getting my PowerMac Quad G5 was simply to call it Titan. That name ended up succeeding into one of our product names for my HPC startup company.
Because the aeronautical information is public information doesn't necessarily mean it doesn't cost. Jeppesen is one of the primary providers of aviation databases and is used in pretty much all Garmin handheld and panel mount GPS units. I can't say that it justifies cost as much as FAA certification, but the databases alone do cost every 56 days should you choose to update them that frequently.
I don't necessarily fly that often, but I have a $1200 handheld GPS that costs about $50 to update the Jeppesen aviation database, an additional $50 to update obstacle databases, and $150 to update the terrain database (which is not anywhere near as often as the others). For IFR GPS units, I think it is about $150 to update the Jeppesen database.
NVIDIA themselves don't make the personal supercomputer. They partner with companies like ours to build and test systems, like our Slipstream S4 that can be sold and branded as a Tesla personal supercomputer. The system I linked in particular will still give you 4 Tesla C1060 GPUs, but for $8,495. Shameless plug? Sure, but it certainly isn't 10k, and is still certified by NVIDIA.
They have hit the Top500. They just recently made 29th position with 170 Tesla S1070 systems in tandem with Xeon cluster nodes. Their overall performance in LINPACK was about 77.48 TFLOPS. This installation was done by Tokyo Tech.
And about $1,800.
Tesla C1060 = $1,700
QuadroFX 5800 = $3,500
You're right about the difference pretty much being the DVI port, but it is a pretty expensive DVI port. Compute professionals didn't want the GeForce series because of lack of support, and they didn't want Quadro because it was too expensive, so the Tesla was NVIDIAs middle ground.
Trust me on this one, just because the article only has an implied no, from someone that wasn't actually the CEO, doesn't mean there wasn't an actual no somewhere. If the author of the article had decided to show up at the CEO's emerging companies summit Q&A session, the quote would have read more like, "No, absolutely not.", then append the the pretty much the jist of the senior VPs quote on the end of that. Not much else to read into it unless you think the guy is lying, rather than batting around the truth.
I haven't done much of the development of the software myself. We have a developer we hired to work with CUDA. From what I've found, the documentation that is available on NVIDIAs site for CUDA is excellent and their developers are active on their forums.
For VMD, it was necessary to have 1 CPU core per GPU. We tested 6 GPUs with 4 cores and we could only spawn 4 threads for GPU processing. The guys at Evolved Machines told me they can use multi GPU off of a single core. If so, I have no idea how. NVIDIA even tells me 1 core per GPU, so that is the gospel I'm following by. Acceleware for some of their stuff even use 2 cores per GPU, but they have their own libraries outside CUDA for GPU stuff, so who knows.
I haven't come across any books on CUDA other than the support manuals, but since it isn't a very mature API, it is only a matter of time.
Too bad this isn't really news. I guess it is news if you consider that someone else has had their application accelerated by NVIDIA GPUs. I guess the only other reason that this could be news is by virtue of having 8 GPU cores.
Unfortunately, this setup won't work ideally for a lot of other CUDA based applications. For the past 6 months, I had a system with 6 GPUs (actual physical GPUs). This is the system that I showed at CES. We are easily able to do 8 physical GPUs, and now I've been solely focused on utilizing Tesla.
Given that NVIDIA released the GX2 series, I was not surprised that someone would announce an 8GPU system. I'm surprised it took this long for someone to do it, and almost equally surprised that slashdot took this long to publish any news that is decent in the realm of GPU super computing. I've been cranking out close to 228 billion atom evals. per second in VMD for months now, versus about 4 billion on dual quad core 3.0GHz Xeons.
The problem with the current Folding@Home implementation is that it is comprised of about 50k lines of Brook+ code. While ideal for ATI boards because their compute abstraction layer (CAL) is implemented using Brook+, this makes it a bit difficult for porting to NVIDIA.
From what I've been hearing, while the purpose of CAL and CUDA are the same, the development environments are different enough to require the FAH guys to do a much more rigorous and in depth code modification in order to properly support CUDA. While I believe they are working on it, they are most likely stumbling upon some non-trivial roadblocks along the way.
One could only wish at this point that Rapidmind just abstract CAL and CUDA in their software... then you can support ATI and NVIDIA out of the box.
Liquid cooling can affect the energy costs in a big way depending on how well integrated the system is. As an example, CoolIT systems had developed a server rack with an integrated liquid cooling system that they had shown off at CES this year. The rack essentially used hydraulic fittings to allow you to hot-swap systems from the chassis, while still keeping the cooling centralized.
They had essentially used the radiator from a Honda Accord, which they found to be able to dissipate between 25 and 35 KW of heat. With a system like this centralizing the area where heat is dumped, fluids can be piped out to a radiator sitting outside, so essentially, a large portion of the heat produced from a rack of computers, can be relocated outside of the data center.
Even without moving the heat outside, you can still save on cooling costs. Because you have the capacity to dissipate so much heat, less AC costs are required simply because you can used a forced air system to move the gobs of hot air out and outside air in. This could potentially save up to 30% in cooling costs alone, let alone if you were to just relocate the exchanger to the exterior of the building.
Just another little nit pick... its 29.27 FPS.
One thing to keep in mind is that more interest in hydrogen vehicles is stemming from greater advancement in production technologies. Hydrogen can be easily produced today from various forms of feedstock in a plasma gasification process. The main advantage in the process today is that newer gasification processes now produce near zero emissions, and can still produce various types of fuels and chemicals.